Datasets:

acmc

/

beamit-full-texts-dataset

like 0

[ "Algorithms", "Amyloid Precursor Protein Secretases", "Amyloid beta-Protein Precursor", "Aspartic Acid Endopeptidases", "Computational Biology", "Drug Design", "Enzyme Inhibitors", "Humans", "Models, Biological", "Models, Chemical", "Structure-Activity Relationship" ]

[ "Background", "Dataset collection", "Diverse conformation generation", "Pharmacophore modeling", "The HypoGen algorithm", "Data analysis", "Pharmacophore validation", "Test set method", "Fischer randomization method", "Decoy set method", "Leave-one-out method", "Database screening", "Molecular docking", "Results and discussion", "Pharmacophore generation", "Statistical data analysis", "Activity prediction and mapping of training set compound on Hypo1", "Validation of Hypo 1", "Test set method", "Fisher randomization method", "Decoy set method", "Leave-one-out method", "Database screening", "Molecular docking", "Conclusion", "Competing interests", "Author’s contributions" ]

[ "Beta-site amyloid precursor protein cleaving enzyme (BACE-1), also known as β-secretase, memapsin-2, or Aspartyl protease-2, is a single-membrane protein belongs to the aspartyl protease class of catabolic enzyme. This is one of the enzymes responsible for the sequential proteolysis of amyloid precursor protein (APP) [1]. The cleavage of APP by BACE-1, which is the rate-limiting step in the amyloid cascade, results in the generation of two peptide fragments Aβ40 and Aβ42. Among two peptide fragments, Aβ42 is the primary species and thought to be causal for the neurotoxicity and amyloid plaque formation that lead to memory and cognitive defects in Alzheimer’s disease (AD) [2]. The AD is a debilitating neurodegenerative disease that results in the irreversible loss of neurons, particularly in the cortex and hippocampus [3]. It is characterized by progressive decline in cognitive function that inevitably leading to incapacitation and death. It also histopathologically characterized by the presence of amyloid plaques and neurofibrillar tangles in the brain. Regardless of the increasing demand for medication, no truly disease-modifying treatment is currently available [4,5]. The BACE knockout study in mice shows a complete absence of Aβ production with no reported side effects [6-8]. Since gene knockout study showed a reduction in AD-like pathology, inhibition of BACE-1 the key enzyme in the production of Aβ peptide has emerged as an attractive therapeutic target for AD [9]. Therefore extensive efforts have been followed in the discovery of potential inhibitors of BACE-1. Most of the designing of BACE-1 inhibitors are based on the transition state mimetic approach, which depends mainly on replacing the scissile amide bond of an appropriate substrate with a stable mimetic of the putative transition-state structure [10].\nThe main aim of our approach, which is discussed in this study is different than the transition state mimetic approach, is to develop an accurate and efficient method for discovering potent BACE-1 inhibitors. A pharmacophore hypothesis was generated based on key structural features of compounds with BACE-1 inhibitory activity. It provides a rational hypothetical representation of the most important chemical features responsible for activity. Herein, a ligand-based 3D pharmacophore hypothesis for BACE-1 inhibitors was constructed based on the structure-activity relationship observed in a set of known BACE-1 inhibitors. The resulted pharmacophore hypotheses were validated by test set, Fischer randomization, leave-one-out, and decoy set methods. The validated pharmacophore hypothesis has been used in in silico screening to identify hits that are highly varied in chemical nature. The retrieved hits were subsequently subjected to a well-defined refining procedure based on estimated activity values, drug-likeness prediction and further by molecular docking study. The identified hits can further be utilized in designing novel and potent BACE-1 inhibitors.", "In a computerized pharmacophore generation process the accurate choice of the training set is a key issue. The built pharmacophore hypothesis can be as good as the input data information. The following criteria should be considered during the selection of data set in order to achieve a significant pharmacophore hypothesis. (1) All compounds used in the training set have to bind to the same receptor in roughly the same fashion. Compounds having more binding interaction with the receptor are more active than those with fewer; (2) the data set must be widely populated covering an activity range of at least 4 orders of magnitude; (3) the most active compounds should inevitably be included in the training set and (4) all biologically relevant data should be obtained by homogenous procedures [11]. Every individual feature in the resulting hypotheses will invade a certain weight that is proportional to its relative contribution to biological activity.\nBy taking these criteria into account, we have collected a total number of 320 BACE-1 inhibitors from various literature resources [12-25] and a database has been created. The 2D structure of the compounds were built using ChemSketch program version 12, and subsequently converted in to 3D structures using Discovery studio 2.5 (DS) [26]. In the next step, 60 compounds were selected as final dataset as the BACE-1 inhibitory activities of these 60 compounds were studied under same biological assay condition. Based on the principle of structural diversity and wide coverage of activity range, 20 compounds were carefully selected as training set compounds and the rest were used as test set in model validation. Here, the IC50 value of the training set compounds was taken into account, the inhibitory activity values of the training set compounds span over a range of four orders of magnitude, from 4 nM to 37 000 nM. The chemical structure and experimental activity of the training set compounds are shown in Figure 1.\nStructure of training set compounds. The 2D chemical structures of 20 compounds of the training set together with their experimental IC50 values in nM.", "Prior to the generation of pharmacophore hypotheses, the training set compounds, which were converted to 3D structure, were used to generate diverse conformations. Diverse Conformation Generation protocol implemented in DS was used to generate conformations using the Best conformation model generation method with CHARMM force field and Poling algorithm to ensure the energy-minimized conformation for each compound. The parameters like maximum number of 250 conformers, the ‘best conformational analysis’ method, and an energy threshold of 20 kcal/mol above the global energy minimum were chosen during conformation generation.", "The training set comprises of 20 compounds was used in pharmacophore hypothesis generation. The HypoGen algorithm available in 3D QSAR Pharmacophore Generation protocol of DS tries to generate hypotheses with features common amongst active molecules and do not reflect the inactive molecules of the training set. The training set compounds were predicted for their inherent chemical features using Feature Mapping protocol implemented in DS. During pharmacophore hypothesis generation a minimum of 4 and a maximum of 5 pharmacophoric features like hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), positive ionizable (PI), ring aromatic (RA) and hydrophobic (HY) were included. These features were selected based on the feature mapping results. All parameters were set to their default values except uncertainty value, which has been changed to 2 instead of 3. An uncertainty value of 2 was more convenient for our dataset because the activity values of the training set spanned exactly the required 4 orders of magnitude; this choice has been further confirmed by preliminary calculations and by other literature evidence [27]. The uncertainty value represents the ratio of the uncertainty range of the actual activity against measured biology activity for each compound.", "With the full range of training set compounds from active to inactive the pharmacophore hypotheses were generated by HypoGen algorithm implemented in DS. This algorithm constructs and ranks the pharmacophore hypotheses that correlate best between 3D spatial arrangement of features in a given training set compounds and their respective experimental activities. This process is accomplished in three steps: the constructive phase, the subtractive phase and the optimization phase [28].\nThe constructive phase identifies the hypotheses that are common amongst the active compounds. HypoGen enumerates all possible pharmacophore configurations using all combinations of pharmacophore features for each of the conformations of the most active compound. In order to consider the left over most active compounds the hypotheses must fit a minimum subset of its features. Hence, a large database of pharmacophore configurations will be generated at the end of the constructive phase.\nThe subtractive phase will remove the pharmacophore configurations that are present in the least active compounds. All compounds whose activity is by default 3.5 orders of magnitude less than that of the most active compound are considered to represent the least active compounds. The value 3.5 is adjustable depending on the activity range of the training set. The optimization phase improves the hypothesis score. These scores of the generated hypotheses depend on the errors in activity estimation from regression and complexity. The optimization involves a variation of features and/or locations to optimize activity prediction via a simulated annealing approach. The total cost parameter will be calculated for every new hypothesis. The HypoGen will quit and reports the 10 top-scoring hypotheses when there is no improvement in the hypothesis score.", "The quality of a pharmacophore hypothesis is best determined by two theoretical cost calculations, which are represented in bit units [29]. One is the ‘fixed cost’ also known as cost of an ideal hypothesis, which represents the simplest model that fits all the data perfectly. The second cost is the ‘null cost’, which represents the highest cost of a pharmacophore with no features that estimates every activity to be the average of the activity data of the training set compounds.\nThe total cost of any pharmacophore hypothesis should always be close to the fixed cost and away from the null cost to be the significant model. The cost difference between fixed and null cost values should be larger for a meaningful pharmacophore hypothesis. A value of 40-60 bits in a pharmacophore hypothesis indicates that it has 75-90% probability of representing a true correlation in the data.\nThe hypotheses are also evaluated based on other cost components. The cost value for every individual hypothesis is the summation of three cost components: the error cost (E), the weight cost (W) and the configuration cost (C). The error cost is the value represents the root-mean-squared difference (RMSD) between experimental and estimated activity value of the training set compounds. The weight cost is a value that increases in a Gaussian form as this function weights in a model deviate from the ideal value of two. The configuration cost or entropy cost measures the entropy of the hypothesis space. If the input training set compounds are too multiplex, e.g. because of too flexible training set compounds, this will result in an effusive number of hypotheses as an outcome of the subtractive phase. This configuration cost should always be less than a maximum value of 17 [30]. The correlation coefficient of the pharmacophore hypothesis should be close to 1.", "The generated pharmacophore hypothesis was validated using test set, Fischer randomization, decoy set and leave-one-out methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.\nA total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.\n[SUBTITLE] Fischer randomization method [SUBSECTION] The main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.\nThe main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.\n[SUBTITLE] Decoy set method [SUBSECTION] An external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.\nAn external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].\nThe pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].", "A total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.", "The main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.", "An external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.", "The pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].", "The validated pharmacophore hypothesis, Hypo 1, was used as a 3D query for screening four different chemical databases. The purpose of this screening is to retrieve novel and potential leads suitable for further development. The chemical databases used were Maybridge, Chembridge, NCI and Asinex. Conformers were generated for each molecule in the database using best conformer generation method that allows a maximum energy of 15 kcal/mol above that of the most stable conformation. The database screening was carried out using Ligand Pharmacophore Mapping protocol implemented in DS with Best/Flexible Search option. The retrieved compounds were filtered by restricting the estimated activity value less than 100 nM and the obtained compounds were further refined using molecular docking study.", "Pharmacophore modeling normally firmly associated with docking procedure, which in a first step flexibly aligns the ligand molecule into a rigid macromolecule environment and then estimates the tightness of the interaction by different scoring functions [35]. The Docking takes all the information from a rigid protein environment and scores several possible interaction modes for different alignments. There are many docking programs available for molecular docking studies. In this study, we used GOLD (Genetic Optimisation for Ligand Docking), a docking program [36] that uses genetic algorithm for docking and performs automated docking with full acyclic ligand flexibility, partial cyclic ligand flexibility and partial protein flexibility in the neighborhood of the protein active site. The crystal structure of BACE-1 complexed with an inhibitor SC7 (PDB ID: 2QP8) was used in molecular docking studies. The inhibitor SC7 was extracted from the active site and the retrieved database hits were docked based on the ligand SC7 coordinates, in to the active site of BACE-1. The water molecules were removed prior to docking because they were not found to play any important roles in BACE1-ligand interaction. The early termination option parameter in GOLD was changed from 3 to 5 and the maximum save conformations was set to 10. All the other parameters were set at their default values.", "[SUBTITLE] Pharmacophore generation [SUBSECTION] We have used the HypoGen algorithm implemented in DS in order to quantitatively correlate the chemical structure of BACE-1 inhibitors to their biological activity. The training set of 20 compounds (Figure 1) with activity values ranging from 4 to 37000 nM was used in pharmacophore model generation. The Feature Mapping protocol resulted in HBA, HBD, RA, PI and HY features. Selecting these features, the pharmacophore generation run was performed along with diverse conformers of training set molecules generated as described in methods section. Ten top-scored pharmacophore hypotheses were generated and in order to choose the best one and also to give an idea about the statistical significance, the pharmacophore hypotheses were subjected to cost analysis. The results of top ten pharmacophore hypotheses and their statistical parameters are given in Table 1. In this study, the first pharmacophore hypothesis (Hypo 1) is the best hypothesis characterized by the large cost difference (121.98 bits), lowest RMSD value (0.804) and a high correlation coefficient of 0.977. All ten hypotheses consist of HBD, PI, RA and HY features. Nine of ten hypotheses were composed of five pharmacophoric features except only one hypothesis, which was of four features. The best pharmacophore hypothesis (Hypo 1), which scored the large cost difference, lowest RMSD, lowest error cost and high correlation coefficient, was made of one HBD, one PI, one RA and two HY features.\nResults of the top 10 pharmacophore hypotheses generated by the HypoGen algorithm\nNull cost = 203.22; fixed cost = 74.77; configuration cost = 15.59.\naCost difference = null cost – total cost.\nbHBD, hydrogen-bond donor; PI, positive ionizable; RA, ring aromatic; HY, hydrophobic.\nWe have used the HypoGen algorithm implemented in DS in order to quantitatively correlate the chemical structure of BACE-1 inhibitors to their biological activity. The training set of 20 compounds (Figure 1) with activity values ranging from 4 to 37000 nM was used in pharmacophore model generation. The Feature Mapping protocol resulted in HBA, HBD, RA, PI and HY features. Selecting these features, the pharmacophore generation run was performed along with diverse conformers of training set molecules generated as described in methods section. Ten top-scored pharmacophore hypotheses were generated and in order to choose the best one and also to give an idea about the statistical significance, the pharmacophore hypotheses were subjected to cost analysis. The results of top ten pharmacophore hypotheses and their statistical parameters are given in Table 1. In this study, the first pharmacophore hypothesis (Hypo 1) is the best hypothesis characterized by the large cost difference (121.98 bits), lowest RMSD value (0.804) and a high correlation coefficient of 0.977. All ten hypotheses consist of HBD, PI, RA and HY features. Nine of ten hypotheses were composed of five pharmacophoric features except only one hypothesis, which was of four features. The best pharmacophore hypothesis (Hypo 1), which scored the large cost difference, lowest RMSD, lowest error cost and high correlation coefficient, was made of one HBD, one PI, one RA and two HY features.\nResults of the top 10 pharmacophore hypotheses generated by the HypoGen algorithm\nNull cost = 203.22; fixed cost = 74.77; configuration cost = 15.59.\naCost difference = null cost – total cost.\nbHBD, hydrogen-bond donor; PI, positive ionizable; RA, ring aromatic; HY, hydrophobic.\n[SUBTITLE] Statistical data analysis [SUBSECTION] The generated hypotheses were subjected to cost analysis. The two main values used for cost analysis are the difference between fixed and null cost and another is the difference between the null cost and the total cost (Δcost). The fixed cost of the run was 74.77 bits, which was well separated from the null cost of 203.22 bits and close to the total cost of 81.24 bits. The large difference (128.45 bits) observed between the fixed cost and null cost value indicates that Hypo 1 has more than 90% statistical significance to be a significant model. All the 10 hypotheses were subjected to further evaluation for their capability to predict the activity of the training set compounds. Configuration cost or entropy value must be less than 17 for which a value of 15.59 was obtained in this study. All hypotheses have scored RMSD values lower than 1.5 and ranging from 0.804 to 1.111, this characterization further emphasizing the good predictive quality of these hypotheses. Based on the rule to select a hypothesis with a lowest total cost, high correlation coefficient, large cost difference and significantly low RMSD value, Hypo 1 gave the best statistical values among other hypotheses. Hence, Hypo 1 with one HBD, one PI, one RA and two HY was chosen as the best hypothesis for further analysis. The inter-feature distance constraints were observed for this five-featured pharmacophore hypothesis (Hypo 1) (Figure 2).\nHypoGen pharmacophore hypothesis for BACE-1 inhibitors. A) The best five feature pharmacophore model Hypo 1 B) 3D spatial arrangement and the distance constraints of the Hypo 1. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.\nThe generated hypotheses were subjected to cost analysis. The two main values used for cost analysis are the difference between fixed and null cost and another is the difference between the null cost and the total cost (Δcost). The fixed cost of the run was 74.77 bits, which was well separated from the null cost of 203.22 bits and close to the total cost of 81.24 bits. The large difference (128.45 bits) observed between the fixed cost and null cost value indicates that Hypo 1 has more than 90% statistical significance to be a significant model. All the 10 hypotheses were subjected to further evaluation for their capability to predict the activity of the training set compounds. Configuration cost or entropy value must be less than 17 for which a value of 15.59 was obtained in this study. All hypotheses have scored RMSD values lower than 1.5 and ranging from 0.804 to 1.111, this characterization further emphasizing the good predictive quality of these hypotheses. Based on the rule to select a hypothesis with a lowest total cost, high correlation coefficient, large cost difference and significantly low RMSD value, Hypo 1 gave the best statistical values among other hypotheses. Hence, Hypo 1 with one HBD, one PI, one RA and two HY was chosen as the best hypothesis for further analysis. The inter-feature distance constraints were observed for this five-featured pharmacophore hypothesis (Hypo 1) (Figure 2).\nHypoGen pharmacophore hypothesis for BACE-1 inhibitors. A) The best five feature pharmacophore model Hypo 1 B) 3D spatial arrangement and the distance constraints of the Hypo 1. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.\n[SUBTITLE] Activity prediction and mapping of training set compound on Hypo1 [SUBSECTION] To verify the predictive ability of Hypo1 on training set compounds, the activity of each training set compound is estimated by regression analysis. The experimental activities of training set compounds were classified into four groups: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). The estimated activity values of training set compounds based on Hypo 1 and the corresponding error values are calculated (Table 2). The error value is the ratio between the estimated and experimental activities. The positive error value indicates that the estimated IC50 value is higher than the experimental activity, whereas the negative error value indicates that the estimated IC50 value is lower than the experimental activity. An error value of less than 10 signifies the prediction of activity lesser than one order of magnitude. Among 20 training set compounds, only one compound had an error value of greater than 3. From Table 2 it is clear that the estimated activity values of most of the training set compounds was predicted with the same activity scale as the experimental activity. Among 20 training set compounds, one most active compound (++++) was estimated as active (+++), one active compound (+++) was estimated as moderately active (++), one moderately active compound (++) was estimated as inactive (+) and two inactive compounds (+) were estimated as moderately active (++). The divergence between the estimated and experimental activity observed in four compounds was only about 1 order of magnitude, which might be an artifact of the program that uses different number of degrees of freedom for these compounds to mismatch the pharmacophore model. Interestingly, for feature fitting, the most active compounds in the training set mapped well on all the chemical features that are one HBD, one PI, one RA and two HY features of Hypo 1 with good fitting score. The active, moderately and inactive compounds have missed at least one of five features. In addition, the most active compounds mapped well on the RA and PI features whereas some active, moderately active and all inactive compounds could not map on the RA and PI features signifying the importance of these two features. The pharmacophore overlay of most active Compound 1 and Hypo 1 has shown a fit value of 9.52. The RA feature corresponds to phenyl ring present in between two amide and a sulfonamide groups, one HBD feature corresponds to nitrogen of amide group located at the branch, PI group corresponds to the only amino nitrogen, one HY feature corresponds to a phenyl group whereas the another HY feature corresponds to alkyl group (Figure 3A). The pharmacophore overlay of least active compound 20 has revealed that it missed two features when mapped on Hypo 1 with a fit value of 5.16. This compound has mapped only the HBD and two HY features in the same manner as most active compound with no mapping over RA and PI features (Figure 3B). Fit value indicates how well the features in the pharmacophore overlaps the chemical features in the molecule and thereby aid in understanding the chemical meaning of the hypothesis [37]. These results emphasized Hypo 1 as a reliable model to accurately estimate the experimental activity of the training set compounds.\nExperimental and estimated IC50 values of the training set compounds based on the pharmacophore hypothesis ‘Hypo 1’.\naPositive value indicates that the estimated IC50 is higher than the experimental IC50; negative value indicates that the estimated IC50 is lower than the experimental IC50.\nbFit value indicates how well the features in the pharmacophore map the chemical features in the compound.\ncActivity scale: most active, ++++, IC50 ≤ 100 nM; active, +++, 100 nM < IC50 ≤ 1000 nM; moderately active, ++, 1000 nM < IC50 ≤ 10,000 nM; inactive, +, IC50 > 10,000 nM.\nPharmacophore mapping of most and least active compounds in the training set. A) Hypo 1 mapped on to the most active Compound 1 B) Hypo 1 mapped on to the least active Compound 20. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.\nTo verify the predictive ability of Hypo1 on training set compounds, the activity of each training set compound is estimated by regression analysis. The experimental activities of training set compounds were classified into four groups: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). The estimated activity values of training set compounds based on Hypo 1 and the corresponding error values are calculated (Table 2). The error value is the ratio between the estimated and experimental activities. The positive error value indicates that the estimated IC50 value is higher than the experimental activity, whereas the negative error value indicates that the estimated IC50 value is lower than the experimental activity. An error value of less than 10 signifies the prediction of activity lesser than one order of magnitude. Among 20 training set compounds, only one compound had an error value of greater than 3. From Table 2 it is clear that the estimated activity values of most of the training set compounds was predicted with the same activity scale as the experimental activity. Among 20 training set compounds, one most active compound (++++) was estimated as active (+++), one active compound (+++) was estimated as moderately active (++), one moderately active compound (++) was estimated as inactive (+) and two inactive compounds (+) were estimated as moderately active (++). The divergence between the estimated and experimental activity observed in four compounds was only about 1 order of magnitude, which might be an artifact of the program that uses different number of degrees of freedom for these compounds to mismatch the pharmacophore model. Interestingly, for feature fitting, the most active compounds in the training set mapped well on all the chemical features that are one HBD, one PI, one RA and two HY features of Hypo 1 with good fitting score. The active, moderately and inactive compounds have missed at least one of five features. In addition, the most active compounds mapped well on the RA and PI features whereas some active, moderately active and all inactive compounds could not map on the RA and PI features signifying the importance of these two features. The pharmacophore overlay of most active Compound 1 and Hypo 1 has shown a fit value of 9.52. The RA feature corresponds to phenyl ring present in between two amide and a sulfonamide groups, one HBD feature corresponds to nitrogen of amide group located at the branch, PI group corresponds to the only amino nitrogen, one HY feature corresponds to a phenyl group whereas the another HY feature corresponds to alkyl group (Figure 3A). The pharmacophore overlay of least active compound 20 has revealed that it missed two features when mapped on Hypo 1 with a fit value of 5.16. This compound has mapped only the HBD and two HY features in the same manner as most active compound with no mapping over RA and PI features (Figure 3B). Fit value indicates how well the features in the pharmacophore overlaps the chemical features in the molecule and thereby aid in understanding the chemical meaning of the hypothesis [37]. These results emphasized Hypo 1 as a reliable model to accurately estimate the experimental activity of the training set compounds.\nExperimental and estimated IC50 values of the training set compounds based on the pharmacophore hypothesis ‘Hypo 1’.\naPositive value indicates that the estimated IC50 is higher than the experimental IC50; negative value indicates that the estimated IC50 is lower than the experimental IC50.\nbFit value indicates how well the features in the pharmacophore map the chemical features in the compound.\ncActivity scale: most active, ++++, IC50 ≤ 100 nM; active, +++, 100 nM < IC50 ≤ 1000 nM; moderately active, ++, 1000 nM < IC50 ≤ 10,000 nM; inactive, +, IC50 > 10,000 nM.\nPharmacophore mapping of most and least active compounds in the training set. A) Hypo 1 mapped on to the most active Compound 1 B) Hypo 1 mapped on to the least active Compound 20. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.\n[SUBTITLE] Validation of Hypo 1 [SUBSECTION] Hypo 1 was further validated by test set, Fischer randomization test, leave-one-out and decoy set methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\nA total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\n[SUBTITLE] Fisher randomization method [SUBSECTION] This method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\nThis method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\n[SUBTITLE] Decoy set method [SUBSECTION] The Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\nThe Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\nThe cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\nHypo 1 was further validated by test set, Fischer randomization test, leave-one-out and decoy set methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\nA total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\n[SUBTITLE] Fisher randomization method [SUBSECTION] This method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\nThis method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\n[SUBTITLE] Decoy set method [SUBSECTION] The Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\nThe Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\nThe cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\n[SUBTITLE] Database screening [SUBSECTION] The validated pharmacophore hypothesis, Hypo1, was used as a 3D structural query for retrieving compounds from chemical databases including MayBridge (59 652 compounds), Chembridge (50 000 compounds), NCI (238 819 compounds) and Asinex (213 462 compounds). As a result of first screening 11 578, 590, 5096 and 63 265 compounds were retrieved from Maybridge, Chembridge, NCI and Asinex respectively. Since the active site of BACE-1 is larger in size, the experimentally known most active inhibitors are also larger in size and violate the first rule of Lipinski’s rule of five. Hence, the retrieved hit compounds were filtered based only on the estimated activity values calculated by Hypo 1. The activity range for the most active compounds is <100 nM. Finally 773 compounds were selected by restricting the minimum estimated activity to <100 nM.\nThe validated pharmacophore hypothesis, Hypo1, was used as a 3D structural query for retrieving compounds from chemical databases including MayBridge (59 652 compounds), Chembridge (50 000 compounds), NCI (238 819 compounds) and Asinex (213 462 compounds). As a result of first screening 11 578, 590, 5096 and 63 265 compounds were retrieved from Maybridge, Chembridge, NCI and Asinex respectively. Since the active site of BACE-1 is larger in size, the experimentally known most active inhibitors are also larger in size and violate the first rule of Lipinski’s rule of five. Hence, the retrieved hit compounds were filtered based only on the estimated activity values calculated by Hypo 1. The activity range for the most active compounds is <100 nM. Finally 773 compounds were selected by restricting the minimum estimated activity to <100 nM.\n[SUBTITLE] Molecular docking [SUBSECTION] To further refine the retrieved hits and also to remove the false positives, these 773 compounds along with the 20 training set compounds were docked into the active site of BACE-1 using GOLD 4.1 program. There are number of crystal structures for BACE-ligand complexes are available in PDB. The crystal structure of BACE-SC7 (PDB ID: 2QP8) complex was taken based on its high resolution. The GOLD fitness score was calculated for all the 793 compounds, it distinguishes molecules based on their interacting ability. The GOLD fitness score for the most active compound in the training set was 53.035. The compounds for further analysis were selected based on the ligand conformations which can satisfy the necessary interactions at the active site and scoring GOLD fitness score greater than 60. Finally 20 compounds from Maybridge and 15 compounds from Asinex have shown the required interaction with BACE-1 as well as good GOLD fitness scores. The compounds with the same chemical scaffolds were filtered carefully based on the molecular interactions observed at the active site. Finally, two compounds with different scaffolds one from Maybridge (RJC01726) and one from Asinex (Asnx-2) were selected as representative compounds. The binding mode of the final hits and the most active Compound 1 in the training set are shown in Figure 5. Figure 5A represents the binding mode of Compound 1 with a GOLD fitness score of 53.035. It has formed hydrogen bond interactions with D93, G95, T133, Q134, G291 and T293 and hydrophobic interactions with Y132, F169, and T292. The GOLD fitness score of RJC01726 was 68.289 and the mode of binding in the active site (Figure 5B) is similar to Compound 1. It has formed hydrogen bond interactions with T133, Q134 and T293 and hydrophobic interactions with D93, G95, F169 and T292. Asnx-2 has shown hydrogen bond interactions with T133, G291 and T293 as well as hydrophobic interactions with D93, Y132, F169 and T292 with a GOLD fitness score of 62.026 (Figure 5C). Figure 5D represents the overlay of most active Compound 1, RJC01726 and Asnx-2 at their binding modes. The pharmacophore overlay of the final hits compounds are shown in Figure 6 and their 2D representations are shown in Figure 7.\nMolecular docking results. Binding orientations of A) Compound 1 of the training set (cyan color), B) RJC01726 (magenta color), C) Asnx-2 (blue color), D) overlay of Compound 1, RJC01726 and Asnx-2 in the active site of BACE-1 protein. Active site residues are shown in stick form and hydrogen bond interactions are indicated with purple dotted lines.\nPharmacophore overlay on final hits. The mapping of pharmacophore hypothesis Hypo 1 on the final hits. A) RJC01726 (red color) B) Asnx-2 (cyan color).\nChemical structure of final hits. 2D representation of the final hits RJC01726 and Asnx-2.\nThe hydrophobic interactions of the final hits compounds were observed using Ligplot program [39]. The novelty of the two hits compounds were confirmed using SciFinder search [40] and PubChem search [41].\nTo further refine the retrieved hits and also to remove the false positives, these 773 compounds along with the 20 training set compounds were docked into the active site of BACE-1 using GOLD 4.1 program. There are number of crystal structures for BACE-ligand complexes are available in PDB. The crystal structure of BACE-SC7 (PDB ID: 2QP8) complex was taken based on its high resolution. The GOLD fitness score was calculated for all the 793 compounds, it distinguishes molecules based on their interacting ability. The GOLD fitness score for the most active compound in the training set was 53.035. The compounds for further analysis were selected based on the ligand conformations which can satisfy the necessary interactions at the active site and scoring GOLD fitness score greater than 60. Finally 20 compounds from Maybridge and 15 compounds from Asinex have shown the required interaction with BACE-1 as well as good GOLD fitness scores. The compounds with the same chemical scaffolds were filtered carefully based on the molecular interactions observed at the active site. Finally, two compounds with different scaffolds one from Maybridge (RJC01726) and one from Asinex (Asnx-2) were selected as representative compounds. The binding mode of the final hits and the most active Compound 1 in the training set are shown in Figure 5. Figure 5A represents the binding mode of Compound 1 with a GOLD fitness score of 53.035. It has formed hydrogen bond interactions with D93, G95, T133, Q134, G291 and T293 and hydrophobic interactions with Y132, F169, and T292. The GOLD fitness score of RJC01726 was 68.289 and the mode of binding in the active site (Figure 5B) is similar to Compound 1. It has formed hydrogen bond interactions with T133, Q134 and T293 and hydrophobic interactions with D93, G95, F169 and T292. Asnx-2 has shown hydrogen bond interactions with T133, G291 and T293 as well as hydrophobic interactions with D93, Y132, F169 and T292 with a GOLD fitness score of 62.026 (Figure 5C). Figure 5D represents the overlay of most active Compound 1, RJC01726 and Asnx-2 at their binding modes. The pharmacophore overlay of the final hits compounds are shown in Figure 6 and their 2D representations are shown in Figure 7.\nMolecular docking results. Binding orientations of A) Compound 1 of the training set (cyan color), B) RJC01726 (magenta color), C) Asnx-2 (blue color), D) overlay of Compound 1, RJC01726 and Asnx-2 in the active site of BACE-1 protein. Active site residues are shown in stick form and hydrogen bond interactions are indicated with purple dotted lines.\nPharmacophore overlay on final hits. The mapping of pharmacophore hypothesis Hypo 1 on the final hits. A) RJC01726 (red color) B) Asnx-2 (cyan color).\nChemical structure of final hits. 2D representation of the final hits RJC01726 and Asnx-2.\nThe hydrophobic interactions of the final hits compounds were observed using Ligplot program [39]. The novelty of the two hits compounds were confirmed using SciFinder search [40] and PubChem search [41].", "We have used the HypoGen algorithm implemented in DS in order to quantitatively correlate the chemical structure of BACE-1 inhibitors to their biological activity. The training set of 20 compounds (Figure 1) with activity values ranging from 4 to 37000 nM was used in pharmacophore model generation. The Feature Mapping protocol resulted in HBA, HBD, RA, PI and HY features. Selecting these features, the pharmacophore generation run was performed along with diverse conformers of training set molecules generated as described in methods section. Ten top-scored pharmacophore hypotheses were generated and in order to choose the best one and also to give an idea about the statistical significance, the pharmacophore hypotheses were subjected to cost analysis. The results of top ten pharmacophore hypotheses and their statistical parameters are given in Table 1. In this study, the first pharmacophore hypothesis (Hypo 1) is the best hypothesis characterized by the large cost difference (121.98 bits), lowest RMSD value (0.804) and a high correlation coefficient of 0.977. All ten hypotheses consist of HBD, PI, RA and HY features. Nine of ten hypotheses were composed of five pharmacophoric features except only one hypothesis, which was of four features. The best pharmacophore hypothesis (Hypo 1), which scored the large cost difference, lowest RMSD, lowest error cost and high correlation coefficient, was made of one HBD, one PI, one RA and two HY features.\nResults of the top 10 pharmacophore hypotheses generated by the HypoGen algorithm\nNull cost = 203.22; fixed cost = 74.77; configuration cost = 15.59.\naCost difference = null cost – total cost.\nbHBD, hydrogen-bond donor; PI, positive ionizable; RA, ring aromatic; HY, hydrophobic.", "The generated hypotheses were subjected to cost analysis. The two main values used for cost analysis are the difference between fixed and null cost and another is the difference between the null cost and the total cost (Δcost). The fixed cost of the run was 74.77 bits, which was well separated from the null cost of 203.22 bits and close to the total cost of 81.24 bits. The large difference (128.45 bits) observed between the fixed cost and null cost value indicates that Hypo 1 has more than 90% statistical significance to be a significant model. All the 10 hypotheses were subjected to further evaluation for their capability to predict the activity of the training set compounds. Configuration cost or entropy value must be less than 17 for which a value of 15.59 was obtained in this study. All hypotheses have scored RMSD values lower than 1.5 and ranging from 0.804 to 1.111, this characterization further emphasizing the good predictive quality of these hypotheses. Based on the rule to select a hypothesis with a lowest total cost, high correlation coefficient, large cost difference and significantly low RMSD value, Hypo 1 gave the best statistical values among other hypotheses. Hence, Hypo 1 with one HBD, one PI, one RA and two HY was chosen as the best hypothesis for further analysis. The inter-feature distance constraints were observed for this five-featured pharmacophore hypothesis (Hypo 1) (Figure 2).\nHypoGen pharmacophore hypothesis for BACE-1 inhibitors. A) The best five feature pharmacophore model Hypo 1 B) 3D spatial arrangement and the distance constraints of the Hypo 1. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.", "To verify the predictive ability of Hypo1 on training set compounds, the activity of each training set compound is estimated by regression analysis. The experimental activities of training set compounds were classified into four groups: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). The estimated activity values of training set compounds based on Hypo 1 and the corresponding error values are calculated (Table 2). The error value is the ratio between the estimated and experimental activities. The positive error value indicates that the estimated IC50 value is higher than the experimental activity, whereas the negative error value indicates that the estimated IC50 value is lower than the experimental activity. An error value of less than 10 signifies the prediction of activity lesser than one order of magnitude. Among 20 training set compounds, only one compound had an error value of greater than 3. From Table 2 it is clear that the estimated activity values of most of the training set compounds was predicted with the same activity scale as the experimental activity. Among 20 training set compounds, one most active compound (++++) was estimated as active (+++), one active compound (+++) was estimated as moderately active (++), one moderately active compound (++) was estimated as inactive (+) and two inactive compounds (+) were estimated as moderately active (++). The divergence between the estimated and experimental activity observed in four compounds was only about 1 order of magnitude, which might be an artifact of the program that uses different number of degrees of freedom for these compounds to mismatch the pharmacophore model. Interestingly, for feature fitting, the most active compounds in the training set mapped well on all the chemical features that are one HBD, one PI, one RA and two HY features of Hypo 1 with good fitting score. The active, moderately and inactive compounds have missed at least one of five features. In addition, the most active compounds mapped well on the RA and PI features whereas some active, moderately active and all inactive compounds could not map on the RA and PI features signifying the importance of these two features. The pharmacophore overlay of most active Compound 1 and Hypo 1 has shown a fit value of 9.52. The RA feature corresponds to phenyl ring present in between two amide and a sulfonamide groups, one HBD feature corresponds to nitrogen of amide group located at the branch, PI group corresponds to the only amino nitrogen, one HY feature corresponds to a phenyl group whereas the another HY feature corresponds to alkyl group (Figure 3A). The pharmacophore overlay of least active compound 20 has revealed that it missed two features when mapped on Hypo 1 with a fit value of 5.16. This compound has mapped only the HBD and two HY features in the same manner as most active compound with no mapping over RA and PI features (Figure 3B). Fit value indicates how well the features in the pharmacophore overlaps the chemical features in the molecule and thereby aid in understanding the chemical meaning of the hypothesis [37]. These results emphasized Hypo 1 as a reliable model to accurately estimate the experimental activity of the training set compounds.\nExperimental and estimated IC50 values of the training set compounds based on the pharmacophore hypothesis ‘Hypo 1’.\naPositive value indicates that the estimated IC50 is higher than the experimental IC50; negative value indicates that the estimated IC50 is lower than the experimental IC50.\nbFit value indicates how well the features in the pharmacophore map the chemical features in the compound.\ncActivity scale: most active, ++++, IC50 ≤ 100 nM; active, +++, 100 nM < IC50 ≤ 1000 nM; moderately active, ++, 1000 nM < IC50 ≤ 10,000 nM; inactive, +, IC50 > 10,000 nM.\nPharmacophore mapping of most and least active compounds in the training set. A) Hypo 1 mapped on to the most active Compound 1 B) Hypo 1 mapped on to the least active Compound 20. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.", "Hypo 1 was further validated by test set, Fischer randomization test, leave-one-out and decoy set methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\nA total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\n[SUBTITLE] Fisher randomization method [SUBSECTION] This method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\nThis method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\n[SUBTITLE] Decoy set method [SUBSECTION] The Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\nThe Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\nThe cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.", "A total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.", "This method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.", "The Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.", "The cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.", "The validated pharmacophore hypothesis, Hypo1, was used as a 3D structural query for retrieving compounds from chemical databases including MayBridge (59 652 compounds), Chembridge (50 000 compounds), NCI (238 819 compounds) and Asinex (213 462 compounds). As a result of first screening 11 578, 590, 5096 and 63 265 compounds were retrieved from Maybridge, Chembridge, NCI and Asinex respectively. Since the active site of BACE-1 is larger in size, the experimentally known most active inhibitors are also larger in size and violate the first rule of Lipinski’s rule of five. Hence, the retrieved hit compounds were filtered based only on the estimated activity values calculated by Hypo 1. The activity range for the most active compounds is <100 nM. Finally 773 compounds were selected by restricting the minimum estimated activity to <100 nM.", "To further refine the retrieved hits and also to remove the false positives, these 773 compounds along with the 20 training set compounds were docked into the active site of BACE-1 using GOLD 4.1 program. There are number of crystal structures for BACE-ligand complexes are available in PDB. The crystal structure of BACE-SC7 (PDB ID: 2QP8) complex was taken based on its high resolution. The GOLD fitness score was calculated for all the 793 compounds, it distinguishes molecules based on their interacting ability. The GOLD fitness score for the most active compound in the training set was 53.035. The compounds for further analysis were selected based on the ligand conformations which can satisfy the necessary interactions at the active site and scoring GOLD fitness score greater than 60. Finally 20 compounds from Maybridge and 15 compounds from Asinex have shown the required interaction with BACE-1 as well as good GOLD fitness scores. The compounds with the same chemical scaffolds were filtered carefully based on the molecular interactions observed at the active site. Finally, two compounds with different scaffolds one from Maybridge (RJC01726) and one from Asinex (Asnx-2) were selected as representative compounds. The binding mode of the final hits and the most active Compound 1 in the training set are shown in Figure 5. Figure 5A represents the binding mode of Compound 1 with a GOLD fitness score of 53.035. It has formed hydrogen bond interactions with D93, G95, T133, Q134, G291 and T293 and hydrophobic interactions with Y132, F169, and T292. The GOLD fitness score of RJC01726 was 68.289 and the mode of binding in the active site (Figure 5B) is similar to Compound 1. It has formed hydrogen bond interactions with T133, Q134 and T293 and hydrophobic interactions with D93, G95, F169 and T292. Asnx-2 has shown hydrogen bond interactions with T133, G291 and T293 as well as hydrophobic interactions with D93, Y132, F169 and T292 with a GOLD fitness score of 62.026 (Figure 5C). Figure 5D represents the overlay of most active Compound 1, RJC01726 and Asnx-2 at their binding modes. The pharmacophore overlay of the final hits compounds are shown in Figure 6 and their 2D representations are shown in Figure 7.\nMolecular docking results. Binding orientations of A) Compound 1 of the training set (cyan color), B) RJC01726 (magenta color), C) Asnx-2 (blue color), D) overlay of Compound 1, RJC01726 and Asnx-2 in the active site of BACE-1 protein. Active site residues are shown in stick form and hydrogen bond interactions are indicated with purple dotted lines.\nPharmacophore overlay on final hits. The mapping of pharmacophore hypothesis Hypo 1 on the final hits. A) RJC01726 (red color) B) Asnx-2 (cyan color).\nChemical structure of final hits. 2D representation of the final hits RJC01726 and Asnx-2.\nThe hydrophobic interactions of the final hits compounds were observed using Ligplot program [39]. The novelty of the two hits compounds were confirmed using SciFinder search [40] and PubChem search [41].", "A chemical feature based 3D pharmacophore hypotheses of BACE-1 inhibitors have been developed using 3D QSAR Pharmacophore Generation protocol available in DS 2.5. The best quantitative pharmacophore model, Hypo 1, was characterized by the highest cost difference (121.98), best correlation coefficient (0.977), lowest total cost value (81.24) and lowest RMSD (0.804). The fixed cost and null cost values were 74.77 and 203.22 bits, respectively. Hypo1 consisted of one HBD, one PI, one RA and two HY features. Hypo1 was further validated by test set, Fischer randomization test, leave-one-out, and decoy set methods. The test set containing 40 compounds was used in investigating the predictive ability of Hypo1 and resulted with a correlation coefficient of 0.917. Other validation methods also have provided reliable results on the strength of Hypo 1. This validated Hypo1 was used as a 3D query in database screening. The database hit compounds were subsequently subjected to filtering by estimated activity value. To further refine the retrieved hits the 793 compounds along with training set were carried out for molecular docking studies. The molecular docking result of all compounds was analyzed based on the GOLD fitness score, binding modes and molecular interactions with essential active site residues. Finally, two hits, namely, RJC01726 and Asnx-2 of different scaffolds with GOLD fitness score of 68.362 and 63.053, respectively, and interactions with important active site residues were chosen as lead candidates. These compounds as such and on further optimization can be used as potential leads in designing new BACE-1 inhibitors.", "The authors declare that they have no competing interests.", "SJ and ST equally involved in designing the work, analyzing the results and writing the manuscript. SS formatted and corrected the manuscript. KWL supervised the work and edited the manuscript. All four authors have read and approved the manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Dataset collection", "Diverse conformation generation", "Pharmacophore modeling", "The HypoGen algorithm", "Data analysis", "Pharmacophore validation", "Test set method", "Fischer randomization method", "Decoy set method", "Leave-one-out method", "Database screening", "Molecular docking", "Results and discussion", "Pharmacophore generation", "Statistical data analysis", "Activity prediction and mapping of training set compound on Hypo1", "Validation of Hypo 1", "Test set method", "Fisher randomization method", "Decoy set method", "Leave-one-out method", "Database screening", "Molecular docking", "Conclusion", "Competing interests", "Author’s contributions", "Supplementary Material" ]

[ "Beta-site amyloid precursor protein cleaving enzyme (BACE-1), also known as β-secretase, memapsin-2, or Aspartyl protease-2, is a single-membrane protein belongs to the aspartyl protease class of catabolic enzyme. This is one of the enzymes responsible for the sequential proteolysis of amyloid precursor protein (APP) [1]. The cleavage of APP by BACE-1, which is the rate-limiting step in the amyloid cascade, results in the generation of two peptide fragments Aβ40 and Aβ42. Among two peptide fragments, Aβ42 is the primary species and thought to be causal for the neurotoxicity and amyloid plaque formation that lead to memory and cognitive defects in Alzheimer’s disease (AD) [2]. The AD is a debilitating neurodegenerative disease that results in the irreversible loss of neurons, particularly in the cortex and hippocampus [3]. It is characterized by progressive decline in cognitive function that inevitably leading to incapacitation and death. It also histopathologically characterized by the presence of amyloid plaques and neurofibrillar tangles in the brain. Regardless of the increasing demand for medication, no truly disease-modifying treatment is currently available [4,5]. The BACE knockout study in mice shows a complete absence of Aβ production with no reported side effects [6-8]. Since gene knockout study showed a reduction in AD-like pathology, inhibition of BACE-1 the key enzyme in the production of Aβ peptide has emerged as an attractive therapeutic target for AD [9]. Therefore extensive efforts have been followed in the discovery of potential inhibitors of BACE-1. Most of the designing of BACE-1 inhibitors are based on the transition state mimetic approach, which depends mainly on replacing the scissile amide bond of an appropriate substrate with a stable mimetic of the putative transition-state structure [10].\nThe main aim of our approach, which is discussed in this study is different than the transition state mimetic approach, is to develop an accurate and efficient method for discovering potent BACE-1 inhibitors. A pharmacophore hypothesis was generated based on key structural features of compounds with BACE-1 inhibitory activity. It provides a rational hypothetical representation of the most important chemical features responsible for activity. Herein, a ligand-based 3D pharmacophore hypothesis for BACE-1 inhibitors was constructed based on the structure-activity relationship observed in a set of known BACE-1 inhibitors. The resulted pharmacophore hypotheses were validated by test set, Fischer randomization, leave-one-out, and decoy set methods. The validated pharmacophore hypothesis has been used in in silico screening to identify hits that are highly varied in chemical nature. The retrieved hits were subsequently subjected to a well-defined refining procedure based on estimated activity values, drug-likeness prediction and further by molecular docking study. The identified hits can further be utilized in designing novel and potent BACE-1 inhibitors.", "[SUBTITLE] Dataset collection [SUBSECTION] In a computerized pharmacophore generation process the accurate choice of the training set is a key issue. The built pharmacophore hypothesis can be as good as the input data information. The following criteria should be considered during the selection of data set in order to achieve a significant pharmacophore hypothesis. (1) All compounds used in the training set have to bind to the same receptor in roughly the same fashion. Compounds having more binding interaction with the receptor are more active than those with fewer; (2) the data set must be widely populated covering an activity range of at least 4 orders of magnitude; (3) the most active compounds should inevitably be included in the training set and (4) all biologically relevant data should be obtained by homogenous procedures [11]. Every individual feature in the resulting hypotheses will invade a certain weight that is proportional to its relative contribution to biological activity.\nBy taking these criteria into account, we have collected a total number of 320 BACE-1 inhibitors from various literature resources [12-25] and a database has been created. The 2D structure of the compounds were built using ChemSketch program version 12, and subsequently converted in to 3D structures using Discovery studio 2.5 (DS) [26]. In the next step, 60 compounds were selected as final dataset as the BACE-1 inhibitory activities of these 60 compounds were studied under same biological assay condition. Based on the principle of structural diversity and wide coverage of activity range, 20 compounds were carefully selected as training set compounds and the rest were used as test set in model validation. Here, the IC50 value of the training set compounds was taken into account, the inhibitory activity values of the training set compounds span over a range of four orders of magnitude, from 4 nM to 37 000 nM. The chemical structure and experimental activity of the training set compounds are shown in Figure 1.\nStructure of training set compounds. The 2D chemical structures of 20 compounds of the training set together with their experimental IC50 values in nM.\nIn a computerized pharmacophore generation process the accurate choice of the training set is a key issue. The built pharmacophore hypothesis can be as good as the input data information. The following criteria should be considered during the selection of data set in order to achieve a significant pharmacophore hypothesis. (1) All compounds used in the training set have to bind to the same receptor in roughly the same fashion. Compounds having more binding interaction with the receptor are more active than those with fewer; (2) the data set must be widely populated covering an activity range of at least 4 orders of magnitude; (3) the most active compounds should inevitably be included in the training set and (4) all biologically relevant data should be obtained by homogenous procedures [11]. Every individual feature in the resulting hypotheses will invade a certain weight that is proportional to its relative contribution to biological activity.\nBy taking these criteria into account, we have collected a total number of 320 BACE-1 inhibitors from various literature resources [12-25] and a database has been created. The 2D structure of the compounds were built using ChemSketch program version 12, and subsequently converted in to 3D structures using Discovery studio 2.5 (DS) [26]. In the next step, 60 compounds were selected as final dataset as the BACE-1 inhibitory activities of these 60 compounds were studied under same biological assay condition. Based on the principle of structural diversity and wide coverage of activity range, 20 compounds were carefully selected as training set compounds and the rest were used as test set in model validation. Here, the IC50 value of the training set compounds was taken into account, the inhibitory activity values of the training set compounds span over a range of four orders of magnitude, from 4 nM to 37 000 nM. The chemical structure and experimental activity of the training set compounds are shown in Figure 1.\nStructure of training set compounds. The 2D chemical structures of 20 compounds of the training set together with their experimental IC50 values in nM.\n[SUBTITLE] Diverse conformation generation [SUBSECTION] Prior to the generation of pharmacophore hypotheses, the training set compounds, which were converted to 3D structure, were used to generate diverse conformations. Diverse Conformation Generation protocol implemented in DS was used to generate conformations using the Best conformation model generation method with CHARMM force field and Poling algorithm to ensure the energy-minimized conformation for each compound. The parameters like maximum number of 250 conformers, the ‘best conformational analysis’ method, and an energy threshold of 20 kcal/mol above the global energy minimum were chosen during conformation generation.\nPrior to the generation of pharmacophore hypotheses, the training set compounds, which were converted to 3D structure, were used to generate diverse conformations. Diverse Conformation Generation protocol implemented in DS was used to generate conformations using the Best conformation model generation method with CHARMM force field and Poling algorithm to ensure the energy-minimized conformation for each compound. The parameters like maximum number of 250 conformers, the ‘best conformational analysis’ method, and an energy threshold of 20 kcal/mol above the global energy minimum were chosen during conformation generation.\n[SUBTITLE] Pharmacophore modeling [SUBSECTION] The training set comprises of 20 compounds was used in pharmacophore hypothesis generation. The HypoGen algorithm available in 3D QSAR Pharmacophore Generation protocol of DS tries to generate hypotheses with features common amongst active molecules and do not reflect the inactive molecules of the training set. The training set compounds were predicted for their inherent chemical features using Feature Mapping protocol implemented in DS. During pharmacophore hypothesis generation a minimum of 4 and a maximum of 5 pharmacophoric features like hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), positive ionizable (PI), ring aromatic (RA) and hydrophobic (HY) were included. These features were selected based on the feature mapping results. All parameters were set to their default values except uncertainty value, which has been changed to 2 instead of 3. An uncertainty value of 2 was more convenient for our dataset because the activity values of the training set spanned exactly the required 4 orders of magnitude; this choice has been further confirmed by preliminary calculations and by other literature evidence [27]. The uncertainty value represents the ratio of the uncertainty range of the actual activity against measured biology activity for each compound.\nThe training set comprises of 20 compounds was used in pharmacophore hypothesis generation. The HypoGen algorithm available in 3D QSAR Pharmacophore Generation protocol of DS tries to generate hypotheses with features common amongst active molecules and do not reflect the inactive molecules of the training set. The training set compounds were predicted for their inherent chemical features using Feature Mapping protocol implemented in DS. During pharmacophore hypothesis generation a minimum of 4 and a maximum of 5 pharmacophoric features like hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), positive ionizable (PI), ring aromatic (RA) and hydrophobic (HY) were included. These features were selected based on the feature mapping results. All parameters were set to their default values except uncertainty value, which has been changed to 2 instead of 3. An uncertainty value of 2 was more convenient for our dataset because the activity values of the training set spanned exactly the required 4 orders of magnitude; this choice has been further confirmed by preliminary calculations and by other literature evidence [27]. The uncertainty value represents the ratio of the uncertainty range of the actual activity against measured biology activity for each compound.\n[SUBTITLE] The HypoGen algorithm [SUBSECTION] With the full range of training set compounds from active to inactive the pharmacophore hypotheses were generated by HypoGen algorithm implemented in DS. This algorithm constructs and ranks the pharmacophore hypotheses that correlate best between 3D spatial arrangement of features in a given training set compounds and their respective experimental activities. This process is accomplished in three steps: the constructive phase, the subtractive phase and the optimization phase [28].\nThe constructive phase identifies the hypotheses that are common amongst the active compounds. HypoGen enumerates all possible pharmacophore configurations using all combinations of pharmacophore features for each of the conformations of the most active compound. In order to consider the left over most active compounds the hypotheses must fit a minimum subset of its features. Hence, a large database of pharmacophore configurations will be generated at the end of the constructive phase.\nThe subtractive phase will remove the pharmacophore configurations that are present in the least active compounds. All compounds whose activity is by default 3.5 orders of magnitude less than that of the most active compound are considered to represent the least active compounds. The value 3.5 is adjustable depending on the activity range of the training set. The optimization phase improves the hypothesis score. These scores of the generated hypotheses depend on the errors in activity estimation from regression and complexity. The optimization involves a variation of features and/or locations to optimize activity prediction via a simulated annealing approach. The total cost parameter will be calculated for every new hypothesis. The HypoGen will quit and reports the 10 top-scoring hypotheses when there is no improvement in the hypothesis score.\nWith the full range of training set compounds from active to inactive the pharmacophore hypotheses were generated by HypoGen algorithm implemented in DS. This algorithm constructs and ranks the pharmacophore hypotheses that correlate best between 3D spatial arrangement of features in a given training set compounds and their respective experimental activities. This process is accomplished in three steps: the constructive phase, the subtractive phase and the optimization phase [28].\nThe constructive phase identifies the hypotheses that are common amongst the active compounds. HypoGen enumerates all possible pharmacophore configurations using all combinations of pharmacophore features for each of the conformations of the most active compound. In order to consider the left over most active compounds the hypotheses must fit a minimum subset of its features. Hence, a large database of pharmacophore configurations will be generated at the end of the constructive phase.\nThe subtractive phase will remove the pharmacophore configurations that are present in the least active compounds. All compounds whose activity is by default 3.5 orders of magnitude less than that of the most active compound are considered to represent the least active compounds. The value 3.5 is adjustable depending on the activity range of the training set. The optimization phase improves the hypothesis score. These scores of the generated hypotheses depend on the errors in activity estimation from regression and complexity. The optimization involves a variation of features and/or locations to optimize activity prediction via a simulated annealing approach. The total cost parameter will be calculated for every new hypothesis. The HypoGen will quit and reports the 10 top-scoring hypotheses when there is no improvement in the hypothesis score.\n[SUBTITLE] Data analysis [SUBSECTION] The quality of a pharmacophore hypothesis is best determined by two theoretical cost calculations, which are represented in bit units [29]. One is the ‘fixed cost’ also known as cost of an ideal hypothesis, which represents the simplest model that fits all the data perfectly. The second cost is the ‘null cost’, which represents the highest cost of a pharmacophore with no features that estimates every activity to be the average of the activity data of the training set compounds.\nThe total cost of any pharmacophore hypothesis should always be close to the fixed cost and away from the null cost to be the significant model. The cost difference between fixed and null cost values should be larger for a meaningful pharmacophore hypothesis. A value of 40-60 bits in a pharmacophore hypothesis indicates that it has 75-90% probability of representing a true correlation in the data.\nThe hypotheses are also evaluated based on other cost components. The cost value for every individual hypothesis is the summation of three cost components: the error cost (E), the weight cost (W) and the configuration cost (C). The error cost is the value represents the root-mean-squared difference (RMSD) between experimental and estimated activity value of the training set compounds. The weight cost is a value that increases in a Gaussian form as this function weights in a model deviate from the ideal value of two. The configuration cost or entropy cost measures the entropy of the hypothesis space. If the input training set compounds are too multiplex, e.g. because of too flexible training set compounds, this will result in an effusive number of hypotheses as an outcome of the subtractive phase. This configuration cost should always be less than a maximum value of 17 [30]. The correlation coefficient of the pharmacophore hypothesis should be close to 1.\nThe quality of a pharmacophore hypothesis is best determined by two theoretical cost calculations, which are represented in bit units [29]. One is the ‘fixed cost’ also known as cost of an ideal hypothesis, which represents the simplest model that fits all the data perfectly. The second cost is the ‘null cost’, which represents the highest cost of a pharmacophore with no features that estimates every activity to be the average of the activity data of the training set compounds.\nThe total cost of any pharmacophore hypothesis should always be close to the fixed cost and away from the null cost to be the significant model. The cost difference between fixed and null cost values should be larger for a meaningful pharmacophore hypothesis. A value of 40-60 bits in a pharmacophore hypothesis indicates that it has 75-90% probability of representing a true correlation in the data.\nThe hypotheses are also evaluated based on other cost components. The cost value for every individual hypothesis is the summation of three cost components: the error cost (E), the weight cost (W) and the configuration cost (C). The error cost is the value represents the root-mean-squared difference (RMSD) between experimental and estimated activity value of the training set compounds. The weight cost is a value that increases in a Gaussian form as this function weights in a model deviate from the ideal value of two. The configuration cost or entropy cost measures the entropy of the hypothesis space. If the input training set compounds are too multiplex, e.g. because of too flexible training set compounds, this will result in an effusive number of hypotheses as an outcome of the subtractive phase. This configuration cost should always be less than a maximum value of 17 [30]. The correlation coefficient of the pharmacophore hypothesis should be close to 1.\n[SUBTITLE] Pharmacophore validation [SUBSECTION] The generated pharmacophore hypothesis was validated using test set, Fischer randomization, decoy set and leave-one-out methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.\nA total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.\n[SUBTITLE] Fischer randomization method [SUBSECTION] The main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.\nThe main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.\n[SUBTITLE] Decoy set method [SUBSECTION] An external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.\nAn external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].\nThe pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].\nThe generated pharmacophore hypothesis was validated using test set, Fischer randomization, decoy set and leave-one-out methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.\nA total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.\n[SUBTITLE] Fischer randomization method [SUBSECTION] The main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.\nThe main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.\n[SUBTITLE] Decoy set method [SUBSECTION] An external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.\nAn external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].\nThe pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].\n[SUBTITLE] Database screening [SUBSECTION] The validated pharmacophore hypothesis, Hypo 1, was used as a 3D query for screening four different chemical databases. The purpose of this screening is to retrieve novel and potential leads suitable for further development. The chemical databases used were Maybridge, Chembridge, NCI and Asinex. Conformers were generated for each molecule in the database using best conformer generation method that allows a maximum energy of 15 kcal/mol above that of the most stable conformation. The database screening was carried out using Ligand Pharmacophore Mapping protocol implemented in DS with Best/Flexible Search option. The retrieved compounds were filtered by restricting the estimated activity value less than 100 nM and the obtained compounds were further refined using molecular docking study.\nThe validated pharmacophore hypothesis, Hypo 1, was used as a 3D query for screening four different chemical databases. The purpose of this screening is to retrieve novel and potential leads suitable for further development. The chemical databases used were Maybridge, Chembridge, NCI and Asinex. Conformers were generated for each molecule in the database using best conformer generation method that allows a maximum energy of 15 kcal/mol above that of the most stable conformation. The database screening was carried out using Ligand Pharmacophore Mapping protocol implemented in DS with Best/Flexible Search option. The retrieved compounds were filtered by restricting the estimated activity value less than 100 nM and the obtained compounds were further refined using molecular docking study.\n[SUBTITLE] Molecular docking [SUBSECTION] Pharmacophore modeling normally firmly associated with docking procedure, which in a first step flexibly aligns the ligand molecule into a rigid macromolecule environment and then estimates the tightness of the interaction by different scoring functions [35]. The Docking takes all the information from a rigid protein environment and scores several possible interaction modes for different alignments. There are many docking programs available for molecular docking studies. In this study, we used GOLD (Genetic Optimisation for Ligand Docking), a docking program [36] that uses genetic algorithm for docking and performs automated docking with full acyclic ligand flexibility, partial cyclic ligand flexibility and partial protein flexibility in the neighborhood of the protein active site. The crystal structure of BACE-1 complexed with an inhibitor SC7 (PDB ID: 2QP8) was used in molecular docking studies. The inhibitor SC7 was extracted from the active site and the retrieved database hits were docked based on the ligand SC7 coordinates, in to the active site of BACE-1. The water molecules were removed prior to docking because they were not found to play any important roles in BACE1-ligand interaction. The early termination option parameter in GOLD was changed from 3 to 5 and the maximum save conformations was set to 10. All the other parameters were set at their default values.\nPharmacophore modeling normally firmly associated with docking procedure, which in a first step flexibly aligns the ligand molecule into a rigid macromolecule environment and then estimates the tightness of the interaction by different scoring functions [35]. The Docking takes all the information from a rigid protein environment and scores several possible interaction modes for different alignments. There are many docking programs available for molecular docking studies. In this study, we used GOLD (Genetic Optimisation for Ligand Docking), a docking program [36] that uses genetic algorithm for docking and performs automated docking with full acyclic ligand flexibility, partial cyclic ligand flexibility and partial protein flexibility in the neighborhood of the protein active site. The crystal structure of BACE-1 complexed with an inhibitor SC7 (PDB ID: 2QP8) was used in molecular docking studies. The inhibitor SC7 was extracted from the active site and the retrieved database hits were docked based on the ligand SC7 coordinates, in to the active site of BACE-1. The water molecules were removed prior to docking because they were not found to play any important roles in BACE1-ligand interaction. The early termination option parameter in GOLD was changed from 3 to 5 and the maximum save conformations was set to 10. All the other parameters were set at their default values.", "In a computerized pharmacophore generation process the accurate choice of the training set is a key issue. The built pharmacophore hypothesis can be as good as the input data information. The following criteria should be considered during the selection of data set in order to achieve a significant pharmacophore hypothesis. (1) All compounds used in the training set have to bind to the same receptor in roughly the same fashion. Compounds having more binding interaction with the receptor are more active than those with fewer; (2) the data set must be widely populated covering an activity range of at least 4 orders of magnitude; (3) the most active compounds should inevitably be included in the training set and (4) all biologically relevant data should be obtained by homogenous procedures [11]. Every individual feature in the resulting hypotheses will invade a certain weight that is proportional to its relative contribution to biological activity.\nBy taking these criteria into account, we have collected a total number of 320 BACE-1 inhibitors from various literature resources [12-25] and a database has been created. The 2D structure of the compounds were built using ChemSketch program version 12, and subsequently converted in to 3D structures using Discovery studio 2.5 (DS) [26]. In the next step, 60 compounds were selected as final dataset as the BACE-1 inhibitory activities of these 60 compounds were studied under same biological assay condition. Based on the principle of structural diversity and wide coverage of activity range, 20 compounds were carefully selected as training set compounds and the rest were used as test set in model validation. Here, the IC50 value of the training set compounds was taken into account, the inhibitory activity values of the training set compounds span over a range of four orders of magnitude, from 4 nM to 37 000 nM. The chemical structure and experimental activity of the training set compounds are shown in Figure 1.\nStructure of training set compounds. The 2D chemical structures of 20 compounds of the training set together with their experimental IC50 values in nM.", "Prior to the generation of pharmacophore hypotheses, the training set compounds, which were converted to 3D structure, were used to generate diverse conformations. Diverse Conformation Generation protocol implemented in DS was used to generate conformations using the Best conformation model generation method with CHARMM force field and Poling algorithm to ensure the energy-minimized conformation for each compound. The parameters like maximum number of 250 conformers, the ‘best conformational analysis’ method, and an energy threshold of 20 kcal/mol above the global energy minimum were chosen during conformation generation.", "The training set comprises of 20 compounds was used in pharmacophore hypothesis generation. The HypoGen algorithm available in 3D QSAR Pharmacophore Generation protocol of DS tries to generate hypotheses with features common amongst active molecules and do not reflect the inactive molecules of the training set. The training set compounds were predicted for their inherent chemical features using Feature Mapping protocol implemented in DS. During pharmacophore hypothesis generation a minimum of 4 and a maximum of 5 pharmacophoric features like hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), positive ionizable (PI), ring aromatic (RA) and hydrophobic (HY) were included. These features were selected based on the feature mapping results. All parameters were set to their default values except uncertainty value, which has been changed to 2 instead of 3. An uncertainty value of 2 was more convenient for our dataset because the activity values of the training set spanned exactly the required 4 orders of magnitude; this choice has been further confirmed by preliminary calculations and by other literature evidence [27]. The uncertainty value represents the ratio of the uncertainty range of the actual activity against measured biology activity for each compound.", "With the full range of training set compounds from active to inactive the pharmacophore hypotheses were generated by HypoGen algorithm implemented in DS. This algorithm constructs and ranks the pharmacophore hypotheses that correlate best between 3D spatial arrangement of features in a given training set compounds and their respective experimental activities. This process is accomplished in three steps: the constructive phase, the subtractive phase and the optimization phase [28].\nThe constructive phase identifies the hypotheses that are common amongst the active compounds. HypoGen enumerates all possible pharmacophore configurations using all combinations of pharmacophore features for each of the conformations of the most active compound. In order to consider the left over most active compounds the hypotheses must fit a minimum subset of its features. Hence, a large database of pharmacophore configurations will be generated at the end of the constructive phase.\nThe subtractive phase will remove the pharmacophore configurations that are present in the least active compounds. All compounds whose activity is by default 3.5 orders of magnitude less than that of the most active compound are considered to represent the least active compounds. The value 3.5 is adjustable depending on the activity range of the training set. The optimization phase improves the hypothesis score. These scores of the generated hypotheses depend on the errors in activity estimation from regression and complexity. The optimization involves a variation of features and/or locations to optimize activity prediction via a simulated annealing approach. The total cost parameter will be calculated for every new hypothesis. The HypoGen will quit and reports the 10 top-scoring hypotheses when there is no improvement in the hypothesis score.", "The quality of a pharmacophore hypothesis is best determined by two theoretical cost calculations, which are represented in bit units [29]. One is the ‘fixed cost’ also known as cost of an ideal hypothesis, which represents the simplest model that fits all the data perfectly. The second cost is the ‘null cost’, which represents the highest cost of a pharmacophore with no features that estimates every activity to be the average of the activity data of the training set compounds.\nThe total cost of any pharmacophore hypothesis should always be close to the fixed cost and away from the null cost to be the significant model. The cost difference between fixed and null cost values should be larger for a meaningful pharmacophore hypothesis. A value of 40-60 bits in a pharmacophore hypothesis indicates that it has 75-90% probability of representing a true correlation in the data.\nThe hypotheses are also evaluated based on other cost components. The cost value for every individual hypothesis is the summation of three cost components: the error cost (E), the weight cost (W) and the configuration cost (C). The error cost is the value represents the root-mean-squared difference (RMSD) between experimental and estimated activity value of the training set compounds. The weight cost is a value that increases in a Gaussian form as this function weights in a model deviate from the ideal value of two. The configuration cost or entropy cost measures the entropy of the hypothesis space. If the input training set compounds are too multiplex, e.g. because of too flexible training set compounds, this will result in an effusive number of hypotheses as an outcome of the subtractive phase. This configuration cost should always be less than a maximum value of 17 [30]. The correlation coefficient of the pharmacophore hypothesis should be close to 1.", "The generated pharmacophore hypothesis was validated using test set, Fischer randomization, decoy set and leave-one-out methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.\nA total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.\n[SUBTITLE] Fischer randomization method [SUBSECTION] The main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.\nThe main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.\n[SUBTITLE] Decoy set method [SUBSECTION] An external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.\nAn external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].\nThe pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].", "A total of 40 compounds with experimental activity data were selected as test set compounds. This method is used to elucidate whether the generated pharmacophore hypothesis is proficient to predict the activities of the compounds other than training set and classify them correctly in their activity scale. The conformation generation for test set compounds was carried out in a similar way like training set compounds using Diverse Conformation Generation protocol in DS. The compounds associated with their conformations were subsequently carried out for pharmacophore mapping using Ligand Pharmacophore Mapping protocol with Best/Flexible Search option available in DS.", "The main purpose of this validation is to verify whether there is a strong correlation existing between the chemical structure and biological activities of compounds. This generates pharmacophore hypotheses by randomizing the activity data of the training set compounds with the same features and parameters used to generate the original pharmacophore hypothesis. The statistical significance is calculated using the following formula: Significance = 100 (1-(1+x/y)), where x represents the total number of hypotheses having a total cost value lower than the original hypothesis, and y represents the total number of HypoGen runs i.e. initial and random runs. The confidence level was set to 99%, where 99 random spreadsheets (random hypotheses) were generated. During the pharmacophore generation, if the randomized data set results in similar or better cost values, RMSD and correlation, it means that the original hypothesis have been generated by chance.", "An external database containing BACE-1 active and inactive compounds was used to evaluate the discriminative ability of Hypo 1 in the separation of active compounds from the inactive compounds. The database was developed using a total of 453 compounds containing 206 actives and 247 inactives. All the compounds were collected from published literature including binding database [12-25,31]. The database screening was carried out using Ligand Pharmacophore Mapping protocol available in DS. A set of statistical parameters [32] like Ht, % yield of actives, Enrichment factor (E), false positives, false negatives and Goodness of Hit (GH) score were calculated.", "The pharmacophore hypothesis is cross validated by leave-one-out method. In this method, one compound is left in the generation of a new pharmacophore model and its affinity is predicted using that new model. The model building and estimation cycle is repeated until each compound was left out once [33]. This test is performed to verify whether the correlation coefficient of the training set compounds is strongly depend on one particular compound or not [34].", "The validated pharmacophore hypothesis, Hypo 1, was used as a 3D query for screening four different chemical databases. The purpose of this screening is to retrieve novel and potential leads suitable for further development. The chemical databases used were Maybridge, Chembridge, NCI and Asinex. Conformers were generated for each molecule in the database using best conformer generation method that allows a maximum energy of 15 kcal/mol above that of the most stable conformation. The database screening was carried out using Ligand Pharmacophore Mapping protocol implemented in DS with Best/Flexible Search option. The retrieved compounds were filtered by restricting the estimated activity value less than 100 nM and the obtained compounds were further refined using molecular docking study.", "Pharmacophore modeling normally firmly associated with docking procedure, which in a first step flexibly aligns the ligand molecule into a rigid macromolecule environment and then estimates the tightness of the interaction by different scoring functions [35]. The Docking takes all the information from a rigid protein environment and scores several possible interaction modes for different alignments. There are many docking programs available for molecular docking studies. In this study, we used GOLD (Genetic Optimisation for Ligand Docking), a docking program [36] that uses genetic algorithm for docking and performs automated docking with full acyclic ligand flexibility, partial cyclic ligand flexibility and partial protein flexibility in the neighborhood of the protein active site. The crystal structure of BACE-1 complexed with an inhibitor SC7 (PDB ID: 2QP8) was used in molecular docking studies. The inhibitor SC7 was extracted from the active site and the retrieved database hits were docked based on the ligand SC7 coordinates, in to the active site of BACE-1. The water molecules were removed prior to docking because they were not found to play any important roles in BACE1-ligand interaction. The early termination option parameter in GOLD was changed from 3 to 5 and the maximum save conformations was set to 10. All the other parameters were set at their default values.", "[SUBTITLE] Pharmacophore generation [SUBSECTION] We have used the HypoGen algorithm implemented in DS in order to quantitatively correlate the chemical structure of BACE-1 inhibitors to their biological activity. The training set of 20 compounds (Figure 1) with activity values ranging from 4 to 37000 nM was used in pharmacophore model generation. The Feature Mapping protocol resulted in HBA, HBD, RA, PI and HY features. Selecting these features, the pharmacophore generation run was performed along with diverse conformers of training set molecules generated as described in methods section. Ten top-scored pharmacophore hypotheses were generated and in order to choose the best one and also to give an idea about the statistical significance, the pharmacophore hypotheses were subjected to cost analysis. The results of top ten pharmacophore hypotheses and their statistical parameters are given in Table 1. In this study, the first pharmacophore hypothesis (Hypo 1) is the best hypothesis characterized by the large cost difference (121.98 bits), lowest RMSD value (0.804) and a high correlation coefficient of 0.977. All ten hypotheses consist of HBD, PI, RA and HY features. Nine of ten hypotheses were composed of five pharmacophoric features except only one hypothesis, which was of four features. The best pharmacophore hypothesis (Hypo 1), which scored the large cost difference, lowest RMSD, lowest error cost and high correlation coefficient, was made of one HBD, one PI, one RA and two HY features.\nResults of the top 10 pharmacophore hypotheses generated by the HypoGen algorithm\nNull cost = 203.22; fixed cost = 74.77; configuration cost = 15.59.\naCost difference = null cost – total cost.\nbHBD, hydrogen-bond donor; PI, positive ionizable; RA, ring aromatic; HY, hydrophobic.\nWe have used the HypoGen algorithm implemented in DS in order to quantitatively correlate the chemical structure of BACE-1 inhibitors to their biological activity. The training set of 20 compounds (Figure 1) with activity values ranging from 4 to 37000 nM was used in pharmacophore model generation. The Feature Mapping protocol resulted in HBA, HBD, RA, PI and HY features. Selecting these features, the pharmacophore generation run was performed along with diverse conformers of training set molecules generated as described in methods section. Ten top-scored pharmacophore hypotheses were generated and in order to choose the best one and also to give an idea about the statistical significance, the pharmacophore hypotheses were subjected to cost analysis. The results of top ten pharmacophore hypotheses and their statistical parameters are given in Table 1. In this study, the first pharmacophore hypothesis (Hypo 1) is the best hypothesis characterized by the large cost difference (121.98 bits), lowest RMSD value (0.804) and a high correlation coefficient of 0.977. All ten hypotheses consist of HBD, PI, RA and HY features. Nine of ten hypotheses were composed of five pharmacophoric features except only one hypothesis, which was of four features. The best pharmacophore hypothesis (Hypo 1), which scored the large cost difference, lowest RMSD, lowest error cost and high correlation coefficient, was made of one HBD, one PI, one RA and two HY features.\nResults of the top 10 pharmacophore hypotheses generated by the HypoGen algorithm\nNull cost = 203.22; fixed cost = 74.77; configuration cost = 15.59.\naCost difference = null cost – total cost.\nbHBD, hydrogen-bond donor; PI, positive ionizable; RA, ring aromatic; HY, hydrophobic.\n[SUBTITLE] Statistical data analysis [SUBSECTION] The generated hypotheses were subjected to cost analysis. The two main values used for cost analysis are the difference between fixed and null cost and another is the difference between the null cost and the total cost (Δcost). The fixed cost of the run was 74.77 bits, which was well separated from the null cost of 203.22 bits and close to the total cost of 81.24 bits. The large difference (128.45 bits) observed between the fixed cost and null cost value indicates that Hypo 1 has more than 90% statistical significance to be a significant model. All the 10 hypotheses were subjected to further evaluation for their capability to predict the activity of the training set compounds. Configuration cost or entropy value must be less than 17 for which a value of 15.59 was obtained in this study. All hypotheses have scored RMSD values lower than 1.5 and ranging from 0.804 to 1.111, this characterization further emphasizing the good predictive quality of these hypotheses. Based on the rule to select a hypothesis with a lowest total cost, high correlation coefficient, large cost difference and significantly low RMSD value, Hypo 1 gave the best statistical values among other hypotheses. Hence, Hypo 1 with one HBD, one PI, one RA and two HY was chosen as the best hypothesis for further analysis. The inter-feature distance constraints were observed for this five-featured pharmacophore hypothesis (Hypo 1) (Figure 2).\nHypoGen pharmacophore hypothesis for BACE-1 inhibitors. A) The best five feature pharmacophore model Hypo 1 B) 3D spatial arrangement and the distance constraints of the Hypo 1. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.\nThe generated hypotheses were subjected to cost analysis. The two main values used for cost analysis are the difference between fixed and null cost and another is the difference between the null cost and the total cost (Δcost). The fixed cost of the run was 74.77 bits, which was well separated from the null cost of 203.22 bits and close to the total cost of 81.24 bits. The large difference (128.45 bits) observed between the fixed cost and null cost value indicates that Hypo 1 has more than 90% statistical significance to be a significant model. All the 10 hypotheses were subjected to further evaluation for their capability to predict the activity of the training set compounds. Configuration cost or entropy value must be less than 17 for which a value of 15.59 was obtained in this study. All hypotheses have scored RMSD values lower than 1.5 and ranging from 0.804 to 1.111, this characterization further emphasizing the good predictive quality of these hypotheses. Based on the rule to select a hypothesis with a lowest total cost, high correlation coefficient, large cost difference and significantly low RMSD value, Hypo 1 gave the best statistical values among other hypotheses. Hence, Hypo 1 with one HBD, one PI, one RA and two HY was chosen as the best hypothesis for further analysis. The inter-feature distance constraints were observed for this five-featured pharmacophore hypothesis (Hypo 1) (Figure 2).\nHypoGen pharmacophore hypothesis for BACE-1 inhibitors. A) The best five feature pharmacophore model Hypo 1 B) 3D spatial arrangement and the distance constraints of the Hypo 1. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.\n[SUBTITLE] Activity prediction and mapping of training set compound on Hypo1 [SUBSECTION] To verify the predictive ability of Hypo1 on training set compounds, the activity of each training set compound is estimated by regression analysis. The experimental activities of training set compounds were classified into four groups: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). The estimated activity values of training set compounds based on Hypo 1 and the corresponding error values are calculated (Table 2). The error value is the ratio between the estimated and experimental activities. The positive error value indicates that the estimated IC50 value is higher than the experimental activity, whereas the negative error value indicates that the estimated IC50 value is lower than the experimental activity. An error value of less than 10 signifies the prediction of activity lesser than one order of magnitude. Among 20 training set compounds, only one compound had an error value of greater than 3. From Table 2 it is clear that the estimated activity values of most of the training set compounds was predicted with the same activity scale as the experimental activity. Among 20 training set compounds, one most active compound (++++) was estimated as active (+++), one active compound (+++) was estimated as moderately active (++), one moderately active compound (++) was estimated as inactive (+) and two inactive compounds (+) were estimated as moderately active (++). The divergence between the estimated and experimental activity observed in four compounds was only about 1 order of magnitude, which might be an artifact of the program that uses different number of degrees of freedom for these compounds to mismatch the pharmacophore model. Interestingly, for feature fitting, the most active compounds in the training set mapped well on all the chemical features that are one HBD, one PI, one RA and two HY features of Hypo 1 with good fitting score. The active, moderately and inactive compounds have missed at least one of five features. In addition, the most active compounds mapped well on the RA and PI features whereas some active, moderately active and all inactive compounds could not map on the RA and PI features signifying the importance of these two features. The pharmacophore overlay of most active Compound 1 and Hypo 1 has shown a fit value of 9.52. The RA feature corresponds to phenyl ring present in between two amide and a sulfonamide groups, one HBD feature corresponds to nitrogen of amide group located at the branch, PI group corresponds to the only amino nitrogen, one HY feature corresponds to a phenyl group whereas the another HY feature corresponds to alkyl group (Figure 3A). The pharmacophore overlay of least active compound 20 has revealed that it missed two features when mapped on Hypo 1 with a fit value of 5.16. This compound has mapped only the HBD and two HY features in the same manner as most active compound with no mapping over RA and PI features (Figure 3B). Fit value indicates how well the features in the pharmacophore overlaps the chemical features in the molecule and thereby aid in understanding the chemical meaning of the hypothesis [37]. These results emphasized Hypo 1 as a reliable model to accurately estimate the experimental activity of the training set compounds.\nExperimental and estimated IC50 values of the training set compounds based on the pharmacophore hypothesis ‘Hypo 1’.\naPositive value indicates that the estimated IC50 is higher than the experimental IC50; negative value indicates that the estimated IC50 is lower than the experimental IC50.\nbFit value indicates how well the features in the pharmacophore map the chemical features in the compound.\ncActivity scale: most active, ++++, IC50 ≤ 100 nM; active, +++, 100 nM < IC50 ≤ 1000 nM; moderately active, ++, 1000 nM < IC50 ≤ 10,000 nM; inactive, +, IC50 > 10,000 nM.\nPharmacophore mapping of most and least active compounds in the training set. A) Hypo 1 mapped on to the most active Compound 1 B) Hypo 1 mapped on to the least active Compound 20. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.\nTo verify the predictive ability of Hypo1 on training set compounds, the activity of each training set compound is estimated by regression analysis. The experimental activities of training set compounds were classified into four groups: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). The estimated activity values of training set compounds based on Hypo 1 and the corresponding error values are calculated (Table 2). The error value is the ratio between the estimated and experimental activities. The positive error value indicates that the estimated IC50 value is higher than the experimental activity, whereas the negative error value indicates that the estimated IC50 value is lower than the experimental activity. An error value of less than 10 signifies the prediction of activity lesser than one order of magnitude. Among 20 training set compounds, only one compound had an error value of greater than 3. From Table 2 it is clear that the estimated activity values of most of the training set compounds was predicted with the same activity scale as the experimental activity. Among 20 training set compounds, one most active compound (++++) was estimated as active (+++), one active compound (+++) was estimated as moderately active (++), one moderately active compound (++) was estimated as inactive (+) and two inactive compounds (+) were estimated as moderately active (++). The divergence between the estimated and experimental activity observed in four compounds was only about 1 order of magnitude, which might be an artifact of the program that uses different number of degrees of freedom for these compounds to mismatch the pharmacophore model. Interestingly, for feature fitting, the most active compounds in the training set mapped well on all the chemical features that are one HBD, one PI, one RA and two HY features of Hypo 1 with good fitting score. The active, moderately and inactive compounds have missed at least one of five features. In addition, the most active compounds mapped well on the RA and PI features whereas some active, moderately active and all inactive compounds could not map on the RA and PI features signifying the importance of these two features. The pharmacophore overlay of most active Compound 1 and Hypo 1 has shown a fit value of 9.52. The RA feature corresponds to phenyl ring present in between two amide and a sulfonamide groups, one HBD feature corresponds to nitrogen of amide group located at the branch, PI group corresponds to the only amino nitrogen, one HY feature corresponds to a phenyl group whereas the another HY feature corresponds to alkyl group (Figure 3A). The pharmacophore overlay of least active compound 20 has revealed that it missed two features when mapped on Hypo 1 with a fit value of 5.16. This compound has mapped only the HBD and two HY features in the same manner as most active compound with no mapping over RA and PI features (Figure 3B). Fit value indicates how well the features in the pharmacophore overlaps the chemical features in the molecule and thereby aid in understanding the chemical meaning of the hypothesis [37]. These results emphasized Hypo 1 as a reliable model to accurately estimate the experimental activity of the training set compounds.\nExperimental and estimated IC50 values of the training set compounds based on the pharmacophore hypothesis ‘Hypo 1’.\naPositive value indicates that the estimated IC50 is higher than the experimental IC50; negative value indicates that the estimated IC50 is lower than the experimental IC50.\nbFit value indicates how well the features in the pharmacophore map the chemical features in the compound.\ncActivity scale: most active, ++++, IC50 ≤ 100 nM; active, +++, 100 nM < IC50 ≤ 1000 nM; moderately active, ++, 1000 nM < IC50 ≤ 10,000 nM; inactive, +, IC50 > 10,000 nM.\nPharmacophore mapping of most and least active compounds in the training set. A) Hypo 1 mapped on to the most active Compound 1 B) Hypo 1 mapped on to the least active Compound 20. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.\n[SUBTITLE] Validation of Hypo 1 [SUBSECTION] Hypo 1 was further validated by test set, Fischer randomization test, leave-one-out and decoy set methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\nA total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\n[SUBTITLE] Fisher randomization method [SUBSECTION] This method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\nThis method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\n[SUBTITLE] Decoy set method [SUBSECTION] The Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\nThe Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\nThe cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\nHypo 1 was further validated by test set, Fischer randomization test, leave-one-out and decoy set methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\nA total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\n[SUBTITLE] Fisher randomization method [SUBSECTION] This method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\nThis method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\n[SUBTITLE] Decoy set method [SUBSECTION] The Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\nThe Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\nThe cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\n[SUBTITLE] Database screening [SUBSECTION] The validated pharmacophore hypothesis, Hypo1, was used as a 3D structural query for retrieving compounds from chemical databases including MayBridge (59 652 compounds), Chembridge (50 000 compounds), NCI (238 819 compounds) and Asinex (213 462 compounds). As a result of first screening 11 578, 590, 5096 and 63 265 compounds were retrieved from Maybridge, Chembridge, NCI and Asinex respectively. Since the active site of BACE-1 is larger in size, the experimentally known most active inhibitors are also larger in size and violate the first rule of Lipinski’s rule of five. Hence, the retrieved hit compounds were filtered based only on the estimated activity values calculated by Hypo 1. The activity range for the most active compounds is <100 nM. Finally 773 compounds were selected by restricting the minimum estimated activity to <100 nM.\nThe validated pharmacophore hypothesis, Hypo1, was used as a 3D structural query for retrieving compounds from chemical databases including MayBridge (59 652 compounds), Chembridge (50 000 compounds), NCI (238 819 compounds) and Asinex (213 462 compounds). As a result of first screening 11 578, 590, 5096 and 63 265 compounds were retrieved from Maybridge, Chembridge, NCI and Asinex respectively. Since the active site of BACE-1 is larger in size, the experimentally known most active inhibitors are also larger in size and violate the first rule of Lipinski’s rule of five. Hence, the retrieved hit compounds were filtered based only on the estimated activity values calculated by Hypo 1. The activity range for the most active compounds is <100 nM. Finally 773 compounds were selected by restricting the minimum estimated activity to <100 nM.\n[SUBTITLE] Molecular docking [SUBSECTION] To further refine the retrieved hits and also to remove the false positives, these 773 compounds along with the 20 training set compounds were docked into the active site of BACE-1 using GOLD 4.1 program. There are number of crystal structures for BACE-ligand complexes are available in PDB. The crystal structure of BACE-SC7 (PDB ID: 2QP8) complex was taken based on its high resolution. The GOLD fitness score was calculated for all the 793 compounds, it distinguishes molecules based on their interacting ability. The GOLD fitness score for the most active compound in the training set was 53.035. The compounds for further analysis were selected based on the ligand conformations which can satisfy the necessary interactions at the active site and scoring GOLD fitness score greater than 60. Finally 20 compounds from Maybridge and 15 compounds from Asinex have shown the required interaction with BACE-1 as well as good GOLD fitness scores. The compounds with the same chemical scaffolds were filtered carefully based on the molecular interactions observed at the active site. Finally, two compounds with different scaffolds one from Maybridge (RJC01726) and one from Asinex (Asnx-2) were selected as representative compounds. The binding mode of the final hits and the most active Compound 1 in the training set are shown in Figure 5. Figure 5A represents the binding mode of Compound 1 with a GOLD fitness score of 53.035. It has formed hydrogen bond interactions with D93, G95, T133, Q134, G291 and T293 and hydrophobic interactions with Y132, F169, and T292. The GOLD fitness score of RJC01726 was 68.289 and the mode of binding in the active site (Figure 5B) is similar to Compound 1. It has formed hydrogen bond interactions with T133, Q134 and T293 and hydrophobic interactions with D93, G95, F169 and T292. Asnx-2 has shown hydrogen bond interactions with T133, G291 and T293 as well as hydrophobic interactions with D93, Y132, F169 and T292 with a GOLD fitness score of 62.026 (Figure 5C). Figure 5D represents the overlay of most active Compound 1, RJC01726 and Asnx-2 at their binding modes. The pharmacophore overlay of the final hits compounds are shown in Figure 6 and their 2D representations are shown in Figure 7.\nMolecular docking results. Binding orientations of A) Compound 1 of the training set (cyan color), B) RJC01726 (magenta color), C) Asnx-2 (blue color), D) overlay of Compound 1, RJC01726 and Asnx-2 in the active site of BACE-1 protein. Active site residues are shown in stick form and hydrogen bond interactions are indicated with purple dotted lines.\nPharmacophore overlay on final hits. The mapping of pharmacophore hypothesis Hypo 1 on the final hits. A) RJC01726 (red color) B) Asnx-2 (cyan color).\nChemical structure of final hits. 2D representation of the final hits RJC01726 and Asnx-2.\nThe hydrophobic interactions of the final hits compounds were observed using Ligplot program [39]. The novelty of the two hits compounds were confirmed using SciFinder search [40] and PubChem search [41].\nTo further refine the retrieved hits and also to remove the false positives, these 773 compounds along with the 20 training set compounds were docked into the active site of BACE-1 using GOLD 4.1 program. There are number of crystal structures for BACE-ligand complexes are available in PDB. The crystal structure of BACE-SC7 (PDB ID: 2QP8) complex was taken based on its high resolution. The GOLD fitness score was calculated for all the 793 compounds, it distinguishes molecules based on their interacting ability. The GOLD fitness score for the most active compound in the training set was 53.035. The compounds for further analysis were selected based on the ligand conformations which can satisfy the necessary interactions at the active site and scoring GOLD fitness score greater than 60. Finally 20 compounds from Maybridge and 15 compounds from Asinex have shown the required interaction with BACE-1 as well as good GOLD fitness scores. The compounds with the same chemical scaffolds were filtered carefully based on the molecular interactions observed at the active site. Finally, two compounds with different scaffolds one from Maybridge (RJC01726) and one from Asinex (Asnx-2) were selected as representative compounds. The binding mode of the final hits and the most active Compound 1 in the training set are shown in Figure 5. Figure 5A represents the binding mode of Compound 1 with a GOLD fitness score of 53.035. It has formed hydrogen bond interactions with D93, G95, T133, Q134, G291 and T293 and hydrophobic interactions with Y132, F169, and T292. The GOLD fitness score of RJC01726 was 68.289 and the mode of binding in the active site (Figure 5B) is similar to Compound 1. It has formed hydrogen bond interactions with T133, Q134 and T293 and hydrophobic interactions with D93, G95, F169 and T292. Asnx-2 has shown hydrogen bond interactions with T133, G291 and T293 as well as hydrophobic interactions with D93, Y132, F169 and T292 with a GOLD fitness score of 62.026 (Figure 5C). Figure 5D represents the overlay of most active Compound 1, RJC01726 and Asnx-2 at their binding modes. The pharmacophore overlay of the final hits compounds are shown in Figure 6 and their 2D representations are shown in Figure 7.\nMolecular docking results. Binding orientations of A) Compound 1 of the training set (cyan color), B) RJC01726 (magenta color), C) Asnx-2 (blue color), D) overlay of Compound 1, RJC01726 and Asnx-2 in the active site of BACE-1 protein. Active site residues are shown in stick form and hydrogen bond interactions are indicated with purple dotted lines.\nPharmacophore overlay on final hits. The mapping of pharmacophore hypothesis Hypo 1 on the final hits. A) RJC01726 (red color) B) Asnx-2 (cyan color).\nChemical structure of final hits. 2D representation of the final hits RJC01726 and Asnx-2.\nThe hydrophobic interactions of the final hits compounds were observed using Ligplot program [39]. The novelty of the two hits compounds were confirmed using SciFinder search [40] and PubChem search [41].", "We have used the HypoGen algorithm implemented in DS in order to quantitatively correlate the chemical structure of BACE-1 inhibitors to their biological activity. The training set of 20 compounds (Figure 1) with activity values ranging from 4 to 37000 nM was used in pharmacophore model generation. The Feature Mapping protocol resulted in HBA, HBD, RA, PI and HY features. Selecting these features, the pharmacophore generation run was performed along with diverse conformers of training set molecules generated as described in methods section. Ten top-scored pharmacophore hypotheses were generated and in order to choose the best one and also to give an idea about the statistical significance, the pharmacophore hypotheses were subjected to cost analysis. The results of top ten pharmacophore hypotheses and their statistical parameters are given in Table 1. In this study, the first pharmacophore hypothesis (Hypo 1) is the best hypothesis characterized by the large cost difference (121.98 bits), lowest RMSD value (0.804) and a high correlation coefficient of 0.977. All ten hypotheses consist of HBD, PI, RA and HY features. Nine of ten hypotheses were composed of five pharmacophoric features except only one hypothesis, which was of four features. The best pharmacophore hypothesis (Hypo 1), which scored the large cost difference, lowest RMSD, lowest error cost and high correlation coefficient, was made of one HBD, one PI, one RA and two HY features.\nResults of the top 10 pharmacophore hypotheses generated by the HypoGen algorithm\nNull cost = 203.22; fixed cost = 74.77; configuration cost = 15.59.\naCost difference = null cost – total cost.\nbHBD, hydrogen-bond donor; PI, positive ionizable; RA, ring aromatic; HY, hydrophobic.", "The generated hypotheses were subjected to cost analysis. The two main values used for cost analysis are the difference between fixed and null cost and another is the difference between the null cost and the total cost (Δcost). The fixed cost of the run was 74.77 bits, which was well separated from the null cost of 203.22 bits and close to the total cost of 81.24 bits. The large difference (128.45 bits) observed between the fixed cost and null cost value indicates that Hypo 1 has more than 90% statistical significance to be a significant model. All the 10 hypotheses were subjected to further evaluation for their capability to predict the activity of the training set compounds. Configuration cost or entropy value must be less than 17 for which a value of 15.59 was obtained in this study. All hypotheses have scored RMSD values lower than 1.5 and ranging from 0.804 to 1.111, this characterization further emphasizing the good predictive quality of these hypotheses. Based on the rule to select a hypothesis with a lowest total cost, high correlation coefficient, large cost difference and significantly low RMSD value, Hypo 1 gave the best statistical values among other hypotheses. Hence, Hypo 1 with one HBD, one PI, one RA and two HY was chosen as the best hypothesis for further analysis. The inter-feature distance constraints were observed for this five-featured pharmacophore hypothesis (Hypo 1) (Figure 2).\nHypoGen pharmacophore hypothesis for BACE-1 inhibitors. A) The best five feature pharmacophore model Hypo 1 B) 3D spatial arrangement and the distance constraints of the Hypo 1. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.", "To verify the predictive ability of Hypo1 on training set compounds, the activity of each training set compound is estimated by regression analysis. The experimental activities of training set compounds were classified into four groups: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). The estimated activity values of training set compounds based on Hypo 1 and the corresponding error values are calculated (Table 2). The error value is the ratio between the estimated and experimental activities. The positive error value indicates that the estimated IC50 value is higher than the experimental activity, whereas the negative error value indicates that the estimated IC50 value is lower than the experimental activity. An error value of less than 10 signifies the prediction of activity lesser than one order of magnitude. Among 20 training set compounds, only one compound had an error value of greater than 3. From Table 2 it is clear that the estimated activity values of most of the training set compounds was predicted with the same activity scale as the experimental activity. Among 20 training set compounds, one most active compound (++++) was estimated as active (+++), one active compound (+++) was estimated as moderately active (++), one moderately active compound (++) was estimated as inactive (+) and two inactive compounds (+) were estimated as moderately active (++). The divergence between the estimated and experimental activity observed in four compounds was only about 1 order of magnitude, which might be an artifact of the program that uses different number of degrees of freedom for these compounds to mismatch the pharmacophore model. Interestingly, for feature fitting, the most active compounds in the training set mapped well on all the chemical features that are one HBD, one PI, one RA and two HY features of Hypo 1 with good fitting score. The active, moderately and inactive compounds have missed at least one of five features. In addition, the most active compounds mapped well on the RA and PI features whereas some active, moderately active and all inactive compounds could not map on the RA and PI features signifying the importance of these two features. The pharmacophore overlay of most active Compound 1 and Hypo 1 has shown a fit value of 9.52. The RA feature corresponds to phenyl ring present in between two amide and a sulfonamide groups, one HBD feature corresponds to nitrogen of amide group located at the branch, PI group corresponds to the only amino nitrogen, one HY feature corresponds to a phenyl group whereas the another HY feature corresponds to alkyl group (Figure 3A). The pharmacophore overlay of least active compound 20 has revealed that it missed two features when mapped on Hypo 1 with a fit value of 5.16. This compound has mapped only the HBD and two HY features in the same manner as most active compound with no mapping over RA and PI features (Figure 3B). Fit value indicates how well the features in the pharmacophore overlaps the chemical features in the molecule and thereby aid in understanding the chemical meaning of the hypothesis [37]. These results emphasized Hypo 1 as a reliable model to accurately estimate the experimental activity of the training set compounds.\nExperimental and estimated IC50 values of the training set compounds based on the pharmacophore hypothesis ‘Hypo 1’.\naPositive value indicates that the estimated IC50 is higher than the experimental IC50; negative value indicates that the estimated IC50 is lower than the experimental IC50.\nbFit value indicates how well the features in the pharmacophore map the chemical features in the compound.\ncActivity scale: most active, ++++, IC50 ≤ 100 nM; active, +++, 100 nM < IC50 ≤ 1000 nM; moderately active, ++, 1000 nM < IC50 ≤ 10,000 nM; inactive, +, IC50 > 10,000 nM.\nPharmacophore mapping of most and least active compounds in the training set. A) Hypo 1 mapped on to the most active Compound 1 B) Hypo 1 mapped on to the least active Compound 20. The features are color coded with magenta, hydrogen bond donor; red, positive ionizable; orange, ring aromatic; cyan, hydrophobic features.", "Hypo 1 was further validated by test set, Fischer randomization test, leave-one-out and decoy set methods.\n[SUBTITLE] Test set method [SUBSECTION] A total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\nA total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.\n[SUBTITLE] Fisher randomization method [SUBSECTION] This method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\nThis method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.\n[SUBTITLE] Decoy set method [SUBSECTION] The Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\nThe Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.\n[SUBTITLE] Leave-one-out method [SUBSECTION] The cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.\nThe cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.", "A total of 40 compounds structurally different from the training set compounds were selected as test set. The test set compounds were prepared in the same way training set compounds were prepared. The top-scored 10 hypotheses was regressed against 40 test set compounds and calculated the correlation coefficient values ranging from 0.917 to 0.859 (Table 1) between experimental and estimated activities. Among 10 hypotheses, Hypo 1 has given a correlation coefficient of 0.917 (Figure 4) indicating a good correlation between the estimated and experimental activities. The predictive ability of the Hypo 1 against test set compounds was considered better than other hypotheses and the estimated activity values along with the experimental and error values based on Hypo 1 are tabulated (See additional file 1: Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.). Most of the test set compounds was estimated correctly to their experimental activity. The test compounds were classified into four groups in a similar way as that of training set: most active (IC50 ≤ 100 nM, ++++), active (100 nM < IC50 ≤ 1000 nM, +++), moderately active (1000 nM < IC50 ≤ 10 000 nM, ++) and inactive (IC50 > 10 000 nM, +). A total of 11 out of 12 active (+++) compounds were estimated correctly as active, but 1 compound was estimated as most active (++++). Interestingly all the six most active (++++) compounds were estimated correctly as most active (++++). Total of twelve active (+++) compounds were estimated correctly as active. Out of thirteen moderately active (++) compounds only one compound was over estimated as active (+++), and 3 compounds were under estimated as inactive (+), whereas nine compounds were estimated correctly as moderately active. Among nine inactive (+) compounds one was over estimated as moderately active (++) compound whereas eight were estimated as inactive compounds. A total of 2 inactive (+) compounds out of 9 were over estimated as moderately active (++) whereas 7 were correctly estimated as inactive compounds. These results suggested that Hypo 1 has a good agreement with the experimental data and able to predict the activities of a wide variety of BACE-1 inhibitors.\nTest set correlation graph. Graph showing the correlation between experimental and Hypo 1 estimated activities of the 40 test set compounds along with 20 training set compounds.", "This method is used to evaluate the statistical significance of Hypo 1 based on the principle of randomizing the activity data of the training set compounds. During validation process random spreadsheets were generated using the training set molecules, and randomly reassigns the activity values to each compound. Subsequently generates the pharmacophore hypotheses using the same features and parameters used in the development of original hypothesis, Hypo 1. A total of 99 random spreadsheets (random hypotheses) required to be generated to achieve a confidence level of 99%. The results of top 10 random spreadsheets along with Hypo 1 are presented in Table 3. None of the top 99 radomly generated hypotheses has scored a total cost lower than the original hypothesis. The statistics of Hypo 1 is far more superior to the top 10 random hypotheses as well as the other 89 random hypotheses. This cross validation results clearly shows that the Hypo 1 was not generated by chance, and has strong confidence to represent a true correlation in the training set.\nFischer’s randomization test results of the pharmacophore hypothesis Hypo 1.", "The Hypo 1 was further validated using an external database for its ability to select BACE-1 inhibitors. This database contains a total (D) of 453 compounds including 206 active (A) compounds. Using Hypo 1, this database screening was carried out, 230 compounds were retrieved as hits (Ht). The results of GH score and E-value calculation are given in Table 4. Among 230 retrieved hit compounds, 197 compounds were from known actives (Ha). The false positive value is 33 and the false negative value is 9. The calculated E value was 1.88 indicates that the model is highly efficient for database screening. The GH value is expected to be greater than 0.7, which indicates a good model [38]. It was observed to be 0.76 for Hypo 1 and proving its ability in predicting the active compounds among inactives.\nStatistical parameters of GH score validation for Hypo 1.\n*[(Ha/4HtA)(3A + Ht) X (1 – (Ht – Ha)/(D –A))]; GH score of 0.6-0.8 indicates a very good model.", "The cross validation of the model was done using the leave-one-out method. This method is progressed by recomputing the pharmacophore hypotheses by leaving one compound at a time from the training set compounds. The importance of this validation is to prove that the correlation of the original pharmacophore hypothesis (Hypo 1) is not depending only on one particular compound. If the activity of each left-out compound is correctly estimated by the corresponding one-missing hypothesis then the test is positive. The feature composition of the pharmacophore, the value of correlation coefficient and the quality of the estimated activity of the left-out compound were used as measures for the assessment of the statistical test. By leaving each one of the 20 training set compounds according to this method, 20 new hypotheses were generated. As a result we did not obtain any meaningful differences between Hypo1 and each hypothesis resulting from the leave-one-out method. This result gives more confidence on Hypo 1 that it does not depend on one particular compound in the training set.", "The validated pharmacophore hypothesis, Hypo1, was used as a 3D structural query for retrieving compounds from chemical databases including MayBridge (59 652 compounds), Chembridge (50 000 compounds), NCI (238 819 compounds) and Asinex (213 462 compounds). As a result of first screening 11 578, 590, 5096 and 63 265 compounds were retrieved from Maybridge, Chembridge, NCI and Asinex respectively. Since the active site of BACE-1 is larger in size, the experimentally known most active inhibitors are also larger in size and violate the first rule of Lipinski’s rule of five. Hence, the retrieved hit compounds were filtered based only on the estimated activity values calculated by Hypo 1. The activity range for the most active compounds is <100 nM. Finally 773 compounds were selected by restricting the minimum estimated activity to <100 nM.", "To further refine the retrieved hits and also to remove the false positives, these 773 compounds along with the 20 training set compounds were docked into the active site of BACE-1 using GOLD 4.1 program. There are number of crystal structures for BACE-ligand complexes are available in PDB. The crystal structure of BACE-SC7 (PDB ID: 2QP8) complex was taken based on its high resolution. The GOLD fitness score was calculated for all the 793 compounds, it distinguishes molecules based on their interacting ability. The GOLD fitness score for the most active compound in the training set was 53.035. The compounds for further analysis were selected based on the ligand conformations which can satisfy the necessary interactions at the active site and scoring GOLD fitness score greater than 60. Finally 20 compounds from Maybridge and 15 compounds from Asinex have shown the required interaction with BACE-1 as well as good GOLD fitness scores. The compounds with the same chemical scaffolds were filtered carefully based on the molecular interactions observed at the active site. Finally, two compounds with different scaffolds one from Maybridge (RJC01726) and one from Asinex (Asnx-2) were selected as representative compounds. The binding mode of the final hits and the most active Compound 1 in the training set are shown in Figure 5. Figure 5A represents the binding mode of Compound 1 with a GOLD fitness score of 53.035. It has formed hydrogen bond interactions with D93, G95, T133, Q134, G291 and T293 and hydrophobic interactions with Y132, F169, and T292. The GOLD fitness score of RJC01726 was 68.289 and the mode of binding in the active site (Figure 5B) is similar to Compound 1. It has formed hydrogen bond interactions with T133, Q134 and T293 and hydrophobic interactions with D93, G95, F169 and T292. Asnx-2 has shown hydrogen bond interactions with T133, G291 and T293 as well as hydrophobic interactions with D93, Y132, F169 and T292 with a GOLD fitness score of 62.026 (Figure 5C). Figure 5D represents the overlay of most active Compound 1, RJC01726 and Asnx-2 at their binding modes. The pharmacophore overlay of the final hits compounds are shown in Figure 6 and their 2D representations are shown in Figure 7.\nMolecular docking results. Binding orientations of A) Compound 1 of the training set (cyan color), B) RJC01726 (magenta color), C) Asnx-2 (blue color), D) overlay of Compound 1, RJC01726 and Asnx-2 in the active site of BACE-1 protein. Active site residues are shown in stick form and hydrogen bond interactions are indicated with purple dotted lines.\nPharmacophore overlay on final hits. The mapping of pharmacophore hypothesis Hypo 1 on the final hits. A) RJC01726 (red color) B) Asnx-2 (cyan color).\nChemical structure of final hits. 2D representation of the final hits RJC01726 and Asnx-2.\nThe hydrophobic interactions of the final hits compounds were observed using Ligplot program [39]. The novelty of the two hits compounds were confirmed using SciFinder search [40] and PubChem search [41].", "A chemical feature based 3D pharmacophore hypotheses of BACE-1 inhibitors have been developed using 3D QSAR Pharmacophore Generation protocol available in DS 2.5. The best quantitative pharmacophore model, Hypo 1, was characterized by the highest cost difference (121.98), best correlation coefficient (0.977), lowest total cost value (81.24) and lowest RMSD (0.804). The fixed cost and null cost values were 74.77 and 203.22 bits, respectively. Hypo1 consisted of one HBD, one PI, one RA and two HY features. Hypo1 was further validated by test set, Fischer randomization test, leave-one-out, and decoy set methods. The test set containing 40 compounds was used in investigating the predictive ability of Hypo1 and resulted with a correlation coefficient of 0.917. Other validation methods also have provided reliable results on the strength of Hypo 1. This validated Hypo1 was used as a 3D query in database screening. The database hit compounds were subsequently subjected to filtering by estimated activity value. To further refine the retrieved hits the 793 compounds along with training set were carried out for molecular docking studies. The molecular docking result of all compounds was analyzed based on the GOLD fitness score, binding modes and molecular interactions with essential active site residues. Finally, two hits, namely, RJC01726 and Asnx-2 of different scaffolds with GOLD fitness score of 68.362 and 63.053, respectively, and interactions with important active site residues were chosen as lead candidates. These compounds as such and on further optimization can be used as potential leads in designing new BACE-1 inhibitors.", "The authors declare that they have no competing interests.", "SJ and ST equally involved in designing the work, analyzing the results and writing the manuscript. SS formatted and corrected the manuscript. KWL supervised the work and edited the manuscript. All four authors have read and approved the manuscript.", "Experimental and estimated IC50 values of the test set compounds based on the pharmacophore hypothesis ‘Hypo 1’.).\nClick here for file" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Algorithms", "Hydrogen Bonding", "Models, Statistical", "Molecular Dynamics Simulation", "Protein Stability", "Protein Structure, Secondary", "Proteins" ]

[ "Background", "I. Problem statement", "II. General approach", "III. Training algorithm", "Results", "I. Experimental setup", "II Generality of models trained on multiple trajectories", "III. Comparison with FIRST-energy model", "IV. Identification of least stable H-bonds", "Discussion", "Conclusions", "Authors' contributions", "Competing interests" ]

[ "A protein is a long sequence of amino-acids, called residues. Under normal physiological conditions, various forces (electrostatic, van der Waals, ...) lead the protein to fold into a compact structure made of secondary structure elements, α-helices and β-strands, connected by bends (called loops). An H-bond corresponds to the attractive electrostatic interaction between a covalent pair D—H of atoms, in which the hydrogen atom H is bonded to a more electronegative donor atom D, and an electronegative acceptor atom A. Due to their strong directional character, short distance ranges, and large number in folded proteins, H-bonds play a key role in both the formation and stabilization of protein structures [1-3]. While H-bonds involving atoms from close residues along the main-chain sequence stabilizes secondary structure elements, H-bonds between atoms in distant residues stabilize the overall 3D arrangement of secondary structure elements and loops.\nH-bonds form and break while the conformation of a protein deforms. For instance, the transition of a folded protein from a non-functional state into a functional (e.g., binding) state may require some H-bonds to break and others to form [4]. So, to better understand the possible deformation of a folded protein, it is desirable to create a reliable model of H-bond stability. Such a model makes it possible to identify rigid groups of atoms in a given protein conformation and determine the remaining degrees of freedom of the structure [7]. Since most H-bonds in a protein conformation are quite stable, it is crucial that the model precisely identifies the least stable bonds. The intrinsic strength of an individual H-bond has been studied before from an energetic viewpoint [5,6]. However, potential energy alone may not be a very good predictor of H-bond stability. Other local interactions may reinforce or weaken an H-bond.", "Let c be the conformation of a protein P at some time considered (with no loss of generality) to be 0 and H be an H-bond present in c. Let M (c) be the set of all physically possible trajectories of P passing through c and π be the probability distribution over this set. We define the stability of H in c over the time interval Δ by:(1)\nwhere I (q, H, t) is a Boolean function that takes value 1 if H is present in the conformation q(t) at time t along trajectory q, and 0 otherwise. The value can be interpreted as the probability that H will be present in the conformation of P at any specified time t ∈ (0, Δ), given that P is at conformation c at time 0. Our goal is to design a method for generating good approximations σ of . We also want these approximations to be protein-independent.", "We use machine learning methods to train a stability model σ from a given set Q of MD simulation trajectories. Each trajectory q ∈ Q is a sequence of conformations of a protein. These conformations are reached at times ti = i × δ, i = 0, 1, 2, …, called ticks, where δ is typically on the order of picoseconds. We detect the H-bonds present in each conformation q(ti) using the geometric criteria given in [8]. Note that an H-bond in a given protein is uniquely identified (across different conformations) by its donor, acceptor, and the hydrogen atom. So, we call the presence of a specific H-bond H in a conformation q(ti) an occurrence of H, denoted by h.\nFor each h, we compute a fixed list of predictors, some numerical, others categorical. Some are time-invariant, like the number of residues along the main-chain between the donor and acceptor atoms. Others are time-dependent. Among them, some describe the geometry of h, e.g., the distance between the hydrogen and the donor. Others describe the local environment of h, e.g., the number of other H-bonds within a certain distance from the mid-point of H.\nWe train σ as a function of these predictors. The predictor list defines a predictor space ∑ and every H-bond occurrence maps to a point in ∑. Given the input set Q of trajectories, we build a data table in which each row corresponds to an occurrence h of an H-bond present in a conformation q(ti) contained in Q. So, many rows may correspond to the same H-bond at different ticks. In our experiments, a typical data table contains several hundred thousand rows. Each column, except the last one, corresponds to a predictor p and the entry (h, p) of the table is the value of p for h. The entry in the last column is the measured stability y of h. More precisely, let H be the H-bond of which h is an occurrence. Let l = Δ/δ, where Δ is the duration over which we wish to predict the stability of h, and m ≤ l be the number of ticks tk, k = i + 1, i + 2,…,i + l, such that H is present in q(tk). The measured stability y of h is the ratio m/l. We chose l = 50 in most of the tests reported below, as this value both provides a ratio m/l large enough for the measured stability to be statistically meaningful, and corresponds to an interesting prediction timescale (50ps). Typically, most H-bond occurrences are quite stable: over 25% have measured stability 1, about 50% higher than 0.8, and only 15% less than 0.3.\nWe build σ as a binary regression tree [9]. This machine learning approach has been one of the most successful in practice. Regression trees are often simple to interpret. The method can work with both categorical and numerical predictors in a unified way, as shown in Section III. Each non-leaf node in a regression tree is a Boolean split. So, each node N of the tree determines a region of ∑ in which all the splits associated with the arcs connecting the root of the tree to N are satisfied. We say that an H-bond occurrence falls into N if it is contained in this region. The predicted stability value stored at a leaf node L is the average of the measured stability values by all the H-bond occurrences in the training data table that fall into L. We expect this average, which is taken over many pieces of trajectories, to approximate well the average defined in Equation (1).\nOnce a regression tree has been generated, it is used as follows. Given an H-bond H in an arbitrary conformation c of an arbitrary protein, the leaf node L of the tree into which H falls is identified by calculating the values of the necessary predictors for H in c. The predicted stability value stored at L is returned.", "We construct a model σ as a binary regression tree using the CART method [9]. The tree is generated recursively in a top-down fashion. When a new node N is created, it is inserted as a leaf of the tree if a predefined depth has been reached or if the number of h falling into N is smaller than a predefined threshold. Otherwise, N is added as an intermediate node, its split is computed, and its left and right children are created. A split s is defined by a pair (p, r), where p is the split predictor and r is the split value. If p is a numerical predictor, then r is a threshold on p, and s ≜ p <r. If p is a categorical predictor, then r is a subset of categories, and s ≜ p ∈ r. We define the score w(p, r) of split s = (p, r) at a node N as the reduction of variance in measured stability that results from s. The algorithm chooses the split—both the predictor and the split value—that has the largest score. Only a relatively small subset of predictors selected by the training algorithm is eventually used in a regression tree.\nTo prevent model overfitting, we limit tree depth to 5 in most of our experiments and limit the minimal number of training samples in an intermediate node to be 10. We further prune the obtained tree using the following adaptive algorithm. We initially set aside a fraction of the training data table called validation subset. Once a tree has been constructed pruning is an iterative process. At each step, one intermediate node N whose split has minimal score becomes a leaf node by removing the sub-tree rooted at N. This process creates a sequence of trees with decreasing numbers of nodes. We compute the mean square error of the predictions made by each tree on the validation subset. The tree with the smallest error is selected.", "[SUBTITLE] I. Experimental setup [SUBSECTION] We used 6 MD simulation trajectories picked from different sources and called hereafter 1c9oA, 1e85A, 1g9oA_1, and 1g9oA_2 from [10], complex from [11], and 1eia (generated by us). In all of them the time interval δ between two successive ticks is 1ps. Table 1 indicates the protein simulated in each trajectory, its number of residues, the force field used by the simulator, and the duration of the trajectory. Each trajectory starts from a folded conformation resolved by X-ray crystallography.\nCharacteristics of the MD simulation trajectories used to create the 6 datasets\nFrom each trajectory we derived a separate data table in which the rows represent H-bond occurrences. Last two columns in Table 1 list the number of distinct H-bonds detected in each trajectory and the total number of H-bond occurrences extracted. Note that complex was generated for a complex of two molecules. All H-bonds occurring in this complex are taken into account in the corresponding data table.\nThe values of the time-varying predictors are subject to thermal noise. Since a model σ will in general be used to predict H-bond stability in a protein conformation sampled using a kinematic model ignoring thermal noise (e.g., by sampling the dihedral angles ϕ, ψ, and χ) [7], we chose to average the values of these predictors over l' ticks to remove thermal noise. More precisely, the value of a predictor stored in the row of the data table corresponding to an H-bond occurrence in q(ti) is the average value of this predictor in , where . Our analysis shows that l' = 50 is near optimal.\nThe performance of a regression model can be measured by the root mean square error (RMSE) of the predictions on a test dataset. For a data table T = {(x1, y1), (x2, y2)},…, (xn, yn)}, where each xi, i = 1,…,n, denotes a vector of predictor values for an H-bond occurrence and yi is the measured stability of the H-bond, and a model σ, the RMSE is defined by: . As RMSE depends not only on the accuracy of σ, but also on the table T, some normalization is necessary to compare results on different tables. So, in our tests we compute the decrease of RMSE relative to a base model σ0. The relative base error decrease (or RBED) is then defined by: In most cases, σ0 is simply defined by , i.e., the average measured stability of all H-bond occurrences in the dataset. In other cases, σ0 is a model based on the H-bond energy.\nWe used 6 MD simulation trajectories picked from different sources and called hereafter 1c9oA, 1e85A, 1g9oA_1, and 1g9oA_2 from [10], complex from [11], and 1eia (generated by us). In all of them the time interval δ between two successive ticks is 1ps. Table 1 indicates the protein simulated in each trajectory, its number of residues, the force field used by the simulator, and the duration of the trajectory. Each trajectory starts from a folded conformation resolved by X-ray crystallography.\nCharacteristics of the MD simulation trajectories used to create the 6 datasets\nFrom each trajectory we derived a separate data table in which the rows represent H-bond occurrences. Last two columns in Table 1 list the number of distinct H-bonds detected in each trajectory and the total number of H-bond occurrences extracted. Note that complex was generated for a complex of two molecules. All H-bonds occurring in this complex are taken into account in the corresponding data table.\nThe values of the time-varying predictors are subject to thermal noise. Since a model σ will in general be used to predict H-bond stability in a protein conformation sampled using a kinematic model ignoring thermal noise (e.g., by sampling the dihedral angles ϕ, ψ, and χ) [7], we chose to average the values of these predictors over l' ticks to remove thermal noise. More precisely, the value of a predictor stored in the row of the data table corresponding to an H-bond occurrence in q(ti) is the average value of this predictor in , where . Our analysis shows that l' = 50 is near optimal.\nThe performance of a regression model can be measured by the root mean square error (RMSE) of the predictions on a test dataset. For a data table T = {(x1, y1), (x2, y2)},…, (xn, yn)}, where each xi, i = 1,…,n, denotes a vector of predictor values for an H-bond occurrence and yi is the measured stability of the H-bond, and a model σ, the RMSE is defined by: . As RMSE depends not only on the accuracy of σ, but also on the table T, some normalization is necessary to compare results on different tables. So, in our tests we compute the decrease of RMSE relative to a base model σ0. The relative base error decrease (or RBED) is then defined by: In most cases, σ0 is simply defined by , i.e., the average measured stability of all H-bond occurrences in the dataset. In other cases, σ0 is a model based on the H-bond energy.\n[SUBTITLE] II Generality of models trained on multiple trajectories [SUBSECTION] Our goal is to train models to predict the stability of H-bonds in any protein. So, we trained models on data tables obtained by mixing subsets of 5 data tables and we tested these models on the remaining data table. For each combination of 5 data tables, we trained 4 groups of models varying in the tree’s maximal depth (5 or 15) and in the fraction of H-bond occurrences randomly taken from each data table (10% or 50%). For each group we trained 10 models. Hence, in total, 240 models were generated. Table 2 shows the mean RBED value for each combination of data tables and each group of models. In columns 3 through 8 we indicate the data table used for testing the models trained on a combination of the 5 other data tables. Figure 1 shows a partial tree trained with combinations of all tables, except 1c9oA.\nMean RBED values of models trained on multiple trajectories\nTop 2 layers of a regression tree trained with combination of all tables, except 1c9oA. The actual tree contains 55 nodes. Each path from the root to a node defines a conjunction of criteria for H-bonds with a certain mean stability. Here, Dist_H_A (the distance between the hydrogen and the acceptor atoms) is the most differentiating predictor. For H-bonds with Dist_H_A≥2.40Å, the mean stability is only 0.38, but it increases to 0.92 if Dist_H_A<2.40Å.\nRBED values show that regression tree model significantly reduces base error and keeps predictive power when applied to a protein not present in the training data. Moreover, the variance of RBED values is very small, meaning that the training process yields models that are stable in performance. Furthermore, the RBED values are lower for models tested on complex. Recall that the trajectory complex was generated for a complex made of a protein and a ligand, while every other trajectory was generated for a single protein. So, it is likely that complex contains H-bonds in situations that did not occur in any of the other trajectories. Both deeper trees and larger data fractions tend to improve model accuracy, but the very small gain is not worth the additional model or computation complexity.\nOur goal is to train models to predict the stability of H-bonds in any protein. So, we trained models on data tables obtained by mixing subsets of 5 data tables and we tested these models on the remaining data table. For each combination of 5 data tables, we trained 4 groups of models varying in the tree’s maximal depth (5 or 15) and in the fraction of H-bond occurrences randomly taken from each data table (10% or 50%). For each group we trained 10 models. Hence, in total, 240 models were generated. Table 2 shows the mean RBED value for each combination of data tables and each group of models. In columns 3 through 8 we indicate the data table used for testing the models trained on a combination of the 5 other data tables. Figure 1 shows a partial tree trained with combinations of all tables, except 1c9oA.\nMean RBED values of models trained on multiple trajectories\nTop 2 layers of a regression tree trained with combination of all tables, except 1c9oA. The actual tree contains 55 nodes. Each path from the root to a node defines a conjunction of criteria for H-bonds with a certain mean stability. Here, Dist_H_A (the distance between the hydrogen and the acceptor atoms) is the most differentiating predictor. For H-bonds with Dist_H_A≥2.40Å, the mean stability is only 0.38, but it increases to 0.92 if Dist_H_A<2.40Å.\nRBED values show that regression tree model significantly reduces base error and keeps predictive power when applied to a protein not present in the training data. Moreover, the variance of RBED values is very small, meaning that the training process yields models that are stable in performance. Furthermore, the RBED values are lower for models tested on complex. Recall that the trajectory complex was generated for a complex made of a protein and a ligand, while every other trajectory was generated for a single protein. So, it is likely that complex contains H-bonds in situations that did not occur in any of the other trajectories. Both deeper trees and larger data fractions tend to improve model accuracy, but the very small gain is not worth the additional model or computation complexity.\n[SUBTITLE] III. Comparison with FIRST-energy model [SUBSECTION] We've checked whether regression models can predict the stability of H-bonds more accurately than potential energy alone. Table 3 presents the mean RBED value for a model obtained in the first row of Table 2 relative to the base model that is a regression tree built from the same training data using FIRST_energy as the only predictor. FIRST_energy is a modified Mayo potential [5] implemented in FIRST (a protein rigidity analysis software) [7]. Comparison on all 6 data tables show that the more complex models are significantly more accurate than the models based on FIRST_energy alone.\nMean RBED values of models using single predictor FIRST_energy\nWe've checked whether regression models can predict the stability of H-bonds more accurately than potential energy alone. Table 3 presents the mean RBED value for a model obtained in the first row of Table 2 relative to the base model that is a regression tree built from the same training data using FIRST_energy as the only predictor. FIRST_energy is a modified Mayo potential [5] implemented in FIRST (a protein rigidity analysis software) [7]. Comparison on all 6 data tables show that the more complex models are significantly more accurate than the models based on FIRST_energy alone.\nMean RBED values of models using single predictor FIRST_energy\n[SUBTITLE] IV. Identification of least stable H-bonds [SUBSECTION] Most H-bond occurrences tend to be stable. So, accurately identifying the weakest ones is important if one wishes to predict the possible deformation of a protein [7]. To evaluate how well our models identify the least stable H-bonds occurrences, we first identify the subset S of the 10% H-bond occurrences with the smallest measured stability in each test table T. Using a regression tree σ obtained in Section II, we sort the H-bond occurrences in T in ascending order of predicted stability and we compute the fraction w ∈ [0,1] of S that is contained in the first 100×u% occurrences in this sorted list, for successive values of u ∈ [0,1]. We call the function w(u) the identification curve of the least stable H-bonds for σ.\nFigure 2 plots the identification curve for 1c9oA. It consists of three curves: the red curve is the (fictitious) ideal identification curve, the blue curve is obtained with one (randomly picked) regression tree computed in Section II, and the green curve is obtained by sorting H-bond occurrences in decreasing values of FIRST_energy. Plots on other proteins present similar curve shapes. For models tested on data tables except complex, about 80% of the 10% H-bond occurrences predicted as the least stable are actually among the 10% truly least stable. The results for complex are less satisfactory because of the reasons discussed in Section II. The regression models are consistently better than the FIRST_energy-only models, though for 1eia the difference is small.\nIdentification curves of the least stable bonds for 1c9oA (see Results, Section IV).\nMost H-bond occurrences tend to be stable. So, accurately identifying the weakest ones is important if one wishes to predict the possible deformation of a protein [7]. To evaluate how well our models identify the least stable H-bonds occurrences, we first identify the subset S of the 10% H-bond occurrences with the smallest measured stability in each test table T. Using a regression tree σ obtained in Section II, we sort the H-bond occurrences in T in ascending order of predicted stability and we compute the fraction w ∈ [0,1] of S that is contained in the first 100×u% occurrences in this sorted list, for successive values of u ∈ [0,1]. We call the function w(u) the identification curve of the least stable H-bonds for σ.\nFigure 2 plots the identification curve for 1c9oA. It consists of three curves: the red curve is the (fictitious) ideal identification curve, the blue curve is obtained with one (randomly picked) regression tree computed in Section II, and the green curve is obtained by sorting H-bond occurrences in decreasing values of FIRST_energy. Plots on other proteins present similar curve shapes. For models tested on data tables except complex, about 80% of the 10% H-bond occurrences predicted as the least stable are actually among the 10% truly least stable. The results for complex are less satisfactory because of the reasons discussed in Section II. The regression models are consistently better than the FIRST_energy-only models, though for 1eia the difference is small.\nIdentification curves of the least stable bonds for 1c9oA (see Results, Section IV).", "We used 6 MD simulation trajectories picked from different sources and called hereafter 1c9oA, 1e85A, 1g9oA_1, and 1g9oA_2 from [10], complex from [11], and 1eia (generated by us). In all of them the time interval δ between two successive ticks is 1ps. Table 1 indicates the protein simulated in each trajectory, its number of residues, the force field used by the simulator, and the duration of the trajectory. Each trajectory starts from a folded conformation resolved by X-ray crystallography.\nCharacteristics of the MD simulation trajectories used to create the 6 datasets\nFrom each trajectory we derived a separate data table in which the rows represent H-bond occurrences. Last two columns in Table 1 list the number of distinct H-bonds detected in each trajectory and the total number of H-bond occurrences extracted. Note that complex was generated for a complex of two molecules. All H-bonds occurring in this complex are taken into account in the corresponding data table.\nThe values of the time-varying predictors are subject to thermal noise. Since a model σ will in general be used to predict H-bond stability in a protein conformation sampled using a kinematic model ignoring thermal noise (e.g., by sampling the dihedral angles ϕ, ψ, and χ) [7], we chose to average the values of these predictors over l' ticks to remove thermal noise. More precisely, the value of a predictor stored in the row of the data table corresponding to an H-bond occurrence in q(ti) is the average value of this predictor in , where . Our analysis shows that l' = 50 is near optimal.\nThe performance of a regression model can be measured by the root mean square error (RMSE) of the predictions on a test dataset. For a data table T = {(x1, y1), (x2, y2)},…, (xn, yn)}, where each xi, i = 1,…,n, denotes a vector of predictor values for an H-bond occurrence and yi is the measured stability of the H-bond, and a model σ, the RMSE is defined by: . As RMSE depends not only on the accuracy of σ, but also on the table T, some normalization is necessary to compare results on different tables. So, in our tests we compute the decrease of RMSE relative to a base model σ0. The relative base error decrease (or RBED) is then defined by: In most cases, σ0 is simply defined by , i.e., the average measured stability of all H-bond occurrences in the dataset. In other cases, σ0 is a model based on the H-bond energy.", "Our goal is to train models to predict the stability of H-bonds in any protein. So, we trained models on data tables obtained by mixing subsets of 5 data tables and we tested these models on the remaining data table. For each combination of 5 data tables, we trained 4 groups of models varying in the tree’s maximal depth (5 or 15) and in the fraction of H-bond occurrences randomly taken from each data table (10% or 50%). For each group we trained 10 models. Hence, in total, 240 models were generated. Table 2 shows the mean RBED value for each combination of data tables and each group of models. In columns 3 through 8 we indicate the data table used for testing the models trained on a combination of the 5 other data tables. Figure 1 shows a partial tree trained with combinations of all tables, except 1c9oA.\nMean RBED values of models trained on multiple trajectories\nTop 2 layers of a regression tree trained with combination of all tables, except 1c9oA. The actual tree contains 55 nodes. Each path from the root to a node defines a conjunction of criteria for H-bonds with a certain mean stability. Here, Dist_H_A (the distance between the hydrogen and the acceptor atoms) is the most differentiating predictor. For H-bonds with Dist_H_A≥2.40Å, the mean stability is only 0.38, but it increases to 0.92 if Dist_H_A<2.40Å.\nRBED values show that regression tree model significantly reduces base error and keeps predictive power when applied to a protein not present in the training data. Moreover, the variance of RBED values is very small, meaning that the training process yields models that are stable in performance. Furthermore, the RBED values are lower for models tested on complex. Recall that the trajectory complex was generated for a complex made of a protein and a ligand, while every other trajectory was generated for a single protein. So, it is likely that complex contains H-bonds in situations that did not occur in any of the other trajectories. Both deeper trees and larger data fractions tend to improve model accuracy, but the very small gain is not worth the additional model or computation complexity.", "We've checked whether regression models can predict the stability of H-bonds more accurately than potential energy alone. Table 3 presents the mean RBED value for a model obtained in the first row of Table 2 relative to the base model that is a regression tree built from the same training data using FIRST_energy as the only predictor. FIRST_energy is a modified Mayo potential [5] implemented in FIRST (a protein rigidity analysis software) [7]. Comparison on all 6 data tables show that the more complex models are significantly more accurate than the models based on FIRST_energy alone.\nMean RBED values of models using single predictor FIRST_energy", "Most H-bond occurrences tend to be stable. So, accurately identifying the weakest ones is important if one wishes to predict the possible deformation of a protein [7]. To evaluate how well our models identify the least stable H-bonds occurrences, we first identify the subset S of the 10% H-bond occurrences with the smallest measured stability in each test table T. Using a regression tree σ obtained in Section II, we sort the H-bond occurrences in T in ascending order of predicted stability and we compute the fraction w ∈ [0,1] of S that is contained in the first 100×u% occurrences in this sorted list, for successive values of u ∈ [0,1]. We call the function w(u) the identification curve of the least stable H-bonds for σ.\nFigure 2 plots the identification curve for 1c9oA. It consists of three curves: the red curve is the (fictitious) ideal identification curve, the blue curve is obtained with one (randomly picked) regression tree computed in Section II, and the green curve is obtained by sorting H-bond occurrences in decreasing values of FIRST_energy. Plots on other proteins present similar curve shapes. For models tested on data tables except complex, about 80% of the 10% H-bond occurrences predicted as the least stable are actually among the 10% truly least stable. The results for complex are less satisfactory because of the reasons discussed in Section II. The regression models are consistently better than the FIRST_energy-only models, though for 1eia the difference is small.\nIdentification curves of the least stable bonds for 1c9oA (see Results, Section IV).", "In all our regression trees the root split was done with predictor Dist_H_A (the distance between the hydrogen and acceptor atoms), which therefore appear as the single most discriminative attribute to predict H-bond stability. This observation is consistent with previous findings. Levitt [6] found that most stable H-bonds have Dist_H_A less than 2.07Å. Jeffrey and Saenger [14] also suggested that Dist_H_A is a key attribute affecting H-bond stability, with a value less than 2.2Å for moderate to strong H-bonds. Consistent with these previous findings, the split values of the deepest Dist_H_A nodes in all our regression trees are around 2.1Å. This distance was observed in [6] to sometimes fluctuate by up to 3Å in stable H-bonds, due to high-frequency atomic vibration. This observation supports our decision to average predictor values over windows of l’ ticks.\nPredictor FIRST_energy is often used in splits close to the root. This is not surprising since it is a function of several other pertinent predictors: Dist_H_A, Angle_D_H_A (the angle between the donor, the hydrogen atom, and the acceptor), Angle_H_A_AA (the angle between the hydrogen atom, the acceptor, and the atom covalently-bonded to the acceptor), and the hybridization state of the bond. Some other distance-based predictors (Dist_D_AA, Dist_D_A, Dist_H_D), angle-based predictors and Ch_type (describing whether the donor and acceptor are from main-chain or side-chain) predictor appear often in regression trees, but closer to the leaf nodes. They nevertheless play a significant role in predicting H-bond stability. For example, as shown in Figure 1, if Angle_H_A_AA is at least 105°, the stability is very high (about 0.96); otherwise, it drops to 0.71. The preference for larger angle matches well with the well-known linearity of H-bonds [14].\nIn order to get a more quantitative measure of the relative impact of the predictors on H-bond stability, we define the importance of a predictor p in a regression tree by: , where Np is the set of nodes where the split is made using p, w(s) is the score of the split s, and n(s) is the number of H-bond occurrences falling into the node where split s is made. We trained 10 models on data tables combining 10% of each of the 6 data tables. Importance scores for each predictor were averaged over these models and then linearly scaled to adjust the score of the least important predictor (with non-zero average importance) equal to 1. The average importance of every predictor appearing in at least one model is shown in Figure 3. The figure confirms that distance-based and angle-based predictors, as well as FIRST_energy, are the most important. It also shows that a number of other predictors—including Resi_name_H, Resi_name_A, and Range (difference in residue numbers of donor and acceptor) —have less, but still significant importance.\nPredictor importance scores\nOverall, we observe that predictors that describe the local environment of an H-bond play a relatively small role in predicting its stability. In particular, we had expected that descriptors such Num_hb_spaceNbr and Num_hb_spaceRgdNbr, which count the number of other H-bonds located in the neighborhood of the analyzed H-bond, would have had more importance. However, this may reflect the fact that the MD simulation trajectories used in our tests are too short to contain enough information to infer the role of such predictors. Indeed, while transitions between meta-stable states are rare in those trajectories, predictors describing local environments may have greater influence on the stability of H-bonds that must break for such transitions to happen. So, longer trajectories may eventually be needed to better model H-bond stability.", "We have described machine learning methods to train protein-independent regression trees modeling H-bond stability in proteins. Test results demonstrate that trained models can predict H-bond stability quite well. In particular, their performance is significantly better (roughly 20% better) than that of a model based on H-bond energy alone. They can accurately identify a large fraction of the least stable H-bonds in a given conformation. However, our results also suggest that better results could be obtained with a richer set of MD simulation trajectories. In particular, the trajectories used in our experiments might be too short to characterize the stability of H-bonds that break and form during a transition between meta-stable states.\nWe believe that the training methods could be improved in several ways:\n- It would be better to averaging predictor values before sub-sampling MD simulation trajectories. This would reduce the risk of filtering out changes in predictor values that are important for H-bond stability. Unfortunately, in our trajectories we only had access to the data after sub-sampling.\n- More sophisticated learning techniques could be used. For example, instead of generating a single tree, we could generate an ensemble of trees, such as Gradient Boosting Trees [16].\n- Finally, the notion of stability itself could be refined, for example by distinguishing between the case where an H-bond frequently switches on and off during a prediction window and the case where it rarely switches.", "All four authors, IC, PY, MM, and JCL, participated in the formulation of the problem, the design of its solution, and the analysis and the interpretation of the results. PY prepared the experimental data. IC adapted a previously available CART software package and ran the experiments. All authors contributed to the writing of the manuscript, read and approved the final manuscript.", "The authors declare that they have no competing interests." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "I. Problem statement", "II. General approach", "III. Training algorithm", "Results", "I. Experimental setup", "II Generality of models trained on multiple trajectories", "III. Comparison with FIRST-energy model", "IV. Identification of least stable H-bonds", "Discussion", "Conclusions", "Authors' contributions", "Competing interests" ]

[ "A protein is a long sequence of amino-acids, called residues. Under normal physiological conditions, various forces (electrostatic, van der Waals, ...) lead the protein to fold into a compact structure made of secondary structure elements, α-helices and β-strands, connected by bends (called loops). An H-bond corresponds to the attractive electrostatic interaction between a covalent pair D—H of atoms, in which the hydrogen atom H is bonded to a more electronegative donor atom D, and an electronegative acceptor atom A. Due to their strong directional character, short distance ranges, and large number in folded proteins, H-bonds play a key role in both the formation and stabilization of protein structures [1-3]. While H-bonds involving atoms from close residues along the main-chain sequence stabilizes secondary structure elements, H-bonds between atoms in distant residues stabilize the overall 3D arrangement of secondary structure elements and loops.\nH-bonds form and break while the conformation of a protein deforms. For instance, the transition of a folded protein from a non-functional state into a functional (e.g., binding) state may require some H-bonds to break and others to form [4]. So, to better understand the possible deformation of a folded protein, it is desirable to create a reliable model of H-bond stability. Such a model makes it possible to identify rigid groups of atoms in a given protein conformation and determine the remaining degrees of freedom of the structure [7]. Since most H-bonds in a protein conformation are quite stable, it is crucial that the model precisely identifies the least stable bonds. The intrinsic strength of an individual H-bond has been studied before from an energetic viewpoint [5,6]. However, potential energy alone may not be a very good predictor of H-bond stability. Other local interactions may reinforce or weaken an H-bond.", "[SUBTITLE] I. Problem statement [SUBSECTION] Let c be the conformation of a protein P at some time considered (with no loss of generality) to be 0 and H be an H-bond present in c. Let M (c) be the set of all physically possible trajectories of P passing through c and π be the probability distribution over this set. We define the stability of H in c over the time interval Δ by:(1)\nwhere I (q, H, t) is a Boolean function that takes value 1 if H is present in the conformation q(t) at time t along trajectory q, and 0 otherwise. The value can be interpreted as the probability that H will be present in the conformation of P at any specified time t ∈ (0, Δ), given that P is at conformation c at time 0. Our goal is to design a method for generating good approximations σ of . We also want these approximations to be protein-independent.\nLet c be the conformation of a protein P at some time considered (with no loss of generality) to be 0 and H be an H-bond present in c. Let M (c) be the set of all physically possible trajectories of P passing through c and π be the probability distribution over this set. We define the stability of H in c over the time interval Δ by:(1)\nwhere I (q, H, t) is a Boolean function that takes value 1 if H is present in the conformation q(t) at time t along trajectory q, and 0 otherwise. The value can be interpreted as the probability that H will be present in the conformation of P at any specified time t ∈ (0, Δ), given that P is at conformation c at time 0. Our goal is to design a method for generating good approximations σ of . We also want these approximations to be protein-independent.\n[SUBTITLE] II. General approach [SUBSECTION] We use machine learning methods to train a stability model σ from a given set Q of MD simulation trajectories. Each trajectory q ∈ Q is a sequence of conformations of a protein. These conformations are reached at times ti = i × δ, i = 0, 1, 2, …, called ticks, where δ is typically on the order of picoseconds. We detect the H-bonds present in each conformation q(ti) using the geometric criteria given in [8]. Note that an H-bond in a given protein is uniquely identified (across different conformations) by its donor, acceptor, and the hydrogen atom. So, we call the presence of a specific H-bond H in a conformation q(ti) an occurrence of H, denoted by h.\nFor each h, we compute a fixed list of predictors, some numerical, others categorical. Some are time-invariant, like the number of residues along the main-chain between the donor and acceptor atoms. Others are time-dependent. Among them, some describe the geometry of h, e.g., the distance between the hydrogen and the donor. Others describe the local environment of h, e.g., the number of other H-bonds within a certain distance from the mid-point of H.\nWe train σ as a function of these predictors. The predictor list defines a predictor space ∑ and every H-bond occurrence maps to a point in ∑. Given the input set Q of trajectories, we build a data table in which each row corresponds to an occurrence h of an H-bond present in a conformation q(ti) contained in Q. So, many rows may correspond to the same H-bond at different ticks. In our experiments, a typical data table contains several hundred thousand rows. Each column, except the last one, corresponds to a predictor p and the entry (h, p) of the table is the value of p for h. The entry in the last column is the measured stability y of h. More precisely, let H be the H-bond of which h is an occurrence. Let l = Δ/δ, where Δ is the duration over which we wish to predict the stability of h, and m ≤ l be the number of ticks tk, k = i + 1, i + 2,…,i + l, such that H is present in q(tk). The measured stability y of h is the ratio m/l. We chose l = 50 in most of the tests reported below, as this value both provides a ratio m/l large enough for the measured stability to be statistically meaningful, and corresponds to an interesting prediction timescale (50ps). Typically, most H-bond occurrences are quite stable: over 25% have measured stability 1, about 50% higher than 0.8, and only 15% less than 0.3.\nWe build σ as a binary regression tree [9]. This machine learning approach has been one of the most successful in practice. Regression trees are often simple to interpret. The method can work with both categorical and numerical predictors in a unified way, as shown in Section III. Each non-leaf node in a regression tree is a Boolean split. So, each node N of the tree determines a region of ∑ in which all the splits associated with the arcs connecting the root of the tree to N are satisfied. We say that an H-bond occurrence falls into N if it is contained in this region. The predicted stability value stored at a leaf node L is the average of the measured stability values by all the H-bond occurrences in the training data table that fall into L. We expect this average, which is taken over many pieces of trajectories, to approximate well the average defined in Equation (1).\nOnce a regression tree has been generated, it is used as follows. Given an H-bond H in an arbitrary conformation c of an arbitrary protein, the leaf node L of the tree into which H falls is identified by calculating the values of the necessary predictors for H in c. The predicted stability value stored at L is returned.\nWe use machine learning methods to train a stability model σ from a given set Q of MD simulation trajectories. Each trajectory q ∈ Q is a sequence of conformations of a protein. These conformations are reached at times ti = i × δ, i = 0, 1, 2, …, called ticks, where δ is typically on the order of picoseconds. We detect the H-bonds present in each conformation q(ti) using the geometric criteria given in [8]. Note that an H-bond in a given protein is uniquely identified (across different conformations) by its donor, acceptor, and the hydrogen atom. So, we call the presence of a specific H-bond H in a conformation q(ti) an occurrence of H, denoted by h.\nFor each h, we compute a fixed list of predictors, some numerical, others categorical. Some are time-invariant, like the number of residues along the main-chain between the donor and acceptor atoms. Others are time-dependent. Among them, some describe the geometry of h, e.g., the distance between the hydrogen and the donor. Others describe the local environment of h, e.g., the number of other H-bonds within a certain distance from the mid-point of H.\nWe train σ as a function of these predictors. The predictor list defines a predictor space ∑ and every H-bond occurrence maps to a point in ∑. Given the input set Q of trajectories, we build a data table in which each row corresponds to an occurrence h of an H-bond present in a conformation q(ti) contained in Q. So, many rows may correspond to the same H-bond at different ticks. In our experiments, a typical data table contains several hundred thousand rows. Each column, except the last one, corresponds to a predictor p and the entry (h, p) of the table is the value of p for h. The entry in the last column is the measured stability y of h. More precisely, let H be the H-bond of which h is an occurrence. Let l = Δ/δ, where Δ is the duration over which we wish to predict the stability of h, and m ≤ l be the number of ticks tk, k = i + 1, i + 2,…,i + l, such that H is present in q(tk). The measured stability y of h is the ratio m/l. We chose l = 50 in most of the tests reported below, as this value both provides a ratio m/l large enough for the measured stability to be statistically meaningful, and corresponds to an interesting prediction timescale (50ps). Typically, most H-bond occurrences are quite stable: over 25% have measured stability 1, about 50% higher than 0.8, and only 15% less than 0.3.\nWe build σ as a binary regression tree [9]. This machine learning approach has been one of the most successful in practice. Regression trees are often simple to interpret. The method can work with both categorical and numerical predictors in a unified way, as shown in Section III. Each non-leaf node in a regression tree is a Boolean split. So, each node N of the tree determines a region of ∑ in which all the splits associated with the arcs connecting the root of the tree to N are satisfied. We say that an H-bond occurrence falls into N if it is contained in this region. The predicted stability value stored at a leaf node L is the average of the measured stability values by all the H-bond occurrences in the training data table that fall into L. We expect this average, which is taken over many pieces of trajectories, to approximate well the average defined in Equation (1).\nOnce a regression tree has been generated, it is used as follows. Given an H-bond H in an arbitrary conformation c of an arbitrary protein, the leaf node L of the tree into which H falls is identified by calculating the values of the necessary predictors for H in c. The predicted stability value stored at L is returned.\n[SUBTITLE] III. Training algorithm [SUBSECTION] We construct a model σ as a binary regression tree using the CART method [9]. The tree is generated recursively in a top-down fashion. When a new node N is created, it is inserted as a leaf of the tree if a predefined depth has been reached or if the number of h falling into N is smaller than a predefined threshold. Otherwise, N is added as an intermediate node, its split is computed, and its left and right children are created. A split s is defined by a pair (p, r), where p is the split predictor and r is the split value. If p is a numerical predictor, then r is a threshold on p, and s ≜ p <r. If p is a categorical predictor, then r is a subset of categories, and s ≜ p ∈ r. We define the score w(p, r) of split s = (p, r) at a node N as the reduction of variance in measured stability that results from s. The algorithm chooses the split—both the predictor and the split value—that has the largest score. Only a relatively small subset of predictors selected by the training algorithm is eventually used in a regression tree.\nTo prevent model overfitting, we limit tree depth to 5 in most of our experiments and limit the minimal number of training samples in an intermediate node to be 10. We further prune the obtained tree using the following adaptive algorithm. We initially set aside a fraction of the training data table called validation subset. Once a tree has been constructed pruning is an iterative process. At each step, one intermediate node N whose split has minimal score becomes a leaf node by removing the sub-tree rooted at N. This process creates a sequence of trees with decreasing numbers of nodes. We compute the mean square error of the predictions made by each tree on the validation subset. The tree with the smallest error is selected.\nWe construct a model σ as a binary regression tree using the CART method [9]. The tree is generated recursively in a top-down fashion. When a new node N is created, it is inserted as a leaf of the tree if a predefined depth has been reached or if the number of h falling into N is smaller than a predefined threshold. Otherwise, N is added as an intermediate node, its split is computed, and its left and right children are created. A split s is defined by a pair (p, r), where p is the split predictor and r is the split value. If p is a numerical predictor, then r is a threshold on p, and s ≜ p <r. If p is a categorical predictor, then r is a subset of categories, and s ≜ p ∈ r. We define the score w(p, r) of split s = (p, r) at a node N as the reduction of variance in measured stability that results from s. The algorithm chooses the split—both the predictor and the split value—that has the largest score. Only a relatively small subset of predictors selected by the training algorithm is eventually used in a regression tree.\nTo prevent model overfitting, we limit tree depth to 5 in most of our experiments and limit the minimal number of training samples in an intermediate node to be 10. We further prune the obtained tree using the following adaptive algorithm. We initially set aside a fraction of the training data table called validation subset. Once a tree has been constructed pruning is an iterative process. At each step, one intermediate node N whose split has minimal score becomes a leaf node by removing the sub-tree rooted at N. This process creates a sequence of trees with decreasing numbers of nodes. We compute the mean square error of the predictions made by each tree on the validation subset. The tree with the smallest error is selected.", "Let c be the conformation of a protein P at some time considered (with no loss of generality) to be 0 and H be an H-bond present in c. Let M (c) be the set of all physically possible trajectories of P passing through c and π be the probability distribution over this set. We define the stability of H in c over the time interval Δ by:(1)\nwhere I (q, H, t) is a Boolean function that takes value 1 if H is present in the conformation q(t) at time t along trajectory q, and 0 otherwise. The value can be interpreted as the probability that H will be present in the conformation of P at any specified time t ∈ (0, Δ), given that P is at conformation c at time 0. Our goal is to design a method for generating good approximations σ of . We also want these approximations to be protein-independent.", "We use machine learning methods to train a stability model σ from a given set Q of MD simulation trajectories. Each trajectory q ∈ Q is a sequence of conformations of a protein. These conformations are reached at times ti = i × δ, i = 0, 1, 2, …, called ticks, where δ is typically on the order of picoseconds. We detect the H-bonds present in each conformation q(ti) using the geometric criteria given in [8]. Note that an H-bond in a given protein is uniquely identified (across different conformations) by its donor, acceptor, and the hydrogen atom. So, we call the presence of a specific H-bond H in a conformation q(ti) an occurrence of H, denoted by h.\nFor each h, we compute a fixed list of predictors, some numerical, others categorical. Some are time-invariant, like the number of residues along the main-chain between the donor and acceptor atoms. Others are time-dependent. Among them, some describe the geometry of h, e.g., the distance between the hydrogen and the donor. Others describe the local environment of h, e.g., the number of other H-bonds within a certain distance from the mid-point of H.\nWe train σ as a function of these predictors. The predictor list defines a predictor space ∑ and every H-bond occurrence maps to a point in ∑. Given the input set Q of trajectories, we build a data table in which each row corresponds to an occurrence h of an H-bond present in a conformation q(ti) contained in Q. So, many rows may correspond to the same H-bond at different ticks. In our experiments, a typical data table contains several hundred thousand rows. Each column, except the last one, corresponds to a predictor p and the entry (h, p) of the table is the value of p for h. The entry in the last column is the measured stability y of h. More precisely, let H be the H-bond of which h is an occurrence. Let l = Δ/δ, where Δ is the duration over which we wish to predict the stability of h, and m ≤ l be the number of ticks tk, k = i + 1, i + 2,…,i + l, such that H is present in q(tk). The measured stability y of h is the ratio m/l. We chose l = 50 in most of the tests reported below, as this value both provides a ratio m/l large enough for the measured stability to be statistically meaningful, and corresponds to an interesting prediction timescale (50ps). Typically, most H-bond occurrences are quite stable: over 25% have measured stability 1, about 50% higher than 0.8, and only 15% less than 0.3.\nWe build σ as a binary regression tree [9]. This machine learning approach has been one of the most successful in practice. Regression trees are often simple to interpret. The method can work with both categorical and numerical predictors in a unified way, as shown in Section III. Each non-leaf node in a regression tree is a Boolean split. So, each node N of the tree determines a region of ∑ in which all the splits associated with the arcs connecting the root of the tree to N are satisfied. We say that an H-bond occurrence falls into N if it is contained in this region. The predicted stability value stored at a leaf node L is the average of the measured stability values by all the H-bond occurrences in the training data table that fall into L. We expect this average, which is taken over many pieces of trajectories, to approximate well the average defined in Equation (1).\nOnce a regression tree has been generated, it is used as follows. Given an H-bond H in an arbitrary conformation c of an arbitrary protein, the leaf node L of the tree into which H falls is identified by calculating the values of the necessary predictors for H in c. The predicted stability value stored at L is returned.", "We construct a model σ as a binary regression tree using the CART method [9]. The tree is generated recursively in a top-down fashion. When a new node N is created, it is inserted as a leaf of the tree if a predefined depth has been reached or if the number of h falling into N is smaller than a predefined threshold. Otherwise, N is added as an intermediate node, its split is computed, and its left and right children are created. A split s is defined by a pair (p, r), where p is the split predictor and r is the split value. If p is a numerical predictor, then r is a threshold on p, and s ≜ p <r. If p is a categorical predictor, then r is a subset of categories, and s ≜ p ∈ r. We define the score w(p, r) of split s = (p, r) at a node N as the reduction of variance in measured stability that results from s. The algorithm chooses the split—both the predictor and the split value—that has the largest score. Only a relatively small subset of predictors selected by the training algorithm is eventually used in a regression tree.\nTo prevent model overfitting, we limit tree depth to 5 in most of our experiments and limit the minimal number of training samples in an intermediate node to be 10. We further prune the obtained tree using the following adaptive algorithm. We initially set aside a fraction of the training data table called validation subset. Once a tree has been constructed pruning is an iterative process. At each step, one intermediate node N whose split has minimal score becomes a leaf node by removing the sub-tree rooted at N. This process creates a sequence of trees with decreasing numbers of nodes. We compute the mean square error of the predictions made by each tree on the validation subset. The tree with the smallest error is selected.", "[SUBTITLE] I. Experimental setup [SUBSECTION] We used 6 MD simulation trajectories picked from different sources and called hereafter 1c9oA, 1e85A, 1g9oA_1, and 1g9oA_2 from [10], complex from [11], and 1eia (generated by us). In all of them the time interval δ between two successive ticks is 1ps. Table 1 indicates the protein simulated in each trajectory, its number of residues, the force field used by the simulator, and the duration of the trajectory. Each trajectory starts from a folded conformation resolved by X-ray crystallography.\nCharacteristics of the MD simulation trajectories used to create the 6 datasets\nFrom each trajectory we derived a separate data table in which the rows represent H-bond occurrences. Last two columns in Table 1 list the number of distinct H-bonds detected in each trajectory and the total number of H-bond occurrences extracted. Note that complex was generated for a complex of two molecules. All H-bonds occurring in this complex are taken into account in the corresponding data table.\nThe values of the time-varying predictors are subject to thermal noise. Since a model σ will in general be used to predict H-bond stability in a protein conformation sampled using a kinematic model ignoring thermal noise (e.g., by sampling the dihedral angles ϕ, ψ, and χ) [7], we chose to average the values of these predictors over l' ticks to remove thermal noise. More precisely, the value of a predictor stored in the row of the data table corresponding to an H-bond occurrence in q(ti) is the average value of this predictor in , where . Our analysis shows that l' = 50 is near optimal.\nThe performance of a regression model can be measured by the root mean square error (RMSE) of the predictions on a test dataset. For a data table T = {(x1, y1), (x2, y2)},…, (xn, yn)}, where each xi, i = 1,…,n, denotes a vector of predictor values for an H-bond occurrence and yi is the measured stability of the H-bond, and a model σ, the RMSE is defined by: . As RMSE depends not only on the accuracy of σ, but also on the table T, some normalization is necessary to compare results on different tables. So, in our tests we compute the decrease of RMSE relative to a base model σ0. The relative base error decrease (or RBED) is then defined by: In most cases, σ0 is simply defined by , i.e., the average measured stability of all H-bond occurrences in the dataset. In other cases, σ0 is a model based on the H-bond energy.\nWe used 6 MD simulation trajectories picked from different sources and called hereafter 1c9oA, 1e85A, 1g9oA_1, and 1g9oA_2 from [10], complex from [11], and 1eia (generated by us). In all of them the time interval δ between two successive ticks is 1ps. Table 1 indicates the protein simulated in each trajectory, its number of residues, the force field used by the simulator, and the duration of the trajectory. Each trajectory starts from a folded conformation resolved by X-ray crystallography.\nCharacteristics of the MD simulation trajectories used to create the 6 datasets\nFrom each trajectory we derived a separate data table in which the rows represent H-bond occurrences. Last two columns in Table 1 list the number of distinct H-bonds detected in each trajectory and the total number of H-bond occurrences extracted. Note that complex was generated for a complex of two molecules. All H-bonds occurring in this complex are taken into account in the corresponding data table.\nThe values of the time-varying predictors are subject to thermal noise. Since a model σ will in general be used to predict H-bond stability in a protein conformation sampled using a kinematic model ignoring thermal noise (e.g., by sampling the dihedral angles ϕ, ψ, and χ) [7], we chose to average the values of these predictors over l' ticks to remove thermal noise. More precisely, the value of a predictor stored in the row of the data table corresponding to an H-bond occurrence in q(ti) is the average value of this predictor in , where . Our analysis shows that l' = 50 is near optimal.\nThe performance of a regression model can be measured by the root mean square error (RMSE) of the predictions on a test dataset. For a data table T = {(x1, y1), (x2, y2)},…, (xn, yn)}, where each xi, i = 1,…,n, denotes a vector of predictor values for an H-bond occurrence and yi is the measured stability of the H-bond, and a model σ, the RMSE is defined by: . As RMSE depends not only on the accuracy of σ, but also on the table T, some normalization is necessary to compare results on different tables. So, in our tests we compute the decrease of RMSE relative to a base model σ0. The relative base error decrease (or RBED) is then defined by: In most cases, σ0 is simply defined by , i.e., the average measured stability of all H-bond occurrences in the dataset. In other cases, σ0 is a model based on the H-bond energy.\n[SUBTITLE] II Generality of models trained on multiple trajectories [SUBSECTION] Our goal is to train models to predict the stability of H-bonds in any protein. So, we trained models on data tables obtained by mixing subsets of 5 data tables and we tested these models on the remaining data table. For each combination of 5 data tables, we trained 4 groups of models varying in the tree’s maximal depth (5 or 15) and in the fraction of H-bond occurrences randomly taken from each data table (10% or 50%). For each group we trained 10 models. Hence, in total, 240 models were generated. Table 2 shows the mean RBED value for each combination of data tables and each group of models. In columns 3 through 8 we indicate the data table used for testing the models trained on a combination of the 5 other data tables. Figure 1 shows a partial tree trained with combinations of all tables, except 1c9oA.\nMean RBED values of models trained on multiple trajectories\nTop 2 layers of a regression tree trained with combination of all tables, except 1c9oA. The actual tree contains 55 nodes. Each path from the root to a node defines a conjunction of criteria for H-bonds with a certain mean stability. Here, Dist_H_A (the distance between the hydrogen and the acceptor atoms) is the most differentiating predictor. For H-bonds with Dist_H_A≥2.40Å, the mean stability is only 0.38, but it increases to 0.92 if Dist_H_A<2.40Å.\nRBED values show that regression tree model significantly reduces base error and keeps predictive power when applied to a protein not present in the training data. Moreover, the variance of RBED values is very small, meaning that the training process yields models that are stable in performance. Furthermore, the RBED values are lower for models tested on complex. Recall that the trajectory complex was generated for a complex made of a protein and a ligand, while every other trajectory was generated for a single protein. So, it is likely that complex contains H-bonds in situations that did not occur in any of the other trajectories. Both deeper trees and larger data fractions tend to improve model accuracy, but the very small gain is not worth the additional model or computation complexity.\nOur goal is to train models to predict the stability of H-bonds in any protein. So, we trained models on data tables obtained by mixing subsets of 5 data tables and we tested these models on the remaining data table. For each combination of 5 data tables, we trained 4 groups of models varying in the tree’s maximal depth (5 or 15) and in the fraction of H-bond occurrences randomly taken from each data table (10% or 50%). For each group we trained 10 models. Hence, in total, 240 models were generated. Table 2 shows the mean RBED value for each combination of data tables and each group of models. In columns 3 through 8 we indicate the data table used for testing the models trained on a combination of the 5 other data tables. Figure 1 shows a partial tree trained with combinations of all tables, except 1c9oA.\nMean RBED values of models trained on multiple trajectories\nTop 2 layers of a regression tree trained with combination of all tables, except 1c9oA. The actual tree contains 55 nodes. Each path from the root to a node defines a conjunction of criteria for H-bonds with a certain mean stability. Here, Dist_H_A (the distance between the hydrogen and the acceptor atoms) is the most differentiating predictor. For H-bonds with Dist_H_A≥2.40Å, the mean stability is only 0.38, but it increases to 0.92 if Dist_H_A<2.40Å.\nRBED values show that regression tree model significantly reduces base error and keeps predictive power when applied to a protein not present in the training data. Moreover, the variance of RBED values is very small, meaning that the training process yields models that are stable in performance. Furthermore, the RBED values are lower for models tested on complex. Recall that the trajectory complex was generated for a complex made of a protein and a ligand, while every other trajectory was generated for a single protein. So, it is likely that complex contains H-bonds in situations that did not occur in any of the other trajectories. Both deeper trees and larger data fractions tend to improve model accuracy, but the very small gain is not worth the additional model or computation complexity.\n[SUBTITLE] III. Comparison with FIRST-energy model [SUBSECTION] We've checked whether regression models can predict the stability of H-bonds more accurately than potential energy alone. Table 3 presents the mean RBED value for a model obtained in the first row of Table 2 relative to the base model that is a regression tree built from the same training data using FIRST_energy as the only predictor. FIRST_energy is a modified Mayo potential [5] implemented in FIRST (a protein rigidity analysis software) [7]. Comparison on all 6 data tables show that the more complex models are significantly more accurate than the models based on FIRST_energy alone.\nMean RBED values of models using single predictor FIRST_energy\nWe've checked whether regression models can predict the stability of H-bonds more accurately than potential energy alone. Table 3 presents the mean RBED value for a model obtained in the first row of Table 2 relative to the base model that is a regression tree built from the same training data using FIRST_energy as the only predictor. FIRST_energy is a modified Mayo potential [5] implemented in FIRST (a protein rigidity analysis software) [7]. Comparison on all 6 data tables show that the more complex models are significantly more accurate than the models based on FIRST_energy alone.\nMean RBED values of models using single predictor FIRST_energy\n[SUBTITLE] IV. Identification of least stable H-bonds [SUBSECTION] Most H-bond occurrences tend to be stable. So, accurately identifying the weakest ones is important if one wishes to predict the possible deformation of a protein [7]. To evaluate how well our models identify the least stable H-bonds occurrences, we first identify the subset S of the 10% H-bond occurrences with the smallest measured stability in each test table T. Using a regression tree σ obtained in Section II, we sort the H-bond occurrences in T in ascending order of predicted stability and we compute the fraction w ∈ [0,1] of S that is contained in the first 100×u% occurrences in this sorted list, for successive values of u ∈ [0,1]. We call the function w(u) the identification curve of the least stable H-bonds for σ.\nFigure 2 plots the identification curve for 1c9oA. It consists of three curves: the red curve is the (fictitious) ideal identification curve, the blue curve is obtained with one (randomly picked) regression tree computed in Section II, and the green curve is obtained by sorting H-bond occurrences in decreasing values of FIRST_energy. Plots on other proteins present similar curve shapes. For models tested on data tables except complex, about 80% of the 10% H-bond occurrences predicted as the least stable are actually among the 10% truly least stable. The results for complex are less satisfactory because of the reasons discussed in Section II. The regression models are consistently better than the FIRST_energy-only models, though for 1eia the difference is small.\nIdentification curves of the least stable bonds for 1c9oA (see Results, Section IV).\nMost H-bond occurrences tend to be stable. So, accurately identifying the weakest ones is important if one wishes to predict the possible deformation of a protein [7]. To evaluate how well our models identify the least stable H-bonds occurrences, we first identify the subset S of the 10% H-bond occurrences with the smallest measured stability in each test table T. Using a regression tree σ obtained in Section II, we sort the H-bond occurrences in T in ascending order of predicted stability and we compute the fraction w ∈ [0,1] of S that is contained in the first 100×u% occurrences in this sorted list, for successive values of u ∈ [0,1]. We call the function w(u) the identification curve of the least stable H-bonds for σ.\nFigure 2 plots the identification curve for 1c9oA. It consists of three curves: the red curve is the (fictitious) ideal identification curve, the blue curve is obtained with one (randomly picked) regression tree computed in Section II, and the green curve is obtained by sorting H-bond occurrences in decreasing values of FIRST_energy. Plots on other proteins present similar curve shapes. For models tested on data tables except complex, about 80% of the 10% H-bond occurrences predicted as the least stable are actually among the 10% truly least stable. The results for complex are less satisfactory because of the reasons discussed in Section II. The regression models are consistently better than the FIRST_energy-only models, though for 1eia the difference is small.\nIdentification curves of the least stable bonds for 1c9oA (see Results, Section IV).", "We used 6 MD simulation trajectories picked from different sources and called hereafter 1c9oA, 1e85A, 1g9oA_1, and 1g9oA_2 from [10], complex from [11], and 1eia (generated by us). In all of them the time interval δ between two successive ticks is 1ps. Table 1 indicates the protein simulated in each trajectory, its number of residues, the force field used by the simulator, and the duration of the trajectory. Each trajectory starts from a folded conformation resolved by X-ray crystallography.\nCharacteristics of the MD simulation trajectories used to create the 6 datasets\nFrom each trajectory we derived a separate data table in which the rows represent H-bond occurrences. Last two columns in Table 1 list the number of distinct H-bonds detected in each trajectory and the total number of H-bond occurrences extracted. Note that complex was generated for a complex of two molecules. All H-bonds occurring in this complex are taken into account in the corresponding data table.\nThe values of the time-varying predictors are subject to thermal noise. Since a model σ will in general be used to predict H-bond stability in a protein conformation sampled using a kinematic model ignoring thermal noise (e.g., by sampling the dihedral angles ϕ, ψ, and χ) [7], we chose to average the values of these predictors over l' ticks to remove thermal noise. More precisely, the value of a predictor stored in the row of the data table corresponding to an H-bond occurrence in q(ti) is the average value of this predictor in , where . Our analysis shows that l' = 50 is near optimal.\nThe performance of a regression model can be measured by the root mean square error (RMSE) of the predictions on a test dataset. For a data table T = {(x1, y1), (x2, y2)},…, (xn, yn)}, where each xi, i = 1,…,n, denotes a vector of predictor values for an H-bond occurrence and yi is the measured stability of the H-bond, and a model σ, the RMSE is defined by: . As RMSE depends not only on the accuracy of σ, but also on the table T, some normalization is necessary to compare results on different tables. So, in our tests we compute the decrease of RMSE relative to a base model σ0. The relative base error decrease (or RBED) is then defined by: In most cases, σ0 is simply defined by , i.e., the average measured stability of all H-bond occurrences in the dataset. In other cases, σ0 is a model based on the H-bond energy.", "Our goal is to train models to predict the stability of H-bonds in any protein. So, we trained models on data tables obtained by mixing subsets of 5 data tables and we tested these models on the remaining data table. For each combination of 5 data tables, we trained 4 groups of models varying in the tree’s maximal depth (5 or 15) and in the fraction of H-bond occurrences randomly taken from each data table (10% or 50%). For each group we trained 10 models. Hence, in total, 240 models were generated. Table 2 shows the mean RBED value for each combination of data tables and each group of models. In columns 3 through 8 we indicate the data table used for testing the models trained on a combination of the 5 other data tables. Figure 1 shows a partial tree trained with combinations of all tables, except 1c9oA.\nMean RBED values of models trained on multiple trajectories\nTop 2 layers of a regression tree trained with combination of all tables, except 1c9oA. The actual tree contains 55 nodes. Each path from the root to a node defines a conjunction of criteria for H-bonds with a certain mean stability. Here, Dist_H_A (the distance between the hydrogen and the acceptor atoms) is the most differentiating predictor. For H-bonds with Dist_H_A≥2.40Å, the mean stability is only 0.38, but it increases to 0.92 if Dist_H_A<2.40Å.\nRBED values show that regression tree model significantly reduces base error and keeps predictive power when applied to a protein not present in the training data. Moreover, the variance of RBED values is very small, meaning that the training process yields models that are stable in performance. Furthermore, the RBED values are lower for models tested on complex. Recall that the trajectory complex was generated for a complex made of a protein and a ligand, while every other trajectory was generated for a single protein. So, it is likely that complex contains H-bonds in situations that did not occur in any of the other trajectories. Both deeper trees and larger data fractions tend to improve model accuracy, but the very small gain is not worth the additional model or computation complexity.", "We've checked whether regression models can predict the stability of H-bonds more accurately than potential energy alone. Table 3 presents the mean RBED value for a model obtained in the first row of Table 2 relative to the base model that is a regression tree built from the same training data using FIRST_energy as the only predictor. FIRST_energy is a modified Mayo potential [5] implemented in FIRST (a protein rigidity analysis software) [7]. Comparison on all 6 data tables show that the more complex models are significantly more accurate than the models based on FIRST_energy alone.\nMean RBED values of models using single predictor FIRST_energy", "Most H-bond occurrences tend to be stable. So, accurately identifying the weakest ones is important if one wishes to predict the possible deformation of a protein [7]. To evaluate how well our models identify the least stable H-bonds occurrences, we first identify the subset S of the 10% H-bond occurrences with the smallest measured stability in each test table T. Using a regression tree σ obtained in Section II, we sort the H-bond occurrences in T in ascending order of predicted stability and we compute the fraction w ∈ [0,1] of S that is contained in the first 100×u% occurrences in this sorted list, for successive values of u ∈ [0,1]. We call the function w(u) the identification curve of the least stable H-bonds for σ.\nFigure 2 plots the identification curve for 1c9oA. It consists of three curves: the red curve is the (fictitious) ideal identification curve, the blue curve is obtained with one (randomly picked) regression tree computed in Section II, and the green curve is obtained by sorting H-bond occurrences in decreasing values of FIRST_energy. Plots on other proteins present similar curve shapes. For models tested on data tables except complex, about 80% of the 10% H-bond occurrences predicted as the least stable are actually among the 10% truly least stable. The results for complex are less satisfactory because of the reasons discussed in Section II. The regression models are consistently better than the FIRST_energy-only models, though for 1eia the difference is small.\nIdentification curves of the least stable bonds for 1c9oA (see Results, Section IV).", "In all our regression trees the root split was done with predictor Dist_H_A (the distance between the hydrogen and acceptor atoms), which therefore appear as the single most discriminative attribute to predict H-bond stability. This observation is consistent with previous findings. Levitt [6] found that most stable H-bonds have Dist_H_A less than 2.07Å. Jeffrey and Saenger [14] also suggested that Dist_H_A is a key attribute affecting H-bond stability, with a value less than 2.2Å for moderate to strong H-bonds. Consistent with these previous findings, the split values of the deepest Dist_H_A nodes in all our regression trees are around 2.1Å. This distance was observed in [6] to sometimes fluctuate by up to 3Å in stable H-bonds, due to high-frequency atomic vibration. This observation supports our decision to average predictor values over windows of l’ ticks.\nPredictor FIRST_energy is often used in splits close to the root. This is not surprising since it is a function of several other pertinent predictors: Dist_H_A, Angle_D_H_A (the angle between the donor, the hydrogen atom, and the acceptor), Angle_H_A_AA (the angle between the hydrogen atom, the acceptor, and the atom covalently-bonded to the acceptor), and the hybridization state of the bond. Some other distance-based predictors (Dist_D_AA, Dist_D_A, Dist_H_D), angle-based predictors and Ch_type (describing whether the donor and acceptor are from main-chain or side-chain) predictor appear often in regression trees, but closer to the leaf nodes. They nevertheless play a significant role in predicting H-bond stability. For example, as shown in Figure 1, if Angle_H_A_AA is at least 105°, the stability is very high (about 0.96); otherwise, it drops to 0.71. The preference for larger angle matches well with the well-known linearity of H-bonds [14].\nIn order to get a more quantitative measure of the relative impact of the predictors on H-bond stability, we define the importance of a predictor p in a regression tree by: , where Np is the set of nodes where the split is made using p, w(s) is the score of the split s, and n(s) is the number of H-bond occurrences falling into the node where split s is made. We trained 10 models on data tables combining 10% of each of the 6 data tables. Importance scores for each predictor were averaged over these models and then linearly scaled to adjust the score of the least important predictor (with non-zero average importance) equal to 1. The average importance of every predictor appearing in at least one model is shown in Figure 3. The figure confirms that distance-based and angle-based predictors, as well as FIRST_energy, are the most important. It also shows that a number of other predictors—including Resi_name_H, Resi_name_A, and Range (difference in residue numbers of donor and acceptor) —have less, but still significant importance.\nPredictor importance scores\nOverall, we observe that predictors that describe the local environment of an H-bond play a relatively small role in predicting its stability. In particular, we had expected that descriptors such Num_hb_spaceNbr and Num_hb_spaceRgdNbr, which count the number of other H-bonds located in the neighborhood of the analyzed H-bond, would have had more importance. However, this may reflect the fact that the MD simulation trajectories used in our tests are too short to contain enough information to infer the role of such predictors. Indeed, while transitions between meta-stable states are rare in those trajectories, predictors describing local environments may have greater influence on the stability of H-bonds that must break for such transitions to happen. So, longer trajectories may eventually be needed to better model H-bond stability.", "We have described machine learning methods to train protein-independent regression trees modeling H-bond stability in proteins. Test results demonstrate that trained models can predict H-bond stability quite well. In particular, their performance is significantly better (roughly 20% better) than that of a model based on H-bond energy alone. They can accurately identify a large fraction of the least stable H-bonds in a given conformation. However, our results also suggest that better results could be obtained with a richer set of MD simulation trajectories. In particular, the trajectories used in our experiments might be too short to characterize the stability of H-bonds that break and form during a transition between meta-stable states.\nWe believe that the training methods could be improved in several ways:\n- It would be better to averaging predictor values before sub-sampling MD simulation trajectories. This would reduce the risk of filtering out changes in predictor values that are important for H-bond stability. Unfortunately, in our trajectories we only had access to the data after sub-sampling.\n- More sophisticated learning techniques could be used. For example, instead of generating a single tree, we could generate an ensemble of trees, such as Gradient Boosting Trees [16].\n- Finally, the notion of stability itself could be refined, for example by distinguishing between the case where an H-bond frequently switches on and off during a prediction window and the case where it rarely switches.", "All four authors, IC, PY, MM, and JCL, participated in the formulation of the problem, the design of its solution, and the analysis and the interpretation of the results. PY prepared the experimental data. IC adapted a previously available CART software package and ran the experiments. All authors contributed to the writing of the manuscript, read and approved the final manuscript.", "The authors declare that they have no competing interests." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Biomarkers", "Cell Line, Tumor", "DNA Copy Number Variations", "Databases, Genetic", "Gene Expression Profiling", "Humans", "Metabolome", "Neoplasms", "Organ Specificity" ]

[ "Background", "Results", "Metabolomic signature of cancer cells from different tissue origins", "Correlation analysis", "Metabolite-gene correlations", "Outlier analysis", "Discussion", "Competing interests", "Authors' contributions" ]

[ "Metabolites are end products of cellular processes. Thus the profile of metabolites provides a snapshot of the physiological state of a cell complementary to its transcriptome and proteome, which manifests the functional status of events which occur more upstream. Since the size of metabolome is about 1-2 orders of magnitude smaller than transcriptome and proteome, the steady state concentration of a metabolite usually reflects the combined effects of multiple upstream factors. Conceivably, this property makes metabolites better biomarkers in some situations. In addition, identifying interrelations between metabolites and genes, proteins, or genome structures can potentially facilitate the elucidation of molecular mechanisms involved in pathophysiological processes.\nPrevious work has revealed significant associations between gene and metabolite expression profiles, which added another layer of quantitative inferences to gene-wise correlations [1-3]. Nam et. al demonstrated that the integrative study of transcriptomics and metabolomics could effectively identify metabolic biomarkers for breast cancer [4]. Nevertheless, the noisy nature of the metabolome data, limited number of metabolites with known structures in metabolome data, the lack of annotation of metabolite-gene relationships despite databases like EHMN and BiGG [5,6], the indirect nature of potential gene-metabolite relationships, and the lack of large scale metabolome study data in the public domain, still limit our knowledge of metabolites and their regulation in normal and disease processes.\nConceivably, cancer samples provide excellent opportunities for identifying metabolic biomarkers and gene-metabome relationships due to the dramatic function alterations at the molecular level in cancer tissues. For example, cancer tissues usually exhibit more than 10-fold changes in the expression level of many genes in numerous microarray studies. Recently, the collaborative NCI60 project from the Developmental Therapeutics Program (DTP) of the National Cancer Institute (NCI) has made extensive measurements of various ‘omics’ data publicly available, including microarray, metabolomics, proteomics, etc. While the number of metabolome of metabolome-related studies and biomarker discoveries associated with cancer is increasing rapidly in targeted studies, there is still no literature on comprehensive analysis of metabolic features and their regulatory mechanisms of different cancer types across multiple ‘omics’ data.\nIn this study, we want to investigate whether different cancer cell lines have distinct metabolic signatures and whether available metabolome data for the NCI-60 cell lines is suitable for cancer subtype classification. We also hope the combined metabolome and transcriptome analysis will reveal some distinct regulatory relationships in some of the NCI-60 dataset which are not otherwise possible by either metabolome or transcriptome study alone. We expect the analysis approach used in this work can be applied to other metabolome-transcriptome datasets when they become available.", "[SUBTITLE] Metabolomic signature of cancer cells from different tissue origins [SUBSECTION] The first question we would like to address is whether cancer cell lines exhibit tissue-origin specific metabolic signatures. We used classification analysis similar to those performed in microarray data [7] to classify 57 cell-lines into 9 cancer types related to their tissue origins. Our initial tests suggest that regardless of method used, the classification error on metabolomics data is always significantly higher than that from microarray (~0.51 for metabolomics and ~0.34 for microarray).\nBesides the fact that NCI-60 metabolome dataset has much less data points but with higher noise levels when compared to microarray data, we speculate that the low number of some cancer types (e.g. only 2 prostate cancer cell lines) and high heterogeneity of some cell lines (breast, ovarian and CNS) may also undermine the performance of the metabolome-based classification. In fact, these 4 cell lines contribute most to the out of bag classification error estimates.\nAfter prostate, breast, ovarian and CNS cell lines are removed from the whole sample, the metabolite classifiers can reach comparable performance with microarray classifiers (Figure 1). The significantly improved metabolome-based classification results after removing cell lines with high variability or heterogeneity strongly suggests that there are indeed cancer subtype-specific (based on tissue of origin) metabolic signatures.\nClassification error. Estimated Out-of-Bag (OOB) Error with regard to progressive cancer class removal. At each step, cancer classes contribute most to classification error were removed and OOB errors were recalculated. B: Breast Cancer. P: Prostate Cancer. O: Ovarian Cancer. C: CNS. The left plot shows OOB errors on entire dataset and the right plot shows OOB errors on subset of classifiers selected by varSelRF. The improvement of variable selection is more remarkable for microarray because of the much larger classifier pool to select from (11961 v.s. 342). Metabolite classifiers can achieve average OOB error of 0.28 when B,P,O,C are removed, reduced from 0.51 from the full set. The OOB error for u133a reduced from 0.34 to 0.18. The average OOB errors from the random cancer class removal are plotted in the orange dashed line.\nSince the removal of cancer classes will also reduce the complexity of the data thus naturally improve classification performance, we also removed the same number of cancer classes at each stage randomly for 500 times to estimate the effect of improvement of OOB errors due to simplification of data structure resulted from sample removal (orange dashed lines in Figure 1). It can be seen that the reductions in OOB errors from random removal are not as significant as those from progressively removal of the prostate, breast, ovarian and CNS cell line classes, suggesting that it is indeed the high heterogeneity and the low sample number of these cell line classes that caused the high OOB error in the original classification. Consequently, our results suggest that although metabolite profiles are not as good as microarray data for the classification of cancer cell lines, metabolites nonetheless contain cell type specific signature and the classification results can be improved if we have more samples or have more data points in the metabolome assays.\nThe first question we would like to address is whether cancer cell lines exhibit tissue-origin specific metabolic signatures. We used classification analysis similar to those performed in microarray data [7] to classify 57 cell-lines into 9 cancer types related to their tissue origins. Our initial tests suggest that regardless of method used, the classification error on metabolomics data is always significantly higher than that from microarray (~0.51 for metabolomics and ~0.34 for microarray).\nBesides the fact that NCI-60 metabolome dataset has much less data points but with higher noise levels when compared to microarray data, we speculate that the low number of some cancer types (e.g. only 2 prostate cancer cell lines) and high heterogeneity of some cell lines (breast, ovarian and CNS) may also undermine the performance of the metabolome-based classification. In fact, these 4 cell lines contribute most to the out of bag classification error estimates.\nAfter prostate, breast, ovarian and CNS cell lines are removed from the whole sample, the metabolite classifiers can reach comparable performance with microarray classifiers (Figure 1). The significantly improved metabolome-based classification results after removing cell lines with high variability or heterogeneity strongly suggests that there are indeed cancer subtype-specific (based on tissue of origin) metabolic signatures.\nClassification error. Estimated Out-of-Bag (OOB) Error with regard to progressive cancer class removal. At each step, cancer classes contribute most to classification error were removed and OOB errors were recalculated. B: Breast Cancer. P: Prostate Cancer. O: Ovarian Cancer. C: CNS. The left plot shows OOB errors on entire dataset and the right plot shows OOB errors on subset of classifiers selected by varSelRF. The improvement of variable selection is more remarkable for microarray because of the much larger classifier pool to select from (11961 v.s. 342). Metabolite classifiers can achieve average OOB error of 0.28 when B,P,O,C are removed, reduced from 0.51 from the full set. The OOB error for u133a reduced from 0.34 to 0.18. The average OOB errors from the random cancer class removal are plotted in the orange dashed line.\nSince the removal of cancer classes will also reduce the complexity of the data thus naturally improve classification performance, we also removed the same number of cancer classes at each stage randomly for 500 times to estimate the effect of improvement of OOB errors due to simplification of data structure resulted from sample removal (orange dashed lines in Figure 1). It can be seen that the reductions in OOB errors from random removal are not as significant as those from progressively removal of the prostate, breast, ovarian and CNS cell line classes, suggesting that it is indeed the high heterogeneity and the low sample number of these cell line classes that caused the high OOB error in the original classification. Consequently, our results suggest that although metabolite profiles are not as good as microarray data for the classification of cancer cell lines, metabolites nonetheless contain cell type specific signature and the classification results can be improved if we have more samples or have more data points in the metabolome assays.\n[SUBTITLE] Correlation analysis [SUBSECTION] Although existing metabolome data do not outperform microarray data in cell line subtype classification, metabolome can still be useful for revealing some basic molecular processes and their alterations in NCI-60 cell lines. We hypothesize that different cell lines should share some basic regulatory and metabolic processes (steady-state) essential for cell growth and metabolism. It is likely that although different cancer cell lines may have very different levels of gene expression and metabolism, the steady-state relationship between genes and metabolites in the same or highly coupled pathways should be conserved across different cell lines in the absence of dramatic genomic changes such as gene mutation and copy number variation (CNV). Identification of such gene-metabolite relationships at steady state will help us better understand the underlying molecular mechanisms related to the metabolic change. They will also help us infer potential metabolic changes based on the expression data or vice versa. On the other hand, if a few cell lines deviated significantly from the gene-metabolite relationships exhibited by the most of the cell lines, the related genes in such outlier cell lines are likely to be silenced (no expression) or agitated (over expression) due to genomic structural changes. Consequently, we are interested in both high correlations which reflect steady state trend over all samples, and outliers which reflect signatures in specific cell lines.\nThe classic method of inspecting the relatedness of two quantitative features is by computing the Pearson Correlation Coefficient (PCC). While PCC have been proven to work well in many gene expression studies, the correlation is often inflated by outliers or clustering effect, which made it not suitable to analyze data with high noise level such as metabolic profiles. In contrast, the robust correlations can provide a much better estimate of the true association in the presence of multi-dimensional outliers if the sample size is sufficiently large. We chose the Pair-wise Quadrant Correlation (PQC) for its good computational performance. However, for small sample size, PQC sometimes may inflate the correlation estimate. Therefore our correlation analysis, we use a novel approach by combining results from both PCC and PQC. Gene-metabolite pairs with high PCC and high PQC tend to demonstrate true linear correlation across all cancer types, which imply a general functional association between gene and metabolite profiles. On the other hand, gene-metabolite pairs with high PCC and low PQC are possibly resulted from a few extreme outliers, which can potentially be linked to cell-line specific signatures.\nAlthough existing metabolome data do not outperform microarray data in cell line subtype classification, metabolome can still be useful for revealing some basic molecular processes and their alterations in NCI-60 cell lines. We hypothesize that different cell lines should share some basic regulatory and metabolic processes (steady-state) essential for cell growth and metabolism. It is likely that although different cancer cell lines may have very different levels of gene expression and metabolism, the steady-state relationship between genes and metabolites in the same or highly coupled pathways should be conserved across different cell lines in the absence of dramatic genomic changes such as gene mutation and copy number variation (CNV). Identification of such gene-metabolite relationships at steady state will help us better understand the underlying molecular mechanisms related to the metabolic change. They will also help us infer potential metabolic changes based on the expression data or vice versa. On the other hand, if a few cell lines deviated significantly from the gene-metabolite relationships exhibited by the most of the cell lines, the related genes in such outlier cell lines are likely to be silenced (no expression) or agitated (over expression) due to genomic structural changes. Consequently, we are interested in both high correlations which reflect steady state trend over all samples, and outliers which reflect signatures in specific cell lines.\nThe classic method of inspecting the relatedness of two quantitative features is by computing the Pearson Correlation Coefficient (PCC). While PCC have been proven to work well in many gene expression studies, the correlation is often inflated by outliers or clustering effect, which made it not suitable to analyze data with high noise level such as metabolic profiles. In contrast, the robust correlations can provide a much better estimate of the true association in the presence of multi-dimensional outliers if the sample size is sufficiently large. We chose the Pair-wise Quadrant Correlation (PQC) for its good computational performance. However, for small sample size, PQC sometimes may inflate the correlation estimate. Therefore our correlation analysis, we use a novel approach by combining results from both PCC and PQC. Gene-metabolite pairs with high PCC and high PQC tend to demonstrate true linear correlation across all cancer types, which imply a general functional association between gene and metabolite profiles. On the other hand, gene-metabolite pairs with high PCC and low PQC are possibly resulted from a few extreme outliers, which can potentially be linked to cell-line specific signatures.\n[SUBTITLE] Metabolite-gene correlations [SUBSECTION] Since metabolite-gene relationships can help us to understand the direct or indirect cause of metabolite changes, we computed PCC and PQC for 11872 gene expression profiles and 253 metabolite measurements over 57 cell-lines.\nTo investigate the biological significances of the correlation analysis, we mapped all the known gene-compound associations from EHMN to our NCI60 analysis. There are a total of 721 EHMN annotated direct metabolite-gene associations, involving 352 genes and 81 known metabolites, that overlap with our NCI-60 metabolite-gene relationships.\nFigure 2 shows a mapping of our current metabolite-compound knowledgebase to the robust correlations. We chose the top 9 pathways in EHMN mapping ordered by number of genes\nHeatmap of gene-metabolite relationship organized according to KEGG pathway. Mapping of EHMN gene-metabolite association data to robust correlation matrix heat map. A. Rows are genes grouped by pathway names, columns are metabolites also grouped by corresponding pathway names. Green circles mark the position of a specific reaction that couples a metabolite and a gene from EHMN. Orange Light cells indicate positive PQC, with black to be 0 and orange to be 1. It can be seen that even though there are patterns of gene-metabolite clustering, very few high correlations can be mapped to known reactions. B. Gene-metabolite correlation in the Glycolysis pathway. C: The gene-gene correlation with in Glycolysis pathway.\nSome of the EHMN gene-compound relationships do show high association across all cell-lines. For example, gene AKR1B1 which reduces L-Arabitol to L-Arabinose with EHMN reaction ID R01758 and R01759, is associated with L-Arabitol with PQC of 0.69 and PCC of 0.36, which implies strong association between metabolite level and gene activity. However, it is surprising that most of the direct enzyme-metabolite relationships cannot be mapped to high correlations (Figure 2A and 2B). There are several possible explanations to the low match-ups of significant correlations to EHMN data. A simple explanation is most of the annotatable gene-metabolite relationships in the NCI-60 dataset may not be the speed limiting factor in the related pathways. It is also possible that the high level of abnormal regulation in the cancer cell lines may have masked global mechanisms such that it is difficult to identify high gene to metabolite correlations across all cell lines.\nInterestingly, the grouping of genes in the same pathway in Figure 2 enables us to detect many metabolites exhibit high correlations with other genes in the same pathway. For example, although phosphoenolpyruvate does not overlap with any of the EHMN annotated direct reaction genes in Figure 2B, it has high correlation with GPI and ALDOA, two of the genes in the glycolysis module that are known to be highly regulated by hypoxia-inducible factor 1alpha and such regulation is related to the aggressive phenotype of hepatocellular carcinoma. Consequently, ALDOA, GPI and other genes highly correlated with phosphoenolpyruvate in our multi-cancer cell line analysis may suggest that these genes has more significant regulatory or speed limiting roles in glycolysis than genes such as ENO1, ENO2 that are directly related to reactions involving phosphoenolpyruvate in these cancer cell lines. Naturally, not all genes in the same pathway strongly correlate with each other since genes in the same pathway are not all (Figure 2C).\nFurther investigation will be worthwhile for high correlation between metabolites and other genes (i.e. those not directly involved in the specific metabolic reaction) in the same pathway, as such un-annotated relationships are likely help us to identify speed limiting enzymes in a pathway, key regulatory genes of related pathways or novel metabolic mechanisms.\nSince metabolite-gene relationships can help us to understand the direct or indirect cause of metabolite changes, we computed PCC and PQC for 11872 gene expression profiles and 253 metabolite measurements over 57 cell-lines.\nTo investigate the biological significances of the correlation analysis, we mapped all the known gene-compound associations from EHMN to our NCI60 analysis. There are a total of 721 EHMN annotated direct metabolite-gene associations, involving 352 genes and 81 known metabolites, that overlap with our NCI-60 metabolite-gene relationships.\nFigure 2 shows a mapping of our current metabolite-compound knowledgebase to the robust correlations. We chose the top 9 pathways in EHMN mapping ordered by number of genes\nHeatmap of gene-metabolite relationship organized according to KEGG pathway. Mapping of EHMN gene-metabolite association data to robust correlation matrix heat map. A. Rows are genes grouped by pathway names, columns are metabolites also grouped by corresponding pathway names. Green circles mark the position of a specific reaction that couples a metabolite and a gene from EHMN. Orange Light cells indicate positive PQC, with black to be 0 and orange to be 1. It can be seen that even though there are patterns of gene-metabolite clustering, very few high correlations can be mapped to known reactions. B. Gene-metabolite correlation in the Glycolysis pathway. C: The gene-gene correlation with in Glycolysis pathway.\nSome of the EHMN gene-compound relationships do show high association across all cell-lines. For example, gene AKR1B1 which reduces L-Arabitol to L-Arabinose with EHMN reaction ID R01758 and R01759, is associated with L-Arabitol with PQC of 0.69 and PCC of 0.36, which implies strong association between metabolite level and gene activity. However, it is surprising that most of the direct enzyme-metabolite relationships cannot be mapped to high correlations (Figure 2A and 2B). There are several possible explanations to the low match-ups of significant correlations to EHMN data. A simple explanation is most of the annotatable gene-metabolite relationships in the NCI-60 dataset may not be the speed limiting factor in the related pathways. It is also possible that the high level of abnormal regulation in the cancer cell lines may have masked global mechanisms such that it is difficult to identify high gene to metabolite correlations across all cell lines.\nInterestingly, the grouping of genes in the same pathway in Figure 2 enables us to detect many metabolites exhibit high correlations with other genes in the same pathway. For example, although phosphoenolpyruvate does not overlap with any of the EHMN annotated direct reaction genes in Figure 2B, it has high correlation with GPI and ALDOA, two of the genes in the glycolysis module that are known to be highly regulated by hypoxia-inducible factor 1alpha and such regulation is related to the aggressive phenotype of hepatocellular carcinoma. Consequently, ALDOA, GPI and other genes highly correlated with phosphoenolpyruvate in our multi-cancer cell line analysis may suggest that these genes has more significant regulatory or speed limiting roles in glycolysis than genes such as ENO1, ENO2 that are directly related to reactions involving phosphoenolpyruvate in these cancer cell lines. Naturally, not all genes in the same pathway strongly correlate with each other since genes in the same pathway are not all (Figure 2C).\nFurther investigation will be worthwhile for high correlation between metabolites and other genes (i.e. those not directly involved in the specific metabolic reaction) in the same pathway, as such un-annotated relationships are likely help us to identify speed limiting enzymes in a pathway, key regulatory genes of related pathways or novel metabolic mechanisms.\n[SUBTITLE] Outlier analysis [SUBSECTION] In our previous analysis we require high PQC in addition to high PCC to identify molecule pairs with true high correlation. We have also identified many cases where cell lines have one or a few gene-metabolite pairs with much higher expression values than the rest of the other samples, which directly produce inflated PCC and very low PQC. To systematically investigate these cases, we used R package mvoulier to detect multidimensional outliers, and recomputed the PCC and PQC scores after outlier removal. Our empirical rule shows that when PCC > 0.6 and PQC < 0.3(Preferably close to 0), and the number of multidimensional outliers is smaller than 3, the high PCC is most likely to be an artifact from very few extreme outliers.\nTo explore the biological significance of these outliers, we compared the outlier cases detected by the criteria mentioned above to the Sanger Cosmic database [8]. Sanger Cosmic contains 177 mutation entries of 28 genes in our 57 NCI60 cell-lines. We ordered all the metabolite-gene correlations of the 28 genes by PCC, and amazingly the top two outliers detected by our approach, NOTCH1~X-2005 in MOLT-4 cell line and KRAS~in X-2690 in OVCAR-5 cell line, are the cell lines with known gene mutation on the exact same genes in Sanger Cosmic database (Figure 3).\nOutlier analysis. Outlier analysis shows extreme gene-metabolite pairs could be resulted from cell specific mutations. Top scatter plots: NOTCH1 ~ X-2005, with outlier in cell line MOLT_4 and KRAS ~ X-2690, with outlier in OVCAR5. It can be seen that the high PCC were both artefacts of extreme outliers. Middle table: PCC and PQC of the corresponding pairs, before and after outlier removal. R_QC: PQC. R_QC_RM: PQC after outlier removal. P_Pearson: PCC. R_P_RM: PCC after outlier removal. Bottom table: annotated mutation records from Sanger Cosmic database, directly matched to these two outlier pairs.\nThe common feature of these two gene-metabolite pairs is high PCC and low PQC before outlier removal and low PCC and low PQC after the outlier removal. From plots in Figure 3 we can verify that the inflated PCC were indeed a product of single cell line outlier. Besides, the sample sizes of these two cases are sufficiently large (22 for NOTCH1~X-2005 and 26 for KRAS~X-2690, respectively) so that the outlier is not likely to be a random fluctuation from small sample size.\nSince our unbiased analysis did not take advantage of any cell line or gene mutation information, the fact that our top two outliers overlaps with documented gene mutation in the Sanger Cosmic dataset suggests that the mutations are likely to be a cause of such outliers, and the associated metabolites may be good candidate biomarkers for such events.\nIn addition to point mutations, we also compared the outlier analysis with Copy Number Variation (CNV) data from Broad institute. Only 34 out of 57 samples from NCI60 have CNV data, and also found some gene-metabolite outlier pairs that are consistent with CNV outliers. For example, gene BRIP1, with CNV of 14.22 in cell line MCF7, has PCC of 0.90 and PQC of 0.008 and one outlier. The corresponding compound, X-3363, may be associated specifically to this copy number variation.\nThe fact that some top ranked gene-metabolite outlier associations matches with known corresponding genomic structural changes in the specific gene strongly suggests the effectiveness of our approach in identifying distinct molecular processes specific to metabolome and transcriptome variations.\nIn our previous analysis we require high PQC in addition to high PCC to identify molecule pairs with true high correlation. We have also identified many cases where cell lines have one or a few gene-metabolite pairs with much higher expression values than the rest of the other samples, which directly produce inflated PCC and very low PQC. To systematically investigate these cases, we used R package mvoulier to detect multidimensional outliers, and recomputed the PCC and PQC scores after outlier removal. Our empirical rule shows that when PCC > 0.6 and PQC < 0.3(Preferably close to 0), and the number of multidimensional outliers is smaller than 3, the high PCC is most likely to be an artifact from very few extreme outliers.\nTo explore the biological significance of these outliers, we compared the outlier cases detected by the criteria mentioned above to the Sanger Cosmic database [8]. Sanger Cosmic contains 177 mutation entries of 28 genes in our 57 NCI60 cell-lines. We ordered all the metabolite-gene correlations of the 28 genes by PCC, and amazingly the top two outliers detected by our approach, NOTCH1~X-2005 in MOLT-4 cell line and KRAS~in X-2690 in OVCAR-5 cell line, are the cell lines with known gene mutation on the exact same genes in Sanger Cosmic database (Figure 3).\nOutlier analysis. Outlier analysis shows extreme gene-metabolite pairs could be resulted from cell specific mutations. Top scatter plots: NOTCH1 ~ X-2005, with outlier in cell line MOLT_4 and KRAS ~ X-2690, with outlier in OVCAR5. It can be seen that the high PCC were both artefacts of extreme outliers. Middle table: PCC and PQC of the corresponding pairs, before and after outlier removal. R_QC: PQC. R_QC_RM: PQC after outlier removal. P_Pearson: PCC. R_P_RM: PCC after outlier removal. Bottom table: annotated mutation records from Sanger Cosmic database, directly matched to these two outlier pairs.\nThe common feature of these two gene-metabolite pairs is high PCC and low PQC before outlier removal and low PCC and low PQC after the outlier removal. From plots in Figure 3 we can verify that the inflated PCC were indeed a product of single cell line outlier. Besides, the sample sizes of these two cases are sufficiently large (22 for NOTCH1~X-2005 and 26 for KRAS~X-2690, respectively) so that the outlier is not likely to be a random fluctuation from small sample size.\nSince our unbiased analysis did not take advantage of any cell line or gene mutation information, the fact that our top two outliers overlaps with documented gene mutation in the Sanger Cosmic dataset suggests that the mutations are likely to be a cause of such outliers, and the associated metabolites may be good candidate biomarkers for such events.\nIn addition to point mutations, we also compared the outlier analysis with Copy Number Variation (CNV) data from Broad institute. Only 34 out of 57 samples from NCI60 have CNV data, and also found some gene-metabolite outlier pairs that are consistent with CNV outliers. For example, gene BRIP1, with CNV of 14.22 in cell line MCF7, has PCC of 0.90 and PQC of 0.008 and one outlier. The corresponding compound, X-3363, may be associated specifically to this copy number variation.\nThe fact that some top ranked gene-metabolite outlier associations matches with known corresponding genomic structural changes in the specific gene strongly suggests the effectiveness of our approach in identifying distinct molecular processes specific to metabolome and transcriptome variations.", "The first question we would like to address is whether cancer cell lines exhibit tissue-origin specific metabolic signatures. We used classification analysis similar to those performed in microarray data [7] to classify 57 cell-lines into 9 cancer types related to their tissue origins. Our initial tests suggest that regardless of method used, the classification error on metabolomics data is always significantly higher than that from microarray (~0.51 for metabolomics and ~0.34 for microarray).\nBesides the fact that NCI-60 metabolome dataset has much less data points but with higher noise levels when compared to microarray data, we speculate that the low number of some cancer types (e.g. only 2 prostate cancer cell lines) and high heterogeneity of some cell lines (breast, ovarian and CNS) may also undermine the performance of the metabolome-based classification. In fact, these 4 cell lines contribute most to the out of bag classification error estimates.\nAfter prostate, breast, ovarian and CNS cell lines are removed from the whole sample, the metabolite classifiers can reach comparable performance with microarray classifiers (Figure 1). The significantly improved metabolome-based classification results after removing cell lines with high variability or heterogeneity strongly suggests that there are indeed cancer subtype-specific (based on tissue of origin) metabolic signatures.\nClassification error. Estimated Out-of-Bag (OOB) Error with regard to progressive cancer class removal. At each step, cancer classes contribute most to classification error were removed and OOB errors were recalculated. B: Breast Cancer. P: Prostate Cancer. O: Ovarian Cancer. C: CNS. The left plot shows OOB errors on entire dataset and the right plot shows OOB errors on subset of classifiers selected by varSelRF. The improvement of variable selection is more remarkable for microarray because of the much larger classifier pool to select from (11961 v.s. 342). Metabolite classifiers can achieve average OOB error of 0.28 when B,P,O,C are removed, reduced from 0.51 from the full set. The OOB error for u133a reduced from 0.34 to 0.18. The average OOB errors from the random cancer class removal are plotted in the orange dashed line.\nSince the removal of cancer classes will also reduce the complexity of the data thus naturally improve classification performance, we also removed the same number of cancer classes at each stage randomly for 500 times to estimate the effect of improvement of OOB errors due to simplification of data structure resulted from sample removal (orange dashed lines in Figure 1). It can be seen that the reductions in OOB errors from random removal are not as significant as those from progressively removal of the prostate, breast, ovarian and CNS cell line classes, suggesting that it is indeed the high heterogeneity and the low sample number of these cell line classes that caused the high OOB error in the original classification. Consequently, our results suggest that although metabolite profiles are not as good as microarray data for the classification of cancer cell lines, metabolites nonetheless contain cell type specific signature and the classification results can be improved if we have more samples or have more data points in the metabolome assays.", "Although existing metabolome data do not outperform microarray data in cell line subtype classification, metabolome can still be useful for revealing some basic molecular processes and their alterations in NCI-60 cell lines. We hypothesize that different cell lines should share some basic regulatory and metabolic processes (steady-state) essential for cell growth and metabolism. It is likely that although different cancer cell lines may have very different levels of gene expression and metabolism, the steady-state relationship between genes and metabolites in the same or highly coupled pathways should be conserved across different cell lines in the absence of dramatic genomic changes such as gene mutation and copy number variation (CNV). Identification of such gene-metabolite relationships at steady state will help us better understand the underlying molecular mechanisms related to the metabolic change. They will also help us infer potential metabolic changes based on the expression data or vice versa. On the other hand, if a few cell lines deviated significantly from the gene-metabolite relationships exhibited by the most of the cell lines, the related genes in such outlier cell lines are likely to be silenced (no expression) or agitated (over expression) due to genomic structural changes. Consequently, we are interested in both high correlations which reflect steady state trend over all samples, and outliers which reflect signatures in specific cell lines.\nThe classic method of inspecting the relatedness of two quantitative features is by computing the Pearson Correlation Coefficient (PCC). While PCC have been proven to work well in many gene expression studies, the correlation is often inflated by outliers or clustering effect, which made it not suitable to analyze data with high noise level such as metabolic profiles. In contrast, the robust correlations can provide a much better estimate of the true association in the presence of multi-dimensional outliers if the sample size is sufficiently large. We chose the Pair-wise Quadrant Correlation (PQC) for its good computational performance. However, for small sample size, PQC sometimes may inflate the correlation estimate. Therefore our correlation analysis, we use a novel approach by combining results from both PCC and PQC. Gene-metabolite pairs with high PCC and high PQC tend to demonstrate true linear correlation across all cancer types, which imply a general functional association between gene and metabolite profiles. On the other hand, gene-metabolite pairs with high PCC and low PQC are possibly resulted from a few extreme outliers, which can potentially be linked to cell-line specific signatures.", "Since metabolite-gene relationships can help us to understand the direct or indirect cause of metabolite changes, we computed PCC and PQC for 11872 gene expression profiles and 253 metabolite measurements over 57 cell-lines.\nTo investigate the biological significances of the correlation analysis, we mapped all the known gene-compound associations from EHMN to our NCI60 analysis. There are a total of 721 EHMN annotated direct metabolite-gene associations, involving 352 genes and 81 known metabolites, that overlap with our NCI-60 metabolite-gene relationships.\nFigure 2 shows a mapping of our current metabolite-compound knowledgebase to the robust correlations. We chose the top 9 pathways in EHMN mapping ordered by number of genes\nHeatmap of gene-metabolite relationship organized according to KEGG pathway. Mapping of EHMN gene-metabolite association data to robust correlation matrix heat map. A. Rows are genes grouped by pathway names, columns are metabolites also grouped by corresponding pathway names. Green circles mark the position of a specific reaction that couples a metabolite and a gene from EHMN. Orange Light cells indicate positive PQC, with black to be 0 and orange to be 1. It can be seen that even though there are patterns of gene-metabolite clustering, very few high correlations can be mapped to known reactions. B. Gene-metabolite correlation in the Glycolysis pathway. C: The gene-gene correlation with in Glycolysis pathway.\nSome of the EHMN gene-compound relationships do show high association across all cell-lines. For example, gene AKR1B1 which reduces L-Arabitol to L-Arabinose with EHMN reaction ID R01758 and R01759, is associated with L-Arabitol with PQC of 0.69 and PCC of 0.36, which implies strong association between metabolite level and gene activity. However, it is surprising that most of the direct enzyme-metabolite relationships cannot be mapped to high correlations (Figure 2A and 2B). There are several possible explanations to the low match-ups of significant correlations to EHMN data. A simple explanation is most of the annotatable gene-metabolite relationships in the NCI-60 dataset may not be the speed limiting factor in the related pathways. It is also possible that the high level of abnormal regulation in the cancer cell lines may have masked global mechanisms such that it is difficult to identify high gene to metabolite correlations across all cell lines.\nInterestingly, the grouping of genes in the same pathway in Figure 2 enables us to detect many metabolites exhibit high correlations with other genes in the same pathway. For example, although phosphoenolpyruvate does not overlap with any of the EHMN annotated direct reaction genes in Figure 2B, it has high correlation with GPI and ALDOA, two of the genes in the glycolysis module that are known to be highly regulated by hypoxia-inducible factor 1alpha and such regulation is related to the aggressive phenotype of hepatocellular carcinoma. Consequently, ALDOA, GPI and other genes highly correlated with phosphoenolpyruvate in our multi-cancer cell line analysis may suggest that these genes has more significant regulatory or speed limiting roles in glycolysis than genes such as ENO1, ENO2 that are directly related to reactions involving phosphoenolpyruvate in these cancer cell lines. Naturally, not all genes in the same pathway strongly correlate with each other since genes in the same pathway are not all (Figure 2C).\nFurther investigation will be worthwhile for high correlation between metabolites and other genes (i.e. those not directly involved in the specific metabolic reaction) in the same pathway, as such un-annotated relationships are likely help us to identify speed limiting enzymes in a pathway, key regulatory genes of related pathways or novel metabolic mechanisms.", "In our previous analysis we require high PQC in addition to high PCC to identify molecule pairs with true high correlation. We have also identified many cases where cell lines have one or a few gene-metabolite pairs with much higher expression values than the rest of the other samples, which directly produce inflated PCC and very low PQC. To systematically investigate these cases, we used R package mvoulier to detect multidimensional outliers, and recomputed the PCC and PQC scores after outlier removal. Our empirical rule shows that when PCC > 0.6 and PQC < 0.3(Preferably close to 0), and the number of multidimensional outliers is smaller than 3, the high PCC is most likely to be an artifact from very few extreme outliers.\nTo explore the biological significance of these outliers, we compared the outlier cases detected by the criteria mentioned above to the Sanger Cosmic database [8]. Sanger Cosmic contains 177 mutation entries of 28 genes in our 57 NCI60 cell-lines. We ordered all the metabolite-gene correlations of the 28 genes by PCC, and amazingly the top two outliers detected by our approach, NOTCH1~X-2005 in MOLT-4 cell line and KRAS~in X-2690 in OVCAR-5 cell line, are the cell lines with known gene mutation on the exact same genes in Sanger Cosmic database (Figure 3).\nOutlier analysis. Outlier analysis shows extreme gene-metabolite pairs could be resulted from cell specific mutations. Top scatter plots: NOTCH1 ~ X-2005, with outlier in cell line MOLT_4 and KRAS ~ X-2690, with outlier in OVCAR5. It can be seen that the high PCC were both artefacts of extreme outliers. Middle table: PCC and PQC of the corresponding pairs, before and after outlier removal. R_QC: PQC. R_QC_RM: PQC after outlier removal. P_Pearson: PCC. R_P_RM: PCC after outlier removal. Bottom table: annotated mutation records from Sanger Cosmic database, directly matched to these two outlier pairs.\nThe common feature of these two gene-metabolite pairs is high PCC and low PQC before outlier removal and low PCC and low PQC after the outlier removal. From plots in Figure 3 we can verify that the inflated PCC were indeed a product of single cell line outlier. Besides, the sample sizes of these two cases are sufficiently large (22 for NOTCH1~X-2005 and 26 for KRAS~X-2690, respectively) so that the outlier is not likely to be a random fluctuation from small sample size.\nSince our unbiased analysis did not take advantage of any cell line or gene mutation information, the fact that our top two outliers overlaps with documented gene mutation in the Sanger Cosmic dataset suggests that the mutations are likely to be a cause of such outliers, and the associated metabolites may be good candidate biomarkers for such events.\nIn addition to point mutations, we also compared the outlier analysis with Copy Number Variation (CNV) data from Broad institute. Only 34 out of 57 samples from NCI60 have CNV data, and also found some gene-metabolite outlier pairs that are consistent with CNV outliers. For example, gene BRIP1, with CNV of 14.22 in cell line MCF7, has PCC of 0.90 and PQC of 0.008 and one outlier. The corresponding compound, X-3363, may be associated specifically to this copy number variation.\nThe fact that some top ranked gene-metabolite outlier associations matches with known corresponding genomic structural changes in the specific gene strongly suggests the effectiveness of our approach in identifying distinct molecular processes specific to metabolome and transcriptome variations.", "Our analysis results on the NCI-60 metabolome and transcriptome data suggest that 1) although the small sample size, the high noise level and intra-cancer class heterogeneity in the current NCI-60 metabolome dataset makes it unsuitable for global cancer subtype classification, there are indeed metabolic signatures associated with cancer subtypes. 2) There are biologically meaningful high correlation gene-metabolite pairs across NCI-60 cell lines, identifiable by robust correlation estimates. 3) Most strikingly, there are several examples of abnormal gene-metabolite coupling that can be directly linked to known gene mutations or copy number variations.\nConceivably, high correlations as well as outliers can be utilized to aid the progressive prediction of unknown metabolites based on annotations in existing pathway databases and literature. For example, the high correlation of an unknown compound with a known gene and in particular, multiple known genes in a pathway can dramatically reduce the search space for the unknown compound, since the most likely candidates will be structurally related molecules or known metabolites from related pathways. We plan to compare the Mass Spectrometer (MS) features of these unknown compounds from the predicted candidate pool and conduct wet-lab experiments for validation. Since we have discovered that many unknown metabolites are strongly associated with each other but not with any known compounds. Correct determination of even a small fraction of them would facilitate the identification of the rest, which also will in turn improve our understanding of the molecular processes and pathways involving these molecules.\nIn summary, the strong biological relevance of our results also suggest that the analysis strategy we developed based on the combination of PCC and PQC correlation values for identifying real correlation and true outliers is a powerful approach for integrative analysis of noisy ‘omics’ datasets. The presentation of gene-metabolite relationship analysis results in a heatmap grouped by genes in the same pathway together with overlay of known gene-metabolite relationship provides a powerful visual exploration approach for identifying both direct and indirect gene-metabolite relationships. The analysis methods we described here will be useful for other integrated analysis of metabolome and transcriptome and the wet lab validation of novel gene-metabolite relationships.", "The authors declare that they have no competing interests.", "GS performed data cleaning, correlation analysis and outlier analysis. CFB helped with correlation analysis and functional inference. CWB cleaned metabolomics data and annotated missing values. GS, BDA evaluated robust correlation methods; FM initialized the study, compiled microarray data and supervised biological validation." ]

[ null, null, null, null, null, null, null, null, null ]

[ "Background", "Results", "Metabolomic signature of cancer cells from different tissue origins", "Correlation analysis", "Metabolite-gene correlations", "Outlier analysis", "Methods", "Discussion", "Competing interests", "Authors' contributions" ]

[ "Metabolites are end products of cellular processes. Thus the profile of metabolites provides a snapshot of the physiological state of a cell complementary to its transcriptome and proteome, which manifests the functional status of events which occur more upstream. Since the size of metabolome is about 1-2 orders of magnitude smaller than transcriptome and proteome, the steady state concentration of a metabolite usually reflects the combined effects of multiple upstream factors. Conceivably, this property makes metabolites better biomarkers in some situations. In addition, identifying interrelations between metabolites and genes, proteins, or genome structures can potentially facilitate the elucidation of molecular mechanisms involved in pathophysiological processes.\nPrevious work has revealed significant associations between gene and metabolite expression profiles, which added another layer of quantitative inferences to gene-wise correlations [1-3]. Nam et. al demonstrated that the integrative study of transcriptomics and metabolomics could effectively identify metabolic biomarkers for breast cancer [4]. Nevertheless, the noisy nature of the metabolome data, limited number of metabolites with known structures in metabolome data, the lack of annotation of metabolite-gene relationships despite databases like EHMN and BiGG [5,6], the indirect nature of potential gene-metabolite relationships, and the lack of large scale metabolome study data in the public domain, still limit our knowledge of metabolites and their regulation in normal and disease processes.\nConceivably, cancer samples provide excellent opportunities for identifying metabolic biomarkers and gene-metabome relationships due to the dramatic function alterations at the molecular level in cancer tissues. For example, cancer tissues usually exhibit more than 10-fold changes in the expression level of many genes in numerous microarray studies. Recently, the collaborative NCI60 project from the Developmental Therapeutics Program (DTP) of the National Cancer Institute (NCI) has made extensive measurements of various ‘omics’ data publicly available, including microarray, metabolomics, proteomics, etc. While the number of metabolome of metabolome-related studies and biomarker discoveries associated with cancer is increasing rapidly in targeted studies, there is still no literature on comprehensive analysis of metabolic features and their regulatory mechanisms of different cancer types across multiple ‘omics’ data.\nIn this study, we want to investigate whether different cancer cell lines have distinct metabolic signatures and whether available metabolome data for the NCI-60 cell lines is suitable for cancer subtype classification. We also hope the combined metabolome and transcriptome analysis will reveal some distinct regulatory relationships in some of the NCI-60 dataset which are not otherwise possible by either metabolome or transcriptome study alone. We expect the analysis approach used in this work can be applied to other metabolome-transcriptome datasets when they become available.", "[SUBTITLE] Metabolomic signature of cancer cells from different tissue origins [SUBSECTION] The first question we would like to address is whether cancer cell lines exhibit tissue-origin specific metabolic signatures. We used classification analysis similar to those performed in microarray data [7] to classify 57 cell-lines into 9 cancer types related to their tissue origins. Our initial tests suggest that regardless of method used, the classification error on metabolomics data is always significantly higher than that from microarray (~0.51 for metabolomics and ~0.34 for microarray).\nBesides the fact that NCI-60 metabolome dataset has much less data points but with higher noise levels when compared to microarray data, we speculate that the low number of some cancer types (e.g. only 2 prostate cancer cell lines) and high heterogeneity of some cell lines (breast, ovarian and CNS) may also undermine the performance of the metabolome-based classification. In fact, these 4 cell lines contribute most to the out of bag classification error estimates.\nAfter prostate, breast, ovarian and CNS cell lines are removed from the whole sample, the metabolite classifiers can reach comparable performance with microarray classifiers (Figure 1). The significantly improved metabolome-based classification results after removing cell lines with high variability or heterogeneity strongly suggests that there are indeed cancer subtype-specific (based on tissue of origin) metabolic signatures.\nClassification error. Estimated Out-of-Bag (OOB) Error with regard to progressive cancer class removal. At each step, cancer classes contribute most to classification error were removed and OOB errors were recalculated. B: Breast Cancer. P: Prostate Cancer. O: Ovarian Cancer. C: CNS. The left plot shows OOB errors on entire dataset and the right plot shows OOB errors on subset of classifiers selected by varSelRF. The improvement of variable selection is more remarkable for microarray because of the much larger classifier pool to select from (11961 v.s. 342). Metabolite classifiers can achieve average OOB error of 0.28 when B,P,O,C are removed, reduced from 0.51 from the full set. The OOB error for u133a reduced from 0.34 to 0.18. The average OOB errors from the random cancer class removal are plotted in the orange dashed line.\nSince the removal of cancer classes will also reduce the complexity of the data thus naturally improve classification performance, we also removed the same number of cancer classes at each stage randomly for 500 times to estimate the effect of improvement of OOB errors due to simplification of data structure resulted from sample removal (orange dashed lines in Figure 1). It can be seen that the reductions in OOB errors from random removal are not as significant as those from progressively removal of the prostate, breast, ovarian and CNS cell line classes, suggesting that it is indeed the high heterogeneity and the low sample number of these cell line classes that caused the high OOB error in the original classification. Consequently, our results suggest that although metabolite profiles are not as good as microarray data for the classification of cancer cell lines, metabolites nonetheless contain cell type specific signature and the classification results can be improved if we have more samples or have more data points in the metabolome assays.\nThe first question we would like to address is whether cancer cell lines exhibit tissue-origin specific metabolic signatures. We used classification analysis similar to those performed in microarray data [7] to classify 57 cell-lines into 9 cancer types related to their tissue origins. Our initial tests suggest that regardless of method used, the classification error on metabolomics data is always significantly higher than that from microarray (~0.51 for metabolomics and ~0.34 for microarray).\nBesides the fact that NCI-60 metabolome dataset has much less data points but with higher noise levels when compared to microarray data, we speculate that the low number of some cancer types (e.g. only 2 prostate cancer cell lines) and high heterogeneity of some cell lines (breast, ovarian and CNS) may also undermine the performance of the metabolome-based classification. In fact, these 4 cell lines contribute most to the out of bag classification error estimates.\nAfter prostate, breast, ovarian and CNS cell lines are removed from the whole sample, the metabolite classifiers can reach comparable performance with microarray classifiers (Figure 1). The significantly improved metabolome-based classification results after removing cell lines with high variability or heterogeneity strongly suggests that there are indeed cancer subtype-specific (based on tissue of origin) metabolic signatures.\nClassification error. Estimated Out-of-Bag (OOB) Error with regard to progressive cancer class removal. At each step, cancer classes contribute most to classification error were removed and OOB errors were recalculated. B: Breast Cancer. P: Prostate Cancer. O: Ovarian Cancer. C: CNS. The left plot shows OOB errors on entire dataset and the right plot shows OOB errors on subset of classifiers selected by varSelRF. The improvement of variable selection is more remarkable for microarray because of the much larger classifier pool to select from (11961 v.s. 342). Metabolite classifiers can achieve average OOB error of 0.28 when B,P,O,C are removed, reduced from 0.51 from the full set. The OOB error for u133a reduced from 0.34 to 0.18. The average OOB errors from the random cancer class removal are plotted in the orange dashed line.\nSince the removal of cancer classes will also reduce the complexity of the data thus naturally improve classification performance, we also removed the same number of cancer classes at each stage randomly for 500 times to estimate the effect of improvement of OOB errors due to simplification of data structure resulted from sample removal (orange dashed lines in Figure 1). It can be seen that the reductions in OOB errors from random removal are not as significant as those from progressively removal of the prostate, breast, ovarian and CNS cell line classes, suggesting that it is indeed the high heterogeneity and the low sample number of these cell line classes that caused the high OOB error in the original classification. Consequently, our results suggest that although metabolite profiles are not as good as microarray data for the classification of cancer cell lines, metabolites nonetheless contain cell type specific signature and the classification results can be improved if we have more samples or have more data points in the metabolome assays.\n[SUBTITLE] Correlation analysis [SUBSECTION] Although existing metabolome data do not outperform microarray data in cell line subtype classification, metabolome can still be useful for revealing some basic molecular processes and their alterations in NCI-60 cell lines. We hypothesize that different cell lines should share some basic regulatory and metabolic processes (steady-state) essential for cell growth and metabolism. It is likely that although different cancer cell lines may have very different levels of gene expression and metabolism, the steady-state relationship between genes and metabolites in the same or highly coupled pathways should be conserved across different cell lines in the absence of dramatic genomic changes such as gene mutation and copy number variation (CNV). Identification of such gene-metabolite relationships at steady state will help us better understand the underlying molecular mechanisms related to the metabolic change. They will also help us infer potential metabolic changes based on the expression data or vice versa. On the other hand, if a few cell lines deviated significantly from the gene-metabolite relationships exhibited by the most of the cell lines, the related genes in such outlier cell lines are likely to be silenced (no expression) or agitated (over expression) due to genomic structural changes. Consequently, we are interested in both high correlations which reflect steady state trend over all samples, and outliers which reflect signatures in specific cell lines.\nThe classic method of inspecting the relatedness of two quantitative features is by computing the Pearson Correlation Coefficient (PCC). While PCC have been proven to work well in many gene expression studies, the correlation is often inflated by outliers or clustering effect, which made it not suitable to analyze data with high noise level such as metabolic profiles. In contrast, the robust correlations can provide a much better estimate of the true association in the presence of multi-dimensional outliers if the sample size is sufficiently large. We chose the Pair-wise Quadrant Correlation (PQC) for its good computational performance. However, for small sample size, PQC sometimes may inflate the correlation estimate. Therefore our correlation analysis, we use a novel approach by combining results from both PCC and PQC. Gene-metabolite pairs with high PCC and high PQC tend to demonstrate true linear correlation across all cancer types, which imply a general functional association between gene and metabolite profiles. On the other hand, gene-metabolite pairs with high PCC and low PQC are possibly resulted from a few extreme outliers, which can potentially be linked to cell-line specific signatures.\nAlthough existing metabolome data do not outperform microarray data in cell line subtype classification, metabolome can still be useful for revealing some basic molecular processes and their alterations in NCI-60 cell lines. We hypothesize that different cell lines should share some basic regulatory and metabolic processes (steady-state) essential for cell growth and metabolism. It is likely that although different cancer cell lines may have very different levels of gene expression and metabolism, the steady-state relationship between genes and metabolites in the same or highly coupled pathways should be conserved across different cell lines in the absence of dramatic genomic changes such as gene mutation and copy number variation (CNV). Identification of such gene-metabolite relationships at steady state will help us better understand the underlying molecular mechanisms related to the metabolic change. They will also help us infer potential metabolic changes based on the expression data or vice versa. On the other hand, if a few cell lines deviated significantly from the gene-metabolite relationships exhibited by the most of the cell lines, the related genes in such outlier cell lines are likely to be silenced (no expression) or agitated (over expression) due to genomic structural changes. Consequently, we are interested in both high correlations which reflect steady state trend over all samples, and outliers which reflect signatures in specific cell lines.\nThe classic method of inspecting the relatedness of two quantitative features is by computing the Pearson Correlation Coefficient (PCC). While PCC have been proven to work well in many gene expression studies, the correlation is often inflated by outliers or clustering effect, which made it not suitable to analyze data with high noise level such as metabolic profiles. In contrast, the robust correlations can provide a much better estimate of the true association in the presence of multi-dimensional outliers if the sample size is sufficiently large. We chose the Pair-wise Quadrant Correlation (PQC) for its good computational performance. However, for small sample size, PQC sometimes may inflate the correlation estimate. Therefore our correlation analysis, we use a novel approach by combining results from both PCC and PQC. Gene-metabolite pairs with high PCC and high PQC tend to demonstrate true linear correlation across all cancer types, which imply a general functional association between gene and metabolite profiles. On the other hand, gene-metabolite pairs with high PCC and low PQC are possibly resulted from a few extreme outliers, which can potentially be linked to cell-line specific signatures.\n[SUBTITLE] Metabolite-gene correlations [SUBSECTION] Since metabolite-gene relationships can help us to understand the direct or indirect cause of metabolite changes, we computed PCC and PQC for 11872 gene expression profiles and 253 metabolite measurements over 57 cell-lines.\nTo investigate the biological significances of the correlation analysis, we mapped all the known gene-compound associations from EHMN to our NCI60 analysis. There are a total of 721 EHMN annotated direct metabolite-gene associations, involving 352 genes and 81 known metabolites, that overlap with our NCI-60 metabolite-gene relationships.\nFigure 2 shows a mapping of our current metabolite-compound knowledgebase to the robust correlations. We chose the top 9 pathways in EHMN mapping ordered by number of genes\nHeatmap of gene-metabolite relationship organized according to KEGG pathway. Mapping of EHMN gene-metabolite association data to robust correlation matrix heat map. A. Rows are genes grouped by pathway names, columns are metabolites also grouped by corresponding pathway names. Green circles mark the position of a specific reaction that couples a metabolite and a gene from EHMN. Orange Light cells indicate positive PQC, with black to be 0 and orange to be 1. It can be seen that even though there are patterns of gene-metabolite clustering, very few high correlations can be mapped to known reactions. B. Gene-metabolite correlation in the Glycolysis pathway. C: The gene-gene correlation with in Glycolysis pathway.\nSome of the EHMN gene-compound relationships do show high association across all cell-lines. For example, gene AKR1B1 which reduces L-Arabitol to L-Arabinose with EHMN reaction ID R01758 and R01759, is associated with L-Arabitol with PQC of 0.69 and PCC of 0.36, which implies strong association between metabolite level and gene activity. However, it is surprising that most of the direct enzyme-metabolite relationships cannot be mapped to high correlations (Figure 2A and 2B). There are several possible explanations to the low match-ups of significant correlations to EHMN data. A simple explanation is most of the annotatable gene-metabolite relationships in the NCI-60 dataset may not be the speed limiting factor in the related pathways. It is also possible that the high level of abnormal regulation in the cancer cell lines may have masked global mechanisms such that it is difficult to identify high gene to metabolite correlations across all cell lines.\nInterestingly, the grouping of genes in the same pathway in Figure 2 enables us to detect many metabolites exhibit high correlations with other genes in the same pathway. For example, although phosphoenolpyruvate does not overlap with any of the EHMN annotated direct reaction genes in Figure 2B, it has high correlation with GPI and ALDOA, two of the genes in the glycolysis module that are known to be highly regulated by hypoxia-inducible factor 1alpha and such regulation is related to the aggressive phenotype of hepatocellular carcinoma. Consequently, ALDOA, GPI and other genes highly correlated with phosphoenolpyruvate in our multi-cancer cell line analysis may suggest that these genes has more significant regulatory or speed limiting roles in glycolysis than genes such as ENO1, ENO2 that are directly related to reactions involving phosphoenolpyruvate in these cancer cell lines. Naturally, not all genes in the same pathway strongly correlate with each other since genes in the same pathway are not all (Figure 2C).\nFurther investigation will be worthwhile for high correlation between metabolites and other genes (i.e. those not directly involved in the specific metabolic reaction) in the same pathway, as such un-annotated relationships are likely help us to identify speed limiting enzymes in a pathway, key regulatory genes of related pathways or novel metabolic mechanisms.\nSince metabolite-gene relationships can help us to understand the direct or indirect cause of metabolite changes, we computed PCC and PQC for 11872 gene expression profiles and 253 metabolite measurements over 57 cell-lines.\nTo investigate the biological significances of the correlation analysis, we mapped all the known gene-compound associations from EHMN to our NCI60 analysis. There are a total of 721 EHMN annotated direct metabolite-gene associations, involving 352 genes and 81 known metabolites, that overlap with our NCI-60 metabolite-gene relationships.\nFigure 2 shows a mapping of our current metabolite-compound knowledgebase to the robust correlations. We chose the top 9 pathways in EHMN mapping ordered by number of genes\nHeatmap of gene-metabolite relationship organized according to KEGG pathway. Mapping of EHMN gene-metabolite association data to robust correlation matrix heat map. A. Rows are genes grouped by pathway names, columns are metabolites also grouped by corresponding pathway names. Green circles mark the position of a specific reaction that couples a metabolite and a gene from EHMN. Orange Light cells indicate positive PQC, with black to be 0 and orange to be 1. It can be seen that even though there are patterns of gene-metabolite clustering, very few high correlations can be mapped to known reactions. B. Gene-metabolite correlation in the Glycolysis pathway. C: The gene-gene correlation with in Glycolysis pathway.\nSome of the EHMN gene-compound relationships do show high association across all cell-lines. For example, gene AKR1B1 which reduces L-Arabitol to L-Arabinose with EHMN reaction ID R01758 and R01759, is associated with L-Arabitol with PQC of 0.69 and PCC of 0.36, which implies strong association between metabolite level and gene activity. However, it is surprising that most of the direct enzyme-metabolite relationships cannot be mapped to high correlations (Figure 2A and 2B). There are several possible explanations to the low match-ups of significant correlations to EHMN data. A simple explanation is most of the annotatable gene-metabolite relationships in the NCI-60 dataset may not be the speed limiting factor in the related pathways. It is also possible that the high level of abnormal regulation in the cancer cell lines may have masked global mechanisms such that it is difficult to identify high gene to metabolite correlations across all cell lines.\nInterestingly, the grouping of genes in the same pathway in Figure 2 enables us to detect many metabolites exhibit high correlations with other genes in the same pathway. For example, although phosphoenolpyruvate does not overlap with any of the EHMN annotated direct reaction genes in Figure 2B, it has high correlation with GPI and ALDOA, two of the genes in the glycolysis module that are known to be highly regulated by hypoxia-inducible factor 1alpha and such regulation is related to the aggressive phenotype of hepatocellular carcinoma. Consequently, ALDOA, GPI and other genes highly correlated with phosphoenolpyruvate in our multi-cancer cell line analysis may suggest that these genes has more significant regulatory or speed limiting roles in glycolysis than genes such as ENO1, ENO2 that are directly related to reactions involving phosphoenolpyruvate in these cancer cell lines. Naturally, not all genes in the same pathway strongly correlate with each other since genes in the same pathway are not all (Figure 2C).\nFurther investigation will be worthwhile for high correlation between metabolites and other genes (i.e. those not directly involved in the specific metabolic reaction) in the same pathway, as such un-annotated relationships are likely help us to identify speed limiting enzymes in a pathway, key regulatory genes of related pathways or novel metabolic mechanisms.\n[SUBTITLE] Outlier analysis [SUBSECTION] In our previous analysis we require high PQC in addition to high PCC to identify molecule pairs with true high correlation. We have also identified many cases where cell lines have one or a few gene-metabolite pairs with much higher expression values than the rest of the other samples, which directly produce inflated PCC and very low PQC. To systematically investigate these cases, we used R package mvoulier to detect multidimensional outliers, and recomputed the PCC and PQC scores after outlier removal. Our empirical rule shows that when PCC > 0.6 and PQC < 0.3(Preferably close to 0), and the number of multidimensional outliers is smaller than 3, the high PCC is most likely to be an artifact from very few extreme outliers.\nTo explore the biological significance of these outliers, we compared the outlier cases detected by the criteria mentioned above to the Sanger Cosmic database [8]. Sanger Cosmic contains 177 mutation entries of 28 genes in our 57 NCI60 cell-lines. We ordered all the metabolite-gene correlations of the 28 genes by PCC, and amazingly the top two outliers detected by our approach, NOTCH1~X-2005 in MOLT-4 cell line and KRAS~in X-2690 in OVCAR-5 cell line, are the cell lines with known gene mutation on the exact same genes in Sanger Cosmic database (Figure 3).\nOutlier analysis. Outlier analysis shows extreme gene-metabolite pairs could be resulted from cell specific mutations. Top scatter plots: NOTCH1 ~ X-2005, with outlier in cell line MOLT_4 and KRAS ~ X-2690, with outlier in OVCAR5. It can be seen that the high PCC were both artefacts of extreme outliers. Middle table: PCC and PQC of the corresponding pairs, before and after outlier removal. R_QC: PQC. R_QC_RM: PQC after outlier removal. P_Pearson: PCC. R_P_RM: PCC after outlier removal. Bottom table: annotated mutation records from Sanger Cosmic database, directly matched to these two outlier pairs.\nThe common feature of these two gene-metabolite pairs is high PCC and low PQC before outlier removal and low PCC and low PQC after the outlier removal. From plots in Figure 3 we can verify that the inflated PCC were indeed a product of single cell line outlier. Besides, the sample sizes of these two cases are sufficiently large (22 for NOTCH1~X-2005 and 26 for KRAS~X-2690, respectively) so that the outlier is not likely to be a random fluctuation from small sample size.\nSince our unbiased analysis did not take advantage of any cell line or gene mutation information, the fact that our top two outliers overlaps with documented gene mutation in the Sanger Cosmic dataset suggests that the mutations are likely to be a cause of such outliers, and the associated metabolites may be good candidate biomarkers for such events.\nIn addition to point mutations, we also compared the outlier analysis with Copy Number Variation (CNV) data from Broad institute. Only 34 out of 57 samples from NCI60 have CNV data, and also found some gene-metabolite outlier pairs that are consistent with CNV outliers. For example, gene BRIP1, with CNV of 14.22 in cell line MCF7, has PCC of 0.90 and PQC of 0.008 and one outlier. The corresponding compound, X-3363, may be associated specifically to this copy number variation.\nThe fact that some top ranked gene-metabolite outlier associations matches with known corresponding genomic structural changes in the specific gene strongly suggests the effectiveness of our approach in identifying distinct molecular processes specific to metabolome and transcriptome variations.\nIn our previous analysis we require high PQC in addition to high PCC to identify molecule pairs with true high correlation. We have also identified many cases where cell lines have one or a few gene-metabolite pairs with much higher expression values than the rest of the other samples, which directly produce inflated PCC and very low PQC. To systematically investigate these cases, we used R package mvoulier to detect multidimensional outliers, and recomputed the PCC and PQC scores after outlier removal. Our empirical rule shows that when PCC > 0.6 and PQC < 0.3(Preferably close to 0), and the number of multidimensional outliers is smaller than 3, the high PCC is most likely to be an artifact from very few extreme outliers.\nTo explore the biological significance of these outliers, we compared the outlier cases detected by the criteria mentioned above to the Sanger Cosmic database [8]. Sanger Cosmic contains 177 mutation entries of 28 genes in our 57 NCI60 cell-lines. We ordered all the metabolite-gene correlations of the 28 genes by PCC, and amazingly the top two outliers detected by our approach, NOTCH1~X-2005 in MOLT-4 cell line and KRAS~in X-2690 in OVCAR-5 cell line, are the cell lines with known gene mutation on the exact same genes in Sanger Cosmic database (Figure 3).\nOutlier analysis. Outlier analysis shows extreme gene-metabolite pairs could be resulted from cell specific mutations. Top scatter plots: NOTCH1 ~ X-2005, with outlier in cell line MOLT_4 and KRAS ~ X-2690, with outlier in OVCAR5. It can be seen that the high PCC were both artefacts of extreme outliers. Middle table: PCC and PQC of the corresponding pairs, before and after outlier removal. R_QC: PQC. R_QC_RM: PQC after outlier removal. P_Pearson: PCC. R_P_RM: PCC after outlier removal. Bottom table: annotated mutation records from Sanger Cosmic database, directly matched to these two outlier pairs.\nThe common feature of these two gene-metabolite pairs is high PCC and low PQC before outlier removal and low PCC and low PQC after the outlier removal. From plots in Figure 3 we can verify that the inflated PCC were indeed a product of single cell line outlier. Besides, the sample sizes of these two cases are sufficiently large (22 for NOTCH1~X-2005 and 26 for KRAS~X-2690, respectively) so that the outlier is not likely to be a random fluctuation from small sample size.\nSince our unbiased analysis did not take advantage of any cell line or gene mutation information, the fact that our top two outliers overlaps with documented gene mutation in the Sanger Cosmic dataset suggests that the mutations are likely to be a cause of such outliers, and the associated metabolites may be good candidate biomarkers for such events.\nIn addition to point mutations, we also compared the outlier analysis with Copy Number Variation (CNV) data from Broad institute. Only 34 out of 57 samples from NCI60 have CNV data, and also found some gene-metabolite outlier pairs that are consistent with CNV outliers. For example, gene BRIP1, with CNV of 14.22 in cell line MCF7, has PCC of 0.90 and PQC of 0.008 and one outlier. The corresponding compound, X-3363, may be associated specifically to this copy number variation.\nThe fact that some top ranked gene-metabolite outlier associations matches with known corresponding genomic structural changes in the specific gene strongly suggests the effectiveness of our approach in identifying distinct molecular processes specific to metabolome and transcriptome variations.", "The first question we would like to address is whether cancer cell lines exhibit tissue-origin specific metabolic signatures. We used classification analysis similar to those performed in microarray data [7] to classify 57 cell-lines into 9 cancer types related to their tissue origins. Our initial tests suggest that regardless of method used, the classification error on metabolomics data is always significantly higher than that from microarray (~0.51 for metabolomics and ~0.34 for microarray).\nBesides the fact that NCI-60 metabolome dataset has much less data points but with higher noise levels when compared to microarray data, we speculate that the low number of some cancer types (e.g. only 2 prostate cancer cell lines) and high heterogeneity of some cell lines (breast, ovarian and CNS) may also undermine the performance of the metabolome-based classification. In fact, these 4 cell lines contribute most to the out of bag classification error estimates.\nAfter prostate, breast, ovarian and CNS cell lines are removed from the whole sample, the metabolite classifiers can reach comparable performance with microarray classifiers (Figure 1). The significantly improved metabolome-based classification results after removing cell lines with high variability or heterogeneity strongly suggests that there are indeed cancer subtype-specific (based on tissue of origin) metabolic signatures.\nClassification error. Estimated Out-of-Bag (OOB) Error with regard to progressive cancer class removal. At each step, cancer classes contribute most to classification error were removed and OOB errors were recalculated. B: Breast Cancer. P: Prostate Cancer. O: Ovarian Cancer. C: CNS. The left plot shows OOB errors on entire dataset and the right plot shows OOB errors on subset of classifiers selected by varSelRF. The improvement of variable selection is more remarkable for microarray because of the much larger classifier pool to select from (11961 v.s. 342). Metabolite classifiers can achieve average OOB error of 0.28 when B,P,O,C are removed, reduced from 0.51 from the full set. The OOB error for u133a reduced from 0.34 to 0.18. The average OOB errors from the random cancer class removal are plotted in the orange dashed line.\nSince the removal of cancer classes will also reduce the complexity of the data thus naturally improve classification performance, we also removed the same number of cancer classes at each stage randomly for 500 times to estimate the effect of improvement of OOB errors due to simplification of data structure resulted from sample removal (orange dashed lines in Figure 1). It can be seen that the reductions in OOB errors from random removal are not as significant as those from progressively removal of the prostate, breast, ovarian and CNS cell line classes, suggesting that it is indeed the high heterogeneity and the low sample number of these cell line classes that caused the high OOB error in the original classification. Consequently, our results suggest that although metabolite profiles are not as good as microarray data for the classification of cancer cell lines, metabolites nonetheless contain cell type specific signature and the classification results can be improved if we have more samples or have more data points in the metabolome assays.", "Although existing metabolome data do not outperform microarray data in cell line subtype classification, metabolome can still be useful for revealing some basic molecular processes and their alterations in NCI-60 cell lines. We hypothesize that different cell lines should share some basic regulatory and metabolic processes (steady-state) essential for cell growth and metabolism. It is likely that although different cancer cell lines may have very different levels of gene expression and metabolism, the steady-state relationship between genes and metabolites in the same or highly coupled pathways should be conserved across different cell lines in the absence of dramatic genomic changes such as gene mutation and copy number variation (CNV). Identification of such gene-metabolite relationships at steady state will help us better understand the underlying molecular mechanisms related to the metabolic change. They will also help us infer potential metabolic changes based on the expression data or vice versa. On the other hand, if a few cell lines deviated significantly from the gene-metabolite relationships exhibited by the most of the cell lines, the related genes in such outlier cell lines are likely to be silenced (no expression) or agitated (over expression) due to genomic structural changes. Consequently, we are interested in both high correlations which reflect steady state trend over all samples, and outliers which reflect signatures in specific cell lines.\nThe classic method of inspecting the relatedness of two quantitative features is by computing the Pearson Correlation Coefficient (PCC). While PCC have been proven to work well in many gene expression studies, the correlation is often inflated by outliers or clustering effect, which made it not suitable to analyze data with high noise level such as metabolic profiles. In contrast, the robust correlations can provide a much better estimate of the true association in the presence of multi-dimensional outliers if the sample size is sufficiently large. We chose the Pair-wise Quadrant Correlation (PQC) for its good computational performance. However, for small sample size, PQC sometimes may inflate the correlation estimate. Therefore our correlation analysis, we use a novel approach by combining results from both PCC and PQC. Gene-metabolite pairs with high PCC and high PQC tend to demonstrate true linear correlation across all cancer types, which imply a general functional association between gene and metabolite profiles. On the other hand, gene-metabolite pairs with high PCC and low PQC are possibly resulted from a few extreme outliers, which can potentially be linked to cell-line specific signatures.", "Since metabolite-gene relationships can help us to understand the direct or indirect cause of metabolite changes, we computed PCC and PQC for 11872 gene expression profiles and 253 metabolite measurements over 57 cell-lines.\nTo investigate the biological significances of the correlation analysis, we mapped all the known gene-compound associations from EHMN to our NCI60 analysis. There are a total of 721 EHMN annotated direct metabolite-gene associations, involving 352 genes and 81 known metabolites, that overlap with our NCI-60 metabolite-gene relationships.\nFigure 2 shows a mapping of our current metabolite-compound knowledgebase to the robust correlations. We chose the top 9 pathways in EHMN mapping ordered by number of genes\nHeatmap of gene-metabolite relationship organized according to KEGG pathway. Mapping of EHMN gene-metabolite association data to robust correlation matrix heat map. A. Rows are genes grouped by pathway names, columns are metabolites also grouped by corresponding pathway names. Green circles mark the position of a specific reaction that couples a metabolite and a gene from EHMN. Orange Light cells indicate positive PQC, with black to be 0 and orange to be 1. It can be seen that even though there are patterns of gene-metabolite clustering, very few high correlations can be mapped to known reactions. B. Gene-metabolite correlation in the Glycolysis pathway. C: The gene-gene correlation with in Glycolysis pathway.\nSome of the EHMN gene-compound relationships do show high association across all cell-lines. For example, gene AKR1B1 which reduces L-Arabitol to L-Arabinose with EHMN reaction ID R01758 and R01759, is associated with L-Arabitol with PQC of 0.69 and PCC of 0.36, which implies strong association between metabolite level and gene activity. However, it is surprising that most of the direct enzyme-metabolite relationships cannot be mapped to high correlations (Figure 2A and 2B). There are several possible explanations to the low match-ups of significant correlations to EHMN data. A simple explanation is most of the annotatable gene-metabolite relationships in the NCI-60 dataset may not be the speed limiting factor in the related pathways. It is also possible that the high level of abnormal regulation in the cancer cell lines may have masked global mechanisms such that it is difficult to identify high gene to metabolite correlations across all cell lines.\nInterestingly, the grouping of genes in the same pathway in Figure 2 enables us to detect many metabolites exhibit high correlations with other genes in the same pathway. For example, although phosphoenolpyruvate does not overlap with any of the EHMN annotated direct reaction genes in Figure 2B, it has high correlation with GPI and ALDOA, two of the genes in the glycolysis module that are known to be highly regulated by hypoxia-inducible factor 1alpha and such regulation is related to the aggressive phenotype of hepatocellular carcinoma. Consequently, ALDOA, GPI and other genes highly correlated with phosphoenolpyruvate in our multi-cancer cell line analysis may suggest that these genes has more significant regulatory or speed limiting roles in glycolysis than genes such as ENO1, ENO2 that are directly related to reactions involving phosphoenolpyruvate in these cancer cell lines. Naturally, not all genes in the same pathway strongly correlate with each other since genes in the same pathway are not all (Figure 2C).\nFurther investigation will be worthwhile for high correlation between metabolites and other genes (i.e. those not directly involved in the specific metabolic reaction) in the same pathway, as such un-annotated relationships are likely help us to identify speed limiting enzymes in a pathway, key regulatory genes of related pathways or novel metabolic mechanisms.", "In our previous analysis we require high PQC in addition to high PCC to identify molecule pairs with true high correlation. We have also identified many cases where cell lines have one or a few gene-metabolite pairs with much higher expression values than the rest of the other samples, which directly produce inflated PCC and very low PQC. To systematically investigate these cases, we used R package mvoulier to detect multidimensional outliers, and recomputed the PCC and PQC scores after outlier removal. Our empirical rule shows that when PCC > 0.6 and PQC < 0.3(Preferably close to 0), and the number of multidimensional outliers is smaller than 3, the high PCC is most likely to be an artifact from very few extreme outliers.\nTo explore the biological significance of these outliers, we compared the outlier cases detected by the criteria mentioned above to the Sanger Cosmic database [8]. Sanger Cosmic contains 177 mutation entries of 28 genes in our 57 NCI60 cell-lines. We ordered all the metabolite-gene correlations of the 28 genes by PCC, and amazingly the top two outliers detected by our approach, NOTCH1~X-2005 in MOLT-4 cell line and KRAS~in X-2690 in OVCAR-5 cell line, are the cell lines with known gene mutation on the exact same genes in Sanger Cosmic database (Figure 3).\nOutlier analysis. Outlier analysis shows extreme gene-metabolite pairs could be resulted from cell specific mutations. Top scatter plots: NOTCH1 ~ X-2005, with outlier in cell line MOLT_4 and KRAS ~ X-2690, with outlier in OVCAR5. It can be seen that the high PCC were both artefacts of extreme outliers. Middle table: PCC and PQC of the corresponding pairs, before and after outlier removal. R_QC: PQC. R_QC_RM: PQC after outlier removal. P_Pearson: PCC. R_P_RM: PCC after outlier removal. Bottom table: annotated mutation records from Sanger Cosmic database, directly matched to these two outlier pairs.\nThe common feature of these two gene-metabolite pairs is high PCC and low PQC before outlier removal and low PCC and low PQC after the outlier removal. From plots in Figure 3 we can verify that the inflated PCC were indeed a product of single cell line outlier. Besides, the sample sizes of these two cases are sufficiently large (22 for NOTCH1~X-2005 and 26 for KRAS~X-2690, respectively) so that the outlier is not likely to be a random fluctuation from small sample size.\nSince our unbiased analysis did not take advantage of any cell line or gene mutation information, the fact that our top two outliers overlaps with documented gene mutation in the Sanger Cosmic dataset suggests that the mutations are likely to be a cause of such outliers, and the associated metabolites may be good candidate biomarkers for such events.\nIn addition to point mutations, we also compared the outlier analysis with Copy Number Variation (CNV) data from Broad institute. Only 34 out of 57 samples from NCI60 have CNV data, and also found some gene-metabolite outlier pairs that are consistent with CNV outliers. For example, gene BRIP1, with CNV of 14.22 in cell line MCF7, has PCC of 0.90 and PQC of 0.008 and one outlier. The corresponding compound, X-3363, may be associated specifically to this copy number variation.\nThe fact that some top ranked gene-metabolite outlier associations matches with known corresponding genomic structural changes in the specific gene strongly suggests the effectiveness of our approach in identifying distinct molecular processes specific to metabolome and transcriptome variations.", "NCI-60 data pre-processing: raw molecular datasets were downloaded from DTP web portal (http://dtp.nci.nih.gov/index.html March 2007 release). There are 57 cell-lines in 9 cancer classes have both microarray and metabolomics data. We used our in-house Entrez-based Custom CDF version 12 to derive gene-level expression data from the NCI-60 Affymetrix Genechip CEL files (http://brainarray.mbni.med.umich.edu)[9]. The metabolite data averaged over triplicate experiments were manually compared with a reference dataset to exclude imputed values (Beecher, unpublished data) as the imputed data could significantly bias the inferences drawn from correlation analysis. All metabolite names were manually compared with KEGG and assigned a KEGG compound ID whenever possible. The pre-processing step produced 11961 and 6089 gene expression profiles from Affymetrix HG-U133A and HG-U133B chips respectively, and quantitative data for 124 known and 218 unknown metabolites.\nAll statistical analyses were performed in R (http://www.r-project.org/). Classification and variable selection were performed by R randomForest (http://cran.r-project.org/web/packages/randomForest/index.html) and varSelRF package [10]. The robust correlations were computed with robust package (http://cran.r-project.org/web/packages/robust/index.html), using the parameter pair-wise Quadrant Correlation (pairwiseQC). Multidimensional outliers were identified by package mvoutlier [11]. We performed the computation on our LINUX cluster of ~100 cores. The results were loaded into an Oracle database for integrative analysis.\nThe annotated gene to metabolite relationship data kindly offered by Edinburgh Human Metabolic Network project is used for identifying biological relevant known gene-metabolite relationships revealed by our analysis [5]. We also compiled a local version of KEGG [12] and DAVID 2008 [13] for similar purposes. In order to associate known mutations and CNVs in NCI-60 cell lines to abnormal gene-metabolite relationships, we have built a local copy of COSMIC [8] dataset from Sanger and Tumorscape [14] from Broad institute for cell-line specific point-mutation and copy number variations (CNV), respectively.", "Our analysis results on the NCI-60 metabolome and transcriptome data suggest that 1) although the small sample size, the high noise level and intra-cancer class heterogeneity in the current NCI-60 metabolome dataset makes it unsuitable for global cancer subtype classification, there are indeed metabolic signatures associated with cancer subtypes. 2) There are biologically meaningful high correlation gene-metabolite pairs across NCI-60 cell lines, identifiable by robust correlation estimates. 3) Most strikingly, there are several examples of abnormal gene-metabolite coupling that can be directly linked to known gene mutations or copy number variations.\nConceivably, high correlations as well as outliers can be utilized to aid the progressive prediction of unknown metabolites based on annotations in existing pathway databases and literature. For example, the high correlation of an unknown compound with a known gene and in particular, multiple known genes in a pathway can dramatically reduce the search space for the unknown compound, since the most likely candidates will be structurally related molecules or known metabolites from related pathways. We plan to compare the Mass Spectrometer (MS) features of these unknown compounds from the predicted candidate pool and conduct wet-lab experiments for validation. Since we have discovered that many unknown metabolites are strongly associated with each other but not with any known compounds. Correct determination of even a small fraction of them would facilitate the identification of the rest, which also will in turn improve our understanding of the molecular processes and pathways involving these molecules.\nIn summary, the strong biological relevance of our results also suggest that the analysis strategy we developed based on the combination of PCC and PQC correlation values for identifying real correlation and true outliers is a powerful approach for integrative analysis of noisy ‘omics’ datasets. The presentation of gene-metabolite relationship analysis results in a heatmap grouped by genes in the same pathway together with overlay of known gene-metabolite relationship provides a powerful visual exploration approach for identifying both direct and indirect gene-metabolite relationships. The analysis methods we described here will be useful for other integrated analysis of metabolome and transcriptome and the wet lab validation of novel gene-metabolite relationships.", "The authors declare that they have no competing interests.", "GS performed data cleaning, correlation analysis and outlier analysis. CFB helped with correlation analysis and functional inference. CWB cleaned metabolomics data and annotated missing values. GS, BDA evaluated robust correlation methods; FM initialized the study, compiled microarray data and supervised biological validation." ]

[ null, null, null, null, null, null, "methods", null, null, null ]

[]

[ "Computational Biology", "Gene Expression Profiling", "Gene Expression Regulation", "Gene Regulatory Networks", "Humans", "MicroRNAs", "Molecular Sequence Annotation", "RNA Processing, Post-Transcriptional", "Transcription Factors" ]

[ "Background", "Regulation relationships", "Identification of significant coregulation relationships", "Functional linkages and networks", "Enriched network motifs", "Analysis of expression profiles", "Functional coregulation pairs", "Functional coregulation networks", "Network motifs for coregulation pairs", "Expression patterns of coregulation pairs", "Discussion and conclusion", "Competing interests", "Authors' contributions" ]

[ "Transcriptional regulatory networks describe the interactions between transcriptional regulatory proteins and their target genes [1-3]. These regulators, known as transcription factors (TFs), are proteins that bind to specific DNA sequences and thereby control the transcription of genetic information encoded in DNA sequences. The interactions between TFs and target genes regulate the transcriptional activities of genome and thus determine the global gene expression program of a living cell.\nIn the last decade, microRNAs (miRNAs) have emerged as another prominent class of gene regulators. miRNAs are endogenous small RNA molecules that are abundant in animals, plants, and some viruses. They can reduce stability and/or translation activity of fully or partially sequence-complementary messenger RNAs (mRNAs), thus regulating gene expression at the post-transcriptional level. It has been found that miRNAs may control many biological processes in development, differentiation, growth, and even cancer development and progression [4-6].\nRecent studies have suggested that miRNAs and TFs are primary metazoan gene regulators, and they seem to function in a similar regulatory logic, such as pleiotropy, combinatorial and cooperative activity, regulation, and even network motifs [7,8]. However, how miRNAs interplay and coordinate with TFs in the regulatory network still remains unclear. Since combinatorial interactions between miRNAs and TFs are complicated and thus hard to be validated by high-throughput experiments, computational modelling may provide a better clue to understand such complex relationships.\nCurrently, to uncover the coregulation interactions between miRNAs and TFs, researchers have to overcome two challenges. One is the incomplete knowledge of regulatory targets. Because the available experimentally verified targets of miRNAs and TFs are far from complete, the regulatory target datasets for global analysis were mainly from computational prediction. The other challenge is about how to integrate transcriptional and post-transcriptional layers to discover highly confident coregulation relationships. To solve these problems, previous studies have developed a bottom-up strategy; that is, they inferred the coordination between two upstream regulators from their downstream shared targets [9,10]. These inferences were basically based on different probabilistic models and statistical tests to measure the significance of shared targets between regulators. Indeed, the methods successfully eliminated those insignificant coregulation interactions occurred merely by chance; however, the biological meanings were ignored in the integration of transcriptional and post-transcriptional regulation interactions.\nHere we proposed a novel framework utilizing functional annotation information to identify significant coregulation between transcriptional and post-transcriptional layers. Based on this model, function-enriched coregulation pairs were discovered, and the regulators were subsequently linked by enriched functions. With these functional linkages, we further constructed functional coregulation networks between regulators and investigated their characteristics. Next, we searched for the network motifs consisting of those function-enriched coregulation pairs, and found that an abundance of pairs were closely connected in their upstream. Finally, the expression patterns of function-enriched coregulation pairs were analyzed. Different coregulation types showed distinct expression correlation trends. More importantly, we found that the disruption of coregulation may be closely related to cancers.", "The transcriptional regulation relationships between human transcription factors and their target genes were collected from TRED (Transcriptional Regulatory Element Database) [11]. The database provides genome-wide promoter annotation and transcription factor binding information from computational prediction and experimental evidence.\nTo collect all human TF-target regulation relationships in TRED, we firstly queried the list of all human TFs in the database. A total of 178 human TFs were obtained by this step. Next, we searched TF target genes for each TF using default parameters (promoter quality from \"known, curated\" to \"with RNA\" and \"all\" binding quality). The results showed that only 133 TFs were found to have at least one target gene by these criteria, and the final number of unique TF-target relationships was 6,764, which were used to construct the human transcriptional regulatory network for our analysis.\nSince the available experimentally verified human miRNA targets are far from complete and thus not enough for global analysis, we used predicted miRNA targets from the TargetScan database (release 4.2) to perform the analysis [12]. In addition, different mature miRNAs may have identical seed regions and thereby target the same binding sites. To eliminate those coregulation interactions among the miRNAs with identical seed regions, we grouped mature miRNAs into families based on the miRNA family information from TargetScan. A total of 162 miRNA families and 7,521 target genes with 44,782 interactions were collected.\nIt is still difficult to predict the promoter region of miRNA genes in the genome. But it has been known that embedded miRNAs frequently coexpress with their host genes [13,14]. Therefore, we extracted miRNA host gene information from miRBase [15] and integrated the embedded miRNAs biogenesis information into the established transcriptional regulation network. A total of 310 premature miRNAs were found embedded in 259 host genes. Most of them (93%) were resided in introns.", "Combing all the potential targets of miRNAs and TFs, we firstly constructed two adjacency matrixes describing the regulator-target interaction for TFs and miRNAs, respectively. Then the two matrixes were combined into three cross-adjacency matrixes representing the shared targets of TF-TF, miRNA-miRNA, and TF-miRNA coregulation pairs. An example of identification of TF-miRNA coregulation pairs is shown in Figure 1.\nExample workflow for identifying significant coregulation\nSecondly, for each group of shared targets, the distribution of Gene Ontology (GO) annotations [16] at the second level in the biological process namespace was calculated. We chose the second level ontology because most of the genes were generally well-annotated at this level and these annotations provided a good balance between the sensitivity and the specificity in the following functional enrichment test. The distributions were considered as the functional profiles or fingerprints for these coregulation pairs.\nNext, we utilized a randomization method to perform a permutation test for functional enrichment. For each group of shared targets, we randomly selected a null group of the same size from whole human genome as background. After 10,000 iterations, the log-likelihood score under multivariate hypergeometric distribution was measured to quantify the significance of functional enrichment. The correction for multiple comparisons was made under 0.05 false discovery rate (FDR) [17]. The final results of significant coregulation pairs were listed in additional file 1.", "For each function-enriched coregulation pair, Fisher’s Exact Test following FDR correction were conducted to identify enriched GO terms. Similarly, we only focused on the second level terms in biological process namespace. A functional linkage was established if a GO term overrepresented in the shared targets of a coregulation pair, implying that the two paired regulators may function coordinately in the specific biological process. Based on these linkages, we further constructed the functional coregulation networks (Figure 2). In order to investigate the specificity of coregulation relationships and provide a global view, only those linkages with relatively high significance that passed FDR-BL correction [17] were used to construct the networks.\nExample workflow for constructing functional coregulation network. For each coregulation pair, a function linkage was established if a GO term enriched in their shared targets. The coregulation network was generated based on these function linkages. Nodes represent regulators, and edge represents GO terms, marked with different colours.", "We searched for network motifs preferentially occurred in function-enriched coregulation pairs rather than in random pairs by a resampling process. The predicted TF-targeting interactions for miRNA genes were collected from miRBase [15] and from literature [9]. In addition, we assumed that those embedded miRNA genes have same transcription units as their host genes and would be regulated together.\nA total of 10,000 background sets of regulator pairs that have the same size as the set of function-enriched pairs were randomly selected from the global network. For each type of network patterns (sub-graphs), the observed frequency from the function-enriched coregulation pairs was first calculated and compared to the background distribution for assessment of significance. Only those network patterns with occurrence probabilities less than 0.001 were considered significant motifs (see additional file 2 for these significant motifs).", "The miRNA and mRNA expression profiles were adopted from a previous study [18]. A total of 217 miRNAs and ~16,000 mRNAs across 8 human tissues (colon, pancreas, kidney, bladder, prostate, uterus, lung, and breast) were measured using miRNA bead-arrays and mRNA microarrays. Both normal and tumor samples were profiled for each tissue. For each type of coregulation, we first generated background distribution by calculating the Pearson's correlation coefficients (PCCs) of expression profiles between the two paired regulators in all possible pairs (i.e., those pairs shared no targets and/or those pairs not identified as function-enriched). After that, the distribution of enriched coregulation pairs was calculated and shown against the background.", "After the integration of miRNA regulation into human transcriptional regulation network, we adopted a novel strategy utilizing functional information to identify function-enriched coregulation pairs, and establish function linkages for each pair. Traditional analysis of functional enrichment was aimed at elucidating the regulatory roles of each individual regulator only, inevitably leaving some significant coregulation hidden in the traditional views. Instead, based on our model, different regulation types involving single regulators or combinations of regulators can all be studied and compared.\nThe distributions of different regulation types were grouped into two diagrams for comparing. Figure 3A shows distributions of individual TF regulation and TF-TF coregulation. The two distributions look similar; however, two biological processes, pigmentation and reproductive process, emerge when it comes to TF-miRNA coregulation, implying that the two biological processes may be the typical processes in which TFs should coordinate with miRNAs to control the expression programs.\nDistributions of enriched biological processes for different regulation types. (A) Distributions for TF-involving regulation: individual TFs, TF-TF pairs, and TF-miRNA pairs. (B) Distributions for miRNA-involving regulation: individual miRNAs, miRNA-miRNA pairs, and TF-miRNA pairs. Note that the same TF-miRNA line are drawn in both (A) and (B) for comparison.\nIn contrast, miRNA-involving regulation shows divergent distributions in Figure 3B. The top ranked biological processes of individual miRNA regulation were biological regulation, cellular process, and developmental process, which were the previously known miRNA-involving processes [4-6]. On the other hand, biological adhesion was relatively high in miRNA-miRNA coregulation, suggesting that miRNAs may regulate this process majorly in a coordination manner.\nMoreover, many biological processes enriched in TF-miRNA coregulation were relatively poor in the regulation involving miRNAs only. In other words, those processes may be the typical processes needed to be coordinately regulated by TFs and miRNAs, and the coordination may provide a mechanism to switch expression programs. More importantly, it suggested that, by coordinating with TFs, miRNAs may engage in a wider diversity of biological processes, and these undiscovered processes were failed to be identified by traditional analysis of functional enrichment for a single regulator.", "In the previous section, different regulators were connected by identified functional linkages, which represented that the two paired regulators may function in coordination with each other in a specific biological process. We further built up functional coregulation networks from these linkages and found interesting properties in the networks.\nFigure 4 shows the three subnetwork examples (the largest connected component) for different coregulation types. Obviously, it appears that the TF-TF coregulation network (Figure 4A) was in high density coregulation (avg. 3.52 links per pair) and full of diversity in edge types (avg. 5.70 edge types per regulator); in other words, TFs may function together in many kinds of biological processes. Likewise, the miRNA-miRNA coregulation network (Figure 4B) showed similar high density coregulation (avg. 3.12 links per pair) but the diversity of involving processes was lower (avg. 3.92 edge types per regulator) than TF-TF coregulation. In contrast, the TF-miRNA coregulation network (Figure 4C) was in higher specificity (avg. 1.48 links per pair; avg. 2.22 edge types per regulator); that is, a TF may coordinate with different neighbour miRNAs in specific biological processes, and vice versa. These results suggested that the cross-layer coregulation may have higher specificity than intra-layer coregulation.\nFunctional coregulation networks. (A) TF-TF coregulation network. (B) miR-miR coregulation network. (C) TF-miR coregulation network. Red nodes represent TFs; green nodes represent miRNAs; and edges represent biological processes. Different edge colours represent different biological processes.", "Many studies have been devoted to understanding network structures in gene regulatory networks, and have found that most networks seem to be largely composed of occurring patterns, called network motifs. The functions associated with common network motifs, such as auto-regulation and feed-forward loops (FFLs), were discovered and revealed by several researches both theoretically and experimentally [1,9,10,19-22].\nUnsurprisingly, the function-enriched coregulation pairs also have preferentially recurring network motifs as shown in Figure 5. Several types of motifs were found in TF-TF coregulation; for example, bidirectional and unidirectional FFLs were explored and these results were consistent with previous studies on network biology. In addition to FFLs, we went further to investigate the upstream regulatory patterns of coregulation pairs and found that the two paired regulators were closely linked in their upstream. For instance, over half of the pairs had common upstream TFs; a significant number of pairs had common TFs and miRNAs; and almost all pairs were cross-regulated in their upstream. Similar results were also found in TF-miRNA coregulation. However, no enriched motif was found in miRNA-miRNA coregulation, probably due to the incomplete knowledge of regulatory regions of miRNA genes.\nNetwork motifs for TF-TF and TF-miRNA coregulation pairs. For each motif, the observation number and the percentage of pairs that have this motif were reported, along with the P-value and one instance coregulation pair.", "Expression data across human normal/tumor tissues have recently become available. A previous study measured miRNA and mRNA expression profiles across 8 tissues (colon, pancreas, kidney, bladder, prostate, uterus, lung, and breast) and each tissue contained both normal and tumor samples [18]. By analyzing the expression profiles, we investigated the correlations between the expression profiles of each coregulation pair in both normal and cancer samples.\nFigure 6 shows the expression correlations of different coregulation types in normal and tumor samples. Interestingly, the three coregulation types show distinct trends in normal tissues. For example, TF-TF had a zero-centered distribution similar to background; TF-miRNA had two tendencies to highly-positive and medium-negative correlations in comparison with its background; miRNA-miRNA showed preference for only positive correlation. The different trends may reflect the diversity of function roles between TFs and miRNAs; that is, TFs can act as activators (+) or repressors (-) in gene regulation; however, miRNAs mainly act as repressors (-) by translation inhibition or transcript destabilization. Thus, it seems that these typical trends may be mechanistically reasonable.\nDistributions of expression correlation for different types of coregulation pairs. Left column for normal tissues; right column for tumor tissues. Rows are in the order of TF-TF, TF-miR, and miR-miR coregulation pairs.\nOn the contrary, all coregulation types turn into an identical trend in tumor tissues. All of them show similar zero-centered distributions resembled to their backgrounds. This trend suggests that the function-enriched coregulation pairs lost their correlation in tumor tissues, implying the disruption of coregulation may be closely associated to cancers. Together these results may support the functionality of identified coregulation pairs.", "We proposed a novel strategy aimed at identifying potential coordinated regulation by utilizing functional annotation information and discovered many biological processes that emerged only in coregulation. Compared to traditional function enrichment analysis, our strategy considered whole function profiles rather than single annotations. In addition, it also solves the restriction of traditional methods that only focus on single regulator. For example, a miRNA can potentially regulate an abundance of target genes. To find enriched functions of the miRNA, all its potential targets will be tested for any enriched function. However, since the target size of a miRNA may be huge, some meaningful biological processes involving only a small subset of genes will be hidden. In fact, these hidden processes may be significantly impacted by miRNAs in coordination with other regulators, namely, other miRNAs or TFs. After all, a biological process may be regulated not only at the transcriptional layer, but also at the posttranscriptional layer [7,8,23,24].\nInterestingly, our results show that pigmentation and reproductive process are two typical biological processes specifically emerging in TF-miRNA coregulation. It is suggested that miRNAs may provide genetic switch mechanisms to essentially inactivate the target genes, thus leading to detectable phenotypic consequences. In model organisms, there have been many studies investigating the switch-like role of miRNAs in pigmentation. For example, miRNAs can regulate the eye pigmentation genes in Drosophila [25]. The influence of miRNAs on pigmentation in zebrafish was also reported [26]. Another study found that miR-434-5p may mediate skin whitening and lightening in mouse [27]. And in melanoma cell lines, it is shown that miR-137 may target a pigmentation regulator [28].\nThe analysis of functional coregulation networks provided other clues. We found that a TF may regulate in coordination with different miRNAs in different biological processes, and vice versa. It suggested that the cross-layer coregulation may have higher specificity than intra-layer coregulation.\nWe also performed network motif analysis to see if any recurring pattern exists in coregulation network structure. Different types of feed-forward loops were found in TF-TF and TF-miRNA coregulation, and these results were consistent with several previous studies on transcriptional network [1,9,10,19-22]. Among these FFLs, a special kind of miRNA-mediated FFLs emerged in TF-miRNA coregulation. In this kind of FFLs, a miRNA may simultaneously repress a TF and its target genes, thus contributing to a switch-like control of expression programs. More importantly, we go further this time to investigate the upstream structure of coregulation pairs and found closely interaction in their upstream. It implies that the network structures of coregulation may have extensive crosstalk in the higher levels.\nFinally, the expression analysis of coregulation discovered distinct trends in different coregulation types; namely, TF-TF showed no correlation, whereas miRNA-miRNA had a preference of positive correlation, and TF-miRNA appeared both positive and negative correlation. A previous study investigated only TF-miRNA correlation and found the same tendencies [9]. The authors rationalized this trend by pointing out the distinct function roles that TFs and miRNAs may play. We further supported this idea by showing the results of TF-TF and miRNA-miRNA coregulation, which were also consistent with the same interpretation. In addition, TF activities are under control at protein level; that is, TFs may be activated or deactivated by a number of mechanisms including phosphorylation, ligand binding, and interaction with other regulatory proteins. Therefore, it is not surprising that co-function TFs may show no correlation in mRNA expression level. Notably, a large proportion of TF-miRNA pairs showed negative correlation in expression profiles, which could be explained by the structure of the miRNA-mediated-FFLs discussed before, supporting the idea that many miRNAs in TF-miRNA coregulation contributed to switch-like regulation.\nMore significantly, by comparing the expression correlations between normal and tumor tissues, we found a common trend in function-enriched coregulation pairs; that is, the function-enriched pairs lost their correlation in tumor tissues. It suggested that the disruption of coregulation may lead to abnormal expression programs and may be directly associated to cancers.\nOur findings shed light on the coregulation of miRNAs in transcriptional regulatory network. Future experimental works will provide more complete knowledge in transcriptional network and miRNA regulation, thus allowing the elucidation of more precise co-regulatory mechanisms.", "The authors declare that they have no competing interests.", "CYC implemented the computational method, carried out the analysis, and drafted the manuscript. CYC and HCH conceived of the study. STC and CSF helped to perform the analysis and participated in discussions of the research. HFJ and HCH participated in the design and coordination of the study, and edited the manuscript. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Regulation relationships", "Identification of significant coregulation relationships", "Functional linkages and networks", "Enriched network motifs", "Analysis of expression profiles", "Results", "Functional coregulation pairs", "Functional coregulation networks", "Network motifs for coregulation pairs", "Expression patterns of coregulation pairs", "Discussion and conclusion", "Competing interests", "Authors' contributions", "Supplementary Material" ]

[ "Transcriptional regulatory networks describe the interactions between transcriptional regulatory proteins and their target genes [1-3]. These regulators, known as transcription factors (TFs), are proteins that bind to specific DNA sequences and thereby control the transcription of genetic information encoded in DNA sequences. The interactions between TFs and target genes regulate the transcriptional activities of genome and thus determine the global gene expression program of a living cell.\nIn the last decade, microRNAs (miRNAs) have emerged as another prominent class of gene regulators. miRNAs are endogenous small RNA molecules that are abundant in animals, plants, and some viruses. They can reduce stability and/or translation activity of fully or partially sequence-complementary messenger RNAs (mRNAs), thus regulating gene expression at the post-transcriptional level. It has been found that miRNAs may control many biological processes in development, differentiation, growth, and even cancer development and progression [4-6].\nRecent studies have suggested that miRNAs and TFs are primary metazoan gene regulators, and they seem to function in a similar regulatory logic, such as pleiotropy, combinatorial and cooperative activity, regulation, and even network motifs [7,8]. However, how miRNAs interplay and coordinate with TFs in the regulatory network still remains unclear. Since combinatorial interactions between miRNAs and TFs are complicated and thus hard to be validated by high-throughput experiments, computational modelling may provide a better clue to understand such complex relationships.\nCurrently, to uncover the coregulation interactions between miRNAs and TFs, researchers have to overcome two challenges. One is the incomplete knowledge of regulatory targets. Because the available experimentally verified targets of miRNAs and TFs are far from complete, the regulatory target datasets for global analysis were mainly from computational prediction. The other challenge is about how to integrate transcriptional and post-transcriptional layers to discover highly confident coregulation relationships. To solve these problems, previous studies have developed a bottom-up strategy; that is, they inferred the coordination between two upstream regulators from their downstream shared targets [9,10]. These inferences were basically based on different probabilistic models and statistical tests to measure the significance of shared targets between regulators. Indeed, the methods successfully eliminated those insignificant coregulation interactions occurred merely by chance; however, the biological meanings were ignored in the integration of transcriptional and post-transcriptional regulation interactions.\nHere we proposed a novel framework utilizing functional annotation information to identify significant coregulation between transcriptional and post-transcriptional layers. Based on this model, function-enriched coregulation pairs were discovered, and the regulators were subsequently linked by enriched functions. With these functional linkages, we further constructed functional coregulation networks between regulators and investigated their characteristics. Next, we searched for the network motifs consisting of those function-enriched coregulation pairs, and found that an abundance of pairs were closely connected in their upstream. Finally, the expression patterns of function-enriched coregulation pairs were analyzed. Different coregulation types showed distinct expression correlation trends. More importantly, we found that the disruption of coregulation may be closely related to cancers.", "[SUBTITLE] Regulation relationships [SUBSECTION] The transcriptional regulation relationships between human transcription factors and their target genes were collected from TRED (Transcriptional Regulatory Element Database) [11]. The database provides genome-wide promoter annotation and transcription factor binding information from computational prediction and experimental evidence.\nTo collect all human TF-target regulation relationships in TRED, we firstly queried the list of all human TFs in the database. A total of 178 human TFs were obtained by this step. Next, we searched TF target genes for each TF using default parameters (promoter quality from \"known, curated\" to \"with RNA\" and \"all\" binding quality). The results showed that only 133 TFs were found to have at least one target gene by these criteria, and the final number of unique TF-target relationships was 6,764, which were used to construct the human transcriptional regulatory network for our analysis.\nSince the available experimentally verified human miRNA targets are far from complete and thus not enough for global analysis, we used predicted miRNA targets from the TargetScan database (release 4.2) to perform the analysis [12]. In addition, different mature miRNAs may have identical seed regions and thereby target the same binding sites. To eliminate those coregulation interactions among the miRNAs with identical seed regions, we grouped mature miRNAs into families based on the miRNA family information from TargetScan. A total of 162 miRNA families and 7,521 target genes with 44,782 interactions were collected.\nIt is still difficult to predict the promoter region of miRNA genes in the genome. But it has been known that embedded miRNAs frequently coexpress with their host genes [13,14]. Therefore, we extracted miRNA host gene information from miRBase [15] and integrated the embedded miRNAs biogenesis information into the established transcriptional regulation network. A total of 310 premature miRNAs were found embedded in 259 host genes. Most of them (93%) were resided in introns.\nThe transcriptional regulation relationships between human transcription factors and their target genes were collected from TRED (Transcriptional Regulatory Element Database) [11]. The database provides genome-wide promoter annotation and transcription factor binding information from computational prediction and experimental evidence.\nTo collect all human TF-target regulation relationships in TRED, we firstly queried the list of all human TFs in the database. A total of 178 human TFs were obtained by this step. Next, we searched TF target genes for each TF using default parameters (promoter quality from \"known, curated\" to \"with RNA\" and \"all\" binding quality). The results showed that only 133 TFs were found to have at least one target gene by these criteria, and the final number of unique TF-target relationships was 6,764, which were used to construct the human transcriptional regulatory network for our analysis.\nSince the available experimentally verified human miRNA targets are far from complete and thus not enough for global analysis, we used predicted miRNA targets from the TargetScan database (release 4.2) to perform the analysis [12]. In addition, different mature miRNAs may have identical seed regions and thereby target the same binding sites. To eliminate those coregulation interactions among the miRNAs with identical seed regions, we grouped mature miRNAs into families based on the miRNA family information from TargetScan. A total of 162 miRNA families and 7,521 target genes with 44,782 interactions were collected.\nIt is still difficult to predict the promoter region of miRNA genes in the genome. But it has been known that embedded miRNAs frequently coexpress with their host genes [13,14]. Therefore, we extracted miRNA host gene information from miRBase [15] and integrated the embedded miRNAs biogenesis information into the established transcriptional regulation network. A total of 310 premature miRNAs were found embedded in 259 host genes. Most of them (93%) were resided in introns.\n[SUBTITLE] Identification of significant coregulation relationships [SUBSECTION] Combing all the potential targets of miRNAs and TFs, we firstly constructed two adjacency matrixes describing the regulator-target interaction for TFs and miRNAs, respectively. Then the two matrixes were combined into three cross-adjacency matrixes representing the shared targets of TF-TF, miRNA-miRNA, and TF-miRNA coregulation pairs. An example of identification of TF-miRNA coregulation pairs is shown in Figure 1.\nExample workflow for identifying significant coregulation\nSecondly, for each group of shared targets, the distribution of Gene Ontology (GO) annotations [16] at the second level in the biological process namespace was calculated. We chose the second level ontology because most of the genes were generally well-annotated at this level and these annotations provided a good balance between the sensitivity and the specificity in the following functional enrichment test. The distributions were considered as the functional profiles or fingerprints for these coregulation pairs.\nNext, we utilized a randomization method to perform a permutation test for functional enrichment. For each group of shared targets, we randomly selected a null group of the same size from whole human genome as background. After 10,000 iterations, the log-likelihood score under multivariate hypergeometric distribution was measured to quantify the significance of functional enrichment. The correction for multiple comparisons was made under 0.05 false discovery rate (FDR) [17]. The final results of significant coregulation pairs were listed in additional file 1.\nCombing all the potential targets of miRNAs and TFs, we firstly constructed two adjacency matrixes describing the regulator-target interaction for TFs and miRNAs, respectively. Then the two matrixes were combined into three cross-adjacency matrixes representing the shared targets of TF-TF, miRNA-miRNA, and TF-miRNA coregulation pairs. An example of identification of TF-miRNA coregulation pairs is shown in Figure 1.\nExample workflow for identifying significant coregulation\nSecondly, for each group of shared targets, the distribution of Gene Ontology (GO) annotations [16] at the second level in the biological process namespace was calculated. We chose the second level ontology because most of the genes were generally well-annotated at this level and these annotations provided a good balance between the sensitivity and the specificity in the following functional enrichment test. The distributions were considered as the functional profiles or fingerprints for these coregulation pairs.\nNext, we utilized a randomization method to perform a permutation test for functional enrichment. For each group of shared targets, we randomly selected a null group of the same size from whole human genome as background. After 10,000 iterations, the log-likelihood score under multivariate hypergeometric distribution was measured to quantify the significance of functional enrichment. The correction for multiple comparisons was made under 0.05 false discovery rate (FDR) [17]. The final results of significant coregulation pairs were listed in additional file 1.\n[SUBTITLE] Functional linkages and networks [SUBSECTION] For each function-enriched coregulation pair, Fisher’s Exact Test following FDR correction were conducted to identify enriched GO terms. Similarly, we only focused on the second level terms in biological process namespace. A functional linkage was established if a GO term overrepresented in the shared targets of a coregulation pair, implying that the two paired regulators may function coordinately in the specific biological process. Based on these linkages, we further constructed the functional coregulation networks (Figure 2). In order to investigate the specificity of coregulation relationships and provide a global view, only those linkages with relatively high significance that passed FDR-BL correction [17] were used to construct the networks.\nExample workflow for constructing functional coregulation network. For each coregulation pair, a function linkage was established if a GO term enriched in their shared targets. The coregulation network was generated based on these function linkages. Nodes represent regulators, and edge represents GO terms, marked with different colours.\nFor each function-enriched coregulation pair, Fisher’s Exact Test following FDR correction were conducted to identify enriched GO terms. Similarly, we only focused on the second level terms in biological process namespace. A functional linkage was established if a GO term overrepresented in the shared targets of a coregulation pair, implying that the two paired regulators may function coordinately in the specific biological process. Based on these linkages, we further constructed the functional coregulation networks (Figure 2). In order to investigate the specificity of coregulation relationships and provide a global view, only those linkages with relatively high significance that passed FDR-BL correction [17] were used to construct the networks.\nExample workflow for constructing functional coregulation network. For each coregulation pair, a function linkage was established if a GO term enriched in their shared targets. The coregulation network was generated based on these function linkages. Nodes represent regulators, and edge represents GO terms, marked with different colours.\n[SUBTITLE] Enriched network motifs [SUBSECTION] We searched for network motifs preferentially occurred in function-enriched coregulation pairs rather than in random pairs by a resampling process. The predicted TF-targeting interactions for miRNA genes were collected from miRBase [15] and from literature [9]. In addition, we assumed that those embedded miRNA genes have same transcription units as their host genes and would be regulated together.\nA total of 10,000 background sets of regulator pairs that have the same size as the set of function-enriched pairs were randomly selected from the global network. For each type of network patterns (sub-graphs), the observed frequency from the function-enriched coregulation pairs was first calculated and compared to the background distribution for assessment of significance. Only those network patterns with occurrence probabilities less than 0.001 were considered significant motifs (see additional file 2 for these significant motifs).\nWe searched for network motifs preferentially occurred in function-enriched coregulation pairs rather than in random pairs by a resampling process. The predicted TF-targeting interactions for miRNA genes were collected from miRBase [15] and from literature [9]. In addition, we assumed that those embedded miRNA genes have same transcription units as their host genes and would be regulated together.\nA total of 10,000 background sets of regulator pairs that have the same size as the set of function-enriched pairs were randomly selected from the global network. For each type of network patterns (sub-graphs), the observed frequency from the function-enriched coregulation pairs was first calculated and compared to the background distribution for assessment of significance. Only those network patterns with occurrence probabilities less than 0.001 were considered significant motifs (see additional file 2 for these significant motifs).\n[SUBTITLE] Analysis of expression profiles [SUBSECTION] The miRNA and mRNA expression profiles were adopted from a previous study [18]. A total of 217 miRNAs and ~16,000 mRNAs across 8 human tissues (colon, pancreas, kidney, bladder, prostate, uterus, lung, and breast) were measured using miRNA bead-arrays and mRNA microarrays. Both normal and tumor samples were profiled for each tissue. For each type of coregulation, we first generated background distribution by calculating the Pearson's correlation coefficients (PCCs) of expression profiles between the two paired regulators in all possible pairs (i.e., those pairs shared no targets and/or those pairs not identified as function-enriched). After that, the distribution of enriched coregulation pairs was calculated and shown against the background.\nThe miRNA and mRNA expression profiles were adopted from a previous study [18]. A total of 217 miRNAs and ~16,000 mRNAs across 8 human tissues (colon, pancreas, kidney, bladder, prostate, uterus, lung, and breast) were measured using miRNA bead-arrays and mRNA microarrays. Both normal and tumor samples were profiled for each tissue. For each type of coregulation, we first generated background distribution by calculating the Pearson's correlation coefficients (PCCs) of expression profiles between the two paired regulators in all possible pairs (i.e., those pairs shared no targets and/or those pairs not identified as function-enriched). After that, the distribution of enriched coregulation pairs was calculated and shown against the background.", "The transcriptional regulation relationships between human transcription factors and their target genes were collected from TRED (Transcriptional Regulatory Element Database) [11]. The database provides genome-wide promoter annotation and transcription factor binding information from computational prediction and experimental evidence.\nTo collect all human TF-target regulation relationships in TRED, we firstly queried the list of all human TFs in the database. A total of 178 human TFs were obtained by this step. Next, we searched TF target genes for each TF using default parameters (promoter quality from \"known, curated\" to \"with RNA\" and \"all\" binding quality). The results showed that only 133 TFs were found to have at least one target gene by these criteria, and the final number of unique TF-target relationships was 6,764, which were used to construct the human transcriptional regulatory network for our analysis.\nSince the available experimentally verified human miRNA targets are far from complete and thus not enough for global analysis, we used predicted miRNA targets from the TargetScan database (release 4.2) to perform the analysis [12]. In addition, different mature miRNAs may have identical seed regions and thereby target the same binding sites. To eliminate those coregulation interactions among the miRNAs with identical seed regions, we grouped mature miRNAs into families based on the miRNA family information from TargetScan. A total of 162 miRNA families and 7,521 target genes with 44,782 interactions were collected.\nIt is still difficult to predict the promoter region of miRNA genes in the genome. But it has been known that embedded miRNAs frequently coexpress with their host genes [13,14]. Therefore, we extracted miRNA host gene information from miRBase [15] and integrated the embedded miRNAs biogenesis information into the established transcriptional regulation network. A total of 310 premature miRNAs were found embedded in 259 host genes. Most of them (93%) were resided in introns.", "Combing all the potential targets of miRNAs and TFs, we firstly constructed two adjacency matrixes describing the regulator-target interaction for TFs and miRNAs, respectively. Then the two matrixes were combined into three cross-adjacency matrixes representing the shared targets of TF-TF, miRNA-miRNA, and TF-miRNA coregulation pairs. An example of identification of TF-miRNA coregulation pairs is shown in Figure 1.\nExample workflow for identifying significant coregulation\nSecondly, for each group of shared targets, the distribution of Gene Ontology (GO) annotations [16] at the second level in the biological process namespace was calculated. We chose the second level ontology because most of the genes were generally well-annotated at this level and these annotations provided a good balance between the sensitivity and the specificity in the following functional enrichment test. The distributions were considered as the functional profiles or fingerprints for these coregulation pairs.\nNext, we utilized a randomization method to perform a permutation test for functional enrichment. For each group of shared targets, we randomly selected a null group of the same size from whole human genome as background. After 10,000 iterations, the log-likelihood score under multivariate hypergeometric distribution was measured to quantify the significance of functional enrichment. The correction for multiple comparisons was made under 0.05 false discovery rate (FDR) [17]. The final results of significant coregulation pairs were listed in additional file 1.", "For each function-enriched coregulation pair, Fisher’s Exact Test following FDR correction were conducted to identify enriched GO terms. Similarly, we only focused on the second level terms in biological process namespace. A functional linkage was established if a GO term overrepresented in the shared targets of a coregulation pair, implying that the two paired regulators may function coordinately in the specific biological process. Based on these linkages, we further constructed the functional coregulation networks (Figure 2). In order to investigate the specificity of coregulation relationships and provide a global view, only those linkages with relatively high significance that passed FDR-BL correction [17] were used to construct the networks.\nExample workflow for constructing functional coregulation network. For each coregulation pair, a function linkage was established if a GO term enriched in their shared targets. The coregulation network was generated based on these function linkages. Nodes represent regulators, and edge represents GO terms, marked with different colours.", "We searched for network motifs preferentially occurred in function-enriched coregulation pairs rather than in random pairs by a resampling process. The predicted TF-targeting interactions for miRNA genes were collected from miRBase [15] and from literature [9]. In addition, we assumed that those embedded miRNA genes have same transcription units as their host genes and would be regulated together.\nA total of 10,000 background sets of regulator pairs that have the same size as the set of function-enriched pairs were randomly selected from the global network. For each type of network patterns (sub-graphs), the observed frequency from the function-enriched coregulation pairs was first calculated and compared to the background distribution for assessment of significance. Only those network patterns with occurrence probabilities less than 0.001 were considered significant motifs (see additional file 2 for these significant motifs).", "The miRNA and mRNA expression profiles were adopted from a previous study [18]. A total of 217 miRNAs and ~16,000 mRNAs across 8 human tissues (colon, pancreas, kidney, bladder, prostate, uterus, lung, and breast) were measured using miRNA bead-arrays and mRNA microarrays. Both normal and tumor samples were profiled for each tissue. For each type of coregulation, we first generated background distribution by calculating the Pearson's correlation coefficients (PCCs) of expression profiles between the two paired regulators in all possible pairs (i.e., those pairs shared no targets and/or those pairs not identified as function-enriched). After that, the distribution of enriched coregulation pairs was calculated and shown against the background.", "[SUBTITLE] Functional coregulation pairs [SUBSECTION] After the integration of miRNA regulation into human transcriptional regulation network, we adopted a novel strategy utilizing functional information to identify function-enriched coregulation pairs, and establish function linkages for each pair. Traditional analysis of functional enrichment was aimed at elucidating the regulatory roles of each individual regulator only, inevitably leaving some significant coregulation hidden in the traditional views. Instead, based on our model, different regulation types involving single regulators or combinations of regulators can all be studied and compared.\nThe distributions of different regulation types were grouped into two diagrams for comparing. Figure 3A shows distributions of individual TF regulation and TF-TF coregulation. The two distributions look similar; however, two biological processes, pigmentation and reproductive process, emerge when it comes to TF-miRNA coregulation, implying that the two biological processes may be the typical processes in which TFs should coordinate with miRNAs to control the expression programs.\nDistributions of enriched biological processes for different regulation types. (A) Distributions for TF-involving regulation: individual TFs, TF-TF pairs, and TF-miRNA pairs. (B) Distributions for miRNA-involving regulation: individual miRNAs, miRNA-miRNA pairs, and TF-miRNA pairs. Note that the same TF-miRNA line are drawn in both (A) and (B) for comparison.\nIn contrast, miRNA-involving regulation shows divergent distributions in Figure 3B. The top ranked biological processes of individual miRNA regulation were biological regulation, cellular process, and developmental process, which were the previously known miRNA-involving processes [4-6]. On the other hand, biological adhesion was relatively high in miRNA-miRNA coregulation, suggesting that miRNAs may regulate this process majorly in a coordination manner.\nMoreover, many biological processes enriched in TF-miRNA coregulation were relatively poor in the regulation involving miRNAs only. In other words, those processes may be the typical processes needed to be coordinately regulated by TFs and miRNAs, and the coordination may provide a mechanism to switch expression programs. More importantly, it suggested that, by coordinating with TFs, miRNAs may engage in a wider diversity of biological processes, and these undiscovered processes were failed to be identified by traditional analysis of functional enrichment for a single regulator.\nAfter the integration of miRNA regulation into human transcriptional regulation network, we adopted a novel strategy utilizing functional information to identify function-enriched coregulation pairs, and establish function linkages for each pair. Traditional analysis of functional enrichment was aimed at elucidating the regulatory roles of each individual regulator only, inevitably leaving some significant coregulation hidden in the traditional views. Instead, based on our model, different regulation types involving single regulators or combinations of regulators can all be studied and compared.\nThe distributions of different regulation types were grouped into two diagrams for comparing. Figure 3A shows distributions of individual TF regulation and TF-TF coregulation. The two distributions look similar; however, two biological processes, pigmentation and reproductive process, emerge when it comes to TF-miRNA coregulation, implying that the two biological processes may be the typical processes in which TFs should coordinate with miRNAs to control the expression programs.\nDistributions of enriched biological processes for different regulation types. (A) Distributions for TF-involving regulation: individual TFs, TF-TF pairs, and TF-miRNA pairs. (B) Distributions for miRNA-involving regulation: individual miRNAs, miRNA-miRNA pairs, and TF-miRNA pairs. Note that the same TF-miRNA line are drawn in both (A) and (B) for comparison.\nIn contrast, miRNA-involving regulation shows divergent distributions in Figure 3B. The top ranked biological processes of individual miRNA regulation were biological regulation, cellular process, and developmental process, which were the previously known miRNA-involving processes [4-6]. On the other hand, biological adhesion was relatively high in miRNA-miRNA coregulation, suggesting that miRNAs may regulate this process majorly in a coordination manner.\nMoreover, many biological processes enriched in TF-miRNA coregulation were relatively poor in the regulation involving miRNAs only. In other words, those processes may be the typical processes needed to be coordinately regulated by TFs and miRNAs, and the coordination may provide a mechanism to switch expression programs. More importantly, it suggested that, by coordinating with TFs, miRNAs may engage in a wider diversity of biological processes, and these undiscovered processes were failed to be identified by traditional analysis of functional enrichment for a single regulator.\n[SUBTITLE] Functional coregulation networks [SUBSECTION] In the previous section, different regulators were connected by identified functional linkages, which represented that the two paired regulators may function in coordination with each other in a specific biological process. We further built up functional coregulation networks from these linkages and found interesting properties in the networks.\nFigure 4 shows the three subnetwork examples (the largest connected component) for different coregulation types. Obviously, it appears that the TF-TF coregulation network (Figure 4A) was in high density coregulation (avg. 3.52 links per pair) and full of diversity in edge types (avg. 5.70 edge types per regulator); in other words, TFs may function together in many kinds of biological processes. Likewise, the miRNA-miRNA coregulation network (Figure 4B) showed similar high density coregulation (avg. 3.12 links per pair) but the diversity of involving processes was lower (avg. 3.92 edge types per regulator) than TF-TF coregulation. In contrast, the TF-miRNA coregulation network (Figure 4C) was in higher specificity (avg. 1.48 links per pair; avg. 2.22 edge types per regulator); that is, a TF may coordinate with different neighbour miRNAs in specific biological processes, and vice versa. These results suggested that the cross-layer coregulation may have higher specificity than intra-layer coregulation.\nFunctional coregulation networks. (A) TF-TF coregulation network. (B) miR-miR coregulation network. (C) TF-miR coregulation network. Red nodes represent TFs; green nodes represent miRNAs; and edges represent biological processes. Different edge colours represent different biological processes.\nIn the previous section, different regulators were connected by identified functional linkages, which represented that the two paired regulators may function in coordination with each other in a specific biological process. We further built up functional coregulation networks from these linkages and found interesting properties in the networks.\nFigure 4 shows the three subnetwork examples (the largest connected component) for different coregulation types. Obviously, it appears that the TF-TF coregulation network (Figure 4A) was in high density coregulation (avg. 3.52 links per pair) and full of diversity in edge types (avg. 5.70 edge types per regulator); in other words, TFs may function together in many kinds of biological processes. Likewise, the miRNA-miRNA coregulation network (Figure 4B) showed similar high density coregulation (avg. 3.12 links per pair) but the diversity of involving processes was lower (avg. 3.92 edge types per regulator) than TF-TF coregulation. In contrast, the TF-miRNA coregulation network (Figure 4C) was in higher specificity (avg. 1.48 links per pair; avg. 2.22 edge types per regulator); that is, a TF may coordinate with different neighbour miRNAs in specific biological processes, and vice versa. These results suggested that the cross-layer coregulation may have higher specificity than intra-layer coregulation.\nFunctional coregulation networks. (A) TF-TF coregulation network. (B) miR-miR coregulation network. (C) TF-miR coregulation network. Red nodes represent TFs; green nodes represent miRNAs; and edges represent biological processes. Different edge colours represent different biological processes.\n[SUBTITLE] Network motifs for coregulation pairs [SUBSECTION] Many studies have been devoted to understanding network structures in gene regulatory networks, and have found that most networks seem to be largely composed of occurring patterns, called network motifs. The functions associated with common network motifs, such as auto-regulation and feed-forward loops (FFLs), were discovered and revealed by several researches both theoretically and experimentally [1,9,10,19-22].\nUnsurprisingly, the function-enriched coregulation pairs also have preferentially recurring network motifs as shown in Figure 5. Several types of motifs were found in TF-TF coregulation; for example, bidirectional and unidirectional FFLs were explored and these results were consistent with previous studies on network biology. In addition to FFLs, we went further to investigate the upstream regulatory patterns of coregulation pairs and found that the two paired regulators were closely linked in their upstream. For instance, over half of the pairs had common upstream TFs; a significant number of pairs had common TFs and miRNAs; and almost all pairs were cross-regulated in their upstream. Similar results were also found in TF-miRNA coregulation. However, no enriched motif was found in miRNA-miRNA coregulation, probably due to the incomplete knowledge of regulatory regions of miRNA genes.\nNetwork motifs for TF-TF and TF-miRNA coregulation pairs. For each motif, the observation number and the percentage of pairs that have this motif were reported, along with the P-value and one instance coregulation pair.\nMany studies have been devoted to understanding network structures in gene regulatory networks, and have found that most networks seem to be largely composed of occurring patterns, called network motifs. The functions associated with common network motifs, such as auto-regulation and feed-forward loops (FFLs), were discovered and revealed by several researches both theoretically and experimentally [1,9,10,19-22].\nUnsurprisingly, the function-enriched coregulation pairs also have preferentially recurring network motifs as shown in Figure 5. Several types of motifs were found in TF-TF coregulation; for example, bidirectional and unidirectional FFLs were explored and these results were consistent with previous studies on network biology. In addition to FFLs, we went further to investigate the upstream regulatory patterns of coregulation pairs and found that the two paired regulators were closely linked in their upstream. For instance, over half of the pairs had common upstream TFs; a significant number of pairs had common TFs and miRNAs; and almost all pairs were cross-regulated in their upstream. Similar results were also found in TF-miRNA coregulation. However, no enriched motif was found in miRNA-miRNA coregulation, probably due to the incomplete knowledge of regulatory regions of miRNA genes.\nNetwork motifs for TF-TF and TF-miRNA coregulation pairs. For each motif, the observation number and the percentage of pairs that have this motif were reported, along with the P-value and one instance coregulation pair.\n[SUBTITLE] Expression patterns of coregulation pairs [SUBSECTION] Expression data across human normal/tumor tissues have recently become available. A previous study measured miRNA and mRNA expression profiles across 8 tissues (colon, pancreas, kidney, bladder, prostate, uterus, lung, and breast) and each tissue contained both normal and tumor samples [18]. By analyzing the expression profiles, we investigated the correlations between the expression profiles of each coregulation pair in both normal and cancer samples.\nFigure 6 shows the expression correlations of different coregulation types in normal and tumor samples. Interestingly, the three coregulation types show distinct trends in normal tissues. For example, TF-TF had a zero-centered distribution similar to background; TF-miRNA had two tendencies to highly-positive and medium-negative correlations in comparison with its background; miRNA-miRNA showed preference for only positive correlation. The different trends may reflect the diversity of function roles between TFs and miRNAs; that is, TFs can act as activators (+) or repressors (-) in gene regulation; however, miRNAs mainly act as repressors (-) by translation inhibition or transcript destabilization. Thus, it seems that these typical trends may be mechanistically reasonable.\nDistributions of expression correlation for different types of coregulation pairs. Left column for normal tissues; right column for tumor tissues. Rows are in the order of TF-TF, TF-miR, and miR-miR coregulation pairs.\nOn the contrary, all coregulation types turn into an identical trend in tumor tissues. All of them show similar zero-centered distributions resembled to their backgrounds. This trend suggests that the function-enriched coregulation pairs lost their correlation in tumor tissues, implying the disruption of coregulation may be closely associated to cancers. Together these results may support the functionality of identified coregulation pairs.\nExpression data across human normal/tumor tissues have recently become available. A previous study measured miRNA and mRNA expression profiles across 8 tissues (colon, pancreas, kidney, bladder, prostate, uterus, lung, and breast) and each tissue contained both normal and tumor samples [18]. By analyzing the expression profiles, we investigated the correlations between the expression profiles of each coregulation pair in both normal and cancer samples.\nFigure 6 shows the expression correlations of different coregulation types in normal and tumor samples. Interestingly, the three coregulation types show distinct trends in normal tissues. For example, TF-TF had a zero-centered distribution similar to background; TF-miRNA had two tendencies to highly-positive and medium-negative correlations in comparison with its background; miRNA-miRNA showed preference for only positive correlation. The different trends may reflect the diversity of function roles between TFs and miRNAs; that is, TFs can act as activators (+) or repressors (-) in gene regulation; however, miRNAs mainly act as repressors (-) by translation inhibition or transcript destabilization. Thus, it seems that these typical trends may be mechanistically reasonable.\nDistributions of expression correlation for different types of coregulation pairs. Left column for normal tissues; right column for tumor tissues. Rows are in the order of TF-TF, TF-miR, and miR-miR coregulation pairs.\nOn the contrary, all coregulation types turn into an identical trend in tumor tissues. All of them show similar zero-centered distributions resembled to their backgrounds. This trend suggests that the function-enriched coregulation pairs lost their correlation in tumor tissues, implying the disruption of coregulation may be closely associated to cancers. Together these results may support the functionality of identified coregulation pairs.", "After the integration of miRNA regulation into human transcriptional regulation network, we adopted a novel strategy utilizing functional information to identify function-enriched coregulation pairs, and establish function linkages for each pair. Traditional analysis of functional enrichment was aimed at elucidating the regulatory roles of each individual regulator only, inevitably leaving some significant coregulation hidden in the traditional views. Instead, based on our model, different regulation types involving single regulators or combinations of regulators can all be studied and compared.\nThe distributions of different regulation types were grouped into two diagrams for comparing. Figure 3A shows distributions of individual TF regulation and TF-TF coregulation. The two distributions look similar; however, two biological processes, pigmentation and reproductive process, emerge when it comes to TF-miRNA coregulation, implying that the two biological processes may be the typical processes in which TFs should coordinate with miRNAs to control the expression programs.\nDistributions of enriched biological processes for different regulation types. (A) Distributions for TF-involving regulation: individual TFs, TF-TF pairs, and TF-miRNA pairs. (B) Distributions for miRNA-involving regulation: individual miRNAs, miRNA-miRNA pairs, and TF-miRNA pairs. Note that the same TF-miRNA line are drawn in both (A) and (B) for comparison.\nIn contrast, miRNA-involving regulation shows divergent distributions in Figure 3B. The top ranked biological processes of individual miRNA regulation were biological regulation, cellular process, and developmental process, which were the previously known miRNA-involving processes [4-6]. On the other hand, biological adhesion was relatively high in miRNA-miRNA coregulation, suggesting that miRNAs may regulate this process majorly in a coordination manner.\nMoreover, many biological processes enriched in TF-miRNA coregulation were relatively poor in the regulation involving miRNAs only. In other words, those processes may be the typical processes needed to be coordinately regulated by TFs and miRNAs, and the coordination may provide a mechanism to switch expression programs. More importantly, it suggested that, by coordinating with TFs, miRNAs may engage in a wider diversity of biological processes, and these undiscovered processes were failed to be identified by traditional analysis of functional enrichment for a single regulator.", "In the previous section, different regulators were connected by identified functional linkages, which represented that the two paired regulators may function in coordination with each other in a specific biological process. We further built up functional coregulation networks from these linkages and found interesting properties in the networks.\nFigure 4 shows the three subnetwork examples (the largest connected component) for different coregulation types. Obviously, it appears that the TF-TF coregulation network (Figure 4A) was in high density coregulation (avg. 3.52 links per pair) and full of diversity in edge types (avg. 5.70 edge types per regulator); in other words, TFs may function together in many kinds of biological processes. Likewise, the miRNA-miRNA coregulation network (Figure 4B) showed similar high density coregulation (avg. 3.12 links per pair) but the diversity of involving processes was lower (avg. 3.92 edge types per regulator) than TF-TF coregulation. In contrast, the TF-miRNA coregulation network (Figure 4C) was in higher specificity (avg. 1.48 links per pair; avg. 2.22 edge types per regulator); that is, a TF may coordinate with different neighbour miRNAs in specific biological processes, and vice versa. These results suggested that the cross-layer coregulation may have higher specificity than intra-layer coregulation.\nFunctional coregulation networks. (A) TF-TF coregulation network. (B) miR-miR coregulation network. (C) TF-miR coregulation network. Red nodes represent TFs; green nodes represent miRNAs; and edges represent biological processes. Different edge colours represent different biological processes.", "Many studies have been devoted to understanding network structures in gene regulatory networks, and have found that most networks seem to be largely composed of occurring patterns, called network motifs. The functions associated with common network motifs, such as auto-regulation and feed-forward loops (FFLs), were discovered and revealed by several researches both theoretically and experimentally [1,9,10,19-22].\nUnsurprisingly, the function-enriched coregulation pairs also have preferentially recurring network motifs as shown in Figure 5. Several types of motifs were found in TF-TF coregulation; for example, bidirectional and unidirectional FFLs were explored and these results were consistent with previous studies on network biology. In addition to FFLs, we went further to investigate the upstream regulatory patterns of coregulation pairs and found that the two paired regulators were closely linked in their upstream. For instance, over half of the pairs had common upstream TFs; a significant number of pairs had common TFs and miRNAs; and almost all pairs were cross-regulated in their upstream. Similar results were also found in TF-miRNA coregulation. However, no enriched motif was found in miRNA-miRNA coregulation, probably due to the incomplete knowledge of regulatory regions of miRNA genes.\nNetwork motifs for TF-TF and TF-miRNA coregulation pairs. For each motif, the observation number and the percentage of pairs that have this motif were reported, along with the P-value and one instance coregulation pair.", "Expression data across human normal/tumor tissues have recently become available. A previous study measured miRNA and mRNA expression profiles across 8 tissues (colon, pancreas, kidney, bladder, prostate, uterus, lung, and breast) and each tissue contained both normal and tumor samples [18]. By analyzing the expression profiles, we investigated the correlations between the expression profiles of each coregulation pair in both normal and cancer samples.\nFigure 6 shows the expression correlations of different coregulation types in normal and tumor samples. Interestingly, the three coregulation types show distinct trends in normal tissues. For example, TF-TF had a zero-centered distribution similar to background; TF-miRNA had two tendencies to highly-positive and medium-negative correlations in comparison with its background; miRNA-miRNA showed preference for only positive correlation. The different trends may reflect the diversity of function roles between TFs and miRNAs; that is, TFs can act as activators (+) or repressors (-) in gene regulation; however, miRNAs mainly act as repressors (-) by translation inhibition or transcript destabilization. Thus, it seems that these typical trends may be mechanistically reasonable.\nDistributions of expression correlation for different types of coregulation pairs. Left column for normal tissues; right column for tumor tissues. Rows are in the order of TF-TF, TF-miR, and miR-miR coregulation pairs.\nOn the contrary, all coregulation types turn into an identical trend in tumor tissues. All of them show similar zero-centered distributions resembled to their backgrounds. This trend suggests that the function-enriched coregulation pairs lost their correlation in tumor tissues, implying the disruption of coregulation may be closely associated to cancers. Together these results may support the functionality of identified coregulation pairs.", "We proposed a novel strategy aimed at identifying potential coordinated regulation by utilizing functional annotation information and discovered many biological processes that emerged only in coregulation. Compared to traditional function enrichment analysis, our strategy considered whole function profiles rather than single annotations. In addition, it also solves the restriction of traditional methods that only focus on single regulator. For example, a miRNA can potentially regulate an abundance of target genes. To find enriched functions of the miRNA, all its potential targets will be tested for any enriched function. However, since the target size of a miRNA may be huge, some meaningful biological processes involving only a small subset of genes will be hidden. In fact, these hidden processes may be significantly impacted by miRNAs in coordination with other regulators, namely, other miRNAs or TFs. After all, a biological process may be regulated not only at the transcriptional layer, but also at the posttranscriptional layer [7,8,23,24].\nInterestingly, our results show that pigmentation and reproductive process are two typical biological processes specifically emerging in TF-miRNA coregulation. It is suggested that miRNAs may provide genetic switch mechanisms to essentially inactivate the target genes, thus leading to detectable phenotypic consequences. In model organisms, there have been many studies investigating the switch-like role of miRNAs in pigmentation. For example, miRNAs can regulate the eye pigmentation genes in Drosophila [25]. The influence of miRNAs on pigmentation in zebrafish was also reported [26]. Another study found that miR-434-5p may mediate skin whitening and lightening in mouse [27]. And in melanoma cell lines, it is shown that miR-137 may target a pigmentation regulator [28].\nThe analysis of functional coregulation networks provided other clues. We found that a TF may regulate in coordination with different miRNAs in different biological processes, and vice versa. It suggested that the cross-layer coregulation may have higher specificity than intra-layer coregulation.\nWe also performed network motif analysis to see if any recurring pattern exists in coregulation network structure. Different types of feed-forward loops were found in TF-TF and TF-miRNA coregulation, and these results were consistent with several previous studies on transcriptional network [1,9,10,19-22]. Among these FFLs, a special kind of miRNA-mediated FFLs emerged in TF-miRNA coregulation. In this kind of FFLs, a miRNA may simultaneously repress a TF and its target genes, thus contributing to a switch-like control of expression programs. More importantly, we go further this time to investigate the upstream structure of coregulation pairs and found closely interaction in their upstream. It implies that the network structures of coregulation may have extensive crosstalk in the higher levels.\nFinally, the expression analysis of coregulation discovered distinct trends in different coregulation types; namely, TF-TF showed no correlation, whereas miRNA-miRNA had a preference of positive correlation, and TF-miRNA appeared both positive and negative correlation. A previous study investigated only TF-miRNA correlation and found the same tendencies [9]. The authors rationalized this trend by pointing out the distinct function roles that TFs and miRNAs may play. We further supported this idea by showing the results of TF-TF and miRNA-miRNA coregulation, which were also consistent with the same interpretation. In addition, TF activities are under control at protein level; that is, TFs may be activated or deactivated by a number of mechanisms including phosphorylation, ligand binding, and interaction with other regulatory proteins. Therefore, it is not surprising that co-function TFs may show no correlation in mRNA expression level. Notably, a large proportion of TF-miRNA pairs showed negative correlation in expression profiles, which could be explained by the structure of the miRNA-mediated-FFLs discussed before, supporting the idea that many miRNAs in TF-miRNA coregulation contributed to switch-like regulation.\nMore significantly, by comparing the expression correlations between normal and tumor tissues, we found a common trend in function-enriched coregulation pairs; that is, the function-enriched pairs lost their correlation in tumor tissues. It suggested that the disruption of coregulation may lead to abnormal expression programs and may be directly associated to cancers.\nOur findings shed light on the coregulation of miRNAs in transcriptional regulatory network. Future experimental works will provide more complete knowledge in transcriptional network and miRNA regulation, thus allowing the elucidation of more precise co-regulatory mechanisms.", "The authors declare that they have no competing interests.", "CYC implemented the computational method, carried out the analysis, and drafted the manuscript. CYC and HCH conceived of the study. STC and CSF helped to perform the analysis and participated in discussions of the research. HFJ and HCH participated in the design and coordination of the study, and edited the manuscript. All authors read and approved the final manuscript.", "\nList of all function-enriched coregulation pairs\nhttp://idv.sinica.edu.tw/joeychen/APBC2011/AdditionalFile1.pdf\n\nClick here for file\n\nList of network motifs for function-enriched coregulation pairs\nhttp://idv.sinica.edu.tw/joeychen/APBC2011/AdditionalFile2.pdf\n\nClick here for file" ]

[ null, "methods", null, null, null, null, null, "results", null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Animals", "Binding Sites", "CpG Islands", "DNA Methylation", "Embryonic Stem Cells", "Epigenomics", "Gene Expression Regulation", "Genome", "Linear Models", "Mice", "Multivariate Analysis", "Protein Binding", "Regulatory Sequences, Nucleic Acid", "Transcription Factors" ]

[ "Background", "Results", "ChIP-seq data is reproduced and extended", "Remapped peaks improve the prediction of gene expression", "Genome-wide locations of TF bindings do not follow exponential distribution", "Density-based regression model outperforms the exponential-based model", "Two gene classes are different in epigenetic patterns", "Epigenetic patterns improve the prediction of gene expression", "TF interactions wired with epigenetic effects", "Discussion", "Data acquisition", "ChIP-seq data and gene expression", "Epigenetic modifications", "Estimation of TF binding density", "Regression model", "Adding epigenetic effects", "Fitting and reducing regression models", "Competing interests" ]

[ "Embryonic stem cells (ESCs) derived from blastocysts are self-renewal and pluripotent [1-3]. To understand the gene regulatory system in ESCs is an important step for uncovering the process of cell fate determination and for promoting regenerative medicine. Considerable recent evidence indicates that several transcription factors (TFs), so-called core TFs, are indispensable to maintain the pluripotency [4,5]. Some of the core TFs reprogram somatic cells back to pluripotent states [6,7]. These observations suggest that the regulatory network of TFs apparently governs the self-renewal and pluripotency [8,9]. On the other hand, many studies have reported that other TFs can functionally substitute for the core TFs [10-13], suggesting that there still exist additional or alternative TFs unrevealed in the network. Epigenetic modifications are also essential for ESCs [14,15]. Their involvement in the maintenance of the pluripotency is still not well clarified.\nTo understand the regulatory mechanism underlying in ESCs, a number of methods have been developed. In particular, massive parallel sequencing [9,16-19] and various in silico approaches [8,9,20,21] have yielded comprehensive recent advances in our understanding. In this study, we focus on predicting the gene expression in ESCs with the massive parallel sequencing data. Although a previous study successfully applied a regression model to the prediction [21], the model is based on a generalized weighting scheme to prepare predictors (explanatory variables). Intuitively, such weighting scheme cannot reflect the nature of the spatial rearrangement of TF-binding.\nHere, we propose a density-based approach that uses the genome-wide bona fide TF binding sites. First, a publicly available ChIP-seq data [9] is reanalyzed. Then, density profiles of TFs estimated from the ChIP-seq data are adopted as predictors in a simple linear regression model to predict the genome-wide gene expression. Predictors are also combined with epigenetic data, such as H3K4me3, H3K27me3, DNA methylation, and CpG island [16,17]. Furthermore, we analyze the regulatory effects of TFs, epigenetic states, and their higher-order interactions by using a pipeline developed in house. We demonstrate the predictive power of the density-based regression model and discuss our findings.", "[SUBTITLE] ChIP-seq data is reproduced and extended [SUBSECTION] To minimize artifacts, we refined the binding signals of 12 core TFs in mouse ESC publicly available [9] (see methods). The ChIP-seq peak datasets generated by various tools are hereafter denoted as FP4_Bowtie, FP4_MAQ, and FP4_Soap2. Also, tag positions mapped by Eland [9] are used for the peak detection (FP4_Eland), and the peak data of Chen et al. is involved (Chen Eland). Thus, we prepared five peak datasets in total.\nDifferences in numbers and positions between the remapped data and the original data were investigated. As a result, relatively larger number of uniquely mapped tags and peaks were gained compared to the original data (Table S1-S3 in Additional file 1). In regard to peaks (Table 1), FP4 with the previously mapped tags (FP4_Eland) covers 85-98% of Chen_Eland, and the intensity of overlapped peaks is strongly correlated. Thus, it is deemed that FP4 has reproduced Chen_Eland and extended it with novel peaks in different genomic locations. In contrast, FP4 with remapped tags shows relatively lower reproducibility, whereas peak intensities are still correlated with Chen_Eland except Esrrb (Figure 1B). Similar observations can be found from an independent study [22].\nReproducibility of newly detected peaks\nFold change is the ratio of newly detected peak number over the original peak number. Overlaps of the original peaks to the newly detected peaks were investigated with 200-bp window. The overlapped peaks were used to calculate the correlation of peak intensity.\nSchematic representation of ChIP-seq data analysis and examples of TF-binding density profiles. (A) Definitions of a peak region and its intensity. (B) Example of intensity correlation between the original data and the reanalyzed data. (C) Density estimation from the genome-wide binding locations of a TF. (D)-(E) Example of density profiles in five peak datasets.\nThe reason why the numbers vary is twofold. First, algorithmic differences in alignment tools cause the different numbers, particularly due to the gapped or ungapped alignment and random indel for mismatches. Second, thresholds for the peak intensity to distinguish experimental noise are different (Table S4 in Additional file 1). That is, Chen et al. used qPCR refinement with small number of peaks, whereas we used Monte Carlo simulation on each chromosome.\nTo minimize artifacts, we refined the binding signals of 12 core TFs in mouse ESC publicly available [9] (see methods). The ChIP-seq peak datasets generated by various tools are hereafter denoted as FP4_Bowtie, FP4_MAQ, and FP4_Soap2. Also, tag positions mapped by Eland [9] are used for the peak detection (FP4_Eland), and the peak data of Chen et al. is involved (Chen Eland). Thus, we prepared five peak datasets in total.\nDifferences in numbers and positions between the remapped data and the original data were investigated. As a result, relatively larger number of uniquely mapped tags and peaks were gained compared to the original data (Table S1-S3 in Additional file 1). In regard to peaks (Table 1), FP4 with the previously mapped tags (FP4_Eland) covers 85-98% of Chen_Eland, and the intensity of overlapped peaks is strongly correlated. Thus, it is deemed that FP4 has reproduced Chen_Eland and extended it with novel peaks in different genomic locations. In contrast, FP4 with remapped tags shows relatively lower reproducibility, whereas peak intensities are still correlated with Chen_Eland except Esrrb (Figure 1B). Similar observations can be found from an independent study [22].\nReproducibility of newly detected peaks\nFold change is the ratio of newly detected peak number over the original peak number. Overlaps of the original peaks to the newly detected peaks were investigated with 200-bp window. The overlapped peaks were used to calculate the correlation of peak intensity.\nSchematic representation of ChIP-seq data analysis and examples of TF-binding density profiles. (A) Definitions of a peak region and its intensity. (B) Example of intensity correlation between the original data and the reanalyzed data. (C) Density estimation from the genome-wide binding locations of a TF. (D)-(E) Example of density profiles in five peak datasets.\nThe reason why the numbers vary is twofold. First, algorithmic differences in alignment tools cause the different numbers, particularly due to the gapped or ungapped alignment and random indel for mismatches. Second, thresholds for the peak intensity to distinguish experimental noise are different (Table S4 in Additional file 1). That is, Chen et al. used qPCR refinement with small number of peaks, whereas we used Monte Carlo simulation on each chromosome.\n[SUBTITLE] Remapped peaks improve the prediction of gene expression [SUBSECTION] To assess the importance of TF bindings, Ouyang et al. [21] successfully applied a regression model to the prediction of absolute gene expression in mouse ESC. We first recover this study. Ouyang et al. used TF association strength (TFAS) by summing up peak intensities that are weighted exponentially according to the relative positions from TSSs. We applied the TFAS data to our simple linear regression model shown in equation (1), namely exponential-based regression model. The predictive power of our model is much higher (CV-R2=0.647) than Ouyang’s model (CV-R2=0.639), suggesting that the simple regression model is comparable to their PC-regression model (Figure 2A).\nPredictive results of density-based linear regression model. (A)-(C) Average correlation coefficient of 10-fold CV in three gene sets. (D) Comparative analysis of two models using ESC-specific gene subsets that co-bound by pluripotent TF pairs. Red solid circles are the cases that the density profiles of a TF pair are significantly different. (E) Distinct binding profiles in two gene classes that are either well explained (C1) or inefficiently explained (C2). (F) Contribution of each TF and epigenetic effects in density-based regression models (Methy: DNA methylation, HistM: histone mark, CpGI: CpG Island).\nNext, we prepared 17060 genes by removing inconsistency between Ouyang’s study and Chen’s study. This procedure is prerequisite for gathering precise TF-binding instances. TFAS data for the genes were calculated by the exactly same procedure of Ouyang et al. As a result, the exponential-based model shows CV-R2=0.495 with Chen_Eland. In contrast, CV-R2 increases to 0.542 (FP4_Eland), 0.587 (FP4_Bowtie), 0.581 (FP4_MAQ), and 0.590 (FP4_Soap2).\nThese results clearly suggest that the proposed simple linear regression model is applicable to the prediction. Furthermore, it has been demonstrated that the peak datasets we remapped give more information for explaining the gene expression.\nTo assess the importance of TF bindings, Ouyang et al. [21] successfully applied a regression model to the prediction of absolute gene expression in mouse ESC. We first recover this study. Ouyang et al. used TF association strength (TFAS) by summing up peak intensities that are weighted exponentially according to the relative positions from TSSs. We applied the TFAS data to our simple linear regression model shown in equation (1), namely exponential-based regression model. The predictive power of our model is much higher (CV-R2=0.647) than Ouyang’s model (CV-R2=0.639), suggesting that the simple regression model is comparable to their PC-regression model (Figure 2A).\nPredictive results of density-based linear regression model. (A)-(C) Average correlation coefficient of 10-fold CV in three gene sets. (D) Comparative analysis of two models using ESC-specific gene subsets that co-bound by pluripotent TF pairs. Red solid circles are the cases that the density profiles of a TF pair are significantly different. (E) Distinct binding profiles in two gene classes that are either well explained (C1) or inefficiently explained (C2). (F) Contribution of each TF and epigenetic effects in density-based regression models (Methy: DNA methylation, HistM: histone mark, CpGI: CpG Island).\nNext, we prepared 17060 genes by removing inconsistency between Ouyang’s study and Chen’s study. This procedure is prerequisite for gathering precise TF-binding instances. TFAS data for the genes were calculated by the exactly same procedure of Ouyang et al. As a result, the exponential-based model shows CV-R2=0.495 with Chen_Eland. In contrast, CV-R2 increases to 0.542 (FP4_Eland), 0.587 (FP4_Bowtie), 0.581 (FP4_MAQ), and 0.590 (FP4_Soap2).\nThese results clearly suggest that the proposed simple linear regression model is applicable to the prediction. Furthermore, it has been demonstrated that the peak datasets we remapped give more information for explaining the gene expression.\n[SUBTITLE] Genome-wide locations of TF bindings do not follow exponential distribution [SUBSECTION] To investigate the characteristics of TF binding sites in ESC, the density profiles of TF-bindings are estimated from each of peak datasets (Figure 1C), then any two density profiles for a TF in different peak datasets are tested by Kolmogorov-Smirnov (KS) test. According to the KS test, the profiles of a TF are almost identical even if the number of mapped tags and peaks are largely different in, say, Esrrb (Figure 1D). The exceptional case is Sox2 in Chen_Eland and FP4_Bowtie (Figure 1E) due in part to the stringent filter used in Chen Eland; e.g. loss of Sox2 peaks in Chen Eland at gene clusters on chromosome X (Figure S1 in Additional file 2).\nImportantly, in the same peak dataset, the profiles are significantly different among TFs, e.g. Oct4 and Smad1 in FP4_Bowtie are shown in Figure S2. It is, therefore, thought that spatial preference of TF-bindings cannot be explained by one generalized distribution. In fact, the binding distributions of Nanog, Smad1, Sox2, and Stat3 definitely do not follow the exponential distribution (Figure S2 in Additional file 3).\nTo investigate the characteristics of TF binding sites in ESC, the density profiles of TF-bindings are estimated from each of peak datasets (Figure 1C), then any two density profiles for a TF in different peak datasets are tested by Kolmogorov-Smirnov (KS) test. According to the KS test, the profiles of a TF are almost identical even if the number of mapped tags and peaks are largely different in, say, Esrrb (Figure 1D). The exceptional case is Sox2 in Chen_Eland and FP4_Bowtie (Figure 1E) due in part to the stringent filter used in Chen Eland; e.g. loss of Sox2 peaks in Chen Eland at gene clusters on chromosome X (Figure S1 in Additional file 2).\nImportantly, in the same peak dataset, the profiles are significantly different among TFs, e.g. Oct4 and Smad1 in FP4_Bowtie are shown in Figure S2. It is, therefore, thought that spatial preference of TF-bindings cannot be explained by one generalized distribution. In fact, the binding distributions of Nanog, Smad1, Sox2, and Stat3 definitely do not follow the exponential distribution (Figure S2 in Additional file 3).\n[SUBTITLE] Density-based regression model outperforms the exponential-based model [SUBSECTION] Our observations from the genome-wide distribution of TF binding sites revealed the distinct binding preference from exponential function (Figure 1E). Thus, we use the density profiles as predictors given as equation (2), which we call the density-based regression model. The predictive power of the density-based model with Chen_Eland (Figure 2B) is slightly higher (CV-R2=0.508) than the exponential-based model (CV-R2=0.495). Similar results were obtained when other peak datasets were used.\nWe suspect that the prediction quality of two regression models may depend on downstream genes that cause specific density profiles. To confirm it, we extracted 4095 ESC-specific genes. E2f1 was excluded here due to its excessive regression coefficient [21]. Then, a subset of 4095 genes that is co-bound by a TF pair was prepared. Since the TFs used are well-known essential regulators in ESCs, the TF pairs, such as Oct4 and Sox2, possibly play an important regulatory role in their downstream ESC-specific genes. All subsets by any combination of two TFs have been prepared.\nFigure 2D illustrates that the density-based regression model outperforms in many cases. Furthermore, 55 gene subsets that are co-bound by TF pairs whose density profiles are significantly different (p < 0.05) were successfully predicted (red solid circles in Figure 2D). These gene subsets cannot be modeled by a generalized exponential function. The results suggest that the spatial preferences of TF bindings are much more dynamically changed in ESC-specific gene subsets rather than observed from all the genes. This is why the density-based model improved the predictive power with respect to the generalized exponential-based model.\nOur observations from the genome-wide distribution of TF binding sites revealed the distinct binding preference from exponential function (Figure 1E). Thus, we use the density profiles as predictors given as equation (2), which we call the density-based regression model. The predictive power of the density-based model with Chen_Eland (Figure 2B) is slightly higher (CV-R2=0.508) than the exponential-based model (CV-R2=0.495). Similar results were obtained when other peak datasets were used.\nWe suspect that the prediction quality of two regression models may depend on downstream genes that cause specific density profiles. To confirm it, we extracted 4095 ESC-specific genes. E2f1 was excluded here due to its excessive regression coefficient [21]. Then, a subset of 4095 genes that is co-bound by a TF pair was prepared. Since the TFs used are well-known essential regulators in ESCs, the TF pairs, such as Oct4 and Sox2, possibly play an important regulatory role in their downstream ESC-specific genes. All subsets by any combination of two TFs have been prepared.\nFigure 2D illustrates that the density-based regression model outperforms in many cases. Furthermore, 55 gene subsets that are co-bound by TF pairs whose density profiles are significantly different (p < 0.05) were successfully predicted (red solid circles in Figure 2D). These gene subsets cannot be modeled by a generalized exponential function. The results suggest that the spatial preferences of TF bindings are much more dynamically changed in ESC-specific gene subsets rather than observed from all the genes. This is why the density-based model improved the predictive power with respect to the generalized exponential-based model.\n[SUBTITLE] Two gene classes are different in epigenetic patterns [SUBSECTION] It was demonstrated previously that the absolute gene expression in ESCs is predictable by the ChIP-seq data of core TFs [21]. We also confirmed the high predictive power of the regression model. However, the results strongly rely on certain genes whose ‘predicted’ expressions are constantly lower, but ‘observed’ expressions are more varied (Figure 2A-C). In Figure 2A, we observed the binomial distribution of predicted expressions that can be partitioned by 1 RPKM (zero on the horizontal axis). We denote C1 for genes where predicted expression is ≥1 RPKM, C2 for the remains. The conspicuous frequency of C2 is also observed from Figure 2B-C. C2 genes in Figure 2C consist of 1205 up- and 1254 down-regulated genes. Further, the subset of C2 (C2′) where observed expression is greater than 1 RPKM consists of 148 up- and 159 down-regulated genes.\nTo characterize the gene classes, we analyzed TF-binding profiles and epigenetic modifications. As a result, in C2 genes, the number of peaks (Figure S3 in Additional file 2) and density profiles (Figure 2E) are apparently different, implying that the small number of TFs bind to distal regions from TSSs. C2 gene promoters are more methylated (Figure 3A). Remarkably, they tend to be absent from CpG islands (Figure 3B), and be marked with neither H2K4me3 nor bivalent domains (Figure 3C). Furthermore, we analyzed gene ontology terms of biological process by DAVID [23]. As a result, C1 was enriched for positive regulation of gene expression (score=8.66), whereas C2 was enriched for neural differentiation (score=34.49). C2′ was enriched for cell morphogenesis (score=2.77).\nEpigenetic modifications in ESC-specific genes. Three epigenetic states observed in genes whose expressions are 4-fold up or down in ESC against EB are considered. Gene class C1 and C2 are well explained and inefficiently explained genes by the regression analysis, respectively.\nC2 genes lack the TF-binding instances, implying less direct regulation by the core TFs. This depletion is due in part to excessive non-CpG DNA methylation [16]. Gene ontology analysis shows that C2 genes are often related to differentiation. Thus, they should be preferentially repressed in ESCs. Interestingly, as the histone marks are relatively rare among C2 genes, they are likely to be controlled by other regulatory pathways connecting to the maintenance of self-renewal. One possibility is the competitive binding of additional TFs not involved in this study because of the global open chromatin conformation in ESC [19]. Other possibilities include additional epigenetic patterns and homeostatic regulation, further investigations are required.\nIt was demonstrated previously that the absolute gene expression in ESCs is predictable by the ChIP-seq data of core TFs [21]. We also confirmed the high predictive power of the regression model. However, the results strongly rely on certain genes whose ‘predicted’ expressions are constantly lower, but ‘observed’ expressions are more varied (Figure 2A-C). In Figure 2A, we observed the binomial distribution of predicted expressions that can be partitioned by 1 RPKM (zero on the horizontal axis). We denote C1 for genes where predicted expression is ≥1 RPKM, C2 for the remains. The conspicuous frequency of C2 is also observed from Figure 2B-C. C2 genes in Figure 2C consist of 1205 up- and 1254 down-regulated genes. Further, the subset of C2 (C2′) where observed expression is greater than 1 RPKM consists of 148 up- and 159 down-regulated genes.\nTo characterize the gene classes, we analyzed TF-binding profiles and epigenetic modifications. As a result, in C2 genes, the number of peaks (Figure S3 in Additional file 2) and density profiles (Figure 2E) are apparently different, implying that the small number of TFs bind to distal regions from TSSs. C2 gene promoters are more methylated (Figure 3A). Remarkably, they tend to be absent from CpG islands (Figure 3B), and be marked with neither H2K4me3 nor bivalent domains (Figure 3C). Furthermore, we analyzed gene ontology terms of biological process by DAVID [23]. As a result, C1 was enriched for positive regulation of gene expression (score=8.66), whereas C2 was enriched for neural differentiation (score=34.49). C2′ was enriched for cell morphogenesis (score=2.77).\nEpigenetic modifications in ESC-specific genes. Three epigenetic states observed in genes whose expressions are 4-fold up or down in ESC against EB are considered. Gene class C1 and C2 are well explained and inefficiently explained genes by the regression analysis, respectively.\nC2 genes lack the TF-binding instances, implying less direct regulation by the core TFs. This depletion is due in part to excessive non-CpG DNA methylation [16]. Gene ontology analysis shows that C2 genes are often related to differentiation. Thus, they should be preferentially repressed in ESCs. Interestingly, as the histone marks are relatively rare among C2 genes, they are likely to be controlled by other regulatory pathways connecting to the maintenance of self-renewal. One possibility is the competitive binding of additional TFs not involved in this study because of the global open chromatin conformation in ESC [19]. Other possibilities include additional epigenetic patterns and homeostatic regulation, further investigations are required.\n[SUBTITLE] Epigenetic patterns improve the prediction of gene expression [SUBSECTION] To further understand the epigenetic effects in gene regulation, we add three epigenetic states to the regression models; histone mark (HistM), DNA methylation (Methy), and CpG island (CpGI). Thus, 14 explanatory variables are used. To identify effective variables in the prediction, we reduced the regression model by using the stepwise model selection. Also, 100 runs of computer simulation that randomly assign the epigenetic states were performed.\nAll models with the epigenetic effects improved CV-R2 with one to three more variables compared with the models without the epigenetic effects (Table 2). The additional variables are the epigenetic effect terms. The results of simulation support that the improvements are not by the chance. In particular, the density-based models with the epigenetic effects are significantly better when remapped peak datasets are used. Furthermore, overall regression coefficients gathered from all the density-based models in Table 2 show the relative importance of epigenetic effects except CpGI (Figure 2F). Note that the positive-biased activities are consistent with the previous study [24].\nEffects of epigenetic patterns in reduced regression models\nTo further understand the epigenetic effects in gene regulation, we add three epigenetic states to the regression models; histone mark (HistM), DNA methylation (Methy), and CpG island (CpGI). Thus, 14 explanatory variables are used. To identify effective variables in the prediction, we reduced the regression model by using the stepwise model selection. Also, 100 runs of computer simulation that randomly assign the epigenetic states were performed.\nAll models with the epigenetic effects improved CV-R2 with one to three more variables compared with the models without the epigenetic effects (Table 2). The additional variables are the epigenetic effect terms. The results of simulation support that the improvements are not by the chance. In particular, the density-based models with the epigenetic effects are significantly better when remapped peak datasets are used. Furthermore, overall regression coefficients gathered from all the density-based models in Table 2 show the relative importance of epigenetic effects except CpGI (Figure 2F). Note that the positive-biased activities are consistent with the previous study [24].\nEffects of epigenetic patterns in reduced regression models\n[SUBTITLE] TF interactions wired with epigenetic effects [SUBSECTION] To investigate the cooperative effects among TFs and epigenetic patterns in gene regulation, we exhaustively searched significant interaction terms from our regression model. First, a subset of ESC-specific genes that are co-bound by a specific TF pair is prepared. Then, the saturated model for the genes is constructed. The model involves 469 variables; 14 main effect terms (11 TFs and 3 epigenetic states) and 455 higher-order interaction terms (all the possible pairwise and triplewise interactions). Finally, our pipeline greedily identifies important variables (see methods). This procedure is independently performed with each of five peak datasets.\nIn total, 215 models were identified in which the predictive power is higher than the models without higher-order terms. These models contained 6-30 variables including at least one interactive term. As an example, the regression model for genes co-bound by Oct4 and Sox2, a well-known pluripotent complex [9,25], contained 15 terms and improved CV-R2=0.4126 from 0.3837 in the model with only 14 main effect terms. This model suggests that 7 interactive terms are important in the explanation of target gene expression. Among them, 3 terms are mediated by the epigenetic effects. The network representation of this model highlights the importance of signaling receptors (Stat3 and Smad1), activating Oct4/Sox2 complex [9] as well as Klf4/CpGI [26], and the interaction of Zfx/Methy newly found here (Figure 4).\nExample of regulatory network of TF interactions with epigenetic effects. This network was generated by the connectivity of nodes in all interaction terms. For example, an interaction term A:B:C is split into three interactions, A:B, A:C, and B:C. Then, the nodes are linked to each other and to the target gene set. A pairwise interaction with an epigenetic effect is treated differently. For example, in the case of B:Methy, B is not linked to the target gene set.\nWith considering the redundancy and conservativity, we represented the interactive terms of 215 models as a network (Figure S4 in Additional file 2). As a result, 19 gene sets covering approximately 86% of genes (3523 out of 4095 genes) were linked by 28 regulatory edges of the epigenetic effects that are commonly found in the five peak datasets (Figure S4 in Additional file 2). These results suggest that the cooperative interactions between TF and the epigenetic state are indispensable to explain the majority of gene expression in ESCs. In addition, we confirmed that the regression coefficients in Figure 2F are dramatically changed in the regression of given gene sets, and also CpGI significantly contributes to the prediction of gene expression (Figure 4).\nTo investigate the cooperative effects among TFs and epigenetic patterns in gene regulation, we exhaustively searched significant interaction terms from our regression model. First, a subset of ESC-specific genes that are co-bound by a specific TF pair is prepared. Then, the saturated model for the genes is constructed. The model involves 469 variables; 14 main effect terms (11 TFs and 3 epigenetic states) and 455 higher-order interaction terms (all the possible pairwise and triplewise interactions). Finally, our pipeline greedily identifies important variables (see methods). This procedure is independently performed with each of five peak datasets.\nIn total, 215 models were identified in which the predictive power is higher than the models without higher-order terms. These models contained 6-30 variables including at least one interactive term. As an example, the regression model for genes co-bound by Oct4 and Sox2, a well-known pluripotent complex [9,25], contained 15 terms and improved CV-R2=0.4126 from 0.3837 in the model with only 14 main effect terms. This model suggests that 7 interactive terms are important in the explanation of target gene expression. Among them, 3 terms are mediated by the epigenetic effects. The network representation of this model highlights the importance of signaling receptors (Stat3 and Smad1), activating Oct4/Sox2 complex [9] as well as Klf4/CpGI [26], and the interaction of Zfx/Methy newly found here (Figure 4).\nExample of regulatory network of TF interactions with epigenetic effects. This network was generated by the connectivity of nodes in all interaction terms. For example, an interaction term A:B:C is split into three interactions, A:B, A:C, and B:C. Then, the nodes are linked to each other and to the target gene set. A pairwise interaction with an epigenetic effect is treated differently. For example, in the case of B:Methy, B is not linked to the target gene set.\nWith considering the redundancy and conservativity, we represented the interactive terms of 215 models as a network (Figure S4 in Additional file 2). As a result, 19 gene sets covering approximately 86% of genes (3523 out of 4095 genes) were linked by 28 regulatory edges of the epigenetic effects that are commonly found in the five peak datasets (Figure S4 in Additional file 2). These results suggest that the cooperative interactions between TF and the epigenetic state are indispensable to explain the majority of gene expression in ESCs. In addition, we confirmed that the regression coefficients in Figure 2F are dramatically changed in the regression of given gene sets, and also CpGI significantly contributes to the prediction of gene expression (Figure 4).", "To minimize artifacts, we refined the binding signals of 12 core TFs in mouse ESC publicly available [9] (see methods). The ChIP-seq peak datasets generated by various tools are hereafter denoted as FP4_Bowtie, FP4_MAQ, and FP4_Soap2. Also, tag positions mapped by Eland [9] are used for the peak detection (FP4_Eland), and the peak data of Chen et al. is involved (Chen Eland). Thus, we prepared five peak datasets in total.\nDifferences in numbers and positions between the remapped data and the original data were investigated. As a result, relatively larger number of uniquely mapped tags and peaks were gained compared to the original data (Table S1-S3 in Additional file 1). In regard to peaks (Table 1), FP4 with the previously mapped tags (FP4_Eland) covers 85-98% of Chen_Eland, and the intensity of overlapped peaks is strongly correlated. Thus, it is deemed that FP4 has reproduced Chen_Eland and extended it with novel peaks in different genomic locations. In contrast, FP4 with remapped tags shows relatively lower reproducibility, whereas peak intensities are still correlated with Chen_Eland except Esrrb (Figure 1B). Similar observations can be found from an independent study [22].\nReproducibility of newly detected peaks\nFold change is the ratio of newly detected peak number over the original peak number. Overlaps of the original peaks to the newly detected peaks were investigated with 200-bp window. The overlapped peaks were used to calculate the correlation of peak intensity.\nSchematic representation of ChIP-seq data analysis and examples of TF-binding density profiles. (A) Definitions of a peak region and its intensity. (B) Example of intensity correlation between the original data and the reanalyzed data. (C) Density estimation from the genome-wide binding locations of a TF. (D)-(E) Example of density profiles in five peak datasets.\nThe reason why the numbers vary is twofold. First, algorithmic differences in alignment tools cause the different numbers, particularly due to the gapped or ungapped alignment and random indel for mismatches. Second, thresholds for the peak intensity to distinguish experimental noise are different (Table S4 in Additional file 1). That is, Chen et al. used qPCR refinement with small number of peaks, whereas we used Monte Carlo simulation on each chromosome.", "To assess the importance of TF bindings, Ouyang et al. [21] successfully applied a regression model to the prediction of absolute gene expression in mouse ESC. We first recover this study. Ouyang et al. used TF association strength (TFAS) by summing up peak intensities that are weighted exponentially according to the relative positions from TSSs. We applied the TFAS data to our simple linear regression model shown in equation (1), namely exponential-based regression model. The predictive power of our model is much higher (CV-R2=0.647) than Ouyang’s model (CV-R2=0.639), suggesting that the simple regression model is comparable to their PC-regression model (Figure 2A).\nPredictive results of density-based linear regression model. (A)-(C) Average correlation coefficient of 10-fold CV in three gene sets. (D) Comparative analysis of two models using ESC-specific gene subsets that co-bound by pluripotent TF pairs. Red solid circles are the cases that the density profiles of a TF pair are significantly different. (E) Distinct binding profiles in two gene classes that are either well explained (C1) or inefficiently explained (C2). (F) Contribution of each TF and epigenetic effects in density-based regression models (Methy: DNA methylation, HistM: histone mark, CpGI: CpG Island).\nNext, we prepared 17060 genes by removing inconsistency between Ouyang’s study and Chen’s study. This procedure is prerequisite for gathering precise TF-binding instances. TFAS data for the genes were calculated by the exactly same procedure of Ouyang et al. As a result, the exponential-based model shows CV-R2=0.495 with Chen_Eland. In contrast, CV-R2 increases to 0.542 (FP4_Eland), 0.587 (FP4_Bowtie), 0.581 (FP4_MAQ), and 0.590 (FP4_Soap2).\nThese results clearly suggest that the proposed simple linear regression model is applicable to the prediction. Furthermore, it has been demonstrated that the peak datasets we remapped give more information for explaining the gene expression.", "To investigate the characteristics of TF binding sites in ESC, the density profiles of TF-bindings are estimated from each of peak datasets (Figure 1C), then any two density profiles for a TF in different peak datasets are tested by Kolmogorov-Smirnov (KS) test. According to the KS test, the profiles of a TF are almost identical even if the number of mapped tags and peaks are largely different in, say, Esrrb (Figure 1D). The exceptional case is Sox2 in Chen_Eland and FP4_Bowtie (Figure 1E) due in part to the stringent filter used in Chen Eland; e.g. loss of Sox2 peaks in Chen Eland at gene clusters on chromosome X (Figure S1 in Additional file 2).\nImportantly, in the same peak dataset, the profiles are significantly different among TFs, e.g. Oct4 and Smad1 in FP4_Bowtie are shown in Figure S2. It is, therefore, thought that spatial preference of TF-bindings cannot be explained by one generalized distribution. In fact, the binding distributions of Nanog, Smad1, Sox2, and Stat3 definitely do not follow the exponential distribution (Figure S2 in Additional file 3).", "Our observations from the genome-wide distribution of TF binding sites revealed the distinct binding preference from exponential function (Figure 1E). Thus, we use the density profiles as predictors given as equation (2), which we call the density-based regression model. The predictive power of the density-based model with Chen_Eland (Figure 2B) is slightly higher (CV-R2=0.508) than the exponential-based model (CV-R2=0.495). Similar results were obtained when other peak datasets were used.\nWe suspect that the prediction quality of two regression models may depend on downstream genes that cause specific density profiles. To confirm it, we extracted 4095 ESC-specific genes. E2f1 was excluded here due to its excessive regression coefficient [21]. Then, a subset of 4095 genes that is co-bound by a TF pair was prepared. Since the TFs used are well-known essential regulators in ESCs, the TF pairs, such as Oct4 and Sox2, possibly play an important regulatory role in their downstream ESC-specific genes. All subsets by any combination of two TFs have been prepared.\nFigure 2D illustrates that the density-based regression model outperforms in many cases. Furthermore, 55 gene subsets that are co-bound by TF pairs whose density profiles are significantly different (p < 0.05) were successfully predicted (red solid circles in Figure 2D). These gene subsets cannot be modeled by a generalized exponential function. The results suggest that the spatial preferences of TF bindings are much more dynamically changed in ESC-specific gene subsets rather than observed from all the genes. This is why the density-based model improved the predictive power with respect to the generalized exponential-based model.", "It was demonstrated previously that the absolute gene expression in ESCs is predictable by the ChIP-seq data of core TFs [21]. We also confirmed the high predictive power of the regression model. However, the results strongly rely on certain genes whose ‘predicted’ expressions are constantly lower, but ‘observed’ expressions are more varied (Figure 2A-C). In Figure 2A, we observed the binomial distribution of predicted expressions that can be partitioned by 1 RPKM (zero on the horizontal axis). We denote C1 for genes where predicted expression is ≥1 RPKM, C2 for the remains. The conspicuous frequency of C2 is also observed from Figure 2B-C. C2 genes in Figure 2C consist of 1205 up- and 1254 down-regulated genes. Further, the subset of C2 (C2′) where observed expression is greater than 1 RPKM consists of 148 up- and 159 down-regulated genes.\nTo characterize the gene classes, we analyzed TF-binding profiles and epigenetic modifications. As a result, in C2 genes, the number of peaks (Figure S3 in Additional file 2) and density profiles (Figure 2E) are apparently different, implying that the small number of TFs bind to distal regions from TSSs. C2 gene promoters are more methylated (Figure 3A). Remarkably, they tend to be absent from CpG islands (Figure 3B), and be marked with neither H2K4me3 nor bivalent domains (Figure 3C). Furthermore, we analyzed gene ontology terms of biological process by DAVID [23]. As a result, C1 was enriched for positive regulation of gene expression (score=8.66), whereas C2 was enriched for neural differentiation (score=34.49). C2′ was enriched for cell morphogenesis (score=2.77).\nEpigenetic modifications in ESC-specific genes. Three epigenetic states observed in genes whose expressions are 4-fold up or down in ESC against EB are considered. Gene class C1 and C2 are well explained and inefficiently explained genes by the regression analysis, respectively.\nC2 genes lack the TF-binding instances, implying less direct regulation by the core TFs. This depletion is due in part to excessive non-CpG DNA methylation [16]. Gene ontology analysis shows that C2 genes are often related to differentiation. Thus, they should be preferentially repressed in ESCs. Interestingly, as the histone marks are relatively rare among C2 genes, they are likely to be controlled by other regulatory pathways connecting to the maintenance of self-renewal. One possibility is the competitive binding of additional TFs not involved in this study because of the global open chromatin conformation in ESC [19]. Other possibilities include additional epigenetic patterns and homeostatic regulation, further investigations are required.", "To further understand the epigenetic effects in gene regulation, we add three epigenetic states to the regression models; histone mark (HistM), DNA methylation (Methy), and CpG island (CpGI). Thus, 14 explanatory variables are used. To identify effective variables in the prediction, we reduced the regression model by using the stepwise model selection. Also, 100 runs of computer simulation that randomly assign the epigenetic states were performed.\nAll models with the epigenetic effects improved CV-R2 with one to three more variables compared with the models without the epigenetic effects (Table 2). The additional variables are the epigenetic effect terms. The results of simulation support that the improvements are not by the chance. In particular, the density-based models with the epigenetic effects are significantly better when remapped peak datasets are used. Furthermore, overall regression coefficients gathered from all the density-based models in Table 2 show the relative importance of epigenetic effects except CpGI (Figure 2F). Note that the positive-biased activities are consistent with the previous study [24].\nEffects of epigenetic patterns in reduced regression models", "To investigate the cooperative effects among TFs and epigenetic patterns in gene regulation, we exhaustively searched significant interaction terms from our regression model. First, a subset of ESC-specific genes that are co-bound by a specific TF pair is prepared. Then, the saturated model for the genes is constructed. The model involves 469 variables; 14 main effect terms (11 TFs and 3 epigenetic states) and 455 higher-order interaction terms (all the possible pairwise and triplewise interactions). Finally, our pipeline greedily identifies important variables (see methods). This procedure is independently performed with each of five peak datasets.\nIn total, 215 models were identified in which the predictive power is higher than the models without higher-order terms. These models contained 6-30 variables including at least one interactive term. As an example, the regression model for genes co-bound by Oct4 and Sox2, a well-known pluripotent complex [9,25], contained 15 terms and improved CV-R2=0.4126 from 0.3837 in the model with only 14 main effect terms. This model suggests that 7 interactive terms are important in the explanation of target gene expression. Among them, 3 terms are mediated by the epigenetic effects. The network representation of this model highlights the importance of signaling receptors (Stat3 and Smad1), activating Oct4/Sox2 complex [9] as well as Klf4/CpGI [26], and the interaction of Zfx/Methy newly found here (Figure 4).\nExample of regulatory network of TF interactions with epigenetic effects. This network was generated by the connectivity of nodes in all interaction terms. For example, an interaction term A:B:C is split into three interactions, A:B, A:C, and B:C. Then, the nodes are linked to each other and to the target gene set. A pairwise interaction with an epigenetic effect is treated differently. For example, in the case of B:Methy, B is not linked to the target gene set.\nWith considering the redundancy and conservativity, we represented the interactive terms of 215 models as a network (Figure S4 in Additional file 2). As a result, 19 gene sets covering approximately 86% of genes (3523 out of 4095 genes) were linked by 28 regulatory edges of the epigenetic effects that are commonly found in the five peak datasets (Figure S4 in Additional file 2). These results suggest that the cooperative interactions between TF and the epigenetic state are indispensable to explain the majority of gene expression in ESCs. In addition, we confirmed that the regression coefficients in Figure 2F are dramatically changed in the regression of given gene sets, and also CpGI significantly contributes to the prediction of gene expression (Figure 4).", "ESCs are the widely accepted source for the study of many biological principles. Despite recent advances in our understanding of biological systems, the gene regulation in ESCs is only incompletely understood. To explore the regulatory mechanism underlying in ESCs, we constructed a predictive model for explaining the absolute gene expression in mouse ESC. This model uses a novel density-based approach to exploit the recent massive parallel sequencing straightforwardly.\nWe first reanalyzed the publicly available ChIP-seq data for 12 well-known pluripotent core TFs [9], and retrieved the reproduced and extended TF-binding sites and intensities (Table 1). Using our regression model based on the exponential function [21], we found that the remapped peaks are more informative to explain the gene expression (Table 2). Therefore, we concluded that the algorithmic differences in computer tools for ChIP-seq data significantly affect the downstream analysis. Analyzing the heterogeneous peak datasets in a comparative manner, we found that the spatial binding preference of each TF is well conserved in all the datasets, whereas the preferences of TFs in a dataset are significantly different from each other (Figure 1D-E). These results imply that density profiles are better explanatory variables than the generalized exponential function. In fact, the predictive power of density-based model is constantly higher than the exponential-based model (Figure 2A-C, Table 2). Even if the density profiles are dynamically changed in certain downstream genes, the proposed model is still outstanding (Figure 2D).\nUnexpectedly, we found two gene classes that are either well explained or inefficiently explained by the regression model. The latter class genes have less binding instances of the pluripotent TFs (Figure 2E), possibly related to excessive DNA methylation (Figure 3A). The gene classes show apparently different characteristics in epigenetic modifications (Figure 3), suggesting that they are likely to be under control in different regulatory mechanisms. In the present study, we simply combined the discrete epigenetic states with the powerful density-based model. This model significantly improved the predictive power (Table 2). Investigating higher-order interactions among the predictors, we found that the cooperative interactions between TF and epigenetic pattern are indispensable for regulating approximately 86% of ESC-specific genes (Figure S4 in Additional file 2). These results suggest that the relative importance of epigenetic effects to regulate the gene expression in ESCs, supporting the general idea [14,15].\nWe proposed a powerful regression model, and uncovered the relative importance of epigenetic regulation in ESCs. Overall prediction quality is still insufficient. As future works, comprehensive representation of epigenetic patterns is required, and additional or alternative TFs in ESCs should be considered.", "[SUBTITLE] ChIP-seq data and gene expression [SUBSECTION] Raw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.\nRaw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.\n[SUBTITLE] Epigenetic modifications [SUBSECTION] DNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.\nDNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.", "Raw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.", "DNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.", "Given a genome-wide location map of a TF-bindings, all peak positions were converted to relative positions to the nearest TSSs. Gaussian kernel density function (bandwidth=300 bps) estimated the density profile of the TF-bindings within ±50K bps. The profile was normalized into range of [0, 1] by dividing by the maximal density height.", "We use a multivariate regression model(1)\nwhere Yi is the expression of gene i, Sij is the score of the jth TF on gene i, wj is the regression coefficient of the jth TF, and ei is the error term. The score Sij is given by(2)\nwhere gk is the perk intensity of the kth binding peak of the jth TF, Fj is the normalized density function for the jth TF, and lk is the relative position of the kth peak to TSS of gene i. Note that a small value is added to Yi for the logarithm.", "Discrete values representing epigenetic states of a gene i are added to the regression model(3)\nwhere H is the type of histone mark (neither mark=1.0, H3K27me3=2.0, bivalent mark=3.0, H3K4me3=4.0), M is the DNA methylation (no annotation=1.0, methylation=2.0, unmethylation=3.0), C is the CpG island (absence=1.0, presence=2.0), and α, β, γ are the regression coefficients for H, M, C, respectively.", "Explanatory variables in a regression model are log-transformed and quantile-normalized. 10 runs of 10-fold cross validation (CV) measure the average correlation coefficient (CV-R) and the average proportion of variation explained by the model (CV-R2). The stepwise model selection is done by stepAIC in R language with the backward and forward procedure. The regression model with higher-order interactions are reduced by a pipeline developed in house; ANOVA in R language first diagnoses the significance of each explanatory variable in the given saturated model. Next, significant variables (p < 0.05 in F-test) are gathered. Finally, the best model is constructed by adding and removing the collected variables one by one in increasing order of p-value until CV-R2 is not improved anymore.", "The authors declare that they have no competing interests." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Results", "ChIP-seq data is reproduced and extended", "Remapped peaks improve the prediction of gene expression", "Genome-wide locations of TF bindings do not follow exponential distribution", "Density-based regression model outperforms the exponential-based model", "Two gene classes are different in epigenetic patterns", "Epigenetic patterns improve the prediction of gene expression", "TF interactions wired with epigenetic effects", "Discussion", "Methods", "Data acquisition", "ChIP-seq data and gene expression", "Epigenetic modifications", "Estimation of TF binding density", "Regression model", "Adding epigenetic effects", "Fitting and reducing regression models", "Competing interests", "Supplementary Material" ]

[ "Embryonic stem cells (ESCs) derived from blastocysts are self-renewal and pluripotent [1-3]. To understand the gene regulatory system in ESCs is an important step for uncovering the process of cell fate determination and for promoting regenerative medicine. Considerable recent evidence indicates that several transcription factors (TFs), so-called core TFs, are indispensable to maintain the pluripotency [4,5]. Some of the core TFs reprogram somatic cells back to pluripotent states [6,7]. These observations suggest that the regulatory network of TFs apparently governs the self-renewal and pluripotency [8,9]. On the other hand, many studies have reported that other TFs can functionally substitute for the core TFs [10-13], suggesting that there still exist additional or alternative TFs unrevealed in the network. Epigenetic modifications are also essential for ESCs [14,15]. Their involvement in the maintenance of the pluripotency is still not well clarified.\nTo understand the regulatory mechanism underlying in ESCs, a number of methods have been developed. In particular, massive parallel sequencing [9,16-19] and various in silico approaches [8,9,20,21] have yielded comprehensive recent advances in our understanding. In this study, we focus on predicting the gene expression in ESCs with the massive parallel sequencing data. Although a previous study successfully applied a regression model to the prediction [21], the model is based on a generalized weighting scheme to prepare predictors (explanatory variables). Intuitively, such weighting scheme cannot reflect the nature of the spatial rearrangement of TF-binding.\nHere, we propose a density-based approach that uses the genome-wide bona fide TF binding sites. First, a publicly available ChIP-seq data [9] is reanalyzed. Then, density profiles of TFs estimated from the ChIP-seq data are adopted as predictors in a simple linear regression model to predict the genome-wide gene expression. Predictors are also combined with epigenetic data, such as H3K4me3, H3K27me3, DNA methylation, and CpG island [16,17]. Furthermore, we analyze the regulatory effects of TFs, epigenetic states, and their higher-order interactions by using a pipeline developed in house. We demonstrate the predictive power of the density-based regression model and discuss our findings.", "[SUBTITLE] ChIP-seq data is reproduced and extended [SUBSECTION] To minimize artifacts, we refined the binding signals of 12 core TFs in mouse ESC publicly available [9] (see methods). The ChIP-seq peak datasets generated by various tools are hereafter denoted as FP4_Bowtie, FP4_MAQ, and FP4_Soap2. Also, tag positions mapped by Eland [9] are used for the peak detection (FP4_Eland), and the peak data of Chen et al. is involved (Chen Eland). Thus, we prepared five peak datasets in total.\nDifferences in numbers and positions between the remapped data and the original data were investigated. As a result, relatively larger number of uniquely mapped tags and peaks were gained compared to the original data (Table S1-S3 in Additional file 1). In regard to peaks (Table 1), FP4 with the previously mapped tags (FP4_Eland) covers 85-98% of Chen_Eland, and the intensity of overlapped peaks is strongly correlated. Thus, it is deemed that FP4 has reproduced Chen_Eland and extended it with novel peaks in different genomic locations. In contrast, FP4 with remapped tags shows relatively lower reproducibility, whereas peak intensities are still correlated with Chen_Eland except Esrrb (Figure 1B). Similar observations can be found from an independent study [22].\nReproducibility of newly detected peaks\nFold change is the ratio of newly detected peak number over the original peak number. Overlaps of the original peaks to the newly detected peaks were investigated with 200-bp window. The overlapped peaks were used to calculate the correlation of peak intensity.\nSchematic representation of ChIP-seq data analysis and examples of TF-binding density profiles. (A) Definitions of a peak region and its intensity. (B) Example of intensity correlation between the original data and the reanalyzed data. (C) Density estimation from the genome-wide binding locations of a TF. (D)-(E) Example of density profiles in five peak datasets.\nThe reason why the numbers vary is twofold. First, algorithmic differences in alignment tools cause the different numbers, particularly due to the gapped or ungapped alignment and random indel for mismatches. Second, thresholds for the peak intensity to distinguish experimental noise are different (Table S4 in Additional file 1). That is, Chen et al. used qPCR refinement with small number of peaks, whereas we used Monte Carlo simulation on each chromosome.\nTo minimize artifacts, we refined the binding signals of 12 core TFs in mouse ESC publicly available [9] (see methods). The ChIP-seq peak datasets generated by various tools are hereafter denoted as FP4_Bowtie, FP4_MAQ, and FP4_Soap2. Also, tag positions mapped by Eland [9] are used for the peak detection (FP4_Eland), and the peak data of Chen et al. is involved (Chen Eland). Thus, we prepared five peak datasets in total.\nDifferences in numbers and positions between the remapped data and the original data were investigated. As a result, relatively larger number of uniquely mapped tags and peaks were gained compared to the original data (Table S1-S3 in Additional file 1). In regard to peaks (Table 1), FP4 with the previously mapped tags (FP4_Eland) covers 85-98% of Chen_Eland, and the intensity of overlapped peaks is strongly correlated. Thus, it is deemed that FP4 has reproduced Chen_Eland and extended it with novel peaks in different genomic locations. In contrast, FP4 with remapped tags shows relatively lower reproducibility, whereas peak intensities are still correlated with Chen_Eland except Esrrb (Figure 1B). Similar observations can be found from an independent study [22].\nReproducibility of newly detected peaks\nFold change is the ratio of newly detected peak number over the original peak number. Overlaps of the original peaks to the newly detected peaks were investigated with 200-bp window. The overlapped peaks were used to calculate the correlation of peak intensity.\nSchematic representation of ChIP-seq data analysis and examples of TF-binding density profiles. (A) Definitions of a peak region and its intensity. (B) Example of intensity correlation between the original data and the reanalyzed data. (C) Density estimation from the genome-wide binding locations of a TF. (D)-(E) Example of density profiles in five peak datasets.\nThe reason why the numbers vary is twofold. First, algorithmic differences in alignment tools cause the different numbers, particularly due to the gapped or ungapped alignment and random indel for mismatches. Second, thresholds for the peak intensity to distinguish experimental noise are different (Table S4 in Additional file 1). That is, Chen et al. used qPCR refinement with small number of peaks, whereas we used Monte Carlo simulation on each chromosome.\n[SUBTITLE] Remapped peaks improve the prediction of gene expression [SUBSECTION] To assess the importance of TF bindings, Ouyang et al. [21] successfully applied a regression model to the prediction of absolute gene expression in mouse ESC. We first recover this study. Ouyang et al. used TF association strength (TFAS) by summing up peak intensities that are weighted exponentially according to the relative positions from TSSs. We applied the TFAS data to our simple linear regression model shown in equation (1), namely exponential-based regression model. The predictive power of our model is much higher (CV-R2=0.647) than Ouyang’s model (CV-R2=0.639), suggesting that the simple regression model is comparable to their PC-regression model (Figure 2A).\nPredictive results of density-based linear regression model. (A)-(C) Average correlation coefficient of 10-fold CV in three gene sets. (D) Comparative analysis of two models using ESC-specific gene subsets that co-bound by pluripotent TF pairs. Red solid circles are the cases that the density profiles of a TF pair are significantly different. (E) Distinct binding profiles in two gene classes that are either well explained (C1) or inefficiently explained (C2). (F) Contribution of each TF and epigenetic effects in density-based regression models (Methy: DNA methylation, HistM: histone mark, CpGI: CpG Island).\nNext, we prepared 17060 genes by removing inconsistency between Ouyang’s study and Chen’s study. This procedure is prerequisite for gathering precise TF-binding instances. TFAS data for the genes were calculated by the exactly same procedure of Ouyang et al. As a result, the exponential-based model shows CV-R2=0.495 with Chen_Eland. In contrast, CV-R2 increases to 0.542 (FP4_Eland), 0.587 (FP4_Bowtie), 0.581 (FP4_MAQ), and 0.590 (FP4_Soap2).\nThese results clearly suggest that the proposed simple linear regression model is applicable to the prediction. Furthermore, it has been demonstrated that the peak datasets we remapped give more information for explaining the gene expression.\nTo assess the importance of TF bindings, Ouyang et al. [21] successfully applied a regression model to the prediction of absolute gene expression in mouse ESC. We first recover this study. Ouyang et al. used TF association strength (TFAS) by summing up peak intensities that are weighted exponentially according to the relative positions from TSSs. We applied the TFAS data to our simple linear regression model shown in equation (1), namely exponential-based regression model. The predictive power of our model is much higher (CV-R2=0.647) than Ouyang’s model (CV-R2=0.639), suggesting that the simple regression model is comparable to their PC-regression model (Figure 2A).\nPredictive results of density-based linear regression model. (A)-(C) Average correlation coefficient of 10-fold CV in three gene sets. (D) Comparative analysis of two models using ESC-specific gene subsets that co-bound by pluripotent TF pairs. Red solid circles are the cases that the density profiles of a TF pair are significantly different. (E) Distinct binding profiles in two gene classes that are either well explained (C1) or inefficiently explained (C2). (F) Contribution of each TF and epigenetic effects in density-based regression models (Methy: DNA methylation, HistM: histone mark, CpGI: CpG Island).\nNext, we prepared 17060 genes by removing inconsistency between Ouyang’s study and Chen’s study. This procedure is prerequisite for gathering precise TF-binding instances. TFAS data for the genes were calculated by the exactly same procedure of Ouyang et al. As a result, the exponential-based model shows CV-R2=0.495 with Chen_Eland. In contrast, CV-R2 increases to 0.542 (FP4_Eland), 0.587 (FP4_Bowtie), 0.581 (FP4_MAQ), and 0.590 (FP4_Soap2).\nThese results clearly suggest that the proposed simple linear regression model is applicable to the prediction. Furthermore, it has been demonstrated that the peak datasets we remapped give more information for explaining the gene expression.\n[SUBTITLE] Genome-wide locations of TF bindings do not follow exponential distribution [SUBSECTION] To investigate the characteristics of TF binding sites in ESC, the density profiles of TF-bindings are estimated from each of peak datasets (Figure 1C), then any two density profiles for a TF in different peak datasets are tested by Kolmogorov-Smirnov (KS) test. According to the KS test, the profiles of a TF are almost identical even if the number of mapped tags and peaks are largely different in, say, Esrrb (Figure 1D). The exceptional case is Sox2 in Chen_Eland and FP4_Bowtie (Figure 1E) due in part to the stringent filter used in Chen Eland; e.g. loss of Sox2 peaks in Chen Eland at gene clusters on chromosome X (Figure S1 in Additional file 2).\nImportantly, in the same peak dataset, the profiles are significantly different among TFs, e.g. Oct4 and Smad1 in FP4_Bowtie are shown in Figure S2. It is, therefore, thought that spatial preference of TF-bindings cannot be explained by one generalized distribution. In fact, the binding distributions of Nanog, Smad1, Sox2, and Stat3 definitely do not follow the exponential distribution (Figure S2 in Additional file 3).\nTo investigate the characteristics of TF binding sites in ESC, the density profiles of TF-bindings are estimated from each of peak datasets (Figure 1C), then any two density profiles for a TF in different peak datasets are tested by Kolmogorov-Smirnov (KS) test. According to the KS test, the profiles of a TF are almost identical even if the number of mapped tags and peaks are largely different in, say, Esrrb (Figure 1D). The exceptional case is Sox2 in Chen_Eland and FP4_Bowtie (Figure 1E) due in part to the stringent filter used in Chen Eland; e.g. loss of Sox2 peaks in Chen Eland at gene clusters on chromosome X (Figure S1 in Additional file 2).\nImportantly, in the same peak dataset, the profiles are significantly different among TFs, e.g. Oct4 and Smad1 in FP4_Bowtie are shown in Figure S2. It is, therefore, thought that spatial preference of TF-bindings cannot be explained by one generalized distribution. In fact, the binding distributions of Nanog, Smad1, Sox2, and Stat3 definitely do not follow the exponential distribution (Figure S2 in Additional file 3).\n[SUBTITLE] Density-based regression model outperforms the exponential-based model [SUBSECTION] Our observations from the genome-wide distribution of TF binding sites revealed the distinct binding preference from exponential function (Figure 1E). Thus, we use the density profiles as predictors given as equation (2), which we call the density-based regression model. The predictive power of the density-based model with Chen_Eland (Figure 2B) is slightly higher (CV-R2=0.508) than the exponential-based model (CV-R2=0.495). Similar results were obtained when other peak datasets were used.\nWe suspect that the prediction quality of two regression models may depend on downstream genes that cause specific density profiles. To confirm it, we extracted 4095 ESC-specific genes. E2f1 was excluded here due to its excessive regression coefficient [21]. Then, a subset of 4095 genes that is co-bound by a TF pair was prepared. Since the TFs used are well-known essential regulators in ESCs, the TF pairs, such as Oct4 and Sox2, possibly play an important regulatory role in their downstream ESC-specific genes. All subsets by any combination of two TFs have been prepared.\nFigure 2D illustrates that the density-based regression model outperforms in many cases. Furthermore, 55 gene subsets that are co-bound by TF pairs whose density profiles are significantly different (p < 0.05) were successfully predicted (red solid circles in Figure 2D). These gene subsets cannot be modeled by a generalized exponential function. The results suggest that the spatial preferences of TF bindings are much more dynamically changed in ESC-specific gene subsets rather than observed from all the genes. This is why the density-based model improved the predictive power with respect to the generalized exponential-based model.\nOur observations from the genome-wide distribution of TF binding sites revealed the distinct binding preference from exponential function (Figure 1E). Thus, we use the density profiles as predictors given as equation (2), which we call the density-based regression model. The predictive power of the density-based model with Chen_Eland (Figure 2B) is slightly higher (CV-R2=0.508) than the exponential-based model (CV-R2=0.495). Similar results were obtained when other peak datasets were used.\nWe suspect that the prediction quality of two regression models may depend on downstream genes that cause specific density profiles. To confirm it, we extracted 4095 ESC-specific genes. E2f1 was excluded here due to its excessive regression coefficient [21]. Then, a subset of 4095 genes that is co-bound by a TF pair was prepared. Since the TFs used are well-known essential regulators in ESCs, the TF pairs, such as Oct4 and Sox2, possibly play an important regulatory role in their downstream ESC-specific genes. All subsets by any combination of two TFs have been prepared.\nFigure 2D illustrates that the density-based regression model outperforms in many cases. Furthermore, 55 gene subsets that are co-bound by TF pairs whose density profiles are significantly different (p < 0.05) were successfully predicted (red solid circles in Figure 2D). These gene subsets cannot be modeled by a generalized exponential function. The results suggest that the spatial preferences of TF bindings are much more dynamically changed in ESC-specific gene subsets rather than observed from all the genes. This is why the density-based model improved the predictive power with respect to the generalized exponential-based model.\n[SUBTITLE] Two gene classes are different in epigenetic patterns [SUBSECTION] It was demonstrated previously that the absolute gene expression in ESCs is predictable by the ChIP-seq data of core TFs [21]. We also confirmed the high predictive power of the regression model. However, the results strongly rely on certain genes whose ‘predicted’ expressions are constantly lower, but ‘observed’ expressions are more varied (Figure 2A-C). In Figure 2A, we observed the binomial distribution of predicted expressions that can be partitioned by 1 RPKM (zero on the horizontal axis). We denote C1 for genes where predicted expression is ≥1 RPKM, C2 for the remains. The conspicuous frequency of C2 is also observed from Figure 2B-C. C2 genes in Figure 2C consist of 1205 up- and 1254 down-regulated genes. Further, the subset of C2 (C2′) where observed expression is greater than 1 RPKM consists of 148 up- and 159 down-regulated genes.\nTo characterize the gene classes, we analyzed TF-binding profiles and epigenetic modifications. As a result, in C2 genes, the number of peaks (Figure S3 in Additional file 2) and density profiles (Figure 2E) are apparently different, implying that the small number of TFs bind to distal regions from TSSs. C2 gene promoters are more methylated (Figure 3A). Remarkably, they tend to be absent from CpG islands (Figure 3B), and be marked with neither H2K4me3 nor bivalent domains (Figure 3C). Furthermore, we analyzed gene ontology terms of biological process by DAVID [23]. As a result, C1 was enriched for positive regulation of gene expression (score=8.66), whereas C2 was enriched for neural differentiation (score=34.49). C2′ was enriched for cell morphogenesis (score=2.77).\nEpigenetic modifications in ESC-specific genes. Three epigenetic states observed in genes whose expressions are 4-fold up or down in ESC against EB are considered. Gene class C1 and C2 are well explained and inefficiently explained genes by the regression analysis, respectively.\nC2 genes lack the TF-binding instances, implying less direct regulation by the core TFs. This depletion is due in part to excessive non-CpG DNA methylation [16]. Gene ontology analysis shows that C2 genes are often related to differentiation. Thus, they should be preferentially repressed in ESCs. Interestingly, as the histone marks are relatively rare among C2 genes, they are likely to be controlled by other regulatory pathways connecting to the maintenance of self-renewal. One possibility is the competitive binding of additional TFs not involved in this study because of the global open chromatin conformation in ESC [19]. Other possibilities include additional epigenetic patterns and homeostatic regulation, further investigations are required.\nIt was demonstrated previously that the absolute gene expression in ESCs is predictable by the ChIP-seq data of core TFs [21]. We also confirmed the high predictive power of the regression model. However, the results strongly rely on certain genes whose ‘predicted’ expressions are constantly lower, but ‘observed’ expressions are more varied (Figure 2A-C). In Figure 2A, we observed the binomial distribution of predicted expressions that can be partitioned by 1 RPKM (zero on the horizontal axis). We denote C1 for genes where predicted expression is ≥1 RPKM, C2 for the remains. The conspicuous frequency of C2 is also observed from Figure 2B-C. C2 genes in Figure 2C consist of 1205 up- and 1254 down-regulated genes. Further, the subset of C2 (C2′) where observed expression is greater than 1 RPKM consists of 148 up- and 159 down-regulated genes.\nTo characterize the gene classes, we analyzed TF-binding profiles and epigenetic modifications. As a result, in C2 genes, the number of peaks (Figure S3 in Additional file 2) and density profiles (Figure 2E) are apparently different, implying that the small number of TFs bind to distal regions from TSSs. C2 gene promoters are more methylated (Figure 3A). Remarkably, they tend to be absent from CpG islands (Figure 3B), and be marked with neither H2K4me3 nor bivalent domains (Figure 3C). Furthermore, we analyzed gene ontology terms of biological process by DAVID [23]. As a result, C1 was enriched for positive regulation of gene expression (score=8.66), whereas C2 was enriched for neural differentiation (score=34.49). C2′ was enriched for cell morphogenesis (score=2.77).\nEpigenetic modifications in ESC-specific genes. Three epigenetic states observed in genes whose expressions are 4-fold up or down in ESC against EB are considered. Gene class C1 and C2 are well explained and inefficiently explained genes by the regression analysis, respectively.\nC2 genes lack the TF-binding instances, implying less direct regulation by the core TFs. This depletion is due in part to excessive non-CpG DNA methylation [16]. Gene ontology analysis shows that C2 genes are often related to differentiation. Thus, they should be preferentially repressed in ESCs. Interestingly, as the histone marks are relatively rare among C2 genes, they are likely to be controlled by other regulatory pathways connecting to the maintenance of self-renewal. One possibility is the competitive binding of additional TFs not involved in this study because of the global open chromatin conformation in ESC [19]. Other possibilities include additional epigenetic patterns and homeostatic regulation, further investigations are required.\n[SUBTITLE] Epigenetic patterns improve the prediction of gene expression [SUBSECTION] To further understand the epigenetic effects in gene regulation, we add three epigenetic states to the regression models; histone mark (HistM), DNA methylation (Methy), and CpG island (CpGI). Thus, 14 explanatory variables are used. To identify effective variables in the prediction, we reduced the regression model by using the stepwise model selection. Also, 100 runs of computer simulation that randomly assign the epigenetic states were performed.\nAll models with the epigenetic effects improved CV-R2 with one to three more variables compared with the models without the epigenetic effects (Table 2). The additional variables are the epigenetic effect terms. The results of simulation support that the improvements are not by the chance. In particular, the density-based models with the epigenetic effects are significantly better when remapped peak datasets are used. Furthermore, overall regression coefficients gathered from all the density-based models in Table 2 show the relative importance of epigenetic effects except CpGI (Figure 2F). Note that the positive-biased activities are consistent with the previous study [24].\nEffects of epigenetic patterns in reduced regression models\nTo further understand the epigenetic effects in gene regulation, we add three epigenetic states to the regression models; histone mark (HistM), DNA methylation (Methy), and CpG island (CpGI). Thus, 14 explanatory variables are used. To identify effective variables in the prediction, we reduced the regression model by using the stepwise model selection. Also, 100 runs of computer simulation that randomly assign the epigenetic states were performed.\nAll models with the epigenetic effects improved CV-R2 with one to three more variables compared with the models without the epigenetic effects (Table 2). The additional variables are the epigenetic effect terms. The results of simulation support that the improvements are not by the chance. In particular, the density-based models with the epigenetic effects are significantly better when remapped peak datasets are used. Furthermore, overall regression coefficients gathered from all the density-based models in Table 2 show the relative importance of epigenetic effects except CpGI (Figure 2F). Note that the positive-biased activities are consistent with the previous study [24].\nEffects of epigenetic patterns in reduced regression models\n[SUBTITLE] TF interactions wired with epigenetic effects [SUBSECTION] To investigate the cooperative effects among TFs and epigenetic patterns in gene regulation, we exhaustively searched significant interaction terms from our regression model. First, a subset of ESC-specific genes that are co-bound by a specific TF pair is prepared. Then, the saturated model for the genes is constructed. The model involves 469 variables; 14 main effect terms (11 TFs and 3 epigenetic states) and 455 higher-order interaction terms (all the possible pairwise and triplewise interactions). Finally, our pipeline greedily identifies important variables (see methods). This procedure is independently performed with each of five peak datasets.\nIn total, 215 models were identified in which the predictive power is higher than the models without higher-order terms. These models contained 6-30 variables including at least one interactive term. As an example, the regression model for genes co-bound by Oct4 and Sox2, a well-known pluripotent complex [9,25], contained 15 terms and improved CV-R2=0.4126 from 0.3837 in the model with only 14 main effect terms. This model suggests that 7 interactive terms are important in the explanation of target gene expression. Among them, 3 terms are mediated by the epigenetic effects. The network representation of this model highlights the importance of signaling receptors (Stat3 and Smad1), activating Oct4/Sox2 complex [9] as well as Klf4/CpGI [26], and the interaction of Zfx/Methy newly found here (Figure 4).\nExample of regulatory network of TF interactions with epigenetic effects. This network was generated by the connectivity of nodes in all interaction terms. For example, an interaction term A:B:C is split into three interactions, A:B, A:C, and B:C. Then, the nodes are linked to each other and to the target gene set. A pairwise interaction with an epigenetic effect is treated differently. For example, in the case of B:Methy, B is not linked to the target gene set.\nWith considering the redundancy and conservativity, we represented the interactive terms of 215 models as a network (Figure S4 in Additional file 2). As a result, 19 gene sets covering approximately 86% of genes (3523 out of 4095 genes) were linked by 28 regulatory edges of the epigenetic effects that are commonly found in the five peak datasets (Figure S4 in Additional file 2). These results suggest that the cooperative interactions between TF and the epigenetic state are indispensable to explain the majority of gene expression in ESCs. In addition, we confirmed that the regression coefficients in Figure 2F are dramatically changed in the regression of given gene sets, and also CpGI significantly contributes to the prediction of gene expression (Figure 4).\nTo investigate the cooperative effects among TFs and epigenetic patterns in gene regulation, we exhaustively searched significant interaction terms from our regression model. First, a subset of ESC-specific genes that are co-bound by a specific TF pair is prepared. Then, the saturated model for the genes is constructed. The model involves 469 variables; 14 main effect terms (11 TFs and 3 epigenetic states) and 455 higher-order interaction terms (all the possible pairwise and triplewise interactions). Finally, our pipeline greedily identifies important variables (see methods). This procedure is independently performed with each of five peak datasets.\nIn total, 215 models were identified in which the predictive power is higher than the models without higher-order terms. These models contained 6-30 variables including at least one interactive term. As an example, the regression model for genes co-bound by Oct4 and Sox2, a well-known pluripotent complex [9,25], contained 15 terms and improved CV-R2=0.4126 from 0.3837 in the model with only 14 main effect terms. This model suggests that 7 interactive terms are important in the explanation of target gene expression. Among them, 3 terms are mediated by the epigenetic effects. The network representation of this model highlights the importance of signaling receptors (Stat3 and Smad1), activating Oct4/Sox2 complex [9] as well as Klf4/CpGI [26], and the interaction of Zfx/Methy newly found here (Figure 4).\nExample of regulatory network of TF interactions with epigenetic effects. This network was generated by the connectivity of nodes in all interaction terms. For example, an interaction term A:B:C is split into three interactions, A:B, A:C, and B:C. Then, the nodes are linked to each other and to the target gene set. A pairwise interaction with an epigenetic effect is treated differently. For example, in the case of B:Methy, B is not linked to the target gene set.\nWith considering the redundancy and conservativity, we represented the interactive terms of 215 models as a network (Figure S4 in Additional file 2). As a result, 19 gene sets covering approximately 86% of genes (3523 out of 4095 genes) were linked by 28 regulatory edges of the epigenetic effects that are commonly found in the five peak datasets (Figure S4 in Additional file 2). These results suggest that the cooperative interactions between TF and the epigenetic state are indispensable to explain the majority of gene expression in ESCs. In addition, we confirmed that the regression coefficients in Figure 2F are dramatically changed in the regression of given gene sets, and also CpGI significantly contributes to the prediction of gene expression (Figure 4).", "To minimize artifacts, we refined the binding signals of 12 core TFs in mouse ESC publicly available [9] (see methods). The ChIP-seq peak datasets generated by various tools are hereafter denoted as FP4_Bowtie, FP4_MAQ, and FP4_Soap2. Also, tag positions mapped by Eland [9] are used for the peak detection (FP4_Eland), and the peak data of Chen et al. is involved (Chen Eland). Thus, we prepared five peak datasets in total.\nDifferences in numbers and positions between the remapped data and the original data were investigated. As a result, relatively larger number of uniquely mapped tags and peaks were gained compared to the original data (Table S1-S3 in Additional file 1). In regard to peaks (Table 1), FP4 with the previously mapped tags (FP4_Eland) covers 85-98% of Chen_Eland, and the intensity of overlapped peaks is strongly correlated. Thus, it is deemed that FP4 has reproduced Chen_Eland and extended it with novel peaks in different genomic locations. In contrast, FP4 with remapped tags shows relatively lower reproducibility, whereas peak intensities are still correlated with Chen_Eland except Esrrb (Figure 1B). Similar observations can be found from an independent study [22].\nReproducibility of newly detected peaks\nFold change is the ratio of newly detected peak number over the original peak number. Overlaps of the original peaks to the newly detected peaks were investigated with 200-bp window. The overlapped peaks were used to calculate the correlation of peak intensity.\nSchematic representation of ChIP-seq data analysis and examples of TF-binding density profiles. (A) Definitions of a peak region and its intensity. (B) Example of intensity correlation between the original data and the reanalyzed data. (C) Density estimation from the genome-wide binding locations of a TF. (D)-(E) Example of density profiles in five peak datasets.\nThe reason why the numbers vary is twofold. First, algorithmic differences in alignment tools cause the different numbers, particularly due to the gapped or ungapped alignment and random indel for mismatches. Second, thresholds for the peak intensity to distinguish experimental noise are different (Table S4 in Additional file 1). That is, Chen et al. used qPCR refinement with small number of peaks, whereas we used Monte Carlo simulation on each chromosome.", "To assess the importance of TF bindings, Ouyang et al. [21] successfully applied a regression model to the prediction of absolute gene expression in mouse ESC. We first recover this study. Ouyang et al. used TF association strength (TFAS) by summing up peak intensities that are weighted exponentially according to the relative positions from TSSs. We applied the TFAS data to our simple linear regression model shown in equation (1), namely exponential-based regression model. The predictive power of our model is much higher (CV-R2=0.647) than Ouyang’s model (CV-R2=0.639), suggesting that the simple regression model is comparable to their PC-regression model (Figure 2A).\nPredictive results of density-based linear regression model. (A)-(C) Average correlation coefficient of 10-fold CV in three gene sets. (D) Comparative analysis of two models using ESC-specific gene subsets that co-bound by pluripotent TF pairs. Red solid circles are the cases that the density profiles of a TF pair are significantly different. (E) Distinct binding profiles in two gene classes that are either well explained (C1) or inefficiently explained (C2). (F) Contribution of each TF and epigenetic effects in density-based regression models (Methy: DNA methylation, HistM: histone mark, CpGI: CpG Island).\nNext, we prepared 17060 genes by removing inconsistency between Ouyang’s study and Chen’s study. This procedure is prerequisite for gathering precise TF-binding instances. TFAS data for the genes were calculated by the exactly same procedure of Ouyang et al. As a result, the exponential-based model shows CV-R2=0.495 with Chen_Eland. In contrast, CV-R2 increases to 0.542 (FP4_Eland), 0.587 (FP4_Bowtie), 0.581 (FP4_MAQ), and 0.590 (FP4_Soap2).\nThese results clearly suggest that the proposed simple linear regression model is applicable to the prediction. Furthermore, it has been demonstrated that the peak datasets we remapped give more information for explaining the gene expression.", "To investigate the characteristics of TF binding sites in ESC, the density profiles of TF-bindings are estimated from each of peak datasets (Figure 1C), then any two density profiles for a TF in different peak datasets are tested by Kolmogorov-Smirnov (KS) test. According to the KS test, the profiles of a TF are almost identical even if the number of mapped tags and peaks are largely different in, say, Esrrb (Figure 1D). The exceptional case is Sox2 in Chen_Eland and FP4_Bowtie (Figure 1E) due in part to the stringent filter used in Chen Eland; e.g. loss of Sox2 peaks in Chen Eland at gene clusters on chromosome X (Figure S1 in Additional file 2).\nImportantly, in the same peak dataset, the profiles are significantly different among TFs, e.g. Oct4 and Smad1 in FP4_Bowtie are shown in Figure S2. It is, therefore, thought that spatial preference of TF-bindings cannot be explained by one generalized distribution. In fact, the binding distributions of Nanog, Smad1, Sox2, and Stat3 definitely do not follow the exponential distribution (Figure S2 in Additional file 3).", "Our observations from the genome-wide distribution of TF binding sites revealed the distinct binding preference from exponential function (Figure 1E). Thus, we use the density profiles as predictors given as equation (2), which we call the density-based regression model. The predictive power of the density-based model with Chen_Eland (Figure 2B) is slightly higher (CV-R2=0.508) than the exponential-based model (CV-R2=0.495). Similar results were obtained when other peak datasets were used.\nWe suspect that the prediction quality of two regression models may depend on downstream genes that cause specific density profiles. To confirm it, we extracted 4095 ESC-specific genes. E2f1 was excluded here due to its excessive regression coefficient [21]. Then, a subset of 4095 genes that is co-bound by a TF pair was prepared. Since the TFs used are well-known essential regulators in ESCs, the TF pairs, such as Oct4 and Sox2, possibly play an important regulatory role in their downstream ESC-specific genes. All subsets by any combination of two TFs have been prepared.\nFigure 2D illustrates that the density-based regression model outperforms in many cases. Furthermore, 55 gene subsets that are co-bound by TF pairs whose density profiles are significantly different (p < 0.05) were successfully predicted (red solid circles in Figure 2D). These gene subsets cannot be modeled by a generalized exponential function. The results suggest that the spatial preferences of TF bindings are much more dynamically changed in ESC-specific gene subsets rather than observed from all the genes. This is why the density-based model improved the predictive power with respect to the generalized exponential-based model.", "It was demonstrated previously that the absolute gene expression in ESCs is predictable by the ChIP-seq data of core TFs [21]. We also confirmed the high predictive power of the regression model. However, the results strongly rely on certain genes whose ‘predicted’ expressions are constantly lower, but ‘observed’ expressions are more varied (Figure 2A-C). In Figure 2A, we observed the binomial distribution of predicted expressions that can be partitioned by 1 RPKM (zero on the horizontal axis). We denote C1 for genes where predicted expression is ≥1 RPKM, C2 for the remains. The conspicuous frequency of C2 is also observed from Figure 2B-C. C2 genes in Figure 2C consist of 1205 up- and 1254 down-regulated genes. Further, the subset of C2 (C2′) where observed expression is greater than 1 RPKM consists of 148 up- and 159 down-regulated genes.\nTo characterize the gene classes, we analyzed TF-binding profiles and epigenetic modifications. As a result, in C2 genes, the number of peaks (Figure S3 in Additional file 2) and density profiles (Figure 2E) are apparently different, implying that the small number of TFs bind to distal regions from TSSs. C2 gene promoters are more methylated (Figure 3A). Remarkably, they tend to be absent from CpG islands (Figure 3B), and be marked with neither H2K4me3 nor bivalent domains (Figure 3C). Furthermore, we analyzed gene ontology terms of biological process by DAVID [23]. As a result, C1 was enriched for positive regulation of gene expression (score=8.66), whereas C2 was enriched for neural differentiation (score=34.49). C2′ was enriched for cell morphogenesis (score=2.77).\nEpigenetic modifications in ESC-specific genes. Three epigenetic states observed in genes whose expressions are 4-fold up or down in ESC against EB are considered. Gene class C1 and C2 are well explained and inefficiently explained genes by the regression analysis, respectively.\nC2 genes lack the TF-binding instances, implying less direct regulation by the core TFs. This depletion is due in part to excessive non-CpG DNA methylation [16]. Gene ontology analysis shows that C2 genes are often related to differentiation. Thus, they should be preferentially repressed in ESCs. Interestingly, as the histone marks are relatively rare among C2 genes, they are likely to be controlled by other regulatory pathways connecting to the maintenance of self-renewal. One possibility is the competitive binding of additional TFs not involved in this study because of the global open chromatin conformation in ESC [19]. Other possibilities include additional epigenetic patterns and homeostatic regulation, further investigations are required.", "To further understand the epigenetic effects in gene regulation, we add three epigenetic states to the regression models; histone mark (HistM), DNA methylation (Methy), and CpG island (CpGI). Thus, 14 explanatory variables are used. To identify effective variables in the prediction, we reduced the regression model by using the stepwise model selection. Also, 100 runs of computer simulation that randomly assign the epigenetic states were performed.\nAll models with the epigenetic effects improved CV-R2 with one to three more variables compared with the models without the epigenetic effects (Table 2). The additional variables are the epigenetic effect terms. The results of simulation support that the improvements are not by the chance. In particular, the density-based models with the epigenetic effects are significantly better when remapped peak datasets are used. Furthermore, overall regression coefficients gathered from all the density-based models in Table 2 show the relative importance of epigenetic effects except CpGI (Figure 2F). Note that the positive-biased activities are consistent with the previous study [24].\nEffects of epigenetic patterns in reduced regression models", "To investigate the cooperative effects among TFs and epigenetic patterns in gene regulation, we exhaustively searched significant interaction terms from our regression model. First, a subset of ESC-specific genes that are co-bound by a specific TF pair is prepared. Then, the saturated model for the genes is constructed. The model involves 469 variables; 14 main effect terms (11 TFs and 3 epigenetic states) and 455 higher-order interaction terms (all the possible pairwise and triplewise interactions). Finally, our pipeline greedily identifies important variables (see methods). This procedure is independently performed with each of five peak datasets.\nIn total, 215 models were identified in which the predictive power is higher than the models without higher-order terms. These models contained 6-30 variables including at least one interactive term. As an example, the regression model for genes co-bound by Oct4 and Sox2, a well-known pluripotent complex [9,25], contained 15 terms and improved CV-R2=0.4126 from 0.3837 in the model with only 14 main effect terms. This model suggests that 7 interactive terms are important in the explanation of target gene expression. Among them, 3 terms are mediated by the epigenetic effects. The network representation of this model highlights the importance of signaling receptors (Stat3 and Smad1), activating Oct4/Sox2 complex [9] as well as Klf4/CpGI [26], and the interaction of Zfx/Methy newly found here (Figure 4).\nExample of regulatory network of TF interactions with epigenetic effects. This network was generated by the connectivity of nodes in all interaction terms. For example, an interaction term A:B:C is split into three interactions, A:B, A:C, and B:C. Then, the nodes are linked to each other and to the target gene set. A pairwise interaction with an epigenetic effect is treated differently. For example, in the case of B:Methy, B is not linked to the target gene set.\nWith considering the redundancy and conservativity, we represented the interactive terms of 215 models as a network (Figure S4 in Additional file 2). As a result, 19 gene sets covering approximately 86% of genes (3523 out of 4095 genes) were linked by 28 regulatory edges of the epigenetic effects that are commonly found in the five peak datasets (Figure S4 in Additional file 2). These results suggest that the cooperative interactions between TF and the epigenetic state are indispensable to explain the majority of gene expression in ESCs. In addition, we confirmed that the regression coefficients in Figure 2F are dramatically changed in the regression of given gene sets, and also CpGI significantly contributes to the prediction of gene expression (Figure 4).", "ESCs are the widely accepted source for the study of many biological principles. Despite recent advances in our understanding of biological systems, the gene regulation in ESCs is only incompletely understood. To explore the regulatory mechanism underlying in ESCs, we constructed a predictive model for explaining the absolute gene expression in mouse ESC. This model uses a novel density-based approach to exploit the recent massive parallel sequencing straightforwardly.\nWe first reanalyzed the publicly available ChIP-seq data for 12 well-known pluripotent core TFs [9], and retrieved the reproduced and extended TF-binding sites and intensities (Table 1). Using our regression model based on the exponential function [21], we found that the remapped peaks are more informative to explain the gene expression (Table 2). Therefore, we concluded that the algorithmic differences in computer tools for ChIP-seq data significantly affect the downstream analysis. Analyzing the heterogeneous peak datasets in a comparative manner, we found that the spatial binding preference of each TF is well conserved in all the datasets, whereas the preferences of TFs in a dataset are significantly different from each other (Figure 1D-E). These results imply that density profiles are better explanatory variables than the generalized exponential function. In fact, the predictive power of density-based model is constantly higher than the exponential-based model (Figure 2A-C, Table 2). Even if the density profiles are dynamically changed in certain downstream genes, the proposed model is still outstanding (Figure 2D).\nUnexpectedly, we found two gene classes that are either well explained or inefficiently explained by the regression model. The latter class genes have less binding instances of the pluripotent TFs (Figure 2E), possibly related to excessive DNA methylation (Figure 3A). The gene classes show apparently different characteristics in epigenetic modifications (Figure 3), suggesting that they are likely to be under control in different regulatory mechanisms. In the present study, we simply combined the discrete epigenetic states with the powerful density-based model. This model significantly improved the predictive power (Table 2). Investigating higher-order interactions among the predictors, we found that the cooperative interactions between TF and epigenetic pattern are indispensable for regulating approximately 86% of ESC-specific genes (Figure S4 in Additional file 2). These results suggest that the relative importance of epigenetic effects to regulate the gene expression in ESCs, supporting the general idea [14,15].\nWe proposed a powerful regression model, and uncovered the relative importance of epigenetic regulation in ESCs. Overall prediction quality is still insufficient. As future works, comprehensive representation of epigenetic patterns is required, and additional or alternative TFs in ESCs should be considered.", "[SUBTITLE] Data acquisition [SUBSECTION] [SUBTITLE] ChIP-seq data and gene expression [SUBSECTION] Raw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.\nRaw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.\n[SUBTITLE] Epigenetic modifications [SUBSECTION] DNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.\nDNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.\n[SUBTITLE] ChIP-seq data and gene expression [SUBSECTION] Raw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.\nRaw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.\n[SUBTITLE] Epigenetic modifications [SUBSECTION] DNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.\nDNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.\n[SUBTITLE] Estimation of TF binding density [SUBSECTION] Given a genome-wide location map of a TF-bindings, all peak positions were converted to relative positions to the nearest TSSs. Gaussian kernel density function (bandwidth=300 bps) estimated the density profile of the TF-bindings within ±50K bps. The profile was normalized into range of [0, 1] by dividing by the maximal density height.\nGiven a genome-wide location map of a TF-bindings, all peak positions were converted to relative positions to the nearest TSSs. Gaussian kernel density function (bandwidth=300 bps) estimated the density profile of the TF-bindings within ±50K bps. The profile was normalized into range of [0, 1] by dividing by the maximal density height.\n[SUBTITLE] Regression model [SUBSECTION] We use a multivariate regression model(1)\nwhere Yi is the expression of gene i, Sij is the score of the jth TF on gene i, wj is the regression coefficient of the jth TF, and ei is the error term. The score Sij is given by(2)\nwhere gk is the perk intensity of the kth binding peak of the jth TF, Fj is the normalized density function for the jth TF, and lk is the relative position of the kth peak to TSS of gene i. Note that a small value is added to Yi for the logarithm.\nWe use a multivariate regression model(1)\nwhere Yi is the expression of gene i, Sij is the score of the jth TF on gene i, wj is the regression coefficient of the jth TF, and ei is the error term. The score Sij is given by(2)\nwhere gk is the perk intensity of the kth binding peak of the jth TF, Fj is the normalized density function for the jth TF, and lk is the relative position of the kth peak to TSS of gene i. Note that a small value is added to Yi for the logarithm.\n[SUBTITLE] Adding epigenetic effects [SUBSECTION] Discrete values representing epigenetic states of a gene i are added to the regression model(3)\nwhere H is the type of histone mark (neither mark=1.0, H3K27me3=2.0, bivalent mark=3.0, H3K4me3=4.0), M is the DNA methylation (no annotation=1.0, methylation=2.0, unmethylation=3.0), C is the CpG island (absence=1.0, presence=2.0), and α, β, γ are the regression coefficients for H, M, C, respectively.\nDiscrete values representing epigenetic states of a gene i are added to the regression model(3)\nwhere H is the type of histone mark (neither mark=1.0, H3K27me3=2.0, bivalent mark=3.0, H3K4me3=4.0), M is the DNA methylation (no annotation=1.0, methylation=2.0, unmethylation=3.0), C is the CpG island (absence=1.0, presence=2.0), and α, β, γ are the regression coefficients for H, M, C, respectively.\n[SUBTITLE] Fitting and reducing regression models [SUBSECTION] Explanatory variables in a regression model are log-transformed and quantile-normalized. 10 runs of 10-fold cross validation (CV) measure the average correlation coefficient (CV-R) and the average proportion of variation explained by the model (CV-R2). The stepwise model selection is done by stepAIC in R language with the backward and forward procedure. The regression model with higher-order interactions are reduced by a pipeline developed in house; ANOVA in R language first diagnoses the significance of each explanatory variable in the given saturated model. Next, significant variables (p < 0.05 in F-test) are gathered. Finally, the best model is constructed by adding and removing the collected variables one by one in increasing order of p-value until CV-R2 is not improved anymore.\nExplanatory variables in a regression model are log-transformed and quantile-normalized. 10 runs of 10-fold cross validation (CV) measure the average correlation coefficient (CV-R) and the average proportion of variation explained by the model (CV-R2). The stepwise model selection is done by stepAIC in R language with the backward and forward procedure. The regression model with higher-order interactions are reduced by a pipeline developed in house; ANOVA in R language first diagnoses the significance of each explanatory variable in the given saturated model. Next, significant variables (p < 0.05 in F-test) are gathered. Finally, the best model is constructed by adding and removing the collected variables one by one in increasing order of p-value until CV-R2 is not improved anymore.", "[SUBTITLE] ChIP-seq data and gene expression [SUBSECTION] Raw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.\nRaw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.\n[SUBTITLE] Epigenetic modifications [SUBSECTION] DNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.\nDNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.", "Raw tag sequences and a control library were downloaded from GEO database (GSE11431). High-quality 26 base pair (bp) tags that have less than three ambiguous bases were mapped to mm8 by Bowtie [27], MAQ [28], and Soap2 [29] with allowing two mismatches. Only uniquely mapped tags were extended to 200-bp virtual fragments (Figure 1A). FP4 (FindPeaks 4.0) [30] detected significant peak regions. Monte Carlo simulation was performed on each chromosome to calculate false discovery rate (FDR). Also, the fold enrichment of tags in each peak region over remapped control tags was measured. Finally, we prepared peaks by criteria, FDR <5% and 5-fold enrichment.\nFor the absolute gene expression, the number of tags per kilobase of exon region per million mapped tags (RPKM) [18] for 18936 mouse genes in ESC and in embryoid body (EB) were prepared from [21]. Positional information of transcription start sites (TSSs) of 17443 Refseq genes in mm8 were prepared from [9]. Removing inconsistent gene IDs between RPKM data and TSS data, we compiled 17060 genes. We prepared 4095 ESC-specific genes whose expressions are 4-fold up- or down-regulated in ESC over EB. The dataset used in here is available at http://www.hgc.jp/~park/research/.", "DNA methylation maps are prepared from two datasets that use different high-throughput detection methods [16,17]. Methylation states of high-CpG-density promoters (GC content ≥0.55) are defined by mean methylation levels; unmethylated if mean ≤0.25, methylated if mean ≥0.75. The genome-wide distribution of CpG islands and histone mark were downloaded from UCSC genome browser. We consider three histone states; histone H3 lysine 4 trimethylation (H3K4me3), an active mark of expression, H3 lysine 27 trimethylation (H3K27me3), a repressive mark, and bivalent domain of H3K4me3 and H3K27me3, a ‘poised’ mark of expression.", "Given a genome-wide location map of a TF-bindings, all peak positions were converted to relative positions to the nearest TSSs. Gaussian kernel density function (bandwidth=300 bps) estimated the density profile of the TF-bindings within ±50K bps. The profile was normalized into range of [0, 1] by dividing by the maximal density height.", "We use a multivariate regression model(1)\nwhere Yi is the expression of gene i, Sij is the score of the jth TF on gene i, wj is the regression coefficient of the jth TF, and ei is the error term. The score Sij is given by(2)\nwhere gk is the perk intensity of the kth binding peak of the jth TF, Fj is the normalized density function for the jth TF, and lk is the relative position of the kth peak to TSS of gene i. Note that a small value is added to Yi for the logarithm.", "Discrete values representing epigenetic states of a gene i are added to the regression model(3)\nwhere H is the type of histone mark (neither mark=1.0, H3K27me3=2.0, bivalent mark=3.0, H3K4me3=4.0), M is the DNA methylation (no annotation=1.0, methylation=2.0, unmethylation=3.0), C is the CpG island (absence=1.0, presence=2.0), and α, β, γ are the regression coefficients for H, M, C, respectively.", "Explanatory variables in a regression model are log-transformed and quantile-normalized. 10 runs of 10-fold cross validation (CV) measure the average correlation coefficient (CV-R) and the average proportion of variation explained by the model (CV-R2). The stepwise model selection is done by stepAIC in R language with the backward and forward procedure. The regression model with higher-order interactions are reduced by a pipeline developed in house; ANOVA in R language first diagnoses the significance of each explanatory variable in the given saturated model. Next, significant variables (p < 0.05 in F-test) are gathered. Finally, the best model is constructed by adding and removing the collected variables one by one in increasing order of p-value until CV-R2 is not improved anymore.", "The authors declare that they have no competing interests.", "Extended analysis of ChIP-seq data This file provides tables including the summary of tag mapping (Table S1), the fold change of remapped tags over the original data (Table S2), the number of peaks in five datasets (Table S3), and the thresholds used to detect significant peaks (Table S4).\nClick here for file\nComprehensive analysis of gene regulation in mouse ESC This file provides figures including an example of peak distributions (Figure S1), TF-binding instances in two gene classes (Figure S3), and the regulatory network of TFs wired with epigenetic effects (Figure S4).\nClick here for file\nDensity profile of 12 TFs This file provides the density profiles of 12 core TFs in five peak datasets (Figure S2).\nClick here for file" ]

[ null, null, null, null, null, null, null, null, null, null, "methods", null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Computational Biology", "Databases, Factual", "Humans", "Information Storage and Retrieval", "Internet", "Microarray Analysis", "Oligonucleotide Array Sequence Analysis", "Semantics" ]

[ "Background", "RDF representation of TMA and DNA microarray data", "Source databases", "Ontology design and data representation as RDF", "System architecture", "Semantically-enabled hypothesis testing", "Hypothesis model", "SPARQL generation", "Statistical test for Fisher’s exact test", "Use case of integrated TMA and DNA microarray database", "Results", "Xperanto-RDF", "Applications for hypothesis testing", "Use case using integrated TMA and DNA microarray database", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Biological databases are collections of scientific experiments, published literatures, and computational analyses organized under a specialized scheme. Biological databases became essential resources to biologists in their daily researches by providing information about biological facts and experimental results and procedures and also by providing management tools for the obtained data. Because these biological databases are designed for specific purposes, and independently managed, and metadata are not provided in many cases, they are neither semantically explicit nor interoperable. To overcome these problems, semantic web technologies such as Resource Description Framework (RDF), Web Ontology Language (OWL) and SPARQL (SPARQL Protocol and RDF Query Language) have been actively accepted in the field of life science for new database design [1-6]. Semantic web repositories are more advantageous than relational databases (RDBs) because metadata are more complete and standardized [7]. Representation of data as RDF makes biological entities semantically explicit and clear so that various tasks can be performed without extensive human interventions. These tasks includes integration of heterogeneous data, applying logic to infer new insights, and publication and sharing of biological findings and models [8]. Several current biological databases provide integrated data structure for knowledge management by applying semantic web technologies [1-4,6,9].\nIn spite of the benefits of semantic web technologies, these databases cannot directly answer to biologists for biologically meaningful questions or hypotheses. For machines to answer these questions, the process of inference based on either logical or statistical relationships of stored data is required. The inference by description logic is a part of semantic web technologies, but statistical inference was not implemented in the current semantic web technologies. These problems, therefore, cannot be solved within the framework of semantic web technologies alone and are rather dependent on the design of an application. However semantic web technologies are still beneficial in those applications.\nTo prove a biological hypothesis, 1) an experiment is designed to test the hypothesis, 2) the experimental data is gathered, and 3) the data is tested by statistical test(s). Because of the increasing amount of high throughput experimental data in biological databases, there is an increasing need of high throughput validation of biological hypotheses. To implement such an application, 1) a hypothesis given by a user should be semantically interpreted, 2) the relevant experimental data should be retrieved from the database, and then 3) the hypothesis should be statistically tested against the retrieved data.\nBesides the fact that tissue microarray (TMA) is being widely used as a high throughput validation tool for the large number of data-driven hypotheses from other genomic technologies, TMA databases is a good candidate for the proof of concept of above mentioned applications. First, most biological hypotheses that can be derived from only TMA experiment are syntactically simple. The hypothesis derived from TMA experiments can be stated as, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Basically a TMA experiment is designed to test dependency between two entities and a hypothesis about the mechanisms of the interactions between two entities cannot be tested unless relevant additional information is provided. There are two important biological entities in TMA, biological samples and markers. In the TMA-validated hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype” [10], for example, ‘colorectal cancer’ is biological condition, and “reduced Apaf-1 expression” and “high-grade phenotype” are entities (B and C). Therefore a dependency-stating hypothesis in a TMA experiment can be considered as a triadic predicate with condition A, entity B, and entity C as parameters. If these parameters are provided, a query for relevant data retrieval can be generated through them. Second, only a few kinds of statistical tests are generally used to test hypothesis in TMA experiments. The main purpose of TMA experiment is to test the statistical relationships between biological entities (or markers) in a population of samples with identical biological condition. The results are determined by the size of the population conforming to the given hypothesis. If more samples show positive relationships with the hypothesis, the hypothesis is more likely to be true.\nFisher’s exact and χ2 test are the most frequently used statistical tests in TMA experiments to test dependency. They are frequently used because most of clinical or histological parameters (e.g., history of hypertension, tumour grade, etc) and the extent or the intensity of the marker expression in a tissue (e.g., 0, 1, 2, 3) have discrete or categorical values and the dependencies between these values are tested by them. For example, if we want to test the above mentioned hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype.” by Fisher’s exact test, we have to investigate each number of cores in slides for four exclusive conditions : a) negative Apaf-1 expression and high-grade phenotype, b) negative Apaf-1 expression and low-grade phenotype, c) weak to strong Apaf-1 expression and high-grade phenotype, and d) weak to strong Apaf-1 expression and low-grade phenotype. Then these four parameters are used for Fisher’s exact or χ2 test to test negative association between Apaf-1 expression and histological grade.\nIn spite of these benefits of TMA database, if it were not semantically explicit, applications for hypothesis testing could not be implemented. TMA data have complex and wide range of semantics, including information for clinical and histopathological features and large amount of metadata should be provided. Semantic web technologies support richer semantics than traditional RDB-based models. It, therefore, is more desirable that the databases for applications for hypothesis testing should be represented as RDF. In addition, SPARQL as a query language is more intuitively understandable to biologists [11]. Lastly, integration with the other databases, including other TMA and DNA microarray databases were considered in the present study and the databases using semantic web technologies are more advantageous in integration.\nWe have created and managed Xperanto-TMA, a web-based TMA database supporting TMA-OM (Tissue Microarray Object Model) [12] and TMA-TAB [13]. Xperanto-TMA uses RDB because technologies supporting object-oriented models were not mature enough to guarantee high performance [12]. Due to the well organized object model, however, clear and rich semantic relationships between entities are well supported. We have also developed and managed a web-based DNA microarray database called Xperanto [14].\nWhen implementing Xperanto-TMA and Xperanto, we extensively investigated semantics of experimental data of TMA and DNA microarray. This experience gave us good resources in constructing semantically rich and explicit TMA and DNA microarray database represented as RDF. Because Xperanto-TMA and Xperanto were independently developed, it was difficult to integrate these two databases. The integration of gene expression microarray and TMA data is a powerful approach to molecular profiling of human cancer [15]. Semantic web gives us opportunity to integrate these two databases through the use of ontology [9]. After integration was achieved, we developed a use case where the usefulness of the integration was found.\nIn this study, we present how we designed semantic TMA database from the existing relational one and then how applications let researchers test hypotheses using semantic web infrastructures. Marker-level integration of TMA and DNA microarray databases will be also described.", "[SUBTITLE] Source databases [SUBSECTION] We represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.\nWe represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.\n[SUBTITLE] Ontology design and data representation as RDF [SUBSECTION] To provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.\nTo provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.\n[SUBTITLE] System architecture [SUBSECTION] In the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF\nIn the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF", "We represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.", "To provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.", "In the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF", "[SUBTITLE] Hypothesis model [SUBSECTION] We investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.\nWe investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.\n[SUBTITLE] SPARQL generation [SUBSECTION] Four SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).\nFour SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).\n[SUBTITLE] Statistical test for Fisher’s exact test [SUBSECTION] After SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.\nAfter SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.", "We investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.", "Four SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).", "After SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.", "We designed a use case using integrated DNA and TMA database, which was to find the distributions of the histologies among the cores of TMA where the intensities of the antibodies corresponding to the markers of interest in DNA microarray were high or low", "[SUBTITLE] Xperanto-RDF [SUBSECTION] As described above, Xperanto-RDF was implemented based on existing TMA and DNA microarray databases (http://clara.snubi.org/Xperanto-RDF). Data from these two databases were represented as RDF triples by a mediator system such that users could have the benefits of semantic web technologies as well as those of RDB.\nAs described above, Xperanto-RDF was implemented based on existing TMA and DNA microarray databases (http://clara.snubi.org/Xperanto-RDF). Data from these two databases were represented as RDF triples by a mediator system such that users could have the benefits of semantic web technologies as well as those of RDB.\n[SUBTITLE] Applications for hypothesis testing [SUBSECTION] A hypothesis model testing dependency can be constructed in the index page of Xperanto-RDF as shown in Fig. 4 (http://clara.snubi.org/Xperanto-RDF). In the context field, a user can choose the shared property among the samples. The attribute of the classifier and the dependent property can be selected by tracking along the tree structure of the entity. Once the classifier or the dependent property is selected, the permissible value set of the attribute is delivered to the select box to the right side of the tree frame. Using the buttons beside, the user can select hypothesis-describing and complementary value sets. Then the relevant SPARQL is generated by the SPARQL generator. On submitting the SPARQLs, the applications perform the query against the stored TMA data, the result sets are delivered to the statistical module, and finally the result of the statistical test including p-value is displayed.\nWeb interface to generate SPARQLs by selecting factors and their value sets\nAs a result of hypothesis testing for the hypothesis in Fig. 3, we found the following results from the TMA database: 26 cores with Apaf-1 = 0 and high-grade phenotype, 4 cores with Apaf-1 ≥ 1 and high-grade phenotype, 6 core with Apaf-1 = 0 and low-grade phenotype, and 19 cores with Apaf-1 ≥ 1 and low-grade phenotype. These results support the given hypothesis because p value for Fisher’s exact text for these data was less than 0.0001.\nTable 1 shows the testing results of all of four hypotheses. P value of the testing result determines whether the hypothesis is accepted or rejected according to Xperanto-RDF. With 0.01 of P value as a threshold for the rejection of null hypothesis, all of four hypotheses were accepted in these procedures. These results indicate that TMA data in Xperanto-RDF supports all of these hypotheses.\nA hypothesis model testing dependency can be constructed in the index page of Xperanto-RDF as shown in Fig. 4 (http://clara.snubi.org/Xperanto-RDF). In the context field, a user can choose the shared property among the samples. The attribute of the classifier and the dependent property can be selected by tracking along the tree structure of the entity. Once the classifier or the dependent property is selected, the permissible value set of the attribute is delivered to the select box to the right side of the tree frame. Using the buttons beside, the user can select hypothesis-describing and complementary value sets. Then the relevant SPARQL is generated by the SPARQL generator. On submitting the SPARQLs, the applications perform the query against the stored TMA data, the result sets are delivered to the statistical module, and finally the result of the statistical test including p-value is displayed.\nWeb interface to generate SPARQLs by selecting factors and their value sets\nAs a result of hypothesis testing for the hypothesis in Fig. 3, we found the following results from the TMA database: 26 cores with Apaf-1 = 0 and high-grade phenotype, 4 cores with Apaf-1 ≥ 1 and high-grade phenotype, 6 core with Apaf-1 = 0 and low-grade phenotype, and 19 cores with Apaf-1 ≥ 1 and low-grade phenotype. These results support the given hypothesis because p value for Fisher’s exact text for these data was less than 0.0001.\nTable 1 shows the testing results of all of four hypotheses. P value of the testing result determines whether the hypothesis is accepted or rejected according to Xperanto-RDF. With 0.01 of P value as a threshold for the rejection of null hypothesis, all of four hypotheses were accepted in these procedures. These results indicate that TMA data in Xperanto-RDF supports all of these hypotheses.\n[SUBTITLE] Use case using integrated TMA and DNA microarray database [SUBSECTION] Xperanto-RDF achieved marker-mediated integration of TMA and DNA microarray data by mapping of DNA probes into antibodies in TMA experiment. By the use of mapping data, we could execute some of semantically integrated queries against data from TMA and DNA microarray experiments. A use case for this system was presented here.\nQuery: In an experiment for glioblastoma using Affymetrix HT Human Genome U133A array, the signal from the probe, “201983_s_at” were interesting to a researcher [17]. He wants to know the distribution of histologies of the samples where the corresponding antibody to the probe was highly expressed (intensity > 1) from our TMA database.\nIf TMA and DNA microarray database management systems had not been semantically integrated, it would have taken many steps to execute the queries and furthermore we needed external data source that showed what the corresponding antibody to the “201983_s_at” probe is. In Xperanto-RDF, the queries can be formulated as in Fig. 5. The query results show that the corresponding antibody is EGFR and the most predominant histological type among the samples where EGFR is highly expressed in TMA is glioblastoma according the data in Xperanto-RDF.\nAn integrated query against TMA and DNA microarray database This query returns a list of histological type of the samples which show strong intensity (>1) for the corresponding antibody to the mRNA probe, “201983_s_at”. The corresponding antibody is EGFR.\nXperanto-RDF achieved marker-mediated integration of TMA and DNA microarray data by mapping of DNA probes into antibodies in TMA experiment. By the use of mapping data, we could execute some of semantically integrated queries against data from TMA and DNA microarray experiments. A use case for this system was presented here.\nQuery: In an experiment for glioblastoma using Affymetrix HT Human Genome U133A array, the signal from the probe, “201983_s_at” were interesting to a researcher [17]. He wants to know the distribution of histologies of the samples where the corresponding antibody to the probe was highly expressed (intensity > 1) from our TMA database.\nIf TMA and DNA microarray database management systems had not been semantically integrated, it would have taken many steps to execute the queries and furthermore we needed external data source that showed what the corresponding antibody to the “201983_s_at” probe is. In Xperanto-RDF, the queries can be formulated as in Fig. 5. The query results show that the corresponding antibody is EGFR and the most predominant histological type among the samples where EGFR is highly expressed in TMA is glioblastoma according the data in Xperanto-RDF.\nAn integrated query against TMA and DNA microarray database This query returns a list of histological type of the samples which show strong intensity (>1) for the corresponding antibody to the mRNA probe, “201983_s_at”. The corresponding antibody is EGFR.", "As described above, Xperanto-RDF was implemented based on existing TMA and DNA microarray databases (http://clara.snubi.org/Xperanto-RDF). Data from these two databases were represented as RDF triples by a mediator system such that users could have the benefits of semantic web technologies as well as those of RDB.", "A hypothesis model testing dependency can be constructed in the index page of Xperanto-RDF as shown in Fig. 4 (http://clara.snubi.org/Xperanto-RDF). In the context field, a user can choose the shared property among the samples. The attribute of the classifier and the dependent property can be selected by tracking along the tree structure of the entity. Once the classifier or the dependent property is selected, the permissible value set of the attribute is delivered to the select box to the right side of the tree frame. Using the buttons beside, the user can select hypothesis-describing and complementary value sets. Then the relevant SPARQL is generated by the SPARQL generator. On submitting the SPARQLs, the applications perform the query against the stored TMA data, the result sets are delivered to the statistical module, and finally the result of the statistical test including p-value is displayed.\nWeb interface to generate SPARQLs by selecting factors and their value sets\nAs a result of hypothesis testing for the hypothesis in Fig. 3, we found the following results from the TMA database: 26 cores with Apaf-1 = 0 and high-grade phenotype, 4 cores with Apaf-1 ≥ 1 and high-grade phenotype, 6 core with Apaf-1 = 0 and low-grade phenotype, and 19 cores with Apaf-1 ≥ 1 and low-grade phenotype. These results support the given hypothesis because p value for Fisher’s exact text for these data was less than 0.0001.\nTable 1 shows the testing results of all of four hypotheses. P value of the testing result determines whether the hypothesis is accepted or rejected according to Xperanto-RDF. With 0.01 of P value as a threshold for the rejection of null hypothesis, all of four hypotheses were accepted in these procedures. These results indicate that TMA data in Xperanto-RDF supports all of these hypotheses.", "Xperanto-RDF achieved marker-mediated integration of TMA and DNA microarray data by mapping of DNA probes into antibodies in TMA experiment. By the use of mapping data, we could execute some of semantically integrated queries against data from TMA and DNA microarray experiments. A use case for this system was presented here.\nQuery: In an experiment for glioblastoma using Affymetrix HT Human Genome U133A array, the signal from the probe, “201983_s_at” were interesting to a researcher [17]. He wants to know the distribution of histologies of the samples where the corresponding antibody to the probe was highly expressed (intensity > 1) from our TMA database.\nIf TMA and DNA microarray database management systems had not been semantically integrated, it would have taken many steps to execute the queries and furthermore we needed external data source that showed what the corresponding antibody to the “201983_s_at” probe is. In Xperanto-RDF, the queries can be formulated as in Fig. 5. The query results show that the corresponding antibody is EGFR and the most predominant histological type among the samples where EGFR is highly expressed in TMA is glioblastoma according the data in Xperanto-RDF.\nAn integrated query against TMA and DNA microarray database This query returns a list of histological type of the samples which show strong intensity (>1) for the corresponding antibody to the mRNA probe, “201983_s_at”. The corresponding antibody is EGFR.", "We presented how a semantic database for TMA and DNA microarray experiment can be developed from the existing RDB implementations using currently available semantic web technologies, how the application for hypothesis testing using TMA experimental data was designed and used, and how TMA database can be integrated with DNA microarray database using semantic web technologies.\nWe successfully implemented applications for hypothesis testing and their usefulness was validated by testing several experimentally validated biological hypotheses by TMA experiments. Although the applications seems useful to biologists, it is still challenging to spread the idea of the hypothesis testing in the biological database using experimental data to the whole community of the biologists. In traditional biomedical research, an experiment is designed to test a hypothesis [18] and it has been believed that only a highly controlled experiment could decide whether the hypothesis was true or false. The value of hypothesis testing in the biological database may be questionable from this point of view. Despite the presence of the weakness, the future of hypothesis testing using biological databases seems bright. More experiments are getting open to the public with detailed metadata about experiments and the experiments tend to be industrialized and standardized. As a result, it seems highly probable that a researcher can find most of the experiments they want from the biological databases in the near future. At present these applications can be coupled with traditional biomedical research in a cost effective way. A researcher can use these applications as a preliminary investigation before the highly controlled experiment is begun.\nThe applications for hypothesis testing can be regarded as a semantically-enhanced meta-analysis tool for TMA experiments in one aspect. One of requirements to perform meta-analysis is the process of data normalization. Our applications do not have the process of data normalization yet because there are no known methods of normalization in TMA experiments. The absence of normalization process do not appear to be greatly influential on the experimental results at present because TMA experiments are usually interpreted by human eyes, crude categorical values are used in the interpretation, and experimental procedures and interpretations are rather standardized.\nWe separated the process of hypothesis model construction, SPARQL execution, and statistical test when implementing the applications for hypothesis testing. This allows us more flexibility to easily manage and develop the whole systems. More statistical modules will be easily implemented to our applications if a hypothesis cannot be tested with current statistical modules. We will implement a statistical module for survival analysis soon.\nPost et al. suggested integrative bioinformatics experimentation, defined as in silico data integration experiment using semantic web technologies, and made a use case to prove a hypothesis [7]. They tested a hypothesis stating that histone modification and gene expression regulation through transcription factor binding sites is correlated, by integrating histone modification data and transcription factor binding sites data derived from the public database using semantic web technologies [7]. Their integrative bioinformatics experimentation cycle consisted of hypothesis definition, experimental design, data integration, extension of data integration experiment, and data analysis and interpretation. Their study was focused on the applicability of semantic web technologies to data integration phase. Compared to this system, our applications process all the procedures to test the hypothesis after the hypothesis is inserted. To the user, the applications give impression as if the machine directly answered for the questions. We believe that we can implement the applications for hypothesis testing because we used TMA experiment data rather than genomic data and we targeted at the syntactically simple hypotheses composed of three factors. We believe that we can develop applications for testing for more complicated hypotheses based on this experience by treating complicated hypotheses as logical combinations of multiple simple hypotheses.\nFrom the beginning our TMA database was developed with consideration on the integration with the other TMA or omics databases. Any TMA database based on TMA-OM can easily exchange the experimental data and the direct implementation of our application for hypothesis test is possible.\nWe also achieved marker-mediated integration of TMA and DNA microarray database. The integration enabled marker-mediated bidirectional data flow between TMA and DNA microarray experimental data. With these integrated databases, the results of DNA microarray experiments can be validated in TMA experiments.\nOur future works are targeted in three directions. First, we are developing hypothesis miner extracting biological hypotheses directly from biomedical literature databases. The miner will feed hypotheses to TMA databases without the insertion procedures by the user. When realized, the integration of literature and experimental databases mediated by the hypotheses can be accomplished. Second, we will test if the same applications can be applied to DNA microarray databases using the same semantic web technologies. Although data-driven approach is predominant in interpreting the DNA microarray experiment results, we believe that traditional hypothesis-driven approach will continue to give us new insights to perceive the biological meaning of the DNA microarray experiment data. Third, we will improve descriptive power of our application-specific ontology and align it with existing ontologies. Our application specific ontology describes two array-based technologies (DNA and tissue microarray). Because DNA microarray is usually describable by MGED ontology, our ontology will become semantically more powerful by relating it with MGED ontology. Meanwhile, a part of our application-specific ontology was made based on TMA-OM and there is no other standard on ontology for TMA to the best of our knowledge. We, therefore, will expand our application-specific ontology so as to fully describe TMA experiment and align it with other neighbourhood ontologies.", "We proved semantic web technologies are beneficial in statistical inference as well as in logical inference based on description logic by implementing an application for hypothesis testing without disrupting previously implemented RDF-based applications. We proposed an application in which biologist’s real interests are reflected. The same scheme we used in developing Xperanto-RDF could be applied to the other omics technologies. However extensive study on the semantic and syntactic structure of the biologic hypothesis will more facilitate the development of applications for hypothesis testing in biological databases.", "All authors declared that there is no conflict of interest in this research.", "YSS and JHK designed and developed the study and wrote the manuscript. CHP and HJC implemented the system and wrote the manuscript. HJS and JK tested and coordinated the system and wrote the manuscript. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "RDF representation of TMA and DNA microarray data", "Source databases", "Ontology design and data representation as RDF", "System architecture", "Semantically-enabled hypothesis testing", "Hypothesis model", "SPARQL generation", "Statistical test for Fisher’s exact test", "Use case of integrated TMA and DNA microarray database", "Results", "Xperanto-RDF", "Applications for hypothesis testing", "Use case using integrated TMA and DNA microarray database", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Biological databases are collections of scientific experiments, published literatures, and computational analyses organized under a specialized scheme. Biological databases became essential resources to biologists in their daily researches by providing information about biological facts and experimental results and procedures and also by providing management tools for the obtained data. Because these biological databases are designed for specific purposes, and independently managed, and metadata are not provided in many cases, they are neither semantically explicit nor interoperable. To overcome these problems, semantic web technologies such as Resource Description Framework (RDF), Web Ontology Language (OWL) and SPARQL (SPARQL Protocol and RDF Query Language) have been actively accepted in the field of life science for new database design [1-6]. Semantic web repositories are more advantageous than relational databases (RDBs) because metadata are more complete and standardized [7]. Representation of data as RDF makes biological entities semantically explicit and clear so that various tasks can be performed without extensive human interventions. These tasks includes integration of heterogeneous data, applying logic to infer new insights, and publication and sharing of biological findings and models [8]. Several current biological databases provide integrated data structure for knowledge management by applying semantic web technologies [1-4,6,9].\nIn spite of the benefits of semantic web technologies, these databases cannot directly answer to biologists for biologically meaningful questions or hypotheses. For machines to answer these questions, the process of inference based on either logical or statistical relationships of stored data is required. The inference by description logic is a part of semantic web technologies, but statistical inference was not implemented in the current semantic web technologies. These problems, therefore, cannot be solved within the framework of semantic web technologies alone and are rather dependent on the design of an application. However semantic web technologies are still beneficial in those applications.\nTo prove a biological hypothesis, 1) an experiment is designed to test the hypothesis, 2) the experimental data is gathered, and 3) the data is tested by statistical test(s). Because of the increasing amount of high throughput experimental data in biological databases, there is an increasing need of high throughput validation of biological hypotheses. To implement such an application, 1) a hypothesis given by a user should be semantically interpreted, 2) the relevant experimental data should be retrieved from the database, and then 3) the hypothesis should be statistically tested against the retrieved data.\nBesides the fact that tissue microarray (TMA) is being widely used as a high throughput validation tool for the large number of data-driven hypotheses from other genomic technologies, TMA databases is a good candidate for the proof of concept of above mentioned applications. First, most biological hypotheses that can be derived from only TMA experiment are syntactically simple. The hypothesis derived from TMA experiments can be stated as, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Basically a TMA experiment is designed to test dependency between two entities and a hypothesis about the mechanisms of the interactions between two entities cannot be tested unless relevant additional information is provided. There are two important biological entities in TMA, biological samples and markers. In the TMA-validated hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype” [10], for example, ‘colorectal cancer’ is biological condition, and “reduced Apaf-1 expression” and “high-grade phenotype” are entities (B and C). Therefore a dependency-stating hypothesis in a TMA experiment can be considered as a triadic predicate with condition A, entity B, and entity C as parameters. If these parameters are provided, a query for relevant data retrieval can be generated through them. Second, only a few kinds of statistical tests are generally used to test hypothesis in TMA experiments. The main purpose of TMA experiment is to test the statistical relationships between biological entities (or markers) in a population of samples with identical biological condition. The results are determined by the size of the population conforming to the given hypothesis. If more samples show positive relationships with the hypothesis, the hypothesis is more likely to be true.\nFisher’s exact and χ2 test are the most frequently used statistical tests in TMA experiments to test dependency. They are frequently used because most of clinical or histological parameters (e.g., history of hypertension, tumour grade, etc) and the extent or the intensity of the marker expression in a tissue (e.g., 0, 1, 2, 3) have discrete or categorical values and the dependencies between these values are tested by them. For example, if we want to test the above mentioned hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype.” by Fisher’s exact test, we have to investigate each number of cores in slides for four exclusive conditions : a) negative Apaf-1 expression and high-grade phenotype, b) negative Apaf-1 expression and low-grade phenotype, c) weak to strong Apaf-1 expression and high-grade phenotype, and d) weak to strong Apaf-1 expression and low-grade phenotype. Then these four parameters are used for Fisher’s exact or χ2 test to test negative association between Apaf-1 expression and histological grade.\nIn spite of these benefits of TMA database, if it were not semantically explicit, applications for hypothesis testing could not be implemented. TMA data have complex and wide range of semantics, including information for clinical and histopathological features and large amount of metadata should be provided. Semantic web technologies support richer semantics than traditional RDB-based models. It, therefore, is more desirable that the databases for applications for hypothesis testing should be represented as RDF. In addition, SPARQL as a query language is more intuitively understandable to biologists [11]. Lastly, integration with the other databases, including other TMA and DNA microarray databases were considered in the present study and the databases using semantic web technologies are more advantageous in integration.\nWe have created and managed Xperanto-TMA, a web-based TMA database supporting TMA-OM (Tissue Microarray Object Model) [12] and TMA-TAB [13]. Xperanto-TMA uses RDB because technologies supporting object-oriented models were not mature enough to guarantee high performance [12]. Due to the well organized object model, however, clear and rich semantic relationships between entities are well supported. We have also developed and managed a web-based DNA microarray database called Xperanto [14].\nWhen implementing Xperanto-TMA and Xperanto, we extensively investigated semantics of experimental data of TMA and DNA microarray. This experience gave us good resources in constructing semantically rich and explicit TMA and DNA microarray database represented as RDF. Because Xperanto-TMA and Xperanto were independently developed, it was difficult to integrate these two databases. The integration of gene expression microarray and TMA data is a powerful approach to molecular profiling of human cancer [15]. Semantic web gives us opportunity to integrate these two databases through the use of ontology [9]. After integration was achieved, we developed a use case where the usefulness of the integration was found.\nIn this study, we present how we designed semantic TMA database from the existing relational one and then how applications let researchers test hypotheses using semantic web infrastructures. Marker-level integration of TMA and DNA microarray databases will be also described.", "[SUBTITLE] RDF representation of TMA and DNA microarray data [SUBSECTION] [SUBTITLE] Source databases [SUBSECTION] We represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.\nWe represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.\n[SUBTITLE] Ontology design and data representation as RDF [SUBSECTION] To provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.\nTo provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.\n[SUBTITLE] System architecture [SUBSECTION] In the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF\nIn the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF\n[SUBTITLE] Source databases [SUBSECTION] We represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.\nWe represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.\n[SUBTITLE] Ontology design and data representation as RDF [SUBSECTION] To provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.\nTo provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.\n[SUBTITLE] System architecture [SUBSECTION] In the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF\nIn the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF\n[SUBTITLE] Semantically-enabled hypothesis testing [SUBSECTION] [SUBTITLE] Hypothesis model [SUBSECTION] We investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.\nWe investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.\n[SUBTITLE] SPARQL generation [SUBSECTION] Four SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).\nFour SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).\n[SUBTITLE] Statistical test for Fisher’s exact test [SUBSECTION] After SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.\nAfter SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.\n[SUBTITLE] Hypothesis model [SUBSECTION] We investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.\nWe investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.\n[SUBTITLE] SPARQL generation [SUBSECTION] Four SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).\nFour SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).\n[SUBTITLE] Statistical test for Fisher’s exact test [SUBSECTION] After SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.\nAfter SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.\n[SUBTITLE] Use case of integrated TMA and DNA microarray database [SUBSECTION] We designed a use case using integrated DNA and TMA database, which was to find the distributions of the histologies among the cores of TMA where the intensities of the antibodies corresponding to the markers of interest in DNA microarray were high or low\nWe designed a use case using integrated DNA and TMA database, which was to find the distributions of the histologies among the cores of TMA where the intensities of the antibodies corresponding to the markers of interest in DNA microarray were high or low", "[SUBTITLE] Source databases [SUBSECTION] We represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.\nWe represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.\n[SUBTITLE] Ontology design and data representation as RDF [SUBSECTION] To provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.\nTo provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.\n[SUBTITLE] System architecture [SUBSECTION] In the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF\nIn the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF", "We represented Xperanto and Xperanto-TMA data as RDF triples. The relational schema were derived from TMA-OM by object-relational mapping [13]. The primary objective of Xperanto-TMA was to describe data and metadata from TMA experiments in a structured format with controlled vocabularies. Xperanto-TMA hosts more than 100 TMA experiments for various human cancers such as colon, gastric, and prostate cancers. For each experiment, detailed descriptions for clinical and histopathologic features including histology, grade, stage, and margin involvement are provided. Staining results are also provided for detailed data elements such as staining intensity and range.\nXperanto, which supports Microarray Gene Expression-Object Model (MAGE) [16], was developed to provide integrated data management and analysis for microarray data using user-friendly web-based interface [14]. Xperanto can store and analyze data and metadata from microRNA and ArrayCGH as well as DNA microarray experiment. It hosts more than 300 experiments for 52 array platforms.\nIn addition to these data, data about the mapping of DNA microarray probes to antibodies used in TMA experiment was provided using Genome Research Informatics Pipeline (GRIP, http://grip.snubi.org) to integrate these two databases.", "To provide RDF with a framework and facilitate the process of integration, an ontology specific for our applications was developed. In ontology design, our previous studies to implement Xperanto and Xperanto-TMA were referred because semantics in TMA and DNA microarray experiments were already extensively analyzed in the previous works [12,14]. Based on the previous works, the ontology was expressed as OWL. Part of the ontology is shown in Fig. 1(a) (Name spaces are omitted for simplicity).\nOntology and data representation as RDF (a) A part of the ontology for main entities represented as RDF graphs. C (pale blue rectangles) represents classes, P (pink rectangles) properties, A anonymous nodes, empty arrows rdfs:subClassOf, pink arrows rdfs:domain, green arrows rdfs:range, and red arrows owl:unionOf. (b) Relationship between data tables in a source database (Xperanto-TMA) and data graphs represented as RDF graphs. Name spaces are omitted for simplicity. Violet ellipses represent classes, orange rectangles object properties, green rectangles data type properties, and dotted arrows class-instance relationships.\nReuse of the existing ontologies such as NCI thesaurus or foundation model of anatomy (FMA) was not considered in this research because our main goal was to prove a hypothesis-testing system could be implemented with existing semantic web technologies and the goal could be achieved with our application-specific ontology.\nAccording to the scheme of the ontology, we made a mapping rule for every data element (not shown) and data in Xperanto-TMA and Xperanto was represented as RDF. As in Fig. 1(b), for example, a row of Experiment table with three columns was transformed into two RDF triples. URIs for a subject were generated using the value of the primary key and the column names, “Name” and “ExpType” became predicates, “hasName” and “ExpType”. The values of the columns became objects of the triples. Each instance of URIs is an instance of an element (a class or a property) of the ontology. Note that a value of a foreign key (ExperimentId in Slide table) was transformed into an object of a triple, which was also a subject of another triple. In this way, all the data in the selected tables of source database could be represented as RDF triples.", "In the actual implementation of RDF representation, we used a mediator system, D2RQ, instead of triple store for transformed RDF triples. When a mediator system is implemented, queries are directed against the mediator and then the mediator translates the queries into multiple SQLs for source RDBs. The use of mediator system have been shown to scale better and be easier to maintain [9]. D2RQ is a java API which mediates RDB and RDF-based applications according to an ontology and a mapping rule. The applications based on D2RQ can use SPARQL queries as if there were triple stores for the SPARQLs. The SPARQLs, then, is conveyed to D2RQ and D2RQ translates it into suitable SQL to the source databases. When query results are returned, D2RQ translate them into answers for the corresponding SPARQL and send it to applications (Fig. 2).\nSystem architecture of Xperanto-RDF", "[SUBTITLE] Hypothesis model [SUBSECTION] We investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.\nWe investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.\n[SUBTITLE] SPARQL generation [SUBSECTION] Four SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).\nFour SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).\n[SUBTITLE] Statistical test for Fisher’s exact test [SUBSECTION] After SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.\nAfter SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.", "We investigated the syntactic and semantic structures of four scientific hypotheses, which were validated by TMA experiments, from biomedical articles and pathologists. The hypotheses could be considered as triadic predicates with three parameters, “In a biological condition A, an entity B is either positively or negatively correlated with an entity C.” Factor A was defined as shared properties among the samples, factor B as an attribute and value set pair for distinct properties of the samples classifying the samples into two groups (the classifier), and factor C as an attribute and value set pair for dependent properties of the samples on the classifier.\nIn a hypothesis, “Reduced expression of Apaf-1 in colon cancer correlates with high-grade phenotype,” a shared property corresponded to colon cancer, a classifier to the reduced expression of Apaf-1 and dependent property to high histological grade. In this hypothesis, ‘reduced expression of Apaf-1’ needed to be explicitly stated through the process of the interpretation by pathologists and it was interpreted as the value of staining intensity of Apaf-1 in tumor cells having “0”. On the contrary, if “increased expression” had been used in the hypothesis, it could have been interpreted as the value having “1”, “2” or “3”, implying the complementary value set against “0”.\nBased on these findings, we made a model for the hypothesis to investigate how a given hypothesis can be analyzed with regards to TMA experimental data and statistical test (Fig. 3(a)). To get relevant data to the hypothesis, we needed data set where context was colon cancer and “Apaf-1 Intensity” and “HistologicGrade” were used as fields. Then we divided the selected data set into two groups, Core Collection A and Core Collection B, whose value for “Apaf-1 Intensity” were “0” and “1-3”, respectively. For each Core Collection A and B, we obtained the distribution of the values for “HistologicGrade”, which were designated as “D1” and “D2”, respectively. The element of distribution, D1 and D2 had a value, one among “High-grade”, “Intermediate-grade”, and “Low-grade”. The hypothesis would be supported if D1 and D2 were significantly different by Fisher’s exact test. The other three hypothesis models were constructed only by replacing the parameters of the above mentioned model (Table 1).\nHypothesis model and procedures to test a hypothesis (a) Model of a hypothesis, “Reduced expression of Apaf-1 in colorectal cancer correlates with high-grade phenotype”. (b) A statistical procedures for the hypothesis testing. Four SPARQLs are generated based on the hypothesis model (the other three SPARQLS not shown). The returned values of four SPARQLs are delivered as four elements for Fisher’s exact test.\nAnalysis of factors and their value sets of hypothesis and the results of statistical test (by Fisher’s exact test ) for the TMA data stored in Xperanto-RDF\nThe complementary value sets to the value sets of the classifier and the dependent property were parts of the hypothesis model, but they were not described in the hypothesis and should be reasoned. If a value set of either the classifier or the dependent property is provided, the complementary value set is generated as the complementary set to the value set of it among the set consisting of permissible value of the attribute of it. Taking an example from the above mentioned hypothesis, each of (“Low-grade”, “Intermediate-grade”) and ( “1”, “2”, “3”) was a complementary value set for the classifier and the dependent property, respectively because the permissible value lists for each factor were (“Low-grade”, “Intermediate-grade”, “High-grade”) and (“0”, “1”, “2”, “3”), and “High-grade” and “0” was described in the hypothesis. This process was implementable because we had already defined permissible value sets for every data element depending on the context when implementing Xperanto-TMA [12,13].\nWith the help of these analyses, we could implement the applications for hypothesis testing, which were operated in three steps. First the applications received inputs for the model construction. Second SPARQLs were generated according to the hypothesis model by a SPARQL generator. Third the applications called a statistical module and performed Fisher’s exact test using the result set of SPARQLs.", "Four SPARQLs were generated from each hypothesis model by a SPARQL generator in the applications to obtain four elements for Fisher’s exact test. For example, the following is a SPARQL from the hypothesis model in Fig. 3(a) to get one element of 2-by-2 matrix for Fisher’s exact test, which is a number of cores with negative Apaf-1 expression and high-grade phenotype in colon cancer (Fig. 3(b)). The SPARQLs have four phrases in the WHERE clause and each phrase is independently generated using a transforming function in the SPARQL generator, Factor2SPARQL() having each factor and either hypothesis-describing or its complementary value set as parameters. The SPARQL is generated by concatenating these four phrases.\nGenerally, SPARQLs are generated as in the following.,\nSPARQLi := “SELECT count(distinct ?cr) WHERE {” + Σphraseij + “?sl xpe:Slide Block ?b. ?sl xpe:Slide_Reporter ?r.}”\nPhraseij := Factor2SPARQLj(Factorj, Valuesetij)\nValuesetij : if j = 1, the shared properties among the samples if ((i = 1, 3) and j = 2) or ((i = 1, 2) and j = 3), hypothesis-describing value set for Factorj, otherwise its complementary value set\n(i = 1, 2, 3,4, j = 1, 2, 3).", "After SPARQLs are processed, the applications send a list having four elements to the statistical module as four elements for Fisher’s exact test. The statistical module was made based on R (version 2.10.1., R Development Core Team, Vienna). After processing, the statistical module shows the p-value of the Fisher’s exact test to the user. Based on this p-value, the user can decide whether the database supports or rejects the given hypothesis.", "We designed a use case using integrated DNA and TMA database, which was to find the distributions of the histologies among the cores of TMA where the intensities of the antibodies corresponding to the markers of interest in DNA microarray were high or low", "[SUBTITLE] Xperanto-RDF [SUBSECTION] As described above, Xperanto-RDF was implemented based on existing TMA and DNA microarray databases (http://clara.snubi.org/Xperanto-RDF). Data from these two databases were represented as RDF triples by a mediator system such that users could have the benefits of semantic web technologies as well as those of RDB.\nAs described above, Xperanto-RDF was implemented based on existing TMA and DNA microarray databases (http://clara.snubi.org/Xperanto-RDF). Data from these two databases were represented as RDF triples by a mediator system such that users could have the benefits of semantic web technologies as well as those of RDB.\n[SUBTITLE] Applications for hypothesis testing [SUBSECTION] A hypothesis model testing dependency can be constructed in the index page of Xperanto-RDF as shown in Fig. 4 (http://clara.snubi.org/Xperanto-RDF). In the context field, a user can choose the shared property among the samples. The attribute of the classifier and the dependent property can be selected by tracking along the tree structure of the entity. Once the classifier or the dependent property is selected, the permissible value set of the attribute is delivered to the select box to the right side of the tree frame. Using the buttons beside, the user can select hypothesis-describing and complementary value sets. Then the relevant SPARQL is generated by the SPARQL generator. On submitting the SPARQLs, the applications perform the query against the stored TMA data, the result sets are delivered to the statistical module, and finally the result of the statistical test including p-value is displayed.\nWeb interface to generate SPARQLs by selecting factors and their value sets\nAs a result of hypothesis testing for the hypothesis in Fig. 3, we found the following results from the TMA database: 26 cores with Apaf-1 = 0 and high-grade phenotype, 4 cores with Apaf-1 ≥ 1 and high-grade phenotype, 6 core with Apaf-1 = 0 and low-grade phenotype, and 19 cores with Apaf-1 ≥ 1 and low-grade phenotype. These results support the given hypothesis because p value for Fisher’s exact text for these data was less than 0.0001.\nTable 1 shows the testing results of all of four hypotheses. P value of the testing result determines whether the hypothesis is accepted or rejected according to Xperanto-RDF. With 0.01 of P value as a threshold for the rejection of null hypothesis, all of four hypotheses were accepted in these procedures. These results indicate that TMA data in Xperanto-RDF supports all of these hypotheses.\nA hypothesis model testing dependency can be constructed in the index page of Xperanto-RDF as shown in Fig. 4 (http://clara.snubi.org/Xperanto-RDF). In the context field, a user can choose the shared property among the samples. The attribute of the classifier and the dependent property can be selected by tracking along the tree structure of the entity. Once the classifier or the dependent property is selected, the permissible value set of the attribute is delivered to the select box to the right side of the tree frame. Using the buttons beside, the user can select hypothesis-describing and complementary value sets. Then the relevant SPARQL is generated by the SPARQL generator. On submitting the SPARQLs, the applications perform the query against the stored TMA data, the result sets are delivered to the statistical module, and finally the result of the statistical test including p-value is displayed.\nWeb interface to generate SPARQLs by selecting factors and their value sets\nAs a result of hypothesis testing for the hypothesis in Fig. 3, we found the following results from the TMA database: 26 cores with Apaf-1 = 0 and high-grade phenotype, 4 cores with Apaf-1 ≥ 1 and high-grade phenotype, 6 core with Apaf-1 = 0 and low-grade phenotype, and 19 cores with Apaf-1 ≥ 1 and low-grade phenotype. These results support the given hypothesis because p value for Fisher’s exact text for these data was less than 0.0001.\nTable 1 shows the testing results of all of four hypotheses. P value of the testing result determines whether the hypothesis is accepted or rejected according to Xperanto-RDF. With 0.01 of P value as a threshold for the rejection of null hypothesis, all of four hypotheses were accepted in these procedures. These results indicate that TMA data in Xperanto-RDF supports all of these hypotheses.\n[SUBTITLE] Use case using integrated TMA and DNA microarray database [SUBSECTION] Xperanto-RDF achieved marker-mediated integration of TMA and DNA microarray data by mapping of DNA probes into antibodies in TMA experiment. By the use of mapping data, we could execute some of semantically integrated queries against data from TMA and DNA microarray experiments. A use case for this system was presented here.\nQuery: In an experiment for glioblastoma using Affymetrix HT Human Genome U133A array, the signal from the probe, “201983_s_at” were interesting to a researcher [17]. He wants to know the distribution of histologies of the samples where the corresponding antibody to the probe was highly expressed (intensity > 1) from our TMA database.\nIf TMA and DNA microarray database management systems had not been semantically integrated, it would have taken many steps to execute the queries and furthermore we needed external data source that showed what the corresponding antibody to the “201983_s_at” probe is. In Xperanto-RDF, the queries can be formulated as in Fig. 5. The query results show that the corresponding antibody is EGFR and the most predominant histological type among the samples where EGFR is highly expressed in TMA is glioblastoma according the data in Xperanto-RDF.\nAn integrated query against TMA and DNA microarray database This query returns a list of histological type of the samples which show strong intensity (>1) for the corresponding antibody to the mRNA probe, “201983_s_at”. The corresponding antibody is EGFR.\nXperanto-RDF achieved marker-mediated integration of TMA and DNA microarray data by mapping of DNA probes into antibodies in TMA experiment. By the use of mapping data, we could execute some of semantically integrated queries against data from TMA and DNA microarray experiments. A use case for this system was presented here.\nQuery: In an experiment for glioblastoma using Affymetrix HT Human Genome U133A array, the signal from the probe, “201983_s_at” were interesting to a researcher [17]. He wants to know the distribution of histologies of the samples where the corresponding antibody to the probe was highly expressed (intensity > 1) from our TMA database.\nIf TMA and DNA microarray database management systems had not been semantically integrated, it would have taken many steps to execute the queries and furthermore we needed external data source that showed what the corresponding antibody to the “201983_s_at” probe is. In Xperanto-RDF, the queries can be formulated as in Fig. 5. The query results show that the corresponding antibody is EGFR and the most predominant histological type among the samples where EGFR is highly expressed in TMA is glioblastoma according the data in Xperanto-RDF.\nAn integrated query against TMA and DNA microarray database This query returns a list of histological type of the samples which show strong intensity (>1) for the corresponding antibody to the mRNA probe, “201983_s_at”. The corresponding antibody is EGFR.", "As described above, Xperanto-RDF was implemented based on existing TMA and DNA microarray databases (http://clara.snubi.org/Xperanto-RDF). Data from these two databases were represented as RDF triples by a mediator system such that users could have the benefits of semantic web technologies as well as those of RDB.", "A hypothesis model testing dependency can be constructed in the index page of Xperanto-RDF as shown in Fig. 4 (http://clara.snubi.org/Xperanto-RDF). In the context field, a user can choose the shared property among the samples. The attribute of the classifier and the dependent property can be selected by tracking along the tree structure of the entity. Once the classifier or the dependent property is selected, the permissible value set of the attribute is delivered to the select box to the right side of the tree frame. Using the buttons beside, the user can select hypothesis-describing and complementary value sets. Then the relevant SPARQL is generated by the SPARQL generator. On submitting the SPARQLs, the applications perform the query against the stored TMA data, the result sets are delivered to the statistical module, and finally the result of the statistical test including p-value is displayed.\nWeb interface to generate SPARQLs by selecting factors and their value sets\nAs a result of hypothesis testing for the hypothesis in Fig. 3, we found the following results from the TMA database: 26 cores with Apaf-1 = 0 and high-grade phenotype, 4 cores with Apaf-1 ≥ 1 and high-grade phenotype, 6 core with Apaf-1 = 0 and low-grade phenotype, and 19 cores with Apaf-1 ≥ 1 and low-grade phenotype. These results support the given hypothesis because p value for Fisher’s exact text for these data was less than 0.0001.\nTable 1 shows the testing results of all of four hypotheses. P value of the testing result determines whether the hypothesis is accepted or rejected according to Xperanto-RDF. With 0.01 of P value as a threshold for the rejection of null hypothesis, all of four hypotheses were accepted in these procedures. These results indicate that TMA data in Xperanto-RDF supports all of these hypotheses.", "Xperanto-RDF achieved marker-mediated integration of TMA and DNA microarray data by mapping of DNA probes into antibodies in TMA experiment. By the use of mapping data, we could execute some of semantically integrated queries against data from TMA and DNA microarray experiments. A use case for this system was presented here.\nQuery: In an experiment for glioblastoma using Affymetrix HT Human Genome U133A array, the signal from the probe, “201983_s_at” were interesting to a researcher [17]. He wants to know the distribution of histologies of the samples where the corresponding antibody to the probe was highly expressed (intensity > 1) from our TMA database.\nIf TMA and DNA microarray database management systems had not been semantically integrated, it would have taken many steps to execute the queries and furthermore we needed external data source that showed what the corresponding antibody to the “201983_s_at” probe is. In Xperanto-RDF, the queries can be formulated as in Fig. 5. The query results show that the corresponding antibody is EGFR and the most predominant histological type among the samples where EGFR is highly expressed in TMA is glioblastoma according the data in Xperanto-RDF.\nAn integrated query against TMA and DNA microarray database This query returns a list of histological type of the samples which show strong intensity (>1) for the corresponding antibody to the mRNA probe, “201983_s_at”. The corresponding antibody is EGFR.", "We presented how a semantic database for TMA and DNA microarray experiment can be developed from the existing RDB implementations using currently available semantic web technologies, how the application for hypothesis testing using TMA experimental data was designed and used, and how TMA database can be integrated with DNA microarray database using semantic web technologies.\nWe successfully implemented applications for hypothesis testing and their usefulness was validated by testing several experimentally validated biological hypotheses by TMA experiments. Although the applications seems useful to biologists, it is still challenging to spread the idea of the hypothesis testing in the biological database using experimental data to the whole community of the biologists. In traditional biomedical research, an experiment is designed to test a hypothesis [18] and it has been believed that only a highly controlled experiment could decide whether the hypothesis was true or false. The value of hypothesis testing in the biological database may be questionable from this point of view. Despite the presence of the weakness, the future of hypothesis testing using biological databases seems bright. More experiments are getting open to the public with detailed metadata about experiments and the experiments tend to be industrialized and standardized. As a result, it seems highly probable that a researcher can find most of the experiments they want from the biological databases in the near future. At present these applications can be coupled with traditional biomedical research in a cost effective way. A researcher can use these applications as a preliminary investigation before the highly controlled experiment is begun.\nThe applications for hypothesis testing can be regarded as a semantically-enhanced meta-analysis tool for TMA experiments in one aspect. One of requirements to perform meta-analysis is the process of data normalization. Our applications do not have the process of data normalization yet because there are no known methods of normalization in TMA experiments. The absence of normalization process do not appear to be greatly influential on the experimental results at present because TMA experiments are usually interpreted by human eyes, crude categorical values are used in the interpretation, and experimental procedures and interpretations are rather standardized.\nWe separated the process of hypothesis model construction, SPARQL execution, and statistical test when implementing the applications for hypothesis testing. This allows us more flexibility to easily manage and develop the whole systems. More statistical modules will be easily implemented to our applications if a hypothesis cannot be tested with current statistical modules. We will implement a statistical module for survival analysis soon.\nPost et al. suggested integrative bioinformatics experimentation, defined as in silico data integration experiment using semantic web technologies, and made a use case to prove a hypothesis [7]. They tested a hypothesis stating that histone modification and gene expression regulation through transcription factor binding sites is correlated, by integrating histone modification data and transcription factor binding sites data derived from the public database using semantic web technologies [7]. Their integrative bioinformatics experimentation cycle consisted of hypothesis definition, experimental design, data integration, extension of data integration experiment, and data analysis and interpretation. Their study was focused on the applicability of semantic web technologies to data integration phase. Compared to this system, our applications process all the procedures to test the hypothesis after the hypothesis is inserted. To the user, the applications give impression as if the machine directly answered for the questions. We believe that we can implement the applications for hypothesis testing because we used TMA experiment data rather than genomic data and we targeted at the syntactically simple hypotheses composed of three factors. We believe that we can develop applications for testing for more complicated hypotheses based on this experience by treating complicated hypotheses as logical combinations of multiple simple hypotheses.\nFrom the beginning our TMA database was developed with consideration on the integration with the other TMA or omics databases. Any TMA database based on TMA-OM can easily exchange the experimental data and the direct implementation of our application for hypothesis test is possible.\nWe also achieved marker-mediated integration of TMA and DNA microarray database. The integration enabled marker-mediated bidirectional data flow between TMA and DNA microarray experimental data. With these integrated databases, the results of DNA microarray experiments can be validated in TMA experiments.\nOur future works are targeted in three directions. First, we are developing hypothesis miner extracting biological hypotheses directly from biomedical literature databases. The miner will feed hypotheses to TMA databases without the insertion procedures by the user. When realized, the integration of literature and experimental databases mediated by the hypotheses can be accomplished. Second, we will test if the same applications can be applied to DNA microarray databases using the same semantic web technologies. Although data-driven approach is predominant in interpreting the DNA microarray experiment results, we believe that traditional hypothesis-driven approach will continue to give us new insights to perceive the biological meaning of the DNA microarray experiment data. Third, we will improve descriptive power of our application-specific ontology and align it with existing ontologies. Our application specific ontology describes two array-based technologies (DNA and tissue microarray). Because DNA microarray is usually describable by MGED ontology, our ontology will become semantically more powerful by relating it with MGED ontology. Meanwhile, a part of our application-specific ontology was made based on TMA-OM and there is no other standard on ontology for TMA to the best of our knowledge. We, therefore, will expand our application-specific ontology so as to fully describe TMA experiment and align it with other neighbourhood ontologies.", "We proved semantic web technologies are beneficial in statistical inference as well as in logical inference based on description logic by implementing an application for hypothesis testing without disrupting previously implemented RDF-based applications. We proposed an application in which biologist’s real interests are reflected. The same scheme we used in developing Xperanto-RDF could be applied to the other omics technologies. However extensive study on the semantic and syntactic structure of the biologic hypothesis will more facilitate the development of applications for hypothesis testing in biological databases.", "All authors declared that there is no conflict of interest in this research.", "YSS and JHK designed and developed the study and wrote the manuscript. CHP and HJC implemented the system and wrote the manuscript. HJS and JK tested and coordinated the system and wrote the manuscript. All authors read and approved the final manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Algorithms", "Biomarkers, Tumor", "Discriminant Analysis", "Gene Expression Profiling", "Neoplasms", "Oligonucleotide Array Sequence Analysis", "Principal Component Analysis", "Sensitivity and Specificity", "Software" ]

[ "Background", "Multi-resolution independent component analysis", "1). Wavelet transforms", "2). Feature selection", "3). Inverse discrete wavelet transforms", "4). Independent component analysis", "5). Subspace decomposition", "Multi-resolution Independent component analysis based support vector machines (MICA-SVM)", "Results", "Cross validations", "Six comparison algorithms", "Multi-resolution independent component analysis based linear discriminant analysis", "Optimal level threshold selections", "MICA-based biomarker discovery", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "With the rapid developments in genomics, high-throughput microarray pattern analysis shows a great potential in cancer diagnosis for its efficiency and cost-effectiveness [1]. However, such a promising technology remains an important research field rather than an applicable clinical-routine. Aside intrinsic factors from microarray profiling technologies, a key issue preventing it from becoming a clinical paradigm is that the relatively low even poor sensitivities and specificities obtained from current pattern recognition methodologies are inadequate to provide a robust clinical support. Moreover, some pattern classification methods may perform reasonably well in some data sets but fail badly in others. Although there is an urgent need in clinical cancer research to develop high-performance pattern recognition methods in gene expression analysis, it is still a challenge in machine learning to attain high-accuracy classification for the special characteristics of gene expression profiles.\nA gene expression profile can be represented by a p×n matrix after preprocessing, each column of which represents gene expression values of all biological samples at a gene; each row of which represents gene expression values of a single biological sample across a genome. The total number of genes is in the order of 103~104, and the total number of biological samples is on the magnitude of tens or hundreds. Since the number of variables (genes) is much greater than the number of samples (observations), some traditional pattern recognition methods (e.g., fisher discriminant analysis) may have instable solutions and lead to a low or poor classification performance. Alternatively, although there are a large number of genes in a profile, only a small portion of them have meaningful contributions to data variations. In addition, the high-dimensional data are not noise-free because preprocessing algorithms may not remove systematic noise contained in raw data completely. Obviously, the data redundancy and noise may inevitably affect the discriminative power of the classification algorithms applied to microarray data.\nIt is clear that feature selection play a critical role in gene expression analysis to decrease dimensionalities, remove noise, and extract meaningful features before performing classification. Feature selection algorithms usually can be categorized into three types: statistical test-based (e.g., two-sample t-tests), wrapper-based (e.g., SVM-based wrappers) [2], and transform-based feature selections. The transform-based feature selection may be mostly used data reduction techniques for their popularity and efficiency. They include principal component analysis (PCA) [3], independent component analysis (ICA) [4], nonnegative matrix factorization (NMF) [5,6], etc, and their different extensions [7,8].\nHowever, these transform-based feature selection algorithms are generally good at selecting global features instead of local features. The global and local features contribute to the global and local characteristics of data and interpret global and local behavior of data respectively. Statistically, the global features consist of high-frequency signals and the local features consist of low-frequency signals. Unlike the global features, the local features are difficult to extract for most feature-selection algorithms, because the low-frequency signals have a lower likelihood to get involved in inferring the ‘new’ low-dimensional data, which are generally the linear combinations of all input variables, than the high-frequency signals. Finally, the low dimensional data obtained from the traditional feature selection methods may miss some local data characteristics described by the local features. For example, PCA is by-nature a global feature selection algorithm: each principal component contains some levels of global characteristics of data and receives contributions from all input variables in the linear combinations. In addition, changes in one variable will inevitably affect all loading vectors globally. However, local features may be a key to attaining high-performance gene expression pattern classification for its subtle data behavior capturing. For example, some benign tumor samples may display very similar global characteristics with malignant tumor samples but with different local characteristics. To attain high-performance diagnosis, it is essential to capture local data characteristics to distinguish these samples with similar global characteristics.\nThe main reason for these algorithms’ global-feature selection mechanism is because they all are single-resolution feature selection methods, where all features are indistinguishably displayed in a single-resolution despite the nature of their frequencies. It inevitably causes global features more likely to be selected than local features and prevents effective local data-characteristics capturing. Mathematically, all variables of the input data are involved in the linear combinations to compute principal components in PCA, independent components in ICA, and basis vectors in NMF respectively. Such a global feature selection mechanism will prevent high-accuracy genomic pattern recognition in the following classification because only the features interpreting global characteristics are involved in training a learning machine (e.g., SVM). The redundant global features may inevitably decrease the generalization of the learning machine and increase the risk of misclassifications or over-fitting. Finally, the learning machines integrated with the global feature-selection algorithms will display instabilities in classifications.\nTo avoid the global feature selection mechanism, it is desirable to distinguish (e.g., sort) features according to their frequencies rather than treat them uniformly, which makes the high-frequency signals dominate the feature selection and the low-frequency signals lose opportunities. A discrete wavelet transform (DWT) [9] can hierarchically organize data in a multi-resolution way by low and high pass filters. The low (high)-pass filters only pass low (high)-frequency signals but attenuate signals with frequencies higher (lower) than a cutoff frequency. Finally, the DWT coefficients at the coarse levels capture global features of the input signals and the coefficients at the fine levels capture local features of the signals, i.e., the low-frequency and high-frequency signals are represented by coefficients in the coarse and fine resolutions respectively. Obviously, the global feature selection mechanism can be relatively easy to overcome after such a ‘multi-resolution feature separation’, by selectively extracting local features and filtering redundant global features.\nIn this study, we propose a novel multi-resolution independent component analysis (MICA) to conduct effective feature selections for high dimensional heterogeneous gene expression data. Then, a multi-resolution independent component analysis based support vector machines (MICA-SVM) are proposed to achieve a high-performance gene expression pattern prediction. We demonstrate its superiority and stability by comparing it with existing state-of-the-art peers on six heterogeneous microarray profiles, in addition to extending MICA to linear discriminant analysis (MICA-LDA). We also develop a MICA-based filter-wrapper biomarker discovery algorithm to further demonstrate the novel feature selection algorithm’s effectiveness in biomarker capturing. Finally, we discuss potential extensions on the multi-resolution independent component analysis in microarray based molecular diagnosis and conclude this paper.", "The goal of the multi-resolution independent component analysis is to seek the statistically independent genomic patterns from a meta-profile computed by suppressing the coarse level coefficients (global features) and maintaining the fine level coefficients (local features) in the DWT of an input profile. As an approximation of the high dimensional input profile, the derived meta-profile captures almost all local features and keeps the most important global features. Unlike independent components in the classic ICA that are mainly retrieved from the global features for their high frequencies, the independent components calculated by our proposed MICA method are statistically independent signals, which contain contributions from almost all local features and the most important global features. As such, the latter is more representative in revealing the latent data structure than the former. Moreover, the redundant global feature suppressing brings MICA an automatic de-noising mechanism: since the coarse level coefficients (e.g., the first level coefficients) in DWT generally contain “contributions” from noise, suppressing coarse level coefficients not only filters unnecessary global features but also removes the noise. The MICA algorithm can be described as following steps.\n[SUBTITLE] 1). Wavelet transforms [SUBSECTION] Given a gene expression profile with p samples across n genes , MICA conducts a L-level discrete wavelet transform for each sample to obtain a sequence of detail coefficient matrices and an approximation coefficient matrix , i.e., , where .\nGiven a gene expression profile with p samples across n genes , MICA conducts a L-level discrete wavelet transform for each sample to obtain a sequence of detail coefficient matrices and an approximation coefficient matrix , i.e., , where .\n[SUBTITLE] 2). Feature selection [SUBSECTION] A level threshold is selected to suppress redundant global features and maintain local features as follows. If , 1) conduct principal component analysis for Dj to obtain its PC matrix: and the corresponding score matrix . 2) reconstruct the original Dj by using the first loading vector u1 in the PC matrix as , where is a vector containing all ‘1’s. If , reconstruct and update each detail coefficient matrix Dj by using the loading vectors with the 100% explained variance percentage and their corresponding vectors in the score matrix: . The explained variance percentage is the ratio between the accumulative variance from the selected data and the total data variance. For example, the explained variance percentage ρr from those first r loading vectors is defined as , where λi is the data variance from the ith loading vector. In the implementation, this step can be ‘lazily’ simplified as: keep all detail coefficient matrices intact to save computing resources.\nA level threshold is selected to suppress redundant global features and maintain local features as follows. If , 1) conduct principal component analysis for Dj to obtain its PC matrix: and the corresponding score matrix . 2) reconstruct the original Dj by using the first loading vector u1 in the PC matrix as , where is a vector containing all ‘1’s. If , reconstruct and update each detail coefficient matrix Dj by using the loading vectors with the 100% explained variance percentage and their corresponding vectors in the score matrix: . The explained variance percentage is the ratio between the accumulative variance from the selected data and the total data variance. For example, the explained variance percentage ρr from those first r loading vectors is defined as , where λi is the data variance from the ith loading vector. In the implementation, this step can be ‘lazily’ simplified as: keep all detail coefficient matrices intact to save computing resources.\n[SUBTITLE] 3). Inverse discrete wavelet transforms [SUBSECTION] Conduct the corresponding inverse discrete wavelet transform using the updated coefficient matrices to get the meta-profile of X:, i.e., .\nConduct the corresponding inverse discrete wavelet transform using the updated coefficient matrices to get the meta-profile of X:, i.e., .\n[SUBTITLE] 4). Independent component analysis [SUBSECTION] Conduct the classic independent component analysis for X* to obtain independent components and the mixing matrix: , where , and .\nConduct the classic independent component analysis for X* to obtain independent components and the mixing matrix: , where , and .\n[SUBTITLE] 5). Subspace decomposition [SUBSECTION] The meta-profile X* is the approximation of the original profile X by removing the redundant global features and retaining almost all local features by selecting features on behalf of their frequencies. It is easy to decompose each sample in the subspace spanned by all independent components . Each independent component is a basis in the subspace., i.e., , where the mixing matrix is , and the independent component matrix is . In other words, each sample can be represented as , where the meta-sample ai is the ith row of the mixing matrix recording the coordinate values of the sample xi in the subspace. As a low dimensional vector, the meta-sample ai retains almost all local features and the most outstanding global features of the original high-dimensional sample xi. Thus it can be called as a data-locality preserved prototype of xi.\nFigure 1 visualizes three controls and cancers of the ‘breast_1’ data (see Table 1 for more information) and their meta-samples obtained from MICA at τ = 3,4,6 with a Daubechies family wavelet ‘db8’, where the control and cancer samples are indicated by red and blue lines respectively. Interestingly, extracted local features and selected important global features make two types of samples display two distinct prototypes in the low-dimension subspace. With the increase of the level thresholds, the two groups of prototypes tend to show more capabilities to separate cancer and control samples. Moreover, two types of meta-samples demonstrate a “self-clustering” mechanism in that the meta-samples belonging to the same type show close spatial proximities. Obviously, the clear sample separation information conveyed by the self-clustering mechanism of the meta-samples is almost impossible to obtain from the original high-dimensional data directly, and the key discriminative features captured by our proposed MICA method would be able to facilitate the subsequent classification step and contribute to high-accuracy disease diagnosis.\nMeta-samples constructed from MICA for six original samples ('breast_1 data'). Meta-samples constructed from multi-resolution independent component analysis for six original samples (three controls and three cancers) in the breast_1 data at the three levels thresholds: τ =3,4,6 with a wavelet ‘db8’. The low-dimensional meta-samples separate two types of samples clearly in visualization.\nSix gene-expression microarray profiles\nThe meta-profile X* is the approximation of the original profile X by removing the redundant global features and retaining almost all local features by selecting features on behalf of their frequencies. It is easy to decompose each sample in the subspace spanned by all independent components . Each independent component is a basis in the subspace., i.e., , where the mixing matrix is , and the independent component matrix is . In other words, each sample can be represented as , where the meta-sample ai is the ith row of the mixing matrix recording the coordinate values of the sample xi in the subspace. As a low dimensional vector, the meta-sample ai retains almost all local features and the most outstanding global features of the original high-dimensional sample xi. Thus it can be called as a data-locality preserved prototype of xi.\nFigure 1 visualizes three controls and cancers of the ‘breast_1’ data (see Table 1 for more information) and their meta-samples obtained from MICA at τ = 3,4,6 with a Daubechies family wavelet ‘db8’, where the control and cancer samples are indicated by red and blue lines respectively. Interestingly, extracted local features and selected important global features make two types of samples display two distinct prototypes in the low-dimension subspace. With the increase of the level thresholds, the two groups of prototypes tend to show more capabilities to separate cancer and control samples. Moreover, two types of meta-samples demonstrate a “self-clustering” mechanism in that the meta-samples belonging to the same type show close spatial proximities. Obviously, the clear sample separation information conveyed by the self-clustering mechanism of the meta-samples is almost impossible to obtain from the original high-dimensional data directly, and the key discriminative features captured by our proposed MICA method would be able to facilitate the subsequent classification step and contribute to high-accuracy disease diagnosis.\nMeta-samples constructed from MICA for six original samples ('breast_1 data'). Meta-samples constructed from multi-resolution independent component analysis for six original samples (three controls and three cancers) in the breast_1 data at the three levels thresholds: τ =3,4,6 with a wavelet ‘db8’. The low-dimensional meta-samples separate two types of samples clearly in visualization.\nSix gene-expression microarray profiles\n[SUBTITLE] Multi-resolution Independent component analysis based support vector machines (MICA-SVM) [SUBSECTION] The MICA-based support vector machines apply the classic support vector machines (C-SVM) [10] to the meta-samples to gain classification information in a low-dimensional space. Unlike the traditional SVM that builds a maximum margin hyper-plane in the original data space ℝn where n~103-104, MICA-SVM separates biological samples by constructing the maximum margin hyperplane in the spanned subspace where k ≤ p ~ 102, using the meta-samples. If we assume the number of support vectors Ns is much less than the training points l, the time complexity of the MICA-SVM is , which is much lower than that of the classic SVM , provided the same number of training points and support vectors. We briefly describe the MICA-SVM classifier for binary classification. Given a training dataset , and sample class type information , where , a meta-dataset is computed by using the multi-resolution independent component analysis. Then, a maximum margin hyper-plane: is constructed to separate the '+1' (‘cancer’) and '-1' (‘control’) types of meta-samples. It is equivalent to solving the following quadratic programming problem,(1)\nA way to solve (1) is through its Lagrangian dual that is also a quadratic programming problem, where are dual variables of the primal variables w and b.(2)\nThe normal of the maximum-margin hyperplane is calculated as . The decision rule is used to determine the class type of a testing sample x′, where are the corresponding meta-samples of samples , computed from MICA respectively. The function is a SVM kernel function that maps these meta-samples into a same-dimensional or high-dimensional feature space. In this work, we only focus on the linear kernel for its simplicity and efficiency in microarray pattern classifications. We will point out in the discussion section that most SVM-based learning machines would encounter overfitting under the standard Gaussian kernel (‘rbf’: radial basis function kernels).\nThe MICA-based support vector machines apply the classic support vector machines (C-SVM) [10] to the meta-samples to gain classification information in a low-dimensional space. Unlike the traditional SVM that builds a maximum margin hyper-plane in the original data space ℝn where n~103-104, MICA-SVM separates biological samples by constructing the maximum margin hyperplane in the spanned subspace where k ≤ p ~ 102, using the meta-samples. If we assume the number of support vectors Ns is much less than the training points l, the time complexity of the MICA-SVM is , which is much lower than that of the classic SVM , provided the same number of training points and support vectors. We briefly describe the MICA-SVM classifier for binary classification. Given a training dataset , and sample class type information , where , a meta-dataset is computed by using the multi-resolution independent component analysis. Then, a maximum margin hyper-plane: is constructed to separate the '+1' (‘cancer’) and '-1' (‘control’) types of meta-samples. It is equivalent to solving the following quadratic programming problem,(1)\nA way to solve (1) is through its Lagrangian dual that is also a quadratic programming problem, where are dual variables of the primal variables w and b.(2)\nThe normal of the maximum-margin hyperplane is calculated as . The decision rule is used to determine the class type of a testing sample x′, where are the corresponding meta-samples of samples , computed from MICA respectively. The function is a SVM kernel function that maps these meta-samples into a same-dimensional or high-dimensional feature space. In this work, we only focus on the linear kernel for its simplicity and efficiency in microarray pattern classifications. We will point out in the discussion section that most SVM-based learning machines would encounter overfitting under the standard Gaussian kernel (‘rbf’: radial basis function kernels).", "Given a gene expression profile with p samples across n genes , MICA conducts a L-level discrete wavelet transform for each sample to obtain a sequence of detail coefficient matrices and an approximation coefficient matrix , i.e., , where .", "A level threshold is selected to suppress redundant global features and maintain local features as follows. If , 1) conduct principal component analysis for Dj to obtain its PC matrix: and the corresponding score matrix . 2) reconstruct the original Dj by using the first loading vector u1 in the PC matrix as , where is a vector containing all ‘1’s. If , reconstruct and update each detail coefficient matrix Dj by using the loading vectors with the 100% explained variance percentage and their corresponding vectors in the score matrix: . The explained variance percentage is the ratio between the accumulative variance from the selected data and the total data variance. For example, the explained variance percentage ρr from those first r loading vectors is defined as , where λi is the data variance from the ith loading vector. In the implementation, this step can be ‘lazily’ simplified as: keep all detail coefficient matrices intact to save computing resources.", "Conduct the corresponding inverse discrete wavelet transform using the updated coefficient matrices to get the meta-profile of X:, i.e., .", "Conduct the classic independent component analysis for X* to obtain independent components and the mixing matrix: , where , and .", "The meta-profile X* is the approximation of the original profile X by removing the redundant global features and retaining almost all local features by selecting features on behalf of their frequencies. It is easy to decompose each sample in the subspace spanned by all independent components . Each independent component is a basis in the subspace., i.e., , where the mixing matrix is , and the independent component matrix is . In other words, each sample can be represented as , where the meta-sample ai is the ith row of the mixing matrix recording the coordinate values of the sample xi in the subspace. As a low dimensional vector, the meta-sample ai retains almost all local features and the most outstanding global features of the original high-dimensional sample xi. Thus it can be called as a data-locality preserved prototype of xi.\nFigure 1 visualizes three controls and cancers of the ‘breast_1’ data (see Table 1 for more information) and their meta-samples obtained from MICA at τ = 3,4,6 with a Daubechies family wavelet ‘db8’, where the control and cancer samples are indicated by red and blue lines respectively. Interestingly, extracted local features and selected important global features make two types of samples display two distinct prototypes in the low-dimension subspace. With the increase of the level thresholds, the two groups of prototypes tend to show more capabilities to separate cancer and control samples. Moreover, two types of meta-samples demonstrate a “self-clustering” mechanism in that the meta-samples belonging to the same type show close spatial proximities. Obviously, the clear sample separation information conveyed by the self-clustering mechanism of the meta-samples is almost impossible to obtain from the original high-dimensional data directly, and the key discriminative features captured by our proposed MICA method would be able to facilitate the subsequent classification step and contribute to high-accuracy disease diagnosis.\nMeta-samples constructed from MICA for six original samples ('breast_1 data'). Meta-samples constructed from multi-resolution independent component analysis for six original samples (three controls and three cancers) in the breast_1 data at the three levels thresholds: τ =3,4,6 with a wavelet ‘db8’. The low-dimensional meta-samples separate two types of samples clearly in visualization.\nSix gene-expression microarray profiles", "The MICA-based support vector machines apply the classic support vector machines (C-SVM) [10] to the meta-samples to gain classification information in a low-dimensional space. Unlike the traditional SVM that builds a maximum margin hyper-plane in the original data space ℝn where n~103-104, MICA-SVM separates biological samples by constructing the maximum margin hyperplane in the spanned subspace where k ≤ p ~ 102, using the meta-samples. If we assume the number of support vectors Ns is much less than the training points l, the time complexity of the MICA-SVM is , which is much lower than that of the classic SVM , provided the same number of training points and support vectors. We briefly describe the MICA-SVM classifier for binary classification. Given a training dataset , and sample class type information , where , a meta-dataset is computed by using the multi-resolution independent component analysis. Then, a maximum margin hyper-plane: is constructed to separate the '+1' (‘cancer’) and '-1' (‘control’) types of meta-samples. It is equivalent to solving the following quadratic programming problem,(1)\nA way to solve (1) is through its Lagrangian dual that is also a quadratic programming problem, where are dual variables of the primal variables w and b.(2)\nThe normal of the maximum-margin hyperplane is calculated as . The decision rule is used to determine the class type of a testing sample x′, where are the corresponding meta-samples of samples , computed from MICA respectively. The function is a SVM kernel function that maps these meta-samples into a same-dimensional or high-dimensional feature space. In this work, we only focus on the linear kernel for its simplicity and efficiency in microarray pattern classifications. We will point out in the discussion section that most SVM-based learning machines would encounter overfitting under the standard Gaussian kernel (‘rbf’: radial basis function kernels).", "We have performed extensive experiments using six publicly available gene expression microarray profiles consisting of five oligonucleotide profiles [11-15] and one cDNA profile [16], in the experiment. Table 1 includes their detailed information. These profiles are heterogeneous data generated from different experimental conditions, different profiling technologies, or even processed by different preprocessing algorithms. For example, the stroma, prostate, glioma, and HCC data only go through basic log2 transforms while the breast_1 data is a dataset obtained by conducting two-sample t-tests from an original dataset going through delicate normalizations [12].\n[SUBTITLE] Cross validations [SUBSECTION] To address our algorithm’s superiority and reproducibility, we compare it with six comparison algorithms in terms of average classification rates, sensitivities, and specificities under the k-fold (k=10) and 100-trial of 50% holdout cross validations. The classification accuracy in the ith classification is the ratio of the correctly classified testing samples over total testing samples: , and the sensitivity and specificity are defined as the ratios: respectively, where tp (tn) is the number of positive (negative) targets correctly classified, and fp (fn) is the number of negative (positive) targets incorrectly classified respectively. In the 100-trial of 50% holdout cross validation (HOCV), all samples in the data set are pooled together and randomly divided into half to get training and testing data. Such a partition is repeated 100 times to get 100 sets of training and testing datasets. In the k-fold cross validation, an input dataset is partitioned into k disjoint equal or approximately equal proportions. One proportion is used for testing and the other k-1 proportions are used for training alternatively in the total k rounds of classifications. Compared with pre-specified training or testing data, the cross validations can decrease potential biases in algorithm performance evaluations.\nTo address our algorithm’s superiority and reproducibility, we compare it with six comparison algorithms in terms of average classification rates, sensitivities, and specificities under the k-fold (k=10) and 100-trial of 50% holdout cross validations. The classification accuracy in the ith classification is the ratio of the correctly classified testing samples over total testing samples: , and the sensitivity and specificity are defined as the ratios: respectively, where tp (tn) is the number of positive (negative) targets correctly classified, and fp (fn) is the number of negative (positive) targets incorrectly classified respectively. In the 100-trial of 50% holdout cross validation (HOCV), all samples in the data set are pooled together and randomly divided into half to get training and testing data. Such a partition is repeated 100 times to get 100 sets of training and testing datasets. In the k-fold cross validation, an input dataset is partitioned into k disjoint equal or approximately equal proportions. One proportion is used for testing and the other k-1 proportions are used for training alternatively in the total k rounds of classifications. Compared with pre-specified training or testing data, the cross validations can decrease potential biases in algorithm performance evaluations.\n[SUBTITLE] Six comparison algorithms [SUBSECTION] The existing six comparison algorithms can be categorized into two types. The first type consists of standard support vector machines (SVM) [10] and linear discriminant analysis (LDA) [17], both of which are state-of-the-art classification algorithms. Especially, SVM is widely employed in gene expression pattern recognition for its popularity. The second type consists of four methods embedding transform-based feature selections in SVM and LDA: they are support vector machines with principal component analysis/independent component analysis/ nonnegative matrix factorization, and linear discriminant analysis with principal component analysis. We refer them as PCA-SVM, ICA-SVM, NMF-SVM, and PCA-LDA conveniently and their related implementation information can be found in Additional file 1.\nWe employ the wavelet ‘db8’ to conduct a 12-level discrete wavelet transform for each data set, and select a level threshold τ = 3 in MICA for all profiles. Although not an optimal level threshold for all data, it guarantees automatic de-noising and ‘fair’ algorithm comparisons. Moreover, we have found that the meta-samples obtained from MICA at τ = 3 can clearly distinguish two types of samples. Although other level threshold selections may be possible, any too ‘coarse’ (e.g.τ = 1) or too ‘fine’ (e.g.τ ≥ 9) level threshold selection may miss some important global or local features and affect following classifications.\nTable 2 and Table 3 illustrate the average performance of the seven algorithms in terms of the classification rates, sensitivities, specificities and their standard deviations under the two types of cross validations respectively. The results of LDA are not included in the two tables for its worst performance. Similarly, the NMF-SVM and ICA-SVM algorithms are excluded from Table 3 for their relatively low performance and high instabilities. Clearly, the proposed MICA-SVM algorithm demonstrates exceptionally leading advantages over its peers in the three classification performance statistics for all datasets. For example, it achieves 98.26%, 99.04%, 99.69%, 98.76%, 98.30% and 97.23% average classification rates on the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 100 trials of 50% HOCV. In addition, MICA-SVM achieves 98.00%, 99.52%, 100.0%, 100.0%, 100.0%, and 99.00% for the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 10-fold CV. All these results indicate that MICA can effectively capture global/local features as well as eliminate the noisy features so that SVM can perform significantly better than the state-of-the-arts. Furthermore, unlike the other methods that display instabilities in classifications, our proposed MICA-SVM algorithm demonstrates a strong stability in attaining high-accuracy detections for all profiles. This observation is also supported by its lower standard deviations of the three classification measures than those of the others.\nAlgorithm average performance comparisons (100 trials of 50% HOCV)\nAlgorithm average performance comparisons (10-fold CV)\nFigure 2 compares the distributions of the classification rates of the four algorithms on the other five profiles under the 100 trials of 50% HOCV. It is obvious that the distributions of classification rates, sensitivities and specificities of MICA-SVM on these data are significantly different from those of the other three peers. Moreover, it seems that there is no statistically significant improvement between SVM and its feature-selection based extensions: ICA-SVM, PCA-SVM, and NMF-SVM, because they achieved the same or slightly lower performance than the standard SVM. The reason for this is rooted in the global feature selection mechanisms of the PCA, ICA, and NMF methods: since biological samples may display very similar global-characteristics and different local-characteristics in their gene expressions, a classification algorithm (e.g., SVM) integrated with the global-feature selection methods will inevitably encounter difficulty in distinguishing these samples. Although extracted by different transform methods, the global features statistically have almost same level contributions to the pattern classifications of a data set. Moreover, the redundant global features brought by the global feature selection mechanism may be involved in the following SVM learning, which limits all the SVM extensions’ generalization and causes their instabilities in classification. However, the local feature capturing and redundant global feature suppressing mechanism in MICA not only attains much better performance than the standard SVM but also maintains algorithm stability in classification. Moreover, Figure 3 shows the MICA-SVM’s leading advantages over the other four peers on behalf of the average classification rates, sensitivities, specificities, and positive prediction ratios under the 10-fold cross validations. All the results directly demonstrate the superiority of MICA to the three general global feature selection algorithms.\nDistributions of the classification rates of four algorithms on five profiles. Distributions of the classification rates of four algorithms: MICA-SVM, ICA-SVM, PCA-SVM, and SVM on five profiles under the 100 trials of 50% holdout cross validations\nComparisons on the five algorithm performance on the six datasets under k-fold cross validations. Comparisons on the five algorithm classification performance on the six datasets under k-fold cross validations. ‘S’ (stroma), ‘B1’ (breast_1), ‘P’ (prostate), ‘G (glioma)’‘H’ (HCC), and ‘B2’ (breast_2). The MICA-SVM algorithm demonstrated exceptional leading performance over the others\nThe existing six comparison algorithms can be categorized into two types. The first type consists of standard support vector machines (SVM) [10] and linear discriminant analysis (LDA) [17], both of which are state-of-the-art classification algorithms. Especially, SVM is widely employed in gene expression pattern recognition for its popularity. The second type consists of four methods embedding transform-based feature selections in SVM and LDA: they are support vector machines with principal component analysis/independent component analysis/ nonnegative matrix factorization, and linear discriminant analysis with principal component analysis. We refer them as PCA-SVM, ICA-SVM, NMF-SVM, and PCA-LDA conveniently and their related implementation information can be found in Additional file 1.\nWe employ the wavelet ‘db8’ to conduct a 12-level discrete wavelet transform for each data set, and select a level threshold τ = 3 in MICA for all profiles. Although not an optimal level threshold for all data, it guarantees automatic de-noising and ‘fair’ algorithm comparisons. Moreover, we have found that the meta-samples obtained from MICA at τ = 3 can clearly distinguish two types of samples. Although other level threshold selections may be possible, any too ‘coarse’ (e.g.τ = 1) or too ‘fine’ (e.g.τ ≥ 9) level threshold selection may miss some important global or local features and affect following classifications.\nTable 2 and Table 3 illustrate the average performance of the seven algorithms in terms of the classification rates, sensitivities, specificities and their standard deviations under the two types of cross validations respectively. The results of LDA are not included in the two tables for its worst performance. Similarly, the NMF-SVM and ICA-SVM algorithms are excluded from Table 3 for their relatively low performance and high instabilities. Clearly, the proposed MICA-SVM algorithm demonstrates exceptionally leading advantages over its peers in the three classification performance statistics for all datasets. For example, it achieves 98.26%, 99.04%, 99.69%, 98.76%, 98.30% and 97.23% average classification rates on the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 100 trials of 50% HOCV. In addition, MICA-SVM achieves 98.00%, 99.52%, 100.0%, 100.0%, 100.0%, and 99.00% for the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 10-fold CV. All these results indicate that MICA can effectively capture global/local features as well as eliminate the noisy features so that SVM can perform significantly better than the state-of-the-arts. Furthermore, unlike the other methods that display instabilities in classifications, our proposed MICA-SVM algorithm demonstrates a strong stability in attaining high-accuracy detections for all profiles. This observation is also supported by its lower standard deviations of the three classification measures than those of the others.\nAlgorithm average performance comparisons (100 trials of 50% HOCV)\nAlgorithm average performance comparisons (10-fold CV)\nFigure 2 compares the distributions of the classification rates of the four algorithms on the other five profiles under the 100 trials of 50% HOCV. It is obvious that the distributions of classification rates, sensitivities and specificities of MICA-SVM on these data are significantly different from those of the other three peers. Moreover, it seems that there is no statistically significant improvement between SVM and its feature-selection based extensions: ICA-SVM, PCA-SVM, and NMF-SVM, because they achieved the same or slightly lower performance than the standard SVM. The reason for this is rooted in the global feature selection mechanisms of the PCA, ICA, and NMF methods: since biological samples may display very similar global-characteristics and different local-characteristics in their gene expressions, a classification algorithm (e.g., SVM) integrated with the global-feature selection methods will inevitably encounter difficulty in distinguishing these samples. Although extracted by different transform methods, the global features statistically have almost same level contributions to the pattern classifications of a data set. Moreover, the redundant global features brought by the global feature selection mechanism may be involved in the following SVM learning, which limits all the SVM extensions’ generalization and causes their instabilities in classification. However, the local feature capturing and redundant global feature suppressing mechanism in MICA not only attains much better performance than the standard SVM but also maintains algorithm stability in classification. Moreover, Figure 3 shows the MICA-SVM’s leading advantages over the other four peers on behalf of the average classification rates, sensitivities, specificities, and positive prediction ratios under the 10-fold cross validations. All the results directly demonstrate the superiority of MICA to the three general global feature selection algorithms.\nDistributions of the classification rates of four algorithms on five profiles. Distributions of the classification rates of four algorithms: MICA-SVM, ICA-SVM, PCA-SVM, and SVM on five profiles under the 100 trials of 50% holdout cross validations\nComparisons on the five algorithm performance on the six datasets under k-fold cross validations. Comparisons on the five algorithm classification performance on the six datasets under k-fold cross validations. ‘S’ (stroma), ‘B1’ (breast_1), ‘P’ (prostate), ‘G (glioma)’‘H’ (HCC), and ‘B2’ (breast_2). The MICA-SVM algorithm demonstrated exceptional leading performance over the others\n[SUBTITLE] Multi-resolution independent component analysis based linear discriminant analysis [SUBSECTION] We also apply MICA to linear discriminant analysis (LDA) to further explore its effectiveness. Similar to the MICA-SVM algorithm, the MICA-based linear discriminant analysis (MICA-LDA) applies the classic LDA to the meta-samples obtained from MICA to gain sample classifications (We skip the detailed algorithm description on MICA-LDA for the space constraint). The MICA-LDA algorithm’s performance on the six profiles can be found in the Additional file 2. To keep consistency with the previous experiments, we still employ the ‘db8’ wavelet and set the level threshold τ = 3 in MICA. Interestingly, the MICA-LDA classifier is only secondary to the MICA-SVM classifier: it outperforms the other comparison algorithms on the five datasets except the prostate data in terms of the average performance under the 100 trials of HOCV and 10-fold CV. This further indicates that MICA’s effective feature selection and its contribution to subsequent classification methods. Figure 4 compares the distribution of classification rates from the three LDA-based algorithms: MICA-LDA, PCA-LDA, and LDA on four data sets under the 100 trials of 50% HOCV. Interestingly, MICA-LDA obviously outperforms PCA-LDA and LDA by its right-skewed classification rate distributions. Although PCA-LDA also demonstrates classification advantages over LDA, MICA-LDA has attained much more impressive improvements than PCA-LDA. On the other hand, this also indicates that the multi-resolution independent component analysis is more effective in the feature selection than principal component analysis, which contributes directly to improving LDA classifier’s performance.\nComparisons of the distributions of algorithm classification rates. Comparisons of the distributions of classification rates of three algorithms on four profiles under the 100 trials of 50% HOCV. LDA classification rates < 50% on the glioma data are not showed in the visualization\nWe also apply MICA to linear discriminant analysis (LDA) to further explore its effectiveness. Similar to the MICA-SVM algorithm, the MICA-based linear discriminant analysis (MICA-LDA) applies the classic LDA to the meta-samples obtained from MICA to gain sample classifications (We skip the detailed algorithm description on MICA-LDA for the space constraint). The MICA-LDA algorithm’s performance on the six profiles can be found in the Additional file 2. To keep consistency with the previous experiments, we still employ the ‘db8’ wavelet and set the level threshold τ = 3 in MICA. Interestingly, the MICA-LDA classifier is only secondary to the MICA-SVM classifier: it outperforms the other comparison algorithms on the five datasets except the prostate data in terms of the average performance under the 100 trials of HOCV and 10-fold CV. This further indicates that MICA’s effective feature selection and its contribution to subsequent classification methods. Figure 4 compares the distribution of classification rates from the three LDA-based algorithms: MICA-LDA, PCA-LDA, and LDA on four data sets under the 100 trials of 50% HOCV. Interestingly, MICA-LDA obviously outperforms PCA-LDA and LDA by its right-skewed classification rate distributions. Although PCA-LDA also demonstrates classification advantages over LDA, MICA-LDA has attained much more impressive improvements than PCA-LDA. On the other hand, this also indicates that the multi-resolution independent component analysis is more effective in the feature selection than principal component analysis, which contributes directly to improving LDA classifier’s performance.\nComparisons of the distributions of algorithm classification rates. Comparisons of the distributions of classification rates of three algorithms on four profiles under the 100 trials of 50% HOCV. LDA classification rates < 50% on the glioma data are not showed in the visualization\n[SUBTITLE] Optimal level threshold selections [SUBSECTION] A remaining question is how to determine the optimal level threshold in MICA so that the following SVM classifier achieves best performance. We employ the condition number of the independent component matrix Z in MICA to resolve it, where Smax and Smin are the maximum and minimum singular values of the matrix Z calculated from MICA. A smaller condition number indicates a more stable matrix that suggests a better status in global and local feature capturing. The level-threshold is counted ‘optimal’ if the condition number δ is the smallest. If the condition numbers from two level thresholds are same numerically, the lower level threshold (which is required to be > 1) is counted as the optimal one. For example, the smallest δ value is achieved at τ = 6 and τ = 7,8,9,10,11 respectively on the HCC data. We choose τ = 6 as the optimal threshold which is corresponding to the best average the average classification rate: 98.77% (STD: 2.26%) with average sensitivity: 99.44% (±2.11%) and specificity are 97.59% (±4.97%) respectively.\nFigure 5 shows the MICA-SVM average classification rates and corresponding condition number δ values under the 100 trials of 50% HOCV on the ‘stroma’, ‘breast_1’, and ‘breast_2’, and ‘HCC’ data, as the level threshold values in MICA are selected from 1 to 11. Obviously, the optimal level threshold can be identified by finding the level threshold corresponding to the minimum condition number. Although the optimal threshold at τ = 8 corresponding to average classification rate 99.11% (±0.89%), which is slightly lower than the actual best average classification rate: 99.20% (±0.92%) achieved at τ = 6, it is ignorable due to possible numerical inaccuracy from the fixed point iteration in MICA. Furthermore, we have found that MICA-SVM has relatively low-level performance at too coarse level thresholds (e.g. τ = 1). Although δ values and MICA-SVM performance show some-level stability under some fine level thresholds, too fine level thresholds (e.g. τ ≥ 8) may decrease classification performance on some data (e.g., stroma data). Also, the optimal level threshold selection method may bring some computing overhead in practical classification. In practice, we suggest the empirical level threshold as for its relative robust performance and automatic de-noising property.\nOptimal threshold selections. Average classification rates and corresponding condition numbers at 11 level thresholds on four profiles under 100 trials of 50% HOCV.\nAlthough only wavelet ‘db8’ is employed in our experiments, there is no other specific requirement in MICA-SVM for a wavelet except it should be orthogonal. To compare effects of different wavelet selections on the algorithm performance, we select four family wavelets: ‘db8’, ‘sym8’, ‘coif4’, and ‘bior4.4’, in the classifications on the six profiles at the level threshold τ = 3. It seems that there is no obvious classification advantage from one wavelet over the other under the 10-fold CV, because the robust prior knowledge and less number of trials may have larger impact factors on the algorithm performance than a wavelet selection. However, we have found that the wavelet ‘db8’ show some advantages over the others under the 100 trials of 50% HOCV. In addition, it is interesting to see that the wavelets ‘coif4’ and ‘sym8’ have almost same-level performance, but the wavelet ‘bior4.4’ has a relatively low performance for the six profiles.\nWe further demonstrate the superiority of MICA-SVM by comparing it with three state-of-the-art partial least square (PLS) based regression methods, which can be found in the Additional file 3. Moreover, we present a novel algorithm stability analysis for the seven classifications and show the advantages of the MICA-SVM and MICA-LDA algorithms over the others (Please see the Additional file 4 for details).\nA remaining question is how to determine the optimal level threshold in MICA so that the following SVM classifier achieves best performance. We employ the condition number of the independent component matrix Z in MICA to resolve it, where Smax and Smin are the maximum and minimum singular values of the matrix Z calculated from MICA. A smaller condition number indicates a more stable matrix that suggests a better status in global and local feature capturing. The level-threshold is counted ‘optimal’ if the condition number δ is the smallest. If the condition numbers from two level thresholds are same numerically, the lower level threshold (which is required to be > 1) is counted as the optimal one. For example, the smallest δ value is achieved at τ = 6 and τ = 7,8,9,10,11 respectively on the HCC data. We choose τ = 6 as the optimal threshold which is corresponding to the best average the average classification rate: 98.77% (STD: 2.26%) with average sensitivity: 99.44% (±2.11%) and specificity are 97.59% (±4.97%) respectively.\nFigure 5 shows the MICA-SVM average classification rates and corresponding condition number δ values under the 100 trials of 50% HOCV on the ‘stroma’, ‘breast_1’, and ‘breast_2’, and ‘HCC’ data, as the level threshold values in MICA are selected from 1 to 11. Obviously, the optimal level threshold can be identified by finding the level threshold corresponding to the minimum condition number. Although the optimal threshold at τ = 8 corresponding to average classification rate 99.11% (±0.89%), which is slightly lower than the actual best average classification rate: 99.20% (±0.92%) achieved at τ = 6, it is ignorable due to possible numerical inaccuracy from the fixed point iteration in MICA. Furthermore, we have found that MICA-SVM has relatively low-level performance at too coarse level thresholds (e.g. τ = 1). Although δ values and MICA-SVM performance show some-level stability under some fine level thresholds, too fine level thresholds (e.g. τ ≥ 8) may decrease classification performance on some data (e.g., stroma data). Also, the optimal level threshold selection method may bring some computing overhead in practical classification. In practice, we suggest the empirical level threshold as for its relative robust performance and automatic de-noising property.\nOptimal threshold selections. Average classification rates and corresponding condition numbers at 11 level thresholds on four profiles under 100 trials of 50% HOCV.\nAlthough only wavelet ‘db8’ is employed in our experiments, there is no other specific requirement in MICA-SVM for a wavelet except it should be orthogonal. To compare effects of different wavelet selections on the algorithm performance, we select four family wavelets: ‘db8’, ‘sym8’, ‘coif4’, and ‘bior4.4’, in the classifications on the six profiles at the level threshold τ = 3. It seems that there is no obvious classification advantage from one wavelet over the other under the 10-fold CV, because the robust prior knowledge and less number of trials may have larger impact factors on the algorithm performance than a wavelet selection. However, we have found that the wavelet ‘db8’ show some advantages over the others under the 100 trials of 50% HOCV. In addition, it is interesting to see that the wavelets ‘coif4’ and ‘sym8’ have almost same-level performance, but the wavelet ‘bior4.4’ has a relatively low performance for the six profiles.\nWe further demonstrate the superiority of MICA-SVM by comparing it with three state-of-the-art partial least square (PLS) based regression methods, which can be found in the Additional file 3. Moreover, we present a novel algorithm stability analysis for the seven classifications and show the advantages of the MICA-SVM and MICA-LDA algorithms over the others (Please see the Additional file 4 for details).\n[SUBTITLE] MICA-based biomarker discovery [SUBSECTION] In addition to classifying large scale heterogeneous tumor profiles with exceptional performance, multi-resolution independent component analysis can be also applied to capture biomarkers for microarray profiles. We present a MICA-based filter-wrapper biomarker capturing algorithm and apply it to the stroma data. The details of this algorithm can be found in the Additional file 5. Table 4 lists the details on all the three biomarkers captured, where the SVM-rate for each biomarker is the classification ratio achieved by a SVM classifier with the ‘rbf’ kernel on the biomarker under leave-one-out cross validations. The order of the three biomarkers in Table 4 is listed according to its order identified in the biomarker discovery process. The SVM accuracy under the three biomarkers is 97.87% and the corresponding sensitivity and specificity are 92.31% and 100% respectively. The first biomarker is gene USP46, which is a broadly expressed gene reported as one gene associated with breast cancer and glioblastomas [18]. The second biomarker is FOSL2, which is one of four members in the Fos gene family. It is responsible for encoding leucine zipper proteins, which is able to dimerize with proteins of the JUN family, and form the transcription factor complex AP-1. As a regulator in cell proliferation, differentiation, and transformation, recent studies [19,20] have showed that it is one of important genes associated with breast cancer, by being involved in the regulation of breast cancer invasion and metastasis. The third biomarker is gene RPL5, which encodes a ribosomal protein that catalyzes protein synthesis. It was reported to associate with biosynthesis and energy utilization that is a cellular function associated with pathogenesis of breast cancer [21]. In addition, it also links to the breast cancer by lowering MDM2, which is a major regulator of p53 levels, preventing p53 ubiquitination and increasing its transcriptional activity [22]. Figure 6 visualizes the 47 samples (13 inflammatory breast cancers ('ibc') and 34 non-inflammatory breast cancers ('non-ibc')) of the stroma data using the three biomarkers. It is interesting to see that two types of cancers are separated into two spatially disjoint sets clearly, though one ‘ibc’ sample is wired in the ‘non-ibc’ samples.\nThree biomarkers discovered for the stroma data\nBiomarker visualization in the stroma data. Visualization of 47 samples in the stroma data by using three biomarkers\nIn addition to classifying large scale heterogeneous tumor profiles with exceptional performance, multi-resolution independent component analysis can be also applied to capture biomarkers for microarray profiles. We present a MICA-based filter-wrapper biomarker capturing algorithm and apply it to the stroma data. The details of this algorithm can be found in the Additional file 5. Table 4 lists the details on all the three biomarkers captured, where the SVM-rate for each biomarker is the classification ratio achieved by a SVM classifier with the ‘rbf’ kernel on the biomarker under leave-one-out cross validations. The order of the three biomarkers in Table 4 is listed according to its order identified in the biomarker discovery process. The SVM accuracy under the three biomarkers is 97.87% and the corresponding sensitivity and specificity are 92.31% and 100% respectively. The first biomarker is gene USP46, which is a broadly expressed gene reported as one gene associated with breast cancer and glioblastomas [18]. The second biomarker is FOSL2, which is one of four members in the Fos gene family. It is responsible for encoding leucine zipper proteins, which is able to dimerize with proteins of the JUN family, and form the transcription factor complex AP-1. As a regulator in cell proliferation, differentiation, and transformation, recent studies [19,20] have showed that it is one of important genes associated with breast cancer, by being involved in the regulation of breast cancer invasion and metastasis. The third biomarker is gene RPL5, which encodes a ribosomal protein that catalyzes protein synthesis. It was reported to associate with biosynthesis and energy utilization that is a cellular function associated with pathogenesis of breast cancer [21]. In addition, it also links to the breast cancer by lowering MDM2, which is a major regulator of p53 levels, preventing p53 ubiquitination and increasing its transcriptional activity [22]. Figure 6 visualizes the 47 samples (13 inflammatory breast cancers ('ibc') and 34 non-inflammatory breast cancers ('non-ibc')) of the stroma data using the three biomarkers. It is interesting to see that two types of cancers are separated into two spatially disjoint sets clearly, though one ‘ibc’ sample is wired in the ‘non-ibc’ samples.\nThree biomarkers discovered for the stroma data\nBiomarker visualization in the stroma data. Visualization of 47 samples in the stroma data by using three biomarkers", "To address our algorithm’s superiority and reproducibility, we compare it with six comparison algorithms in terms of average classification rates, sensitivities, and specificities under the k-fold (k=10) and 100-trial of 50% holdout cross validations. The classification accuracy in the ith classification is the ratio of the correctly classified testing samples over total testing samples: , and the sensitivity and specificity are defined as the ratios: respectively, where tp (tn) is the number of positive (negative) targets correctly classified, and fp (fn) is the number of negative (positive) targets incorrectly classified respectively. In the 100-trial of 50% holdout cross validation (HOCV), all samples in the data set are pooled together and randomly divided into half to get training and testing data. Such a partition is repeated 100 times to get 100 sets of training and testing datasets. In the k-fold cross validation, an input dataset is partitioned into k disjoint equal or approximately equal proportions. One proportion is used for testing and the other k-1 proportions are used for training alternatively in the total k rounds of classifications. Compared with pre-specified training or testing data, the cross validations can decrease potential biases in algorithm performance evaluations.", "The existing six comparison algorithms can be categorized into two types. The first type consists of standard support vector machines (SVM) [10] and linear discriminant analysis (LDA) [17], both of which are state-of-the-art classification algorithms. Especially, SVM is widely employed in gene expression pattern recognition for its popularity. The second type consists of four methods embedding transform-based feature selections in SVM and LDA: they are support vector machines with principal component analysis/independent component analysis/ nonnegative matrix factorization, and linear discriminant analysis with principal component analysis. We refer them as PCA-SVM, ICA-SVM, NMF-SVM, and PCA-LDA conveniently and their related implementation information can be found in Additional file 1.\nWe employ the wavelet ‘db8’ to conduct a 12-level discrete wavelet transform for each data set, and select a level threshold τ = 3 in MICA for all profiles. Although not an optimal level threshold for all data, it guarantees automatic de-noising and ‘fair’ algorithm comparisons. Moreover, we have found that the meta-samples obtained from MICA at τ = 3 can clearly distinguish two types of samples. Although other level threshold selections may be possible, any too ‘coarse’ (e.g.τ = 1) or too ‘fine’ (e.g.τ ≥ 9) level threshold selection may miss some important global or local features and affect following classifications.\nTable 2 and Table 3 illustrate the average performance of the seven algorithms in terms of the classification rates, sensitivities, specificities and their standard deviations under the two types of cross validations respectively. The results of LDA are not included in the two tables for its worst performance. Similarly, the NMF-SVM and ICA-SVM algorithms are excluded from Table 3 for their relatively low performance and high instabilities. Clearly, the proposed MICA-SVM algorithm demonstrates exceptionally leading advantages over its peers in the three classification performance statistics for all datasets. For example, it achieves 98.26%, 99.04%, 99.69%, 98.76%, 98.30% and 97.23% average classification rates on the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 100 trials of 50% HOCV. In addition, MICA-SVM achieves 98.00%, 99.52%, 100.0%, 100.0%, 100.0%, and 99.00% for the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 10-fold CV. All these results indicate that MICA can effectively capture global/local features as well as eliminate the noisy features so that SVM can perform significantly better than the state-of-the-arts. Furthermore, unlike the other methods that display instabilities in classifications, our proposed MICA-SVM algorithm demonstrates a strong stability in attaining high-accuracy detections for all profiles. This observation is also supported by its lower standard deviations of the three classification measures than those of the others.\nAlgorithm average performance comparisons (100 trials of 50% HOCV)\nAlgorithm average performance comparisons (10-fold CV)\nFigure 2 compares the distributions of the classification rates of the four algorithms on the other five profiles under the 100 trials of 50% HOCV. It is obvious that the distributions of classification rates, sensitivities and specificities of MICA-SVM on these data are significantly different from those of the other three peers. Moreover, it seems that there is no statistically significant improvement between SVM and its feature-selection based extensions: ICA-SVM, PCA-SVM, and NMF-SVM, because they achieved the same or slightly lower performance than the standard SVM. The reason for this is rooted in the global feature selection mechanisms of the PCA, ICA, and NMF methods: since biological samples may display very similar global-characteristics and different local-characteristics in their gene expressions, a classification algorithm (e.g., SVM) integrated with the global-feature selection methods will inevitably encounter difficulty in distinguishing these samples. Although extracted by different transform methods, the global features statistically have almost same level contributions to the pattern classifications of a data set. Moreover, the redundant global features brought by the global feature selection mechanism may be involved in the following SVM learning, which limits all the SVM extensions’ generalization and causes their instabilities in classification. However, the local feature capturing and redundant global feature suppressing mechanism in MICA not only attains much better performance than the standard SVM but also maintains algorithm stability in classification. Moreover, Figure 3 shows the MICA-SVM’s leading advantages over the other four peers on behalf of the average classification rates, sensitivities, specificities, and positive prediction ratios under the 10-fold cross validations. All the results directly demonstrate the superiority of MICA to the three general global feature selection algorithms.\nDistributions of the classification rates of four algorithms on five profiles. Distributions of the classification rates of four algorithms: MICA-SVM, ICA-SVM, PCA-SVM, and SVM on five profiles under the 100 trials of 50% holdout cross validations\nComparisons on the five algorithm performance on the six datasets under k-fold cross validations. Comparisons on the five algorithm classification performance on the six datasets under k-fold cross validations. ‘S’ (stroma), ‘B1’ (breast_1), ‘P’ (prostate), ‘G (glioma)’‘H’ (HCC), and ‘B2’ (breast_2). The MICA-SVM algorithm demonstrated exceptional leading performance over the others", "We also apply MICA to linear discriminant analysis (LDA) to further explore its effectiveness. Similar to the MICA-SVM algorithm, the MICA-based linear discriminant analysis (MICA-LDA) applies the classic LDA to the meta-samples obtained from MICA to gain sample classifications (We skip the detailed algorithm description on MICA-LDA for the space constraint). The MICA-LDA algorithm’s performance on the six profiles can be found in the Additional file 2. To keep consistency with the previous experiments, we still employ the ‘db8’ wavelet and set the level threshold τ = 3 in MICA. Interestingly, the MICA-LDA classifier is only secondary to the MICA-SVM classifier: it outperforms the other comparison algorithms on the five datasets except the prostate data in terms of the average performance under the 100 trials of HOCV and 10-fold CV. This further indicates that MICA’s effective feature selection and its contribution to subsequent classification methods. Figure 4 compares the distribution of classification rates from the three LDA-based algorithms: MICA-LDA, PCA-LDA, and LDA on four data sets under the 100 trials of 50% HOCV. Interestingly, MICA-LDA obviously outperforms PCA-LDA and LDA by its right-skewed classification rate distributions. Although PCA-LDA also demonstrates classification advantages over LDA, MICA-LDA has attained much more impressive improvements than PCA-LDA. On the other hand, this also indicates that the multi-resolution independent component analysis is more effective in the feature selection than principal component analysis, which contributes directly to improving LDA classifier’s performance.\nComparisons of the distributions of algorithm classification rates. Comparisons of the distributions of classification rates of three algorithms on four profiles under the 100 trials of 50% HOCV. LDA classification rates < 50% on the glioma data are not showed in the visualization", "A remaining question is how to determine the optimal level threshold in MICA so that the following SVM classifier achieves best performance. We employ the condition number of the independent component matrix Z in MICA to resolve it, where Smax and Smin are the maximum and minimum singular values of the matrix Z calculated from MICA. A smaller condition number indicates a more stable matrix that suggests a better status in global and local feature capturing. The level-threshold is counted ‘optimal’ if the condition number δ is the smallest. If the condition numbers from two level thresholds are same numerically, the lower level threshold (which is required to be > 1) is counted as the optimal one. For example, the smallest δ value is achieved at τ = 6 and τ = 7,8,9,10,11 respectively on the HCC data. We choose τ = 6 as the optimal threshold which is corresponding to the best average the average classification rate: 98.77% (STD: 2.26%) with average sensitivity: 99.44% (±2.11%) and specificity are 97.59% (±4.97%) respectively.\nFigure 5 shows the MICA-SVM average classification rates and corresponding condition number δ values under the 100 trials of 50% HOCV on the ‘stroma’, ‘breast_1’, and ‘breast_2’, and ‘HCC’ data, as the level threshold values in MICA are selected from 1 to 11. Obviously, the optimal level threshold can be identified by finding the level threshold corresponding to the minimum condition number. Although the optimal threshold at τ = 8 corresponding to average classification rate 99.11% (±0.89%), which is slightly lower than the actual best average classification rate: 99.20% (±0.92%) achieved at τ = 6, it is ignorable due to possible numerical inaccuracy from the fixed point iteration in MICA. Furthermore, we have found that MICA-SVM has relatively low-level performance at too coarse level thresholds (e.g. τ = 1). Although δ values and MICA-SVM performance show some-level stability under some fine level thresholds, too fine level thresholds (e.g. τ ≥ 8) may decrease classification performance on some data (e.g., stroma data). Also, the optimal level threshold selection method may bring some computing overhead in practical classification. In practice, we suggest the empirical level threshold as for its relative robust performance and automatic de-noising property.\nOptimal threshold selections. Average classification rates and corresponding condition numbers at 11 level thresholds on four profiles under 100 trials of 50% HOCV.\nAlthough only wavelet ‘db8’ is employed in our experiments, there is no other specific requirement in MICA-SVM for a wavelet except it should be orthogonal. To compare effects of different wavelet selections on the algorithm performance, we select four family wavelets: ‘db8’, ‘sym8’, ‘coif4’, and ‘bior4.4’, in the classifications on the six profiles at the level threshold τ = 3. It seems that there is no obvious classification advantage from one wavelet over the other under the 10-fold CV, because the robust prior knowledge and less number of trials may have larger impact factors on the algorithm performance than a wavelet selection. However, we have found that the wavelet ‘db8’ show some advantages over the others under the 100 trials of 50% HOCV. In addition, it is interesting to see that the wavelets ‘coif4’ and ‘sym8’ have almost same-level performance, but the wavelet ‘bior4.4’ has a relatively low performance for the six profiles.\nWe further demonstrate the superiority of MICA-SVM by comparing it with three state-of-the-art partial least square (PLS) based regression methods, which can be found in the Additional file 3. Moreover, we present a novel algorithm stability analysis for the seven classifications and show the advantages of the MICA-SVM and MICA-LDA algorithms over the others (Please see the Additional file 4 for details).", "In addition to classifying large scale heterogeneous tumor profiles with exceptional performance, multi-resolution independent component analysis can be also applied to capture biomarkers for microarray profiles. We present a MICA-based filter-wrapper biomarker capturing algorithm and apply it to the stroma data. The details of this algorithm can be found in the Additional file 5. Table 4 lists the details on all the three biomarkers captured, where the SVM-rate for each biomarker is the classification ratio achieved by a SVM classifier with the ‘rbf’ kernel on the biomarker under leave-one-out cross validations. The order of the three biomarkers in Table 4 is listed according to its order identified in the biomarker discovery process. The SVM accuracy under the three biomarkers is 97.87% and the corresponding sensitivity and specificity are 92.31% and 100% respectively. The first biomarker is gene USP46, which is a broadly expressed gene reported as one gene associated with breast cancer and glioblastomas [18]. The second biomarker is FOSL2, which is one of four members in the Fos gene family. It is responsible for encoding leucine zipper proteins, which is able to dimerize with proteins of the JUN family, and form the transcription factor complex AP-1. As a regulator in cell proliferation, differentiation, and transformation, recent studies [19,20] have showed that it is one of important genes associated with breast cancer, by being involved in the regulation of breast cancer invasion and metastasis. The third biomarker is gene RPL5, which encodes a ribosomal protein that catalyzes protein synthesis. It was reported to associate with biosynthesis and energy utilization that is a cellular function associated with pathogenesis of breast cancer [21]. In addition, it also links to the breast cancer by lowering MDM2, which is a major regulator of p53 levels, preventing p53 ubiquitination and increasing its transcriptional activity [22]. Figure 6 visualizes the 47 samples (13 inflammatory breast cancers ('ibc') and 34 non-inflammatory breast cancers ('non-ibc')) of the stroma data using the three biomarkers. It is interesting to see that two types of cancers are separated into two spatially disjoint sets clearly, though one ‘ibc’ sample is wired in the ‘non-ibc’ samples.\nThree biomarkers discovered for the stroma data\nBiomarker visualization in the stroma data. Visualization of 47 samples in the stroma data by using three biomarkers", "It is worthy to note that independent component analysis is a necessary step to achieve a good classification performance. A similar multi-resolution principal component analysis based SVM algorithm is not able to reach comparable performance as our algorithm because of the loss of statistical independence in the feature selection. Also, MICA-SVM encounters overfitting as SVM, PCA-SVM, ICA-SVM classifiers under the standard Gaussian kernel (‘rbf’), where each learning machine can only recognize the majority type samples of the training data in classification despite the testing sample type. Moreover, we have tried kernel ICA [23] based support vector machines (KICA-SVM) in our experiments in addition to the previous nine comparison algorithms. However, The KICA-SVM classifier generally has a lower performance level than the standard SVM classifier. Furthermore, the KICA-SVM not only shows a strong instability in classification but also inevitably encounters overfitting under the standard Gaussian kernel like the other learning machines. It seems to suggest that kernel based data reduction may not be a desirable approach in effective feature selection for high dimensional heterogeneous gene profiles. Similar results can be also found in kernel PCA [24] based support vector machine (KPCA-SVM) classifications: a KPCA-SVM classifier is essentially the PCA-SVM classifier when its two kernels are selected as ‘linear’, otherwise, it encounters overfitting under the standard Gaussian kernel. In our ongoing project, in addition to further polishing our algorithm by comparing them with other state-of-the-art methods (e.g., SVM-RFE [2]), we are interested in theoretically validating the MICA-SVM‘s advantages over the classic SVM classifier from the viewpoint of Vapnik–Chervonenkis (VC) dimension theory [10].", "In this study, we present a novel multi-resolution feature selection algorithm: multi-resolution independent component analysis for effective feature selection for high-dimensional heterogeneous gene expression profiles, propose a high-performance MICA-SVM classification algorithm, and demonstrate its superiority and stability by comparing it with the nine state-of-the-art algorithms. Our algorithm not only consistently demonstrates the high-accuracy or clinical-level cancer diagnosis by treating an input profile a whole biomarker but also shows effectiveness in meaningful biomarker discovery. It suggests a great potential to facilitate high-throughput microarray technology into a clinical routine, especially, current classification methods have relative low even poor performance on the gene expression data. In addition, the multi-resolution data analysis based redundant global feature suppressing and effective local feature extraction will have a positive impact on large scale ‘omics’ data mining. In our future work, we plan to further explore MICA-SVM’s potential in other platform gene expression data, SNP, and protein expression data classification.", "The authors declare that they have no competing interests.", "HEY collects and processes the data, designs algorithms, implements the methods, and drafts paper. LXL participates in discussion and provides help to polish the paper. HEY and LXL jointly finalize the paper." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Multi-resolution independent component analysis", "1). Wavelet transforms", "2). Feature selection", "3). Inverse discrete wavelet transforms", "4). Independent component analysis", "5). Subspace decomposition", "Multi-resolution Independent component analysis based support vector machines (MICA-SVM)", "Results", "Cross validations", "Six comparison algorithms", "Multi-resolution independent component analysis based linear discriminant analysis", "Optimal level threshold selections", "MICA-based biomarker discovery", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Supplementary Material" ]

[ "With the rapid developments in genomics, high-throughput microarray pattern analysis shows a great potential in cancer diagnosis for its efficiency and cost-effectiveness [1]. However, such a promising technology remains an important research field rather than an applicable clinical-routine. Aside intrinsic factors from microarray profiling technologies, a key issue preventing it from becoming a clinical paradigm is that the relatively low even poor sensitivities and specificities obtained from current pattern recognition methodologies are inadequate to provide a robust clinical support. Moreover, some pattern classification methods may perform reasonably well in some data sets but fail badly in others. Although there is an urgent need in clinical cancer research to develop high-performance pattern recognition methods in gene expression analysis, it is still a challenge in machine learning to attain high-accuracy classification for the special characteristics of gene expression profiles.\nA gene expression profile can be represented by a p×n matrix after preprocessing, each column of which represents gene expression values of all biological samples at a gene; each row of which represents gene expression values of a single biological sample across a genome. The total number of genes is in the order of 103~104, and the total number of biological samples is on the magnitude of tens or hundreds. Since the number of variables (genes) is much greater than the number of samples (observations), some traditional pattern recognition methods (e.g., fisher discriminant analysis) may have instable solutions and lead to a low or poor classification performance. Alternatively, although there are a large number of genes in a profile, only a small portion of them have meaningful contributions to data variations. In addition, the high-dimensional data are not noise-free because preprocessing algorithms may not remove systematic noise contained in raw data completely. Obviously, the data redundancy and noise may inevitably affect the discriminative power of the classification algorithms applied to microarray data.\nIt is clear that feature selection play a critical role in gene expression analysis to decrease dimensionalities, remove noise, and extract meaningful features before performing classification. Feature selection algorithms usually can be categorized into three types: statistical test-based (e.g., two-sample t-tests), wrapper-based (e.g., SVM-based wrappers) [2], and transform-based feature selections. The transform-based feature selection may be mostly used data reduction techniques for their popularity and efficiency. They include principal component analysis (PCA) [3], independent component analysis (ICA) [4], nonnegative matrix factorization (NMF) [5,6], etc, and their different extensions [7,8].\nHowever, these transform-based feature selection algorithms are generally good at selecting global features instead of local features. The global and local features contribute to the global and local characteristics of data and interpret global and local behavior of data respectively. Statistically, the global features consist of high-frequency signals and the local features consist of low-frequency signals. Unlike the global features, the local features are difficult to extract for most feature-selection algorithms, because the low-frequency signals have a lower likelihood to get involved in inferring the ‘new’ low-dimensional data, which are generally the linear combinations of all input variables, than the high-frequency signals. Finally, the low dimensional data obtained from the traditional feature selection methods may miss some local data characteristics described by the local features. For example, PCA is by-nature a global feature selection algorithm: each principal component contains some levels of global characteristics of data and receives contributions from all input variables in the linear combinations. In addition, changes in one variable will inevitably affect all loading vectors globally. However, local features may be a key to attaining high-performance gene expression pattern classification for its subtle data behavior capturing. For example, some benign tumor samples may display very similar global characteristics with malignant tumor samples but with different local characteristics. To attain high-performance diagnosis, it is essential to capture local data characteristics to distinguish these samples with similar global characteristics.\nThe main reason for these algorithms’ global-feature selection mechanism is because they all are single-resolution feature selection methods, where all features are indistinguishably displayed in a single-resolution despite the nature of their frequencies. It inevitably causes global features more likely to be selected than local features and prevents effective local data-characteristics capturing. Mathematically, all variables of the input data are involved in the linear combinations to compute principal components in PCA, independent components in ICA, and basis vectors in NMF respectively. Such a global feature selection mechanism will prevent high-accuracy genomic pattern recognition in the following classification because only the features interpreting global characteristics are involved in training a learning machine (e.g., SVM). The redundant global features may inevitably decrease the generalization of the learning machine and increase the risk of misclassifications or over-fitting. Finally, the learning machines integrated with the global feature-selection algorithms will display instabilities in classifications.\nTo avoid the global feature selection mechanism, it is desirable to distinguish (e.g., sort) features according to their frequencies rather than treat them uniformly, which makes the high-frequency signals dominate the feature selection and the low-frequency signals lose opportunities. A discrete wavelet transform (DWT) [9] can hierarchically organize data in a multi-resolution way by low and high pass filters. The low (high)-pass filters only pass low (high)-frequency signals but attenuate signals with frequencies higher (lower) than a cutoff frequency. Finally, the DWT coefficients at the coarse levels capture global features of the input signals and the coefficients at the fine levels capture local features of the signals, i.e., the low-frequency and high-frequency signals are represented by coefficients in the coarse and fine resolutions respectively. Obviously, the global feature selection mechanism can be relatively easy to overcome after such a ‘multi-resolution feature separation’, by selectively extracting local features and filtering redundant global features.\nIn this study, we propose a novel multi-resolution independent component analysis (MICA) to conduct effective feature selections for high dimensional heterogeneous gene expression data. Then, a multi-resolution independent component analysis based support vector machines (MICA-SVM) are proposed to achieve a high-performance gene expression pattern prediction. We demonstrate its superiority and stability by comparing it with existing state-of-the-art peers on six heterogeneous microarray profiles, in addition to extending MICA to linear discriminant analysis (MICA-LDA). We also develop a MICA-based filter-wrapper biomarker discovery algorithm to further demonstrate the novel feature selection algorithm’s effectiveness in biomarker capturing. Finally, we discuss potential extensions on the multi-resolution independent component analysis in microarray based molecular diagnosis and conclude this paper.", "Multi-resolution independent component analysis is based on the discrete wavelet transform (DWT) and independent component analysis (ICA). A discrete wavelet transform decomposes input data in a multi-resolution form by using a wavelet and scaling function. The coefficients at the coarse and fine levels describe the global and local behavior of data respectively. Mathematically, DWT is equivalent to multiplying the input data by a set of orthogonal matrices block-wisely. On the other hand, ICA seeks to represent input data as the linear combination of a set of statistically independent components by minimizing their mutual information. Theoretically, it is equivalent to inverting the central limit theorem (CLT) by searching maximally non-normal projections of the original data distribution. More information about the DWT and ICA methods can be found in [4,9].", "The goal of the multi-resolution independent component analysis is to seek the statistically independent genomic patterns from a meta-profile computed by suppressing the coarse level coefficients (global features) and maintaining the fine level coefficients (local features) in the DWT of an input profile. As an approximation of the high dimensional input profile, the derived meta-profile captures almost all local features and keeps the most important global features. Unlike independent components in the classic ICA that are mainly retrieved from the global features for their high frequencies, the independent components calculated by our proposed MICA method are statistically independent signals, which contain contributions from almost all local features and the most important global features. As such, the latter is more representative in revealing the latent data structure than the former. Moreover, the redundant global feature suppressing brings MICA an automatic de-noising mechanism: since the coarse level coefficients (e.g., the first level coefficients) in DWT generally contain “contributions” from noise, suppressing coarse level coefficients not only filters unnecessary global features but also removes the noise. The MICA algorithm can be described as following steps.\n[SUBTITLE] 1). Wavelet transforms [SUBSECTION] Given a gene expression profile with p samples across n genes , MICA conducts a L-level discrete wavelet transform for each sample to obtain a sequence of detail coefficient matrices and an approximation coefficient matrix , i.e., , where .\nGiven a gene expression profile with p samples across n genes , MICA conducts a L-level discrete wavelet transform for each sample to obtain a sequence of detail coefficient matrices and an approximation coefficient matrix , i.e., , where .\n[SUBTITLE] 2). Feature selection [SUBSECTION] A level threshold is selected to suppress redundant global features and maintain local features as follows. If , 1) conduct principal component analysis for Dj to obtain its PC matrix: and the corresponding score matrix . 2) reconstruct the original Dj by using the first loading vector u1 in the PC matrix as , where is a vector containing all ‘1’s. If , reconstruct and update each detail coefficient matrix Dj by using the loading vectors with the 100% explained variance percentage and their corresponding vectors in the score matrix: . The explained variance percentage is the ratio between the accumulative variance from the selected data and the total data variance. For example, the explained variance percentage ρr from those first r loading vectors is defined as , where λi is the data variance from the ith loading vector. In the implementation, this step can be ‘lazily’ simplified as: keep all detail coefficient matrices intact to save computing resources.\nA level threshold is selected to suppress redundant global features and maintain local features as follows. If , 1) conduct principal component analysis for Dj to obtain its PC matrix: and the corresponding score matrix . 2) reconstruct the original Dj by using the first loading vector u1 in the PC matrix as , where is a vector containing all ‘1’s. If , reconstruct and update each detail coefficient matrix Dj by using the loading vectors with the 100% explained variance percentage and their corresponding vectors in the score matrix: . The explained variance percentage is the ratio between the accumulative variance from the selected data and the total data variance. For example, the explained variance percentage ρr from those first r loading vectors is defined as , where λi is the data variance from the ith loading vector. In the implementation, this step can be ‘lazily’ simplified as: keep all detail coefficient matrices intact to save computing resources.\n[SUBTITLE] 3). Inverse discrete wavelet transforms [SUBSECTION] Conduct the corresponding inverse discrete wavelet transform using the updated coefficient matrices to get the meta-profile of X:, i.e., .\nConduct the corresponding inverse discrete wavelet transform using the updated coefficient matrices to get the meta-profile of X:, i.e., .\n[SUBTITLE] 4). Independent component analysis [SUBSECTION] Conduct the classic independent component analysis for X* to obtain independent components and the mixing matrix: , where , and .\nConduct the classic independent component analysis for X* to obtain independent components and the mixing matrix: , where , and .\n[SUBTITLE] 5). Subspace decomposition [SUBSECTION] The meta-profile X* is the approximation of the original profile X by removing the redundant global features and retaining almost all local features by selecting features on behalf of their frequencies. It is easy to decompose each sample in the subspace spanned by all independent components . Each independent component is a basis in the subspace., i.e., , where the mixing matrix is , and the independent component matrix is . In other words, each sample can be represented as , where the meta-sample ai is the ith row of the mixing matrix recording the coordinate values of the sample xi in the subspace. As a low dimensional vector, the meta-sample ai retains almost all local features and the most outstanding global features of the original high-dimensional sample xi. Thus it can be called as a data-locality preserved prototype of xi.\nFigure 1 visualizes three controls and cancers of the ‘breast_1’ data (see Table 1 for more information) and their meta-samples obtained from MICA at τ = 3,4,6 with a Daubechies family wavelet ‘db8’, where the control and cancer samples are indicated by red and blue lines respectively. Interestingly, extracted local features and selected important global features make two types of samples display two distinct prototypes in the low-dimension subspace. With the increase of the level thresholds, the two groups of prototypes tend to show more capabilities to separate cancer and control samples. Moreover, two types of meta-samples demonstrate a “self-clustering” mechanism in that the meta-samples belonging to the same type show close spatial proximities. Obviously, the clear sample separation information conveyed by the self-clustering mechanism of the meta-samples is almost impossible to obtain from the original high-dimensional data directly, and the key discriminative features captured by our proposed MICA method would be able to facilitate the subsequent classification step and contribute to high-accuracy disease diagnosis.\nMeta-samples constructed from MICA for six original samples ('breast_1 data'). Meta-samples constructed from multi-resolution independent component analysis for six original samples (three controls and three cancers) in the breast_1 data at the three levels thresholds: τ =3,4,6 with a wavelet ‘db8’. The low-dimensional meta-samples separate two types of samples clearly in visualization.\nSix gene-expression microarray profiles\nThe meta-profile X* is the approximation of the original profile X by removing the redundant global features and retaining almost all local features by selecting features on behalf of their frequencies. It is easy to decompose each sample in the subspace spanned by all independent components . Each independent component is a basis in the subspace., i.e., , where the mixing matrix is , and the independent component matrix is . In other words, each sample can be represented as , where the meta-sample ai is the ith row of the mixing matrix recording the coordinate values of the sample xi in the subspace. As a low dimensional vector, the meta-sample ai retains almost all local features and the most outstanding global features of the original high-dimensional sample xi. Thus it can be called as a data-locality preserved prototype of xi.\nFigure 1 visualizes three controls and cancers of the ‘breast_1’ data (see Table 1 for more information) and their meta-samples obtained from MICA at τ = 3,4,6 with a Daubechies family wavelet ‘db8’, where the control and cancer samples are indicated by red and blue lines respectively. Interestingly, extracted local features and selected important global features make two types of samples display two distinct prototypes in the low-dimension subspace. With the increase of the level thresholds, the two groups of prototypes tend to show more capabilities to separate cancer and control samples. Moreover, two types of meta-samples demonstrate a “self-clustering” mechanism in that the meta-samples belonging to the same type show close spatial proximities. Obviously, the clear sample separation information conveyed by the self-clustering mechanism of the meta-samples is almost impossible to obtain from the original high-dimensional data directly, and the key discriminative features captured by our proposed MICA method would be able to facilitate the subsequent classification step and contribute to high-accuracy disease diagnosis.\nMeta-samples constructed from MICA for six original samples ('breast_1 data'). Meta-samples constructed from multi-resolution independent component analysis for six original samples (three controls and three cancers) in the breast_1 data at the three levels thresholds: τ =3,4,6 with a wavelet ‘db8’. The low-dimensional meta-samples separate two types of samples clearly in visualization.\nSix gene-expression microarray profiles\n[SUBTITLE] Multi-resolution Independent component analysis based support vector machines (MICA-SVM) [SUBSECTION] The MICA-based support vector machines apply the classic support vector machines (C-SVM) [10] to the meta-samples to gain classification information in a low-dimensional space. Unlike the traditional SVM that builds a maximum margin hyper-plane in the original data space ℝn where n~103-104, MICA-SVM separates biological samples by constructing the maximum margin hyperplane in the spanned subspace where k ≤ p ~ 102, using the meta-samples. If we assume the number of support vectors Ns is much less than the training points l, the time complexity of the MICA-SVM is , which is much lower than that of the classic SVM , provided the same number of training points and support vectors. We briefly describe the MICA-SVM classifier for binary classification. Given a training dataset , and sample class type information , where , a meta-dataset is computed by using the multi-resolution independent component analysis. Then, a maximum margin hyper-plane: is constructed to separate the '+1' (‘cancer’) and '-1' (‘control’) types of meta-samples. It is equivalent to solving the following quadratic programming problem,(1)\nA way to solve (1) is through its Lagrangian dual that is also a quadratic programming problem, where are dual variables of the primal variables w and b.(2)\nThe normal of the maximum-margin hyperplane is calculated as . The decision rule is used to determine the class type of a testing sample x′, where are the corresponding meta-samples of samples , computed from MICA respectively. The function is a SVM kernel function that maps these meta-samples into a same-dimensional or high-dimensional feature space. In this work, we only focus on the linear kernel for its simplicity and efficiency in microarray pattern classifications. We will point out in the discussion section that most SVM-based learning machines would encounter overfitting under the standard Gaussian kernel (‘rbf’: radial basis function kernels).\nThe MICA-based support vector machines apply the classic support vector machines (C-SVM) [10] to the meta-samples to gain classification information in a low-dimensional space. Unlike the traditional SVM that builds a maximum margin hyper-plane in the original data space ℝn where n~103-104, MICA-SVM separates biological samples by constructing the maximum margin hyperplane in the spanned subspace where k ≤ p ~ 102, using the meta-samples. If we assume the number of support vectors Ns is much less than the training points l, the time complexity of the MICA-SVM is , which is much lower than that of the classic SVM , provided the same number of training points and support vectors. We briefly describe the MICA-SVM classifier for binary classification. Given a training dataset , and sample class type information , where , a meta-dataset is computed by using the multi-resolution independent component analysis. Then, a maximum margin hyper-plane: is constructed to separate the '+1' (‘cancer’) and '-1' (‘control’) types of meta-samples. It is equivalent to solving the following quadratic programming problem,(1)\nA way to solve (1) is through its Lagrangian dual that is also a quadratic programming problem, where are dual variables of the primal variables w and b.(2)\nThe normal of the maximum-margin hyperplane is calculated as . The decision rule is used to determine the class type of a testing sample x′, where are the corresponding meta-samples of samples , computed from MICA respectively. The function is a SVM kernel function that maps these meta-samples into a same-dimensional or high-dimensional feature space. In this work, we only focus on the linear kernel for its simplicity and efficiency in microarray pattern classifications. We will point out in the discussion section that most SVM-based learning machines would encounter overfitting under the standard Gaussian kernel (‘rbf’: radial basis function kernels).", "Given a gene expression profile with p samples across n genes , MICA conducts a L-level discrete wavelet transform for each sample to obtain a sequence of detail coefficient matrices and an approximation coefficient matrix , i.e., , where .", "A level threshold is selected to suppress redundant global features and maintain local features as follows. If , 1) conduct principal component analysis for Dj to obtain its PC matrix: and the corresponding score matrix . 2) reconstruct the original Dj by using the first loading vector u1 in the PC matrix as , where is a vector containing all ‘1’s. If , reconstruct and update each detail coefficient matrix Dj by using the loading vectors with the 100% explained variance percentage and their corresponding vectors in the score matrix: . The explained variance percentage is the ratio between the accumulative variance from the selected data and the total data variance. For example, the explained variance percentage ρr from those first r loading vectors is defined as , where λi is the data variance from the ith loading vector. In the implementation, this step can be ‘lazily’ simplified as: keep all detail coefficient matrices intact to save computing resources.", "Conduct the corresponding inverse discrete wavelet transform using the updated coefficient matrices to get the meta-profile of X:, i.e., .", "Conduct the classic independent component analysis for X* to obtain independent components and the mixing matrix: , where , and .", "The meta-profile X* is the approximation of the original profile X by removing the redundant global features and retaining almost all local features by selecting features on behalf of their frequencies. It is easy to decompose each sample in the subspace spanned by all independent components . Each independent component is a basis in the subspace., i.e., , where the mixing matrix is , and the independent component matrix is . In other words, each sample can be represented as , where the meta-sample ai is the ith row of the mixing matrix recording the coordinate values of the sample xi in the subspace. As a low dimensional vector, the meta-sample ai retains almost all local features and the most outstanding global features of the original high-dimensional sample xi. Thus it can be called as a data-locality preserved prototype of xi.\nFigure 1 visualizes three controls and cancers of the ‘breast_1’ data (see Table 1 for more information) and their meta-samples obtained from MICA at τ = 3,4,6 with a Daubechies family wavelet ‘db8’, where the control and cancer samples are indicated by red and blue lines respectively. Interestingly, extracted local features and selected important global features make two types of samples display two distinct prototypes in the low-dimension subspace. With the increase of the level thresholds, the two groups of prototypes tend to show more capabilities to separate cancer and control samples. Moreover, two types of meta-samples demonstrate a “self-clustering” mechanism in that the meta-samples belonging to the same type show close spatial proximities. Obviously, the clear sample separation information conveyed by the self-clustering mechanism of the meta-samples is almost impossible to obtain from the original high-dimensional data directly, and the key discriminative features captured by our proposed MICA method would be able to facilitate the subsequent classification step and contribute to high-accuracy disease diagnosis.\nMeta-samples constructed from MICA for six original samples ('breast_1 data'). Meta-samples constructed from multi-resolution independent component analysis for six original samples (three controls and three cancers) in the breast_1 data at the three levels thresholds: τ =3,4,6 with a wavelet ‘db8’. The low-dimensional meta-samples separate two types of samples clearly in visualization.\nSix gene-expression microarray profiles", "The MICA-based support vector machines apply the classic support vector machines (C-SVM) [10] to the meta-samples to gain classification information in a low-dimensional space. Unlike the traditional SVM that builds a maximum margin hyper-plane in the original data space ℝn where n~103-104, MICA-SVM separates biological samples by constructing the maximum margin hyperplane in the spanned subspace where k ≤ p ~ 102, using the meta-samples. If we assume the number of support vectors Ns is much less than the training points l, the time complexity of the MICA-SVM is , which is much lower than that of the classic SVM , provided the same number of training points and support vectors. We briefly describe the MICA-SVM classifier for binary classification. Given a training dataset , and sample class type information , where , a meta-dataset is computed by using the multi-resolution independent component analysis. Then, a maximum margin hyper-plane: is constructed to separate the '+1' (‘cancer’) and '-1' (‘control’) types of meta-samples. It is equivalent to solving the following quadratic programming problem,(1)\nA way to solve (1) is through its Lagrangian dual that is also a quadratic programming problem, where are dual variables of the primal variables w and b.(2)\nThe normal of the maximum-margin hyperplane is calculated as . The decision rule is used to determine the class type of a testing sample x′, where are the corresponding meta-samples of samples , computed from MICA respectively. The function is a SVM kernel function that maps these meta-samples into a same-dimensional or high-dimensional feature space. In this work, we only focus on the linear kernel for its simplicity and efficiency in microarray pattern classifications. We will point out in the discussion section that most SVM-based learning machines would encounter overfitting under the standard Gaussian kernel (‘rbf’: radial basis function kernels).", "We have performed extensive experiments using six publicly available gene expression microarray profiles consisting of five oligonucleotide profiles [11-15] and one cDNA profile [16], in the experiment. Table 1 includes their detailed information. These profiles are heterogeneous data generated from different experimental conditions, different profiling technologies, or even processed by different preprocessing algorithms. For example, the stroma, prostate, glioma, and HCC data only go through basic log2 transforms while the breast_1 data is a dataset obtained by conducting two-sample t-tests from an original dataset going through delicate normalizations [12].\n[SUBTITLE] Cross validations [SUBSECTION] To address our algorithm’s superiority and reproducibility, we compare it with six comparison algorithms in terms of average classification rates, sensitivities, and specificities under the k-fold (k=10) and 100-trial of 50% holdout cross validations. The classification accuracy in the ith classification is the ratio of the correctly classified testing samples over total testing samples: , and the sensitivity and specificity are defined as the ratios: respectively, where tp (tn) is the number of positive (negative) targets correctly classified, and fp (fn) is the number of negative (positive) targets incorrectly classified respectively. In the 100-trial of 50% holdout cross validation (HOCV), all samples in the data set are pooled together and randomly divided into half to get training and testing data. Such a partition is repeated 100 times to get 100 sets of training and testing datasets. In the k-fold cross validation, an input dataset is partitioned into k disjoint equal or approximately equal proportions. One proportion is used for testing and the other k-1 proportions are used for training alternatively in the total k rounds of classifications. Compared with pre-specified training or testing data, the cross validations can decrease potential biases in algorithm performance evaluations.\nTo address our algorithm’s superiority and reproducibility, we compare it with six comparison algorithms in terms of average classification rates, sensitivities, and specificities under the k-fold (k=10) and 100-trial of 50% holdout cross validations. The classification accuracy in the ith classification is the ratio of the correctly classified testing samples over total testing samples: , and the sensitivity and specificity are defined as the ratios: respectively, where tp (tn) is the number of positive (negative) targets correctly classified, and fp (fn) is the number of negative (positive) targets incorrectly classified respectively. In the 100-trial of 50% holdout cross validation (HOCV), all samples in the data set are pooled together and randomly divided into half to get training and testing data. Such a partition is repeated 100 times to get 100 sets of training and testing datasets. In the k-fold cross validation, an input dataset is partitioned into k disjoint equal or approximately equal proportions. One proportion is used for testing and the other k-1 proportions are used for training alternatively in the total k rounds of classifications. Compared with pre-specified training or testing data, the cross validations can decrease potential biases in algorithm performance evaluations.\n[SUBTITLE] Six comparison algorithms [SUBSECTION] The existing six comparison algorithms can be categorized into two types. The first type consists of standard support vector machines (SVM) [10] and linear discriminant analysis (LDA) [17], both of which are state-of-the-art classification algorithms. Especially, SVM is widely employed in gene expression pattern recognition for its popularity. The second type consists of four methods embedding transform-based feature selections in SVM and LDA: they are support vector machines with principal component analysis/independent component analysis/ nonnegative matrix factorization, and linear discriminant analysis with principal component analysis. We refer them as PCA-SVM, ICA-SVM, NMF-SVM, and PCA-LDA conveniently and their related implementation information can be found in Additional file 1.\nWe employ the wavelet ‘db8’ to conduct a 12-level discrete wavelet transform for each data set, and select a level threshold τ = 3 in MICA for all profiles. Although not an optimal level threshold for all data, it guarantees automatic de-noising and ‘fair’ algorithm comparisons. Moreover, we have found that the meta-samples obtained from MICA at τ = 3 can clearly distinguish two types of samples. Although other level threshold selections may be possible, any too ‘coarse’ (e.g.τ = 1) or too ‘fine’ (e.g.τ ≥ 9) level threshold selection may miss some important global or local features and affect following classifications.\nTable 2 and Table 3 illustrate the average performance of the seven algorithms in terms of the classification rates, sensitivities, specificities and their standard deviations under the two types of cross validations respectively. The results of LDA are not included in the two tables for its worst performance. Similarly, the NMF-SVM and ICA-SVM algorithms are excluded from Table 3 for their relatively low performance and high instabilities. Clearly, the proposed MICA-SVM algorithm demonstrates exceptionally leading advantages over its peers in the three classification performance statistics for all datasets. For example, it achieves 98.26%, 99.04%, 99.69%, 98.76%, 98.30% and 97.23% average classification rates on the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 100 trials of 50% HOCV. In addition, MICA-SVM achieves 98.00%, 99.52%, 100.0%, 100.0%, 100.0%, and 99.00% for the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 10-fold CV. All these results indicate that MICA can effectively capture global/local features as well as eliminate the noisy features so that SVM can perform significantly better than the state-of-the-arts. Furthermore, unlike the other methods that display instabilities in classifications, our proposed MICA-SVM algorithm demonstrates a strong stability in attaining high-accuracy detections for all profiles. This observation is also supported by its lower standard deviations of the three classification measures than those of the others.\nAlgorithm average performance comparisons (100 trials of 50% HOCV)\nAlgorithm average performance comparisons (10-fold CV)\nFigure 2 compares the distributions of the classification rates of the four algorithms on the other five profiles under the 100 trials of 50% HOCV. It is obvious that the distributions of classification rates, sensitivities and specificities of MICA-SVM on these data are significantly different from those of the other three peers. Moreover, it seems that there is no statistically significant improvement between SVM and its feature-selection based extensions: ICA-SVM, PCA-SVM, and NMF-SVM, because they achieved the same or slightly lower performance than the standard SVM. The reason for this is rooted in the global feature selection mechanisms of the PCA, ICA, and NMF methods: since biological samples may display very similar global-characteristics and different local-characteristics in their gene expressions, a classification algorithm (e.g., SVM) integrated with the global-feature selection methods will inevitably encounter difficulty in distinguishing these samples. Although extracted by different transform methods, the global features statistically have almost same level contributions to the pattern classifications of a data set. Moreover, the redundant global features brought by the global feature selection mechanism may be involved in the following SVM learning, which limits all the SVM extensions’ generalization and causes their instabilities in classification. However, the local feature capturing and redundant global feature suppressing mechanism in MICA not only attains much better performance than the standard SVM but also maintains algorithm stability in classification. Moreover, Figure 3 shows the MICA-SVM’s leading advantages over the other four peers on behalf of the average classification rates, sensitivities, specificities, and positive prediction ratios under the 10-fold cross validations. All the results directly demonstrate the superiority of MICA to the three general global feature selection algorithms.\nDistributions of the classification rates of four algorithms on five profiles. Distributions of the classification rates of four algorithms: MICA-SVM, ICA-SVM, PCA-SVM, and SVM on five profiles under the 100 trials of 50% holdout cross validations\nComparisons on the five algorithm performance on the six datasets under k-fold cross validations. Comparisons on the five algorithm classification performance on the six datasets under k-fold cross validations. ‘S’ (stroma), ‘B1’ (breast_1), ‘P’ (prostate), ‘G (glioma)’‘H’ (HCC), and ‘B2’ (breast_2). The MICA-SVM algorithm demonstrated exceptional leading performance over the others\nThe existing six comparison algorithms can be categorized into two types. The first type consists of standard support vector machines (SVM) [10] and linear discriminant analysis (LDA) [17], both of which are state-of-the-art classification algorithms. Especially, SVM is widely employed in gene expression pattern recognition for its popularity. The second type consists of four methods embedding transform-based feature selections in SVM and LDA: they are support vector machines with principal component analysis/independent component analysis/ nonnegative matrix factorization, and linear discriminant analysis with principal component analysis. We refer them as PCA-SVM, ICA-SVM, NMF-SVM, and PCA-LDA conveniently and their related implementation information can be found in Additional file 1.\nWe employ the wavelet ‘db8’ to conduct a 12-level discrete wavelet transform for each data set, and select a level threshold τ = 3 in MICA for all profiles. Although not an optimal level threshold for all data, it guarantees automatic de-noising and ‘fair’ algorithm comparisons. Moreover, we have found that the meta-samples obtained from MICA at τ = 3 can clearly distinguish two types of samples. Although other level threshold selections may be possible, any too ‘coarse’ (e.g.τ = 1) or too ‘fine’ (e.g.τ ≥ 9) level threshold selection may miss some important global or local features and affect following classifications.\nTable 2 and Table 3 illustrate the average performance of the seven algorithms in terms of the classification rates, sensitivities, specificities and their standard deviations under the two types of cross validations respectively. The results of LDA are not included in the two tables for its worst performance. Similarly, the NMF-SVM and ICA-SVM algorithms are excluded from Table 3 for their relatively low performance and high instabilities. Clearly, the proposed MICA-SVM algorithm demonstrates exceptionally leading advantages over its peers in the three classification performance statistics for all datasets. For example, it achieves 98.26%, 99.04%, 99.69%, 98.76%, 98.30% and 97.23% average classification rates on the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 100 trials of 50% HOCV. In addition, MICA-SVM achieves 98.00%, 99.52%, 100.0%, 100.0%, 100.0%, and 99.00% for the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 10-fold CV. All these results indicate that MICA can effectively capture global/local features as well as eliminate the noisy features so that SVM can perform significantly better than the state-of-the-arts. Furthermore, unlike the other methods that display instabilities in classifications, our proposed MICA-SVM algorithm demonstrates a strong stability in attaining high-accuracy detections for all profiles. This observation is also supported by its lower standard deviations of the three classification measures than those of the others.\nAlgorithm average performance comparisons (100 trials of 50% HOCV)\nAlgorithm average performance comparisons (10-fold CV)\nFigure 2 compares the distributions of the classification rates of the four algorithms on the other five profiles under the 100 trials of 50% HOCV. It is obvious that the distributions of classification rates, sensitivities and specificities of MICA-SVM on these data are significantly different from those of the other three peers. Moreover, it seems that there is no statistically significant improvement between SVM and its feature-selection based extensions: ICA-SVM, PCA-SVM, and NMF-SVM, because they achieved the same or slightly lower performance than the standard SVM. The reason for this is rooted in the global feature selection mechanisms of the PCA, ICA, and NMF methods: since biological samples may display very similar global-characteristics and different local-characteristics in their gene expressions, a classification algorithm (e.g., SVM) integrated with the global-feature selection methods will inevitably encounter difficulty in distinguishing these samples. Although extracted by different transform methods, the global features statistically have almost same level contributions to the pattern classifications of a data set. Moreover, the redundant global features brought by the global feature selection mechanism may be involved in the following SVM learning, which limits all the SVM extensions’ generalization and causes their instabilities in classification. However, the local feature capturing and redundant global feature suppressing mechanism in MICA not only attains much better performance than the standard SVM but also maintains algorithm stability in classification. Moreover, Figure 3 shows the MICA-SVM’s leading advantages over the other four peers on behalf of the average classification rates, sensitivities, specificities, and positive prediction ratios under the 10-fold cross validations. All the results directly demonstrate the superiority of MICA to the three general global feature selection algorithms.\nDistributions of the classification rates of four algorithms on five profiles. Distributions of the classification rates of four algorithms: MICA-SVM, ICA-SVM, PCA-SVM, and SVM on five profiles under the 100 trials of 50% holdout cross validations\nComparisons on the five algorithm performance on the six datasets under k-fold cross validations. Comparisons on the five algorithm classification performance on the six datasets under k-fold cross validations. ‘S’ (stroma), ‘B1’ (breast_1), ‘P’ (prostate), ‘G (glioma)’‘H’ (HCC), and ‘B2’ (breast_2). The MICA-SVM algorithm demonstrated exceptional leading performance over the others\n[SUBTITLE] Multi-resolution independent component analysis based linear discriminant analysis [SUBSECTION] We also apply MICA to linear discriminant analysis (LDA) to further explore its effectiveness. Similar to the MICA-SVM algorithm, the MICA-based linear discriminant analysis (MICA-LDA) applies the classic LDA to the meta-samples obtained from MICA to gain sample classifications (We skip the detailed algorithm description on MICA-LDA for the space constraint). The MICA-LDA algorithm’s performance on the six profiles can be found in the Additional file 2. To keep consistency with the previous experiments, we still employ the ‘db8’ wavelet and set the level threshold τ = 3 in MICA. Interestingly, the MICA-LDA classifier is only secondary to the MICA-SVM classifier: it outperforms the other comparison algorithms on the five datasets except the prostate data in terms of the average performance under the 100 trials of HOCV and 10-fold CV. This further indicates that MICA’s effective feature selection and its contribution to subsequent classification methods. Figure 4 compares the distribution of classification rates from the three LDA-based algorithms: MICA-LDA, PCA-LDA, and LDA on four data sets under the 100 trials of 50% HOCV. Interestingly, MICA-LDA obviously outperforms PCA-LDA and LDA by its right-skewed classification rate distributions. Although PCA-LDA also demonstrates classification advantages over LDA, MICA-LDA has attained much more impressive improvements than PCA-LDA. On the other hand, this also indicates that the multi-resolution independent component analysis is more effective in the feature selection than principal component analysis, which contributes directly to improving LDA classifier’s performance.\nComparisons of the distributions of algorithm classification rates. Comparisons of the distributions of classification rates of three algorithms on four profiles under the 100 trials of 50% HOCV. LDA classification rates < 50% on the glioma data are not showed in the visualization\nWe also apply MICA to linear discriminant analysis (LDA) to further explore its effectiveness. Similar to the MICA-SVM algorithm, the MICA-based linear discriminant analysis (MICA-LDA) applies the classic LDA to the meta-samples obtained from MICA to gain sample classifications (We skip the detailed algorithm description on MICA-LDA for the space constraint). The MICA-LDA algorithm’s performance on the six profiles can be found in the Additional file 2. To keep consistency with the previous experiments, we still employ the ‘db8’ wavelet and set the level threshold τ = 3 in MICA. Interestingly, the MICA-LDA classifier is only secondary to the MICA-SVM classifier: it outperforms the other comparison algorithms on the five datasets except the prostate data in terms of the average performance under the 100 trials of HOCV and 10-fold CV. This further indicates that MICA’s effective feature selection and its contribution to subsequent classification methods. Figure 4 compares the distribution of classification rates from the three LDA-based algorithms: MICA-LDA, PCA-LDA, and LDA on four data sets under the 100 trials of 50% HOCV. Interestingly, MICA-LDA obviously outperforms PCA-LDA and LDA by its right-skewed classification rate distributions. Although PCA-LDA also demonstrates classification advantages over LDA, MICA-LDA has attained much more impressive improvements than PCA-LDA. On the other hand, this also indicates that the multi-resolution independent component analysis is more effective in the feature selection than principal component analysis, which contributes directly to improving LDA classifier’s performance.\nComparisons of the distributions of algorithm classification rates. Comparisons of the distributions of classification rates of three algorithms on four profiles under the 100 trials of 50% HOCV. LDA classification rates < 50% on the glioma data are not showed in the visualization\n[SUBTITLE] Optimal level threshold selections [SUBSECTION] A remaining question is how to determine the optimal level threshold in MICA so that the following SVM classifier achieves best performance. We employ the condition number of the independent component matrix Z in MICA to resolve it, where Smax and Smin are the maximum and minimum singular values of the matrix Z calculated from MICA. A smaller condition number indicates a more stable matrix that suggests a better status in global and local feature capturing. The level-threshold is counted ‘optimal’ if the condition number δ is the smallest. If the condition numbers from two level thresholds are same numerically, the lower level threshold (which is required to be > 1) is counted as the optimal one. For example, the smallest δ value is achieved at τ = 6 and τ = 7,8,9,10,11 respectively on the HCC data. We choose τ = 6 as the optimal threshold which is corresponding to the best average the average classification rate: 98.77% (STD: 2.26%) with average sensitivity: 99.44% (±2.11%) and specificity are 97.59% (±4.97%) respectively.\nFigure 5 shows the MICA-SVM average classification rates and corresponding condition number δ values under the 100 trials of 50% HOCV on the ‘stroma’, ‘breast_1’, and ‘breast_2’, and ‘HCC’ data, as the level threshold values in MICA are selected from 1 to 11. Obviously, the optimal level threshold can be identified by finding the level threshold corresponding to the minimum condition number. Although the optimal threshold at τ = 8 corresponding to average classification rate 99.11% (±0.89%), which is slightly lower than the actual best average classification rate: 99.20% (±0.92%) achieved at τ = 6, it is ignorable due to possible numerical inaccuracy from the fixed point iteration in MICA. Furthermore, we have found that MICA-SVM has relatively low-level performance at too coarse level thresholds (e.g. τ = 1). Although δ values and MICA-SVM performance show some-level stability under some fine level thresholds, too fine level thresholds (e.g. τ ≥ 8) may decrease classification performance on some data (e.g., stroma data). Also, the optimal level threshold selection method may bring some computing overhead in practical classification. In practice, we suggest the empirical level threshold as for its relative robust performance and automatic de-noising property.\nOptimal threshold selections. Average classification rates and corresponding condition numbers at 11 level thresholds on four profiles under 100 trials of 50% HOCV.\nAlthough only wavelet ‘db8’ is employed in our experiments, there is no other specific requirement in MICA-SVM for a wavelet except it should be orthogonal. To compare effects of different wavelet selections on the algorithm performance, we select four family wavelets: ‘db8’, ‘sym8’, ‘coif4’, and ‘bior4.4’, in the classifications on the six profiles at the level threshold τ = 3. It seems that there is no obvious classification advantage from one wavelet over the other under the 10-fold CV, because the robust prior knowledge and less number of trials may have larger impact factors on the algorithm performance than a wavelet selection. However, we have found that the wavelet ‘db8’ show some advantages over the others under the 100 trials of 50% HOCV. In addition, it is interesting to see that the wavelets ‘coif4’ and ‘sym8’ have almost same-level performance, but the wavelet ‘bior4.4’ has a relatively low performance for the six profiles.\nWe further demonstrate the superiority of MICA-SVM by comparing it with three state-of-the-art partial least square (PLS) based regression methods, which can be found in the Additional file 3. Moreover, we present a novel algorithm stability analysis for the seven classifications and show the advantages of the MICA-SVM and MICA-LDA algorithms over the others (Please see the Additional file 4 for details).\nA remaining question is how to determine the optimal level threshold in MICA so that the following SVM classifier achieves best performance. We employ the condition number of the independent component matrix Z in MICA to resolve it, where Smax and Smin are the maximum and minimum singular values of the matrix Z calculated from MICA. A smaller condition number indicates a more stable matrix that suggests a better status in global and local feature capturing. The level-threshold is counted ‘optimal’ if the condition number δ is the smallest. If the condition numbers from two level thresholds are same numerically, the lower level threshold (which is required to be > 1) is counted as the optimal one. For example, the smallest δ value is achieved at τ = 6 and τ = 7,8,9,10,11 respectively on the HCC data. We choose τ = 6 as the optimal threshold which is corresponding to the best average the average classification rate: 98.77% (STD: 2.26%) with average sensitivity: 99.44% (±2.11%) and specificity are 97.59% (±4.97%) respectively.\nFigure 5 shows the MICA-SVM average classification rates and corresponding condition number δ values under the 100 trials of 50% HOCV on the ‘stroma’, ‘breast_1’, and ‘breast_2’, and ‘HCC’ data, as the level threshold values in MICA are selected from 1 to 11. Obviously, the optimal level threshold can be identified by finding the level threshold corresponding to the minimum condition number. Although the optimal threshold at τ = 8 corresponding to average classification rate 99.11% (±0.89%), which is slightly lower than the actual best average classification rate: 99.20% (±0.92%) achieved at τ = 6, it is ignorable due to possible numerical inaccuracy from the fixed point iteration in MICA. Furthermore, we have found that MICA-SVM has relatively low-level performance at too coarse level thresholds (e.g. τ = 1). Although δ values and MICA-SVM performance show some-level stability under some fine level thresholds, too fine level thresholds (e.g. τ ≥ 8) may decrease classification performance on some data (e.g., stroma data). Also, the optimal level threshold selection method may bring some computing overhead in practical classification. In practice, we suggest the empirical level threshold as for its relative robust performance and automatic de-noising property.\nOptimal threshold selections. Average classification rates and corresponding condition numbers at 11 level thresholds on four profiles under 100 trials of 50% HOCV.\nAlthough only wavelet ‘db8’ is employed in our experiments, there is no other specific requirement in MICA-SVM for a wavelet except it should be orthogonal. To compare effects of different wavelet selections on the algorithm performance, we select four family wavelets: ‘db8’, ‘sym8’, ‘coif4’, and ‘bior4.4’, in the classifications on the six profiles at the level threshold τ = 3. It seems that there is no obvious classification advantage from one wavelet over the other under the 10-fold CV, because the robust prior knowledge and less number of trials may have larger impact factors on the algorithm performance than a wavelet selection. However, we have found that the wavelet ‘db8’ show some advantages over the others under the 100 trials of 50% HOCV. In addition, it is interesting to see that the wavelets ‘coif4’ and ‘sym8’ have almost same-level performance, but the wavelet ‘bior4.4’ has a relatively low performance for the six profiles.\nWe further demonstrate the superiority of MICA-SVM by comparing it with three state-of-the-art partial least square (PLS) based regression methods, which can be found in the Additional file 3. Moreover, we present a novel algorithm stability analysis for the seven classifications and show the advantages of the MICA-SVM and MICA-LDA algorithms over the others (Please see the Additional file 4 for details).\n[SUBTITLE] MICA-based biomarker discovery [SUBSECTION] In addition to classifying large scale heterogeneous tumor profiles with exceptional performance, multi-resolution independent component analysis can be also applied to capture biomarkers for microarray profiles. We present a MICA-based filter-wrapper biomarker capturing algorithm and apply it to the stroma data. The details of this algorithm can be found in the Additional file 5. Table 4 lists the details on all the three biomarkers captured, where the SVM-rate for each biomarker is the classification ratio achieved by a SVM classifier with the ‘rbf’ kernel on the biomarker under leave-one-out cross validations. The order of the three biomarkers in Table 4 is listed according to its order identified in the biomarker discovery process. The SVM accuracy under the three biomarkers is 97.87% and the corresponding sensitivity and specificity are 92.31% and 100% respectively. The first biomarker is gene USP46, which is a broadly expressed gene reported as one gene associated with breast cancer and glioblastomas [18]. The second biomarker is FOSL2, which is one of four members in the Fos gene family. It is responsible for encoding leucine zipper proteins, which is able to dimerize with proteins of the JUN family, and form the transcription factor complex AP-1. As a regulator in cell proliferation, differentiation, and transformation, recent studies [19,20] have showed that it is one of important genes associated with breast cancer, by being involved in the regulation of breast cancer invasion and metastasis. The third biomarker is gene RPL5, which encodes a ribosomal protein that catalyzes protein synthesis. It was reported to associate with biosynthesis and energy utilization that is a cellular function associated with pathogenesis of breast cancer [21]. In addition, it also links to the breast cancer by lowering MDM2, which is a major regulator of p53 levels, preventing p53 ubiquitination and increasing its transcriptional activity [22]. Figure 6 visualizes the 47 samples (13 inflammatory breast cancers ('ibc') and 34 non-inflammatory breast cancers ('non-ibc')) of the stroma data using the three biomarkers. It is interesting to see that two types of cancers are separated into two spatially disjoint sets clearly, though one ‘ibc’ sample is wired in the ‘non-ibc’ samples.\nThree biomarkers discovered for the stroma data\nBiomarker visualization in the stroma data. Visualization of 47 samples in the stroma data by using three biomarkers\nIn addition to classifying large scale heterogeneous tumor profiles with exceptional performance, multi-resolution independent component analysis can be also applied to capture biomarkers for microarray profiles. We present a MICA-based filter-wrapper biomarker capturing algorithm and apply it to the stroma data. The details of this algorithm can be found in the Additional file 5. Table 4 lists the details on all the three biomarkers captured, where the SVM-rate for each biomarker is the classification ratio achieved by a SVM classifier with the ‘rbf’ kernel on the biomarker under leave-one-out cross validations. The order of the three biomarkers in Table 4 is listed according to its order identified in the biomarker discovery process. The SVM accuracy under the three biomarkers is 97.87% and the corresponding sensitivity and specificity are 92.31% and 100% respectively. The first biomarker is gene USP46, which is a broadly expressed gene reported as one gene associated with breast cancer and glioblastomas [18]. The second biomarker is FOSL2, which is one of four members in the Fos gene family. It is responsible for encoding leucine zipper proteins, which is able to dimerize with proteins of the JUN family, and form the transcription factor complex AP-1. As a regulator in cell proliferation, differentiation, and transformation, recent studies [19,20] have showed that it is one of important genes associated with breast cancer, by being involved in the regulation of breast cancer invasion and metastasis. The third biomarker is gene RPL5, which encodes a ribosomal protein that catalyzes protein synthesis. It was reported to associate with biosynthesis and energy utilization that is a cellular function associated with pathogenesis of breast cancer [21]. In addition, it also links to the breast cancer by lowering MDM2, which is a major regulator of p53 levels, preventing p53 ubiquitination and increasing its transcriptional activity [22]. Figure 6 visualizes the 47 samples (13 inflammatory breast cancers ('ibc') and 34 non-inflammatory breast cancers ('non-ibc')) of the stroma data using the three biomarkers. It is interesting to see that two types of cancers are separated into two spatially disjoint sets clearly, though one ‘ibc’ sample is wired in the ‘non-ibc’ samples.\nThree biomarkers discovered for the stroma data\nBiomarker visualization in the stroma data. Visualization of 47 samples in the stroma data by using three biomarkers", "To address our algorithm’s superiority and reproducibility, we compare it with six comparison algorithms in terms of average classification rates, sensitivities, and specificities under the k-fold (k=10) and 100-trial of 50% holdout cross validations. The classification accuracy in the ith classification is the ratio of the correctly classified testing samples over total testing samples: , and the sensitivity and specificity are defined as the ratios: respectively, where tp (tn) is the number of positive (negative) targets correctly classified, and fp (fn) is the number of negative (positive) targets incorrectly classified respectively. In the 100-trial of 50% holdout cross validation (HOCV), all samples in the data set are pooled together and randomly divided into half to get training and testing data. Such a partition is repeated 100 times to get 100 sets of training and testing datasets. In the k-fold cross validation, an input dataset is partitioned into k disjoint equal or approximately equal proportions. One proportion is used for testing and the other k-1 proportions are used for training alternatively in the total k rounds of classifications. Compared with pre-specified training or testing data, the cross validations can decrease potential biases in algorithm performance evaluations.", "The existing six comparison algorithms can be categorized into two types. The first type consists of standard support vector machines (SVM) [10] and linear discriminant analysis (LDA) [17], both of which are state-of-the-art classification algorithms. Especially, SVM is widely employed in gene expression pattern recognition for its popularity. The second type consists of four methods embedding transform-based feature selections in SVM and LDA: they are support vector machines with principal component analysis/independent component analysis/ nonnegative matrix factorization, and linear discriminant analysis with principal component analysis. We refer them as PCA-SVM, ICA-SVM, NMF-SVM, and PCA-LDA conveniently and their related implementation information can be found in Additional file 1.\nWe employ the wavelet ‘db8’ to conduct a 12-level discrete wavelet transform for each data set, and select a level threshold τ = 3 in MICA for all profiles. Although not an optimal level threshold for all data, it guarantees automatic de-noising and ‘fair’ algorithm comparisons. Moreover, we have found that the meta-samples obtained from MICA at τ = 3 can clearly distinguish two types of samples. Although other level threshold selections may be possible, any too ‘coarse’ (e.g.τ = 1) or too ‘fine’ (e.g.τ ≥ 9) level threshold selection may miss some important global or local features and affect following classifications.\nTable 2 and Table 3 illustrate the average performance of the seven algorithms in terms of the classification rates, sensitivities, specificities and their standard deviations under the two types of cross validations respectively. The results of LDA are not included in the two tables for its worst performance. Similarly, the NMF-SVM and ICA-SVM algorithms are excluded from Table 3 for their relatively low performance and high instabilities. Clearly, the proposed MICA-SVM algorithm demonstrates exceptionally leading advantages over its peers in the three classification performance statistics for all datasets. For example, it achieves 98.26%, 99.04%, 99.69%, 98.76%, 98.30% and 97.23% average classification rates on the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 100 trials of 50% HOCV. In addition, MICA-SVM achieves 98.00%, 99.52%, 100.0%, 100.0%, 100.0%, and 99.00% for the stroma, breast_1, prostate, glioma, HCC, and breast_2 data respectively under the 10-fold CV. All these results indicate that MICA can effectively capture global/local features as well as eliminate the noisy features so that SVM can perform significantly better than the state-of-the-arts. Furthermore, unlike the other methods that display instabilities in classifications, our proposed MICA-SVM algorithm demonstrates a strong stability in attaining high-accuracy detections for all profiles. This observation is also supported by its lower standard deviations of the three classification measures than those of the others.\nAlgorithm average performance comparisons (100 trials of 50% HOCV)\nAlgorithm average performance comparisons (10-fold CV)\nFigure 2 compares the distributions of the classification rates of the four algorithms on the other five profiles under the 100 trials of 50% HOCV. It is obvious that the distributions of classification rates, sensitivities and specificities of MICA-SVM on these data are significantly different from those of the other three peers. Moreover, it seems that there is no statistically significant improvement between SVM and its feature-selection based extensions: ICA-SVM, PCA-SVM, and NMF-SVM, because they achieved the same or slightly lower performance than the standard SVM. The reason for this is rooted in the global feature selection mechanisms of the PCA, ICA, and NMF methods: since biological samples may display very similar global-characteristics and different local-characteristics in their gene expressions, a classification algorithm (e.g., SVM) integrated with the global-feature selection methods will inevitably encounter difficulty in distinguishing these samples. Although extracted by different transform methods, the global features statistically have almost same level contributions to the pattern classifications of a data set. Moreover, the redundant global features brought by the global feature selection mechanism may be involved in the following SVM learning, which limits all the SVM extensions’ generalization and causes their instabilities in classification. However, the local feature capturing and redundant global feature suppressing mechanism in MICA not only attains much better performance than the standard SVM but also maintains algorithm stability in classification. Moreover, Figure 3 shows the MICA-SVM’s leading advantages over the other four peers on behalf of the average classification rates, sensitivities, specificities, and positive prediction ratios under the 10-fold cross validations. All the results directly demonstrate the superiority of MICA to the three general global feature selection algorithms.\nDistributions of the classification rates of four algorithms on five profiles. Distributions of the classification rates of four algorithms: MICA-SVM, ICA-SVM, PCA-SVM, and SVM on five profiles under the 100 trials of 50% holdout cross validations\nComparisons on the five algorithm performance on the six datasets under k-fold cross validations. Comparisons on the five algorithm classification performance on the six datasets under k-fold cross validations. ‘S’ (stroma), ‘B1’ (breast_1), ‘P’ (prostate), ‘G (glioma)’‘H’ (HCC), and ‘B2’ (breast_2). The MICA-SVM algorithm demonstrated exceptional leading performance over the others", "We also apply MICA to linear discriminant analysis (LDA) to further explore its effectiveness. Similar to the MICA-SVM algorithm, the MICA-based linear discriminant analysis (MICA-LDA) applies the classic LDA to the meta-samples obtained from MICA to gain sample classifications (We skip the detailed algorithm description on MICA-LDA for the space constraint). The MICA-LDA algorithm’s performance on the six profiles can be found in the Additional file 2. To keep consistency with the previous experiments, we still employ the ‘db8’ wavelet and set the level threshold τ = 3 in MICA. Interestingly, the MICA-LDA classifier is only secondary to the MICA-SVM classifier: it outperforms the other comparison algorithms on the five datasets except the prostate data in terms of the average performance under the 100 trials of HOCV and 10-fold CV. This further indicates that MICA’s effective feature selection and its contribution to subsequent classification methods. Figure 4 compares the distribution of classification rates from the three LDA-based algorithms: MICA-LDA, PCA-LDA, and LDA on four data sets under the 100 trials of 50% HOCV. Interestingly, MICA-LDA obviously outperforms PCA-LDA and LDA by its right-skewed classification rate distributions. Although PCA-LDA also demonstrates classification advantages over LDA, MICA-LDA has attained much more impressive improvements than PCA-LDA. On the other hand, this also indicates that the multi-resolution independent component analysis is more effective in the feature selection than principal component analysis, which contributes directly to improving LDA classifier’s performance.\nComparisons of the distributions of algorithm classification rates. Comparisons of the distributions of classification rates of three algorithms on four profiles under the 100 trials of 50% HOCV. LDA classification rates < 50% on the glioma data are not showed in the visualization", "A remaining question is how to determine the optimal level threshold in MICA so that the following SVM classifier achieves best performance. We employ the condition number of the independent component matrix Z in MICA to resolve it, where Smax and Smin are the maximum and minimum singular values of the matrix Z calculated from MICA. A smaller condition number indicates a more stable matrix that suggests a better status in global and local feature capturing. The level-threshold is counted ‘optimal’ if the condition number δ is the smallest. If the condition numbers from two level thresholds are same numerically, the lower level threshold (which is required to be > 1) is counted as the optimal one. For example, the smallest δ value is achieved at τ = 6 and τ = 7,8,9,10,11 respectively on the HCC data. We choose τ = 6 as the optimal threshold which is corresponding to the best average the average classification rate: 98.77% (STD: 2.26%) with average sensitivity: 99.44% (±2.11%) and specificity are 97.59% (±4.97%) respectively.\nFigure 5 shows the MICA-SVM average classification rates and corresponding condition number δ values under the 100 trials of 50% HOCV on the ‘stroma’, ‘breast_1’, and ‘breast_2’, and ‘HCC’ data, as the level threshold values in MICA are selected from 1 to 11. Obviously, the optimal level threshold can be identified by finding the level threshold corresponding to the minimum condition number. Although the optimal threshold at τ = 8 corresponding to average classification rate 99.11% (±0.89%), which is slightly lower than the actual best average classification rate: 99.20% (±0.92%) achieved at τ = 6, it is ignorable due to possible numerical inaccuracy from the fixed point iteration in MICA. Furthermore, we have found that MICA-SVM has relatively low-level performance at too coarse level thresholds (e.g. τ = 1). Although δ values and MICA-SVM performance show some-level stability under some fine level thresholds, too fine level thresholds (e.g. τ ≥ 8) may decrease classification performance on some data (e.g., stroma data). Also, the optimal level threshold selection method may bring some computing overhead in practical classification. In practice, we suggest the empirical level threshold as for its relative robust performance and automatic de-noising property.\nOptimal threshold selections. Average classification rates and corresponding condition numbers at 11 level thresholds on four profiles under 100 trials of 50% HOCV.\nAlthough only wavelet ‘db8’ is employed in our experiments, there is no other specific requirement in MICA-SVM for a wavelet except it should be orthogonal. To compare effects of different wavelet selections on the algorithm performance, we select four family wavelets: ‘db8’, ‘sym8’, ‘coif4’, and ‘bior4.4’, in the classifications on the six profiles at the level threshold τ = 3. It seems that there is no obvious classification advantage from one wavelet over the other under the 10-fold CV, because the robust prior knowledge and less number of trials may have larger impact factors on the algorithm performance than a wavelet selection. However, we have found that the wavelet ‘db8’ show some advantages over the others under the 100 trials of 50% HOCV. In addition, it is interesting to see that the wavelets ‘coif4’ and ‘sym8’ have almost same-level performance, but the wavelet ‘bior4.4’ has a relatively low performance for the six profiles.\nWe further demonstrate the superiority of MICA-SVM by comparing it with three state-of-the-art partial least square (PLS) based regression methods, which can be found in the Additional file 3. Moreover, we present a novel algorithm stability analysis for the seven classifications and show the advantages of the MICA-SVM and MICA-LDA algorithms over the others (Please see the Additional file 4 for details).", "In addition to classifying large scale heterogeneous tumor profiles with exceptional performance, multi-resolution independent component analysis can be also applied to capture biomarkers for microarray profiles. We present a MICA-based filter-wrapper biomarker capturing algorithm and apply it to the stroma data. The details of this algorithm can be found in the Additional file 5. Table 4 lists the details on all the three biomarkers captured, where the SVM-rate for each biomarker is the classification ratio achieved by a SVM classifier with the ‘rbf’ kernel on the biomarker under leave-one-out cross validations. The order of the three biomarkers in Table 4 is listed according to its order identified in the biomarker discovery process. The SVM accuracy under the three biomarkers is 97.87% and the corresponding sensitivity and specificity are 92.31% and 100% respectively. The first biomarker is gene USP46, which is a broadly expressed gene reported as one gene associated with breast cancer and glioblastomas [18]. The second biomarker is FOSL2, which is one of four members in the Fos gene family. It is responsible for encoding leucine zipper proteins, which is able to dimerize with proteins of the JUN family, and form the transcription factor complex AP-1. As a regulator in cell proliferation, differentiation, and transformation, recent studies [19,20] have showed that it is one of important genes associated with breast cancer, by being involved in the regulation of breast cancer invasion and metastasis. The third biomarker is gene RPL5, which encodes a ribosomal protein that catalyzes protein synthesis. It was reported to associate with biosynthesis and energy utilization that is a cellular function associated with pathogenesis of breast cancer [21]. In addition, it also links to the breast cancer by lowering MDM2, which is a major regulator of p53 levels, preventing p53 ubiquitination and increasing its transcriptional activity [22]. Figure 6 visualizes the 47 samples (13 inflammatory breast cancers ('ibc') and 34 non-inflammatory breast cancers ('non-ibc')) of the stroma data using the three biomarkers. It is interesting to see that two types of cancers are separated into two spatially disjoint sets clearly, though one ‘ibc’ sample is wired in the ‘non-ibc’ samples.\nThree biomarkers discovered for the stroma data\nBiomarker visualization in the stroma data. Visualization of 47 samples in the stroma data by using three biomarkers", "It is worthy to note that independent component analysis is a necessary step to achieve a good classification performance. A similar multi-resolution principal component analysis based SVM algorithm is not able to reach comparable performance as our algorithm because of the loss of statistical independence in the feature selection. Also, MICA-SVM encounters overfitting as SVM, PCA-SVM, ICA-SVM classifiers under the standard Gaussian kernel (‘rbf’), where each learning machine can only recognize the majority type samples of the training data in classification despite the testing sample type. Moreover, we have tried kernel ICA [23] based support vector machines (KICA-SVM) in our experiments in addition to the previous nine comparison algorithms. However, The KICA-SVM classifier generally has a lower performance level than the standard SVM classifier. Furthermore, the KICA-SVM not only shows a strong instability in classification but also inevitably encounters overfitting under the standard Gaussian kernel like the other learning machines. It seems to suggest that kernel based data reduction may not be a desirable approach in effective feature selection for high dimensional heterogeneous gene profiles. Similar results can be also found in kernel PCA [24] based support vector machine (KPCA-SVM) classifications: a KPCA-SVM classifier is essentially the PCA-SVM classifier when its two kernels are selected as ‘linear’, otherwise, it encounters overfitting under the standard Gaussian kernel. In our ongoing project, in addition to further polishing our algorithm by comparing them with other state-of-the-art methods (e.g., SVM-RFE [2]), we are interested in theoretically validating the MICA-SVM‘s advantages over the classic SVM classifier from the viewpoint of Vapnik–Chervonenkis (VC) dimension theory [10].", "In this study, we present a novel multi-resolution feature selection algorithm: multi-resolution independent component analysis for effective feature selection for high-dimensional heterogeneous gene expression profiles, propose a high-performance MICA-SVM classification algorithm, and demonstrate its superiority and stability by comparing it with the nine state-of-the-art algorithms. Our algorithm not only consistently demonstrates the high-accuracy or clinical-level cancer diagnosis by treating an input profile a whole biomarker but also shows effectiveness in meaningful biomarker discovery. It suggests a great potential to facilitate high-throughput microarray technology into a clinical routine, especially, current classification methods have relative low even poor performance on the gene expression data. In addition, the multi-resolution data analysis based redundant global feature suppressing and effective local feature extraction will have a positive impact on large scale ‘omics’ data mining. In our future work, we plan to further explore MICA-SVM’s potential in other platform gene expression data, SNP, and protein expression data classification.", "The authors declare that they have no competing interests.", "HEY collects and processes the data, designs algorithms, implements the methods, and drafts paper. LXL participates in discussion and provides help to polish the paper. HEY and LXL jointly finalize the paper.", "Implementations of the four comparison algorithms PCA-LDA, PCA-SVM, ICA-SVM, and NMF-SVM algorithm implementations\nClick here for file\nMICA-LDA performance MICA-LDA performance under 100 trials of 50% HOCV and 10-fold CV\nClick here for file\nComparing MICA-SVM with PLS-based regression methods Comparisons MICA-SVM with three partial least square (PLS) based regression methods\nClick here for file\nAlgorithmic stability analysis\nClick here for file\nMICA-based biomarker discovery algorithms\nClick here for file" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Chromosome Mapping", "DNA, Complementary", "Exons", "Genome, Viral", "High-Throughput Nucleotide Sequencing", "RNA Splicing", "Sequence Alignment", "Software", "User-Computer Interface" ]

[ "Background", "Collection of genomes and sequences reads", "Extraction of exon-exon junction sequences", "Sequence reads processing and mapping", "Prediction of viral miRNAs", "cDNA preparation", "Experimental validation of discrete EEJ", "Results and discussion", "UMARS pipeline and interface", "Case study and demonstration of UMARS:Vir", "Case study and demonstration of UMARS:EEJ", "Authors’ contributions", "Competing interests" ]

[ "Biomedical research has been greatly accelerated by the advances in sequencing technologies, especially genomic research. Recently, next-generation sequencing (NGS) technology, including Roche 454, Illumina GA and ABI SOLiD platforms, has emerged as a powerful tool for generating high-throughput sequencing data. Systematic evaluation revealed that these three platforms could possess high sequencing sensitivity because of the large number of reads obtained [1]. Therefore, NGS technology has been applied in many studies, including transcriptome profiling [2-4], SNP identification [5,6], genome sequencing and re-sequencing [7,8], biomarker detection [9], and metagenomics [10,11]. NGS technology was also applied in miRNA identification and profiling studies. Morin and colleagues identified 104 novel human miRNA genes and made a list of miRNAs differentially expressed between embryo cell libraries [12]. Glazov discovered 449 new chicken miRNAs and 39 mirtrons [13]. In addition, Wheeler not only sequenced miRNAs from several metazoan genomes but also studied miRNA’s evolution status [14].\nIn a typical analysis pipeline, the generated NGS sequence reads are first subject to adaptor trimming and then mapping back to reference sequences, including genomes, scaffolds or transcripts. Several tools, including blast [15], Razers [16], SeqMap [17], SOAP2 [18], BWA [19], MAQ [20] and Bowtie [21], have been used for such mapping. Following the mapping step, the NGS reads are further processed to meet specific experimental interrogations. While it is essential to process the mappable reads in subsequent studies, a fraction of sequence reads cannot be mapped back to reference sequences. In many cases, these un-mappable reads are imputed to sequencing errors and discarded without further consideration. With the rapid increase of NGS reads, we intend to examine the possible biological relevance and possible sources of un-mappable reads. Therefore, we have developed the Un-MAppable Reads Solution (UMARS) pipeline in this study. Although un-mappable reads could originate from platform-specific technique errors, there have been reports demonstrating the possibilities of viral genomic sequences or cryptic splicing isoforms in NGS data [22,23].\nEukaryotic organisms are often infected by different viruses, leading to stable symbiosis or parasitism. As parasites, the infecting viruses rule the infected cells to produce their own genetic materials. Therefore, the collected RNA samples could be contaminated by viral transcripts when tissue or cells are lysed, which produces un-mappable reads when only the host cell genome is used for mapping. Kreuze et al detected virus infection by deep sequencing of viral small RNAs [22]. They concluded that NGS technology can be a method for diagnosis and discovery of virus infections. Wu et al also reached a similar conclusion [24]. In Kreuze’s study, in addition to the expected infecting viruses, unexpected novel virus reads and unidentified sequence reads also accounted for a large fraction of all reads. The results from these studies demonstrate that the genomic sequences from infecting viruses may contribute to un-mappable reads, and NGS technology is useful for systematic examination of putative viral genomes.\nAnother possible source of un-mappable reads is cryptic splicing isoforms. During gene expression, eukaryotic genes usually undergo mRNA splicing by removing introns and merging exons. The sequence reads located at the exon-exon junctions of novel alternative splicing isoforms can be mapped back neither to the genome nor to reference mRNAs. For example, Trapnell et al. could identify novel wobble splicing junctions from NGS reads [23]. However, there are no specific tools for discovering cryptic alternative splicing exon-exon junctions from large numbers of NGS reads.\nAt present, there is no biological user-friendly bioinformatic tool or service available focusing on the scanning of viral genomic regions or novel alternative splicing exon-exon junctions from un-mappable reads. We believe that such a tool would be beneficial for biological science researchers.", "For mapping sequence reads back to viral genomes, we first downloaded 3602 viral genomic sequences from NCBI RefSeq 40 [25]. According to the categories of their hosts, these viral genomes were classified into five classes, including animals, plants, fungi, protozoan plus algae and bacteria plus archaea. We also downloaded the genomic sequences of several animal species from the UCSC genome browser database [26] for extracting exon-exon junctions. The genomic versions of these species are listed in Additional file 1. In this study, sequence reads from NGS technology of several libraries were used. The sequencing platform, RNA source species, and RNA source tissue of these libraries are listed in Additional file 2.", "During maturation, eukaryotic genes usually undergo messenger RNA splicing, producing many alternative splicing isoforms from one gene. UCSC mapped these splicing isoforms back to genomes, and determined the boundaries and genomic coordinates of exons, recording the information in refFlat files. As shown in Fig. 1, from such coordinate information, we may exactly define the boundaries of exons and introns. Further, we may also define the exon-end fragments at exon’s both termini, start (S) or end (E). By extracting and assembling the exon-end fragments, from continuous or discrete exons, we collected 60-nt exon-exon junctions (EEJs). Therefore, they are either continuous or discrete EEJs, where the former denote known splicing patterns and the latter denote novel ones. As a result, the number of 60-nt EEJs is for each transcript, where n is the number of exon in each transcript. By doing so, we collected 60-nt EEJs for 21 species. The refFlat versions, number of EEJs and scientific names of these 21 species are available in Additional file 1.\nCollection of exon-exon junctions. By our definition, the EEJs can be continuous or discrete. The former represent known alternative splicing products. The latter, however, represent novel alternative splicing isoforms.", "NGS technologies have produced millions of reads, some of which may occur with high frequency. Such high-occurrence reads cause redundancy problems, and should be solved first. Therefore, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve this redundancy problem. NRP identifies unique sequence reads from input data, assigned a new ID and tabulates the occurrence frequency (copy number) of each unique read. After NRP processing, non-redundant un-mappable reads may be mapped back to viral genomes or EEJs by UMARS. In the studies involved in mapping sequence reads back to genomes, 100% identity is usually demanded [12,13]. Because viral genomes usually have higher mutation rates than eukaryotic ones, we allowed one nucleotide variation, including mismatch and gap, when mapping back to viral genomes or to EEJ sequences. The mapping procedures in this study were done with blast [15].", "After the mapping procedure, the viral genomic loci mapped by sequence reads are considered as candidate miRNAs. These genomic loci and their flanking sequences were extracted, followed by alignment using miRNA identification pipeline [27]. For each candidate miRNA, the pipeline first calculated the values of ten features, which serve as discrimination indices in a Support Vector Machine (SVM) algorithm. Then, the SVM was used as a classifier to classify candidate miRNAs into positive or negative sets.", "In this study, we used sequence reads from L2 library (Additional file 2) to scan the EEJs of novel alternative splicing isoforms, followed by experimental validation of the detected EEJs in 23 human tissues. Bellow we described how to prepare cDNAs from these tissues. Human tissue poly(A) RNAs (5μg) or total RNAs (40μg) purchased from Clontech (Clontech, Palo Alto, CA) were reverse-transcribed by Transcriptor reverse transcriptase (Roche Applied Science), primed by oligo (dT)15 according to the supplier's instructions. After the reverse transcription reaction, the mixtures were phenol-extracted once, followed by chloroform extraction. Excess primers were removed by applying the mixtures to Chroma Spin-200 (Clontech) gel filtration column. The purified cDNAs were properly diluted and subjected to Polymerase Chain Reaction (PCR) as the amplification templates. In this study, cDNAs from 23 tissues were investigated and they were labeled as follow: M: DNA marker, 1: blood, 2: bone marrow, 3: brain, 4: colon, 5: heart, 6: kidney, 7: liver, 8: lung, 9: ovary, 10: pancreas, 11: placenta, 12: skeletal muscle, 13: small intestine, 14: stomach, 15: testis, 16: whole fetus, 17: breast tumor, 18: cervix tumor, 19: colon tumor, 20: kidney tumor, 21: lung tumor, 22: ovary tumor, 23: gastric tumor, 24: PCR no-template control.", "The dtetcted EEJs were verified by PCR amplification, followed by capillary sequencing confirmation. Primer pair sequences were picked from each couple of the “discrete” EEJ-spanning exons and were listed in Additional file 3. PCR components include mainly 1mM dNTP, 1μM primer separately, 0.1U Takara Taq DNA polymerase (Takara) per 10μL reaction volume, and the diluted cDNA. The thermal reaction was set at 94°C for 3 minutes, 40 cycles (GAPDH was run specifically for 30 cycles) of denaturing at 94°C for 20 seconds, annealing at 58°C for 30 seconds, and extension at 72°C for 30 seconds, finally at 72°C for 10 minutes. PCR products were separated by 3% NuSieve (Lonza, Rockland, ME) conventional TAE-Agarose gel, and visualized through the ultraviolet light source. The detected and the estimated target size regions of the gel were cut-out and the nucleic-acid contents were purified by Viogene Gel Purification reagents. Minor bands eluted were further subjected to additional 30 PCR cycles with the same pair of primer. The amplified nucleic acid fragments were directly sequenced by ABI 3730xl DNA Analyzer (Applied Biosystems).", "[SUBTITLE] UMARS pipeline and interface [SUBSECTION] The main concept of UMARS is to provide a user-friendly solution for biologists to retrieve valuable information from discarded NGS reads that could not be mapped back to a reference genome. We have initially produced two major datasets for interrogation, a virus genome sequence collection and an animal exon-exon junction sequence collection. Since many NGS studies intend to study miRNAs, we also provided an additional portal for a miRNA discovery pipeline. The overall UMARS schema is illustrated in Fig. 2. The first step of UMARS deals with the redundancy problem of NGS reads. For convenience and efficiency in data processing and network traffic, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve the problem. The reads uploaded to UMARS must be processed by NRP in advance or be presented in the format of NRP output. NRP accepts the files containing unmapped read sequences in the FASTA format. An NRP standalone program can be downloaded from the UMARS website. Next, the uploaded non-redundant NGS reads could be further processed by either UMARS:EEJ or UMARS:Vir.\nFlowchart of UMARS service. The UMARS service can be divided into two subservices: for discovering viral genomic regions (UMARS:Vir); and discovering novel alternative splicing exon-exon junctions (UMARS:EEJ). Un-mappable reads must be processed by NRP to solving redundancy problem before uploaded to UMARS.\nThe purpose of UMARS:EEJ is to identify novel alternative splicing exon-exon junctions (EEJs) from un-mappable reads. The sequences of all possible EEJs of 21 species were collected in advance. In UMARS:EEJ, uploaded reads are mapped to EEJs. To avoid random sequence matches, besides our mapping criteria (see Materials and methods), a mapping match must overlap both exons for at least five nucleotides, not skewing too much to either exon. Following the mapping procedure, UMARS tabulats detected EEJs and their expression levels. The detected EEJs are reported as either continuous or discrete EEJs. Continuous EEJs represent known mRNA transcripts. However, discrete EEJs could represent novel splicing isoforms.\nThe purpose of UMARS:Vir is to identify possible virus genomic regions from un-mappable NGS reads. In UMARS:Vir, the uploaded NGS reads are mapped to all 3,602 known virus genomes. Following the mapping procedure, UMARS tabulates detected virus species and their expression levels. The detected viral genomic regions may locate at intergenic, protein-coding gene, pre-miRNA regions and so on according to the annotations of RefSeq 40 and miRBase 15. Such information of genomic annotation is also provided by the UMARS service. Several viruses are reported to encode viral miRNAs, regulating expression of host genes and playing important roles in host immune misfunctions [28-30]. Therefore, UMARS:Vir may further have the option to detect viral miRNAs by an additional miRNA identification pipeline from viral intergenic genomic regions.\nThe UMARS service is freely available to the academic community. Users may access UMARS via http://musk.ibms.sinica.edu.tw/UMARS/. The non-redundant reads uploaded into UMARS must not exceed 10 MB in size. Such size limitation has nothing to do with pipeline performance but reduces the network load. As shown in Fig. 3a, in UMARS:Vir, users can select the host category (animals, plants, fungi, protozoan or bacteria) of their expected virus. In UMARS:EEJ, users can select the species corresponding to the EEJs (Fig. 3b). The UMARS results will be sent back to the users via e-mail.\nInterface of UMARS service. UMARS:Vir and UMARS:EEJ have individual parameters. Users must adjust the parameters according to their specific data source to obtain optimal results. In UMARS:Vir service, users must specify the value, Standard or Loose, of “Parameter Criteria” parameter. Specifying Standard outputs only the viruses with CN >= 100 and RN >= 10 (see Table 1); while specifying Loose outputs all virus with CN >= 1 and RNA >= 1. This is an empirical criterion to reduce random match.\nSummary of viruses detected from L1 un-mappable reads.\nCN denotes copy number of total reads mappable to the virus; RN denotes the number of distinct genomic regions mapped by reads. The values in parentheses denote the corresponding values when no mismatch was allowed in the mapping procedure. Because their high similarity, most of the reads of type 1 EBV were also mapped back to type 2 EBV.\nThe main concept of UMARS is to provide a user-friendly solution for biologists to retrieve valuable information from discarded NGS reads that could not be mapped back to a reference genome. We have initially produced two major datasets for interrogation, a virus genome sequence collection and an animal exon-exon junction sequence collection. Since many NGS studies intend to study miRNAs, we also provided an additional portal for a miRNA discovery pipeline. The overall UMARS schema is illustrated in Fig. 2. The first step of UMARS deals with the redundancy problem of NGS reads. For convenience and efficiency in data processing and network traffic, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve the problem. The reads uploaded to UMARS must be processed by NRP in advance or be presented in the format of NRP output. NRP accepts the files containing unmapped read sequences in the FASTA format. An NRP standalone program can be downloaded from the UMARS website. Next, the uploaded non-redundant NGS reads could be further processed by either UMARS:EEJ or UMARS:Vir.\nFlowchart of UMARS service. The UMARS service can be divided into two subservices: for discovering viral genomic regions (UMARS:Vir); and discovering novel alternative splicing exon-exon junctions (UMARS:EEJ). Un-mappable reads must be processed by NRP to solving redundancy problem before uploaded to UMARS.\nThe purpose of UMARS:EEJ is to identify novel alternative splicing exon-exon junctions (EEJs) from un-mappable reads. The sequences of all possible EEJs of 21 species were collected in advance. In UMARS:EEJ, uploaded reads are mapped to EEJs. To avoid random sequence matches, besides our mapping criteria (see Materials and methods), a mapping match must overlap both exons for at least five nucleotides, not skewing too much to either exon. Following the mapping procedure, UMARS tabulats detected EEJs and their expression levels. The detected EEJs are reported as either continuous or discrete EEJs. Continuous EEJs represent known mRNA transcripts. However, discrete EEJs could represent novel splicing isoforms.\nThe purpose of UMARS:Vir is to identify possible virus genomic regions from un-mappable NGS reads. In UMARS:Vir, the uploaded NGS reads are mapped to all 3,602 known virus genomes. Following the mapping procedure, UMARS tabulates detected virus species and their expression levels. The detected viral genomic regions may locate at intergenic, protein-coding gene, pre-miRNA regions and so on according to the annotations of RefSeq 40 and miRBase 15. Such information of genomic annotation is also provided by the UMARS service. Several viruses are reported to encode viral miRNAs, regulating expression of host genes and playing important roles in host immune misfunctions [28-30]. Therefore, UMARS:Vir may further have the option to detect viral miRNAs by an additional miRNA identification pipeline from viral intergenic genomic regions.\nThe UMARS service is freely available to the academic community. Users may access UMARS via http://musk.ibms.sinica.edu.tw/UMARS/. The non-redundant reads uploaded into UMARS must not exceed 10 MB in size. Such size limitation has nothing to do with pipeline performance but reduces the network load. As shown in Fig. 3a, in UMARS:Vir, users can select the host category (animals, plants, fungi, protozoan or bacteria) of their expected virus. In UMARS:EEJ, users can select the species corresponding to the EEJs (Fig. 3b). The UMARS results will be sent back to the users via e-mail.\nInterface of UMARS service. UMARS:Vir and UMARS:EEJ have individual parameters. Users must adjust the parameters according to their specific data source to obtain optimal results. In UMARS:Vir service, users must specify the value, Standard or Loose, of “Parameter Criteria” parameter. Specifying Standard outputs only the viruses with CN >= 100 and RN >= 10 (see Table 1); while specifying Loose outputs all virus with CN >= 1 and RNA >= 1. This is an empirical criterion to reduce random match.\nSummary of viruses detected from L1 un-mappable reads.\nCN denotes copy number of total reads mappable to the virus; RN denotes the number of distinct genomic regions mapped by reads. The values in parentheses denote the corresponding values when no mismatch was allowed in the mapping procedure. Because their high similarity, most of the reads of type 1 EBV were also mapped back to type 2 EBV.\n[SUBTITLE] Case study and demonstration of UMARS:Vir [SUBSECTION] To demonstrate the utility of UMARS, we have analyzed NGS reads using UMARS:Vir. In the first case, we investigated the un-mappable reads from the human NPC cells (L1 library) infected with Epstein-Barr virus (EBV, also named human herpes virus 4 type 1). We examined whether the expected EBV genome could be detected by UMARS:Vir. As a result, eight viruses were detected under our mapping criteria. As shown in Table 1, the expected EBV matches dominated over other un-expected viruses in terms of expression level, which shows that UMARS:Vir can be used to detect infections by a specific virus from un-mappable reads. Besides EBV, there were seven un-expected viruses detected, most of which infect primates, and all of them belong to the herpes virus family.\nAs demonstrated in Table 1, 105,325 reads were mapped back to EBV under our mapping criteria. These EBV reads spanned the EBV genome at 629 genomic regions. The genomic contents, intergenic or intragenic, and detailed information about these regions are listed in Additional file 4. As shown in Table 2 and Additional file 4, 30.1% of these regions are located at the inter-genic regions; 60.1% at the EBV pre-miRNA regions; and only 9.1% at the protein-coding regions. As well, 21 out of 98 protein-coding genes and 22 (excluding mir-BHRF1-1, mir-BHRF1-2, mir-BHRF1-3) out of 25 pre-miRNAs were detected, respectively. This miRNA dominance phenomenon was observed because the L1 sample RNA was extracted with a small RNA kit rather than with a transcriptome kit.\nEBV genomic regions mapped by reads.\nThe regions mapped by reads were classified into three categories according to the annotation of RefSeq 40 and miRBase 15. pre-miRNA regions dominated over other regions in terms of copy number and region number.\nBecause of the strong sequencing intensity, additional mature miRNAs from the same precursors, including isomiRs at the same arm and minor forms of mature miRNA at the opposite arm, are usually detected from NGS reads [12,13,31]. After arranging miRNA reads in order within their corresponding pre-miRNAs, we observed many isomiRs at all of the 22 detected pre-miRNAs (Additional file 5). Compared with EBV pre-miRNAs in miRBase 15, the reference mature miRNAs do not always represent the most abundant reads. In addition, according to miRBase 15 annotation, mir-BART12, mir-BART16 and mir-BART22 encode mature miRNAs only at their 3’, 5’ and 3’ arm respectively. However, we detected additional mature miRNAs at the 5’ arm of mir-BART12, the 3’ arm of mir-BART16 and the 5’ arm of mir-BART22. Moreover, the 5’ arm of mir-BART12 and the 3’ arm of mir-BART16 encode more reads than the original arms. This result is similar to that in Wheeler’s report [14] and should be noted in future data updates of the miRBase.\nTo demonstrate the utility of UMARS, we have analyzed NGS reads using UMARS:Vir. In the first case, we investigated the un-mappable reads from the human NPC cells (L1 library) infected with Epstein-Barr virus (EBV, also named human herpes virus 4 type 1). We examined whether the expected EBV genome could be detected by UMARS:Vir. As a result, eight viruses were detected under our mapping criteria. As shown in Table 1, the expected EBV matches dominated over other un-expected viruses in terms of expression level, which shows that UMARS:Vir can be used to detect infections by a specific virus from un-mappable reads. Besides EBV, there were seven un-expected viruses detected, most of which infect primates, and all of them belong to the herpes virus family.\nAs demonstrated in Table 1, 105,325 reads were mapped back to EBV under our mapping criteria. These EBV reads spanned the EBV genome at 629 genomic regions. The genomic contents, intergenic or intragenic, and detailed information about these regions are listed in Additional file 4. As shown in Table 2 and Additional file 4, 30.1% of these regions are located at the inter-genic regions; 60.1% at the EBV pre-miRNA regions; and only 9.1% at the protein-coding regions. As well, 21 out of 98 protein-coding genes and 22 (excluding mir-BHRF1-1, mir-BHRF1-2, mir-BHRF1-3) out of 25 pre-miRNAs were detected, respectively. This miRNA dominance phenomenon was observed because the L1 sample RNA was extracted with a small RNA kit rather than with a transcriptome kit.\nEBV genomic regions mapped by reads.\nThe regions mapped by reads were classified into three categories according to the annotation of RefSeq 40 and miRBase 15. pre-miRNA regions dominated over other regions in terms of copy number and region number.\nBecause of the strong sequencing intensity, additional mature miRNAs from the same precursors, including isomiRs at the same arm and minor forms of mature miRNA at the opposite arm, are usually detected from NGS reads [12,13,31]. After arranging miRNA reads in order within their corresponding pre-miRNAs, we observed many isomiRs at all of the 22 detected pre-miRNAs (Additional file 5). Compared with EBV pre-miRNAs in miRBase 15, the reference mature miRNAs do not always represent the most abundant reads. In addition, according to miRBase 15 annotation, mir-BART12, mir-BART16 and mir-BART22 encode mature miRNAs only at their 3’, 5’ and 3’ arm respectively. However, we detected additional mature miRNAs at the 5’ arm of mir-BART12, the 3’ arm of mir-BART16 and the 5’ arm of mir-BART22. Moreover, the 5’ arm of mir-BART12 and the 3’ arm of mir-BART16 encode more reads than the original arms. This result is similar to that in Wheeler’s report [14] and should be noted in future data updates of the miRBase.\n[SUBTITLE] Case study and demonstration of UMARS:EEJ [SUBSECTION] In addition to virus infections, splicing exon-exon junctions of gene transcripts may also contribute to un-mappable reads. We, therefore, had the un-mappable reads from human H23 cells (L2 library) processed by UMARS:EEJ. As a result, 35,915 un-mappable reads were mapped back to 4,254 EEJs, extracted from 2,738 transcripts (Table 3). On further examination, 70.2% of the detected EEJs were continuous ones, which indicated they were the major mRNA transcription forms.\nSummary of EEJs detected from L2 un-mappable reads.\nThe digits in parenthesis denote the values when no mismatch was allowed in the mapping procedure. Continuous EEJ dominated over discrete ones in terms of read copy number and EST abundance.\nHowever, there were 1,269 discrete EEJs detected by UMARS:EEJ from the L2 library. We randomly selected five detected EEJs for PCR validation and three of them were successfully validated. Referring to the UCSC Genome Browser, some discrete EEJs had EST evidence, although these splicing isoforms were not deposited in RefSeq. As shown in Fig. 4a, four ESTs, including BU934587, D80989, D52190 and D53867, matched the detected EEJ from NM_007108 (4:2-4), which was assembled from the exon-end fragments of exon 2 and exon 4. We further designed a PCR experiment in which the forward and reverse primers were located at exon 2 and exon 4, respectively (see Materials and methods). As a result, we succeeded in detecting the expected fragment (d transcript) of the alternative transcript in most of the 23 tissues (Fig. 4b and Fig. 4c), which demonstrated that the EEJ, NM_007108 (4:2-4), occurred almost universally. The sequencing result also proved the authenticity of this discrete transcript (Fig. 4d). Another two detected EEJs, NM_021019(7:2-4) and NM_178580(13:10-13), were also successfully validated (Additional file 6 and Additional file 7).\nPCR experimental validation of the detected EEJ, NM_007108 (4:2-4). The marker lane is labeled with M and is presented as 50bp ladder. The central bright band in the marker lane is equal to 350 bp. (a) UCSC Genome Browser shows that four ESTs match the detected EEJ. (b, c) The PCR result showed that the expected EEJ (d transcript) can be experimentally detected. c and d denotes continuous and discrete transcripts, respectively. (d) The sequencing result provided the authenticity of the detected EEJs.\nTable 4 demonstrates the detailed output sample of UMARS:EEJ alignment. The first value “19:6-8” in the “EEJ pattern” column demonstrates that NM_001009566 has 19 exons and the detected EEJ was assembled from the exon-end fragments of exon 6 and exon 8, leading to a discrete EEJ. The first value “14 + 5” in the “Mapping pattern” column demonstrates that the read was split into 14-nt and 5-nt fragments; the former mapped to exon 6 and the latter to exon 8. CN and MM denote the value of the expression level and the number of variations. Because one-nucleotide variation was allowed, three unique reads were mapped back to the same position of the same EEJ. NM_014944 (18:5-7) was also detected owing to alternative splicing from the same gene. All results of our L2 library case are listed in Additional file 8.\nDetected EEJs from CLSTN1.\nNM_001009566 and NM_014944 are splicing isoforms of CLSTN1. The EEJ (19:6-8) from NM_001009566 is identical to the EEJ (18:5-7) from NM_014944.\nMYL6 (myosin light polypeptide 6) encodes a myosin alkali light chain and is associated with cell migration [32]. It was also reported that fibroblasts promote the growth of breast tumor cells by enhancing the expression of several genes, including MYL6 [33]. In this study, c alternative transcript and d transcript of MYL6 have similar expression levels in most normal tissues (Additional file 6). However, the d transcript dominates over c alternative transcript in most tumor tissues, including breast tumor (17th lane in Additional file 6a). It is possible that these alternative splicing isoforms function differently with each other and are associated with tumor genesis.\nIn addition to virus infections, splicing exon-exon junctions of gene transcripts may also contribute to un-mappable reads. We, therefore, had the un-mappable reads from human H23 cells (L2 library) processed by UMARS:EEJ. As a result, 35,915 un-mappable reads were mapped back to 4,254 EEJs, extracted from 2,738 transcripts (Table 3). On further examination, 70.2% of the detected EEJs were continuous ones, which indicated they were the major mRNA transcription forms.\nSummary of EEJs detected from L2 un-mappable reads.\nThe digits in parenthesis denote the values when no mismatch was allowed in the mapping procedure. Continuous EEJ dominated over discrete ones in terms of read copy number and EST abundance.\nHowever, there were 1,269 discrete EEJs detected by UMARS:EEJ from the L2 library. We randomly selected five detected EEJs for PCR validation and three of them were successfully validated. Referring to the UCSC Genome Browser, some discrete EEJs had EST evidence, although these splicing isoforms were not deposited in RefSeq. As shown in Fig. 4a, four ESTs, including BU934587, D80989, D52190 and D53867, matched the detected EEJ from NM_007108 (4:2-4), which was assembled from the exon-end fragments of exon 2 and exon 4. We further designed a PCR experiment in which the forward and reverse primers were located at exon 2 and exon 4, respectively (see Materials and methods). As a result, we succeeded in detecting the expected fragment (d transcript) of the alternative transcript in most of the 23 tissues (Fig. 4b and Fig. 4c), which demonstrated that the EEJ, NM_007108 (4:2-4), occurred almost universally. The sequencing result also proved the authenticity of this discrete transcript (Fig. 4d). Another two detected EEJs, NM_021019(7:2-4) and NM_178580(13:10-13), were also successfully validated (Additional file 6 and Additional file 7).\nPCR experimental validation of the detected EEJ, NM_007108 (4:2-4). The marker lane is labeled with M and is presented as 50bp ladder. The central bright band in the marker lane is equal to 350 bp. (a) UCSC Genome Browser shows that four ESTs match the detected EEJ. (b, c) The PCR result showed that the expected EEJ (d transcript) can be experimentally detected. c and d denotes continuous and discrete transcripts, respectively. (d) The sequencing result provided the authenticity of the detected EEJs.\nTable 4 demonstrates the detailed output sample of UMARS:EEJ alignment. The first value “19:6-8” in the “EEJ pattern” column demonstrates that NM_001009566 has 19 exons and the detected EEJ was assembled from the exon-end fragments of exon 6 and exon 8, leading to a discrete EEJ. The first value “14 + 5” in the “Mapping pattern” column demonstrates that the read was split into 14-nt and 5-nt fragments; the former mapped to exon 6 and the latter to exon 8. CN and MM denote the value of the expression level and the number of variations. Because one-nucleotide variation was allowed, three unique reads were mapped back to the same position of the same EEJ. NM_014944 (18:5-7) was also detected owing to alternative splicing from the same gene. All results of our L2 library case are listed in Additional file 8.\nDetected EEJs from CLSTN1.\nNM_001009566 and NM_014944 are splicing isoforms of CLSTN1. The EEJ (19:6-8) from NM_001009566 is identical to the EEJ (18:5-7) from NM_014944.\nMYL6 (myosin light polypeptide 6) encodes a myosin alkali light chain and is associated with cell migration [32]. It was also reported that fibroblasts promote the growth of breast tumor cells by enhancing the expression of several genes, including MYL6 [33]. In this study, c alternative transcript and d transcript of MYL6 have similar expression levels in most normal tissues (Additional file 6). However, the d transcript dominates over c alternative transcript in most tumor tissues, including breast tumor (17th lane in Additional file 6a). It is possible that these alternative splicing isoforms function differently with each other and are associated with tumor genesis.", "The main concept of UMARS is to provide a user-friendly solution for biologists to retrieve valuable information from discarded NGS reads that could not be mapped back to a reference genome. We have initially produced two major datasets for interrogation, a virus genome sequence collection and an animal exon-exon junction sequence collection. Since many NGS studies intend to study miRNAs, we also provided an additional portal for a miRNA discovery pipeline. The overall UMARS schema is illustrated in Fig. 2. The first step of UMARS deals with the redundancy problem of NGS reads. For convenience and efficiency in data processing and network traffic, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve the problem. The reads uploaded to UMARS must be processed by NRP in advance or be presented in the format of NRP output. NRP accepts the files containing unmapped read sequences in the FASTA format. An NRP standalone program can be downloaded from the UMARS website. Next, the uploaded non-redundant NGS reads could be further processed by either UMARS:EEJ or UMARS:Vir.\nFlowchart of UMARS service. The UMARS service can be divided into two subservices: for discovering viral genomic regions (UMARS:Vir); and discovering novel alternative splicing exon-exon junctions (UMARS:EEJ). Un-mappable reads must be processed by NRP to solving redundancy problem before uploaded to UMARS.\nThe purpose of UMARS:EEJ is to identify novel alternative splicing exon-exon junctions (EEJs) from un-mappable reads. The sequences of all possible EEJs of 21 species were collected in advance. In UMARS:EEJ, uploaded reads are mapped to EEJs. To avoid random sequence matches, besides our mapping criteria (see Materials and methods), a mapping match must overlap both exons for at least five nucleotides, not skewing too much to either exon. Following the mapping procedure, UMARS tabulats detected EEJs and their expression levels. The detected EEJs are reported as either continuous or discrete EEJs. Continuous EEJs represent known mRNA transcripts. However, discrete EEJs could represent novel splicing isoforms.\nThe purpose of UMARS:Vir is to identify possible virus genomic regions from un-mappable NGS reads. In UMARS:Vir, the uploaded NGS reads are mapped to all 3,602 known virus genomes. Following the mapping procedure, UMARS tabulates detected virus species and their expression levels. The detected viral genomic regions may locate at intergenic, protein-coding gene, pre-miRNA regions and so on according to the annotations of RefSeq 40 and miRBase 15. Such information of genomic annotation is also provided by the UMARS service. Several viruses are reported to encode viral miRNAs, regulating expression of host genes and playing important roles in host immune misfunctions [28-30]. Therefore, UMARS:Vir may further have the option to detect viral miRNAs by an additional miRNA identification pipeline from viral intergenic genomic regions.\nThe UMARS service is freely available to the academic community. Users may access UMARS via http://musk.ibms.sinica.edu.tw/UMARS/. The non-redundant reads uploaded into UMARS must not exceed 10 MB in size. Such size limitation has nothing to do with pipeline performance but reduces the network load. As shown in Fig. 3a, in UMARS:Vir, users can select the host category (animals, plants, fungi, protozoan or bacteria) of their expected virus. In UMARS:EEJ, users can select the species corresponding to the EEJs (Fig. 3b). The UMARS results will be sent back to the users via e-mail.\nInterface of UMARS service. UMARS:Vir and UMARS:EEJ have individual parameters. Users must adjust the parameters according to their specific data source to obtain optimal results. In UMARS:Vir service, users must specify the value, Standard or Loose, of “Parameter Criteria” parameter. Specifying Standard outputs only the viruses with CN >= 100 and RN >= 10 (see Table 1); while specifying Loose outputs all virus with CN >= 1 and RNA >= 1. This is an empirical criterion to reduce random match.\nSummary of viruses detected from L1 un-mappable reads.\nCN denotes copy number of total reads mappable to the virus; RN denotes the number of distinct genomic regions mapped by reads. The values in parentheses denote the corresponding values when no mismatch was allowed in the mapping procedure. Because their high similarity, most of the reads of type 1 EBV were also mapped back to type 2 EBV.", "To demonstrate the utility of UMARS, we have analyzed NGS reads using UMARS:Vir. In the first case, we investigated the un-mappable reads from the human NPC cells (L1 library) infected with Epstein-Barr virus (EBV, also named human herpes virus 4 type 1). We examined whether the expected EBV genome could be detected by UMARS:Vir. As a result, eight viruses were detected under our mapping criteria. As shown in Table 1, the expected EBV matches dominated over other un-expected viruses in terms of expression level, which shows that UMARS:Vir can be used to detect infections by a specific virus from un-mappable reads. Besides EBV, there were seven un-expected viruses detected, most of which infect primates, and all of them belong to the herpes virus family.\nAs demonstrated in Table 1, 105,325 reads were mapped back to EBV under our mapping criteria. These EBV reads spanned the EBV genome at 629 genomic regions. The genomic contents, intergenic or intragenic, and detailed information about these regions are listed in Additional file 4. As shown in Table 2 and Additional file 4, 30.1% of these regions are located at the inter-genic regions; 60.1% at the EBV pre-miRNA regions; and only 9.1% at the protein-coding regions. As well, 21 out of 98 protein-coding genes and 22 (excluding mir-BHRF1-1, mir-BHRF1-2, mir-BHRF1-3) out of 25 pre-miRNAs were detected, respectively. This miRNA dominance phenomenon was observed because the L1 sample RNA was extracted with a small RNA kit rather than with a transcriptome kit.\nEBV genomic regions mapped by reads.\nThe regions mapped by reads were classified into three categories according to the annotation of RefSeq 40 and miRBase 15. pre-miRNA regions dominated over other regions in terms of copy number and region number.\nBecause of the strong sequencing intensity, additional mature miRNAs from the same precursors, including isomiRs at the same arm and minor forms of mature miRNA at the opposite arm, are usually detected from NGS reads [12,13,31]. After arranging miRNA reads in order within their corresponding pre-miRNAs, we observed many isomiRs at all of the 22 detected pre-miRNAs (Additional file 5). Compared with EBV pre-miRNAs in miRBase 15, the reference mature miRNAs do not always represent the most abundant reads. In addition, according to miRBase 15 annotation, mir-BART12, mir-BART16 and mir-BART22 encode mature miRNAs only at their 3’, 5’ and 3’ arm respectively. However, we detected additional mature miRNAs at the 5’ arm of mir-BART12, the 3’ arm of mir-BART16 and the 5’ arm of mir-BART22. Moreover, the 5’ arm of mir-BART12 and the 3’ arm of mir-BART16 encode more reads than the original arms. This result is similar to that in Wheeler’s report [14] and should be noted in future data updates of the miRBase.", "In addition to virus infections, splicing exon-exon junctions of gene transcripts may also contribute to un-mappable reads. We, therefore, had the un-mappable reads from human H23 cells (L2 library) processed by UMARS:EEJ. As a result, 35,915 un-mappable reads were mapped back to 4,254 EEJs, extracted from 2,738 transcripts (Table 3). On further examination, 70.2% of the detected EEJs were continuous ones, which indicated they were the major mRNA transcription forms.\nSummary of EEJs detected from L2 un-mappable reads.\nThe digits in parenthesis denote the values when no mismatch was allowed in the mapping procedure. Continuous EEJ dominated over discrete ones in terms of read copy number and EST abundance.\nHowever, there were 1,269 discrete EEJs detected by UMARS:EEJ from the L2 library. We randomly selected five detected EEJs for PCR validation and three of them were successfully validated. Referring to the UCSC Genome Browser, some discrete EEJs had EST evidence, although these splicing isoforms were not deposited in RefSeq. As shown in Fig. 4a, four ESTs, including BU934587, D80989, D52190 and D53867, matched the detected EEJ from NM_007108 (4:2-4), which was assembled from the exon-end fragments of exon 2 and exon 4. We further designed a PCR experiment in which the forward and reverse primers were located at exon 2 and exon 4, respectively (see Materials and methods). As a result, we succeeded in detecting the expected fragment (d transcript) of the alternative transcript in most of the 23 tissues (Fig. 4b and Fig. 4c), which demonstrated that the EEJ, NM_007108 (4:2-4), occurred almost universally. The sequencing result also proved the authenticity of this discrete transcript (Fig. 4d). Another two detected EEJs, NM_021019(7:2-4) and NM_178580(13:10-13), were also successfully validated (Additional file 6 and Additional file 7).\nPCR experimental validation of the detected EEJ, NM_007108 (4:2-4). The marker lane is labeled with M and is presented as 50bp ladder. The central bright band in the marker lane is equal to 350 bp. (a) UCSC Genome Browser shows that four ESTs match the detected EEJ. (b, c) The PCR result showed that the expected EEJ (d transcript) can be experimentally detected. c and d denotes continuous and discrete transcripts, respectively. (d) The sequencing result provided the authenticity of the detected EEJs.\nTable 4 demonstrates the detailed output sample of UMARS:EEJ alignment. The first value “19:6-8” in the “EEJ pattern” column demonstrates that NM_001009566 has 19 exons and the detected EEJ was assembled from the exon-end fragments of exon 6 and exon 8, leading to a discrete EEJ. The first value “14 + 5” in the “Mapping pattern” column demonstrates that the read was split into 14-nt and 5-nt fragments; the former mapped to exon 6 and the latter to exon 8. CN and MM denote the value of the expression level and the number of variations. Because one-nucleotide variation was allowed, three unique reads were mapped back to the same position of the same EEJ. NM_014944 (18:5-7) was also detected owing to alternative splicing from the same gene. All results of our L2 library case are listed in Additional file 8.\nDetected EEJs from CLSTN1.\nNM_001009566 and NM_014944 are splicing isoforms of CLSTN1. The EEJ (19:6-8) from NM_001009566 is identical to the EEJ (18:5-7) from NM_014944.\nMYL6 (myosin light polypeptide 6) encodes a myosin alkali light chain and is associated with cell migration [32]. It was also reported that fibroblasts promote the growth of breast tumor cells by enhancing the expression of several genes, including MYL6 [33]. In this study, c alternative transcript and d transcript of MYL6 have similar expression levels in most normal tissues (Additional file 6). However, the d transcript dominates over c alternative transcript in most tumor tissues, including breast tumor (17th lane in Additional file 6a). It is possible that these alternative splicing isoforms function differently with each other and are associated with tumor genesis.", "SCL and WCC executed this study and wrote the draft of this manuscript. WCC and CNH were responsible for Construction of UMARS interface. CHL did the jobs of cDNA preparation and PCR experiments. YSJ and HCC provided the SOLiD read data. KWT and CHC participated in discussion on research design. WCL supervised the study and edited the manuscript.", "The authors declare that they have no competing interests." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods and materials", "Collection of genomes and sequences reads", "Extraction of exon-exon junction sequences", "Sequence reads processing and mapping", "Prediction of viral miRNAs", "cDNA preparation", "Experimental validation of discrete EEJ", "Results and discussion", "UMARS pipeline and interface", "Case study and demonstration of UMARS:Vir", "Case study and demonstration of UMARS:EEJ", "Conclusion", "Authors’ contributions", "Competing interests", "Supplementary Material" ]

[ "Biomedical research has been greatly accelerated by the advances in sequencing technologies, especially genomic research. Recently, next-generation sequencing (NGS) technology, including Roche 454, Illumina GA and ABI SOLiD platforms, has emerged as a powerful tool for generating high-throughput sequencing data. Systematic evaluation revealed that these three platforms could possess high sequencing sensitivity because of the large number of reads obtained [1]. Therefore, NGS technology has been applied in many studies, including transcriptome profiling [2-4], SNP identification [5,6], genome sequencing and re-sequencing [7,8], biomarker detection [9], and metagenomics [10,11]. NGS technology was also applied in miRNA identification and profiling studies. Morin and colleagues identified 104 novel human miRNA genes and made a list of miRNAs differentially expressed between embryo cell libraries [12]. Glazov discovered 449 new chicken miRNAs and 39 mirtrons [13]. In addition, Wheeler not only sequenced miRNAs from several metazoan genomes but also studied miRNA’s evolution status [14].\nIn a typical analysis pipeline, the generated NGS sequence reads are first subject to adaptor trimming and then mapping back to reference sequences, including genomes, scaffolds or transcripts. Several tools, including blast [15], Razers [16], SeqMap [17], SOAP2 [18], BWA [19], MAQ [20] and Bowtie [21], have been used for such mapping. Following the mapping step, the NGS reads are further processed to meet specific experimental interrogations. While it is essential to process the mappable reads in subsequent studies, a fraction of sequence reads cannot be mapped back to reference sequences. In many cases, these un-mappable reads are imputed to sequencing errors and discarded without further consideration. With the rapid increase of NGS reads, we intend to examine the possible biological relevance and possible sources of un-mappable reads. Therefore, we have developed the Un-MAppable Reads Solution (UMARS) pipeline in this study. Although un-mappable reads could originate from platform-specific technique errors, there have been reports demonstrating the possibilities of viral genomic sequences or cryptic splicing isoforms in NGS data [22,23].\nEukaryotic organisms are often infected by different viruses, leading to stable symbiosis or parasitism. As parasites, the infecting viruses rule the infected cells to produce their own genetic materials. Therefore, the collected RNA samples could be contaminated by viral transcripts when tissue or cells are lysed, which produces un-mappable reads when only the host cell genome is used for mapping. Kreuze et al detected virus infection by deep sequencing of viral small RNAs [22]. They concluded that NGS technology can be a method for diagnosis and discovery of virus infections. Wu et al also reached a similar conclusion [24]. In Kreuze’s study, in addition to the expected infecting viruses, unexpected novel virus reads and unidentified sequence reads also accounted for a large fraction of all reads. The results from these studies demonstrate that the genomic sequences from infecting viruses may contribute to un-mappable reads, and NGS technology is useful for systematic examination of putative viral genomes.\nAnother possible source of un-mappable reads is cryptic splicing isoforms. During gene expression, eukaryotic genes usually undergo mRNA splicing by removing introns and merging exons. The sequence reads located at the exon-exon junctions of novel alternative splicing isoforms can be mapped back neither to the genome nor to reference mRNAs. For example, Trapnell et al. could identify novel wobble splicing junctions from NGS reads [23]. However, there are no specific tools for discovering cryptic alternative splicing exon-exon junctions from large numbers of NGS reads.\nAt present, there is no biological user-friendly bioinformatic tool or service available focusing on the scanning of viral genomic regions or novel alternative splicing exon-exon junctions from un-mappable reads. We believe that such a tool would be beneficial for biological science researchers.", "[SUBTITLE] Collection of genomes and sequences reads [SUBSECTION] For mapping sequence reads back to viral genomes, we first downloaded 3602 viral genomic sequences from NCBI RefSeq 40 [25]. According to the categories of their hosts, these viral genomes were classified into five classes, including animals, plants, fungi, protozoan plus algae and bacteria plus archaea. We also downloaded the genomic sequences of several animal species from the UCSC genome browser database [26] for extracting exon-exon junctions. The genomic versions of these species are listed in Additional file 1. In this study, sequence reads from NGS technology of several libraries were used. The sequencing platform, RNA source species, and RNA source tissue of these libraries are listed in Additional file 2.\nFor mapping sequence reads back to viral genomes, we first downloaded 3602 viral genomic sequences from NCBI RefSeq 40 [25]. According to the categories of their hosts, these viral genomes were classified into five classes, including animals, plants, fungi, protozoan plus algae and bacteria plus archaea. We also downloaded the genomic sequences of several animal species from the UCSC genome browser database [26] for extracting exon-exon junctions. The genomic versions of these species are listed in Additional file 1. In this study, sequence reads from NGS technology of several libraries were used. The sequencing platform, RNA source species, and RNA source tissue of these libraries are listed in Additional file 2.\n[SUBTITLE] Extraction of exon-exon junction sequences [SUBSECTION] During maturation, eukaryotic genes usually undergo messenger RNA splicing, producing many alternative splicing isoforms from one gene. UCSC mapped these splicing isoforms back to genomes, and determined the boundaries and genomic coordinates of exons, recording the information in refFlat files. As shown in Fig. 1, from such coordinate information, we may exactly define the boundaries of exons and introns. Further, we may also define the exon-end fragments at exon’s both termini, start (S) or end (E). By extracting and assembling the exon-end fragments, from continuous or discrete exons, we collected 60-nt exon-exon junctions (EEJs). Therefore, they are either continuous or discrete EEJs, where the former denote known splicing patterns and the latter denote novel ones. As a result, the number of 60-nt EEJs is for each transcript, where n is the number of exon in each transcript. By doing so, we collected 60-nt EEJs for 21 species. The refFlat versions, number of EEJs and scientific names of these 21 species are available in Additional file 1.\nCollection of exon-exon junctions. By our definition, the EEJs can be continuous or discrete. The former represent known alternative splicing products. The latter, however, represent novel alternative splicing isoforms.\nDuring maturation, eukaryotic genes usually undergo messenger RNA splicing, producing many alternative splicing isoforms from one gene. UCSC mapped these splicing isoforms back to genomes, and determined the boundaries and genomic coordinates of exons, recording the information in refFlat files. As shown in Fig. 1, from such coordinate information, we may exactly define the boundaries of exons and introns. Further, we may also define the exon-end fragments at exon’s both termini, start (S) or end (E). By extracting and assembling the exon-end fragments, from continuous or discrete exons, we collected 60-nt exon-exon junctions (EEJs). Therefore, they are either continuous or discrete EEJs, where the former denote known splicing patterns and the latter denote novel ones. As a result, the number of 60-nt EEJs is for each transcript, where n is the number of exon in each transcript. By doing so, we collected 60-nt EEJs for 21 species. The refFlat versions, number of EEJs and scientific names of these 21 species are available in Additional file 1.\nCollection of exon-exon junctions. By our definition, the EEJs can be continuous or discrete. The former represent known alternative splicing products. The latter, however, represent novel alternative splicing isoforms.\n[SUBTITLE] Sequence reads processing and mapping [SUBSECTION] NGS technologies have produced millions of reads, some of which may occur with high frequency. Such high-occurrence reads cause redundancy problems, and should be solved first. Therefore, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve this redundancy problem. NRP identifies unique sequence reads from input data, assigned a new ID and tabulates the occurrence frequency (copy number) of each unique read. After NRP processing, non-redundant un-mappable reads may be mapped back to viral genomes or EEJs by UMARS. In the studies involved in mapping sequence reads back to genomes, 100% identity is usually demanded [12,13]. Because viral genomes usually have higher mutation rates than eukaryotic ones, we allowed one nucleotide variation, including mismatch and gap, when mapping back to viral genomes or to EEJ sequences. The mapping procedures in this study were done with blast [15].\nNGS technologies have produced millions of reads, some of which may occur with high frequency. Such high-occurrence reads cause redundancy problems, and should be solved first. Therefore, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve this redundancy problem. NRP identifies unique sequence reads from input data, assigned a new ID and tabulates the occurrence frequency (copy number) of each unique read. After NRP processing, non-redundant un-mappable reads may be mapped back to viral genomes or EEJs by UMARS. In the studies involved in mapping sequence reads back to genomes, 100% identity is usually demanded [12,13]. Because viral genomes usually have higher mutation rates than eukaryotic ones, we allowed one nucleotide variation, including mismatch and gap, when mapping back to viral genomes or to EEJ sequences. The mapping procedures in this study were done with blast [15].\n[SUBTITLE] Prediction of viral miRNAs [SUBSECTION] After the mapping procedure, the viral genomic loci mapped by sequence reads are considered as candidate miRNAs. These genomic loci and their flanking sequences were extracted, followed by alignment using miRNA identification pipeline [27]. For each candidate miRNA, the pipeline first calculated the values of ten features, which serve as discrimination indices in a Support Vector Machine (SVM) algorithm. Then, the SVM was used as a classifier to classify candidate miRNAs into positive or negative sets.\nAfter the mapping procedure, the viral genomic loci mapped by sequence reads are considered as candidate miRNAs. These genomic loci and their flanking sequences were extracted, followed by alignment using miRNA identification pipeline [27]. For each candidate miRNA, the pipeline first calculated the values of ten features, which serve as discrimination indices in a Support Vector Machine (SVM) algorithm. Then, the SVM was used as a classifier to classify candidate miRNAs into positive or negative sets.\n[SUBTITLE] cDNA preparation [SUBSECTION] In this study, we used sequence reads from L2 library (Additional file 2) to scan the EEJs of novel alternative splicing isoforms, followed by experimental validation of the detected EEJs in 23 human tissues. Bellow we described how to prepare cDNAs from these tissues. Human tissue poly(A) RNAs (5μg) or total RNAs (40μg) purchased from Clontech (Clontech, Palo Alto, CA) were reverse-transcribed by Transcriptor reverse transcriptase (Roche Applied Science), primed by oligo (dT)15 according to the supplier's instructions. After the reverse transcription reaction, the mixtures were phenol-extracted once, followed by chloroform extraction. Excess primers were removed by applying the mixtures to Chroma Spin-200 (Clontech) gel filtration column. The purified cDNAs were properly diluted and subjected to Polymerase Chain Reaction (PCR) as the amplification templates. In this study, cDNAs from 23 tissues were investigated and they were labeled as follow: M: DNA marker, 1: blood, 2: bone marrow, 3: brain, 4: colon, 5: heart, 6: kidney, 7: liver, 8: lung, 9: ovary, 10: pancreas, 11: placenta, 12: skeletal muscle, 13: small intestine, 14: stomach, 15: testis, 16: whole fetus, 17: breast tumor, 18: cervix tumor, 19: colon tumor, 20: kidney tumor, 21: lung tumor, 22: ovary tumor, 23: gastric tumor, 24: PCR no-template control.\nIn this study, we used sequence reads from L2 library (Additional file 2) to scan the EEJs of novel alternative splicing isoforms, followed by experimental validation of the detected EEJs in 23 human tissues. Bellow we described how to prepare cDNAs from these tissues. Human tissue poly(A) RNAs (5μg) or total RNAs (40μg) purchased from Clontech (Clontech, Palo Alto, CA) were reverse-transcribed by Transcriptor reverse transcriptase (Roche Applied Science), primed by oligo (dT)15 according to the supplier's instructions. After the reverse transcription reaction, the mixtures were phenol-extracted once, followed by chloroform extraction. Excess primers were removed by applying the mixtures to Chroma Spin-200 (Clontech) gel filtration column. The purified cDNAs were properly diluted and subjected to Polymerase Chain Reaction (PCR) as the amplification templates. In this study, cDNAs from 23 tissues were investigated and they were labeled as follow: M: DNA marker, 1: blood, 2: bone marrow, 3: brain, 4: colon, 5: heart, 6: kidney, 7: liver, 8: lung, 9: ovary, 10: pancreas, 11: placenta, 12: skeletal muscle, 13: small intestine, 14: stomach, 15: testis, 16: whole fetus, 17: breast tumor, 18: cervix tumor, 19: colon tumor, 20: kidney tumor, 21: lung tumor, 22: ovary tumor, 23: gastric tumor, 24: PCR no-template control.\n[SUBTITLE] Experimental validation of discrete EEJ [SUBSECTION] The dtetcted EEJs were verified by PCR amplification, followed by capillary sequencing confirmation. Primer pair sequences were picked from each couple of the “discrete” EEJ-spanning exons and were listed in Additional file 3. PCR components include mainly 1mM dNTP, 1μM primer separately, 0.1U Takara Taq DNA polymerase (Takara) per 10μL reaction volume, and the diluted cDNA. The thermal reaction was set at 94°C for 3 minutes, 40 cycles (GAPDH was run specifically for 30 cycles) of denaturing at 94°C for 20 seconds, annealing at 58°C for 30 seconds, and extension at 72°C for 30 seconds, finally at 72°C for 10 minutes. PCR products were separated by 3% NuSieve (Lonza, Rockland, ME) conventional TAE-Agarose gel, and visualized through the ultraviolet light source. The detected and the estimated target size regions of the gel were cut-out and the nucleic-acid contents were purified by Viogene Gel Purification reagents. Minor bands eluted were further subjected to additional 30 PCR cycles with the same pair of primer. The amplified nucleic acid fragments were directly sequenced by ABI 3730xl DNA Analyzer (Applied Biosystems).\nThe dtetcted EEJs were verified by PCR amplification, followed by capillary sequencing confirmation. Primer pair sequences were picked from each couple of the “discrete” EEJ-spanning exons and were listed in Additional file 3. PCR components include mainly 1mM dNTP, 1μM primer separately, 0.1U Takara Taq DNA polymerase (Takara) per 10μL reaction volume, and the diluted cDNA. The thermal reaction was set at 94°C for 3 minutes, 40 cycles (GAPDH was run specifically for 30 cycles) of denaturing at 94°C for 20 seconds, annealing at 58°C for 30 seconds, and extension at 72°C for 30 seconds, finally at 72°C for 10 minutes. PCR products were separated by 3% NuSieve (Lonza, Rockland, ME) conventional TAE-Agarose gel, and visualized through the ultraviolet light source. The detected and the estimated target size regions of the gel were cut-out and the nucleic-acid contents were purified by Viogene Gel Purification reagents. Minor bands eluted were further subjected to additional 30 PCR cycles with the same pair of primer. The amplified nucleic acid fragments were directly sequenced by ABI 3730xl DNA Analyzer (Applied Biosystems).", "For mapping sequence reads back to viral genomes, we first downloaded 3602 viral genomic sequences from NCBI RefSeq 40 [25]. According to the categories of their hosts, these viral genomes were classified into five classes, including animals, plants, fungi, protozoan plus algae and bacteria plus archaea. We also downloaded the genomic sequences of several animal species from the UCSC genome browser database [26] for extracting exon-exon junctions. The genomic versions of these species are listed in Additional file 1. In this study, sequence reads from NGS technology of several libraries were used. The sequencing platform, RNA source species, and RNA source tissue of these libraries are listed in Additional file 2.", "During maturation, eukaryotic genes usually undergo messenger RNA splicing, producing many alternative splicing isoforms from one gene. UCSC mapped these splicing isoforms back to genomes, and determined the boundaries and genomic coordinates of exons, recording the information in refFlat files. As shown in Fig. 1, from such coordinate information, we may exactly define the boundaries of exons and introns. Further, we may also define the exon-end fragments at exon’s both termini, start (S) or end (E). By extracting and assembling the exon-end fragments, from continuous or discrete exons, we collected 60-nt exon-exon junctions (EEJs). Therefore, they are either continuous or discrete EEJs, where the former denote known splicing patterns and the latter denote novel ones. As a result, the number of 60-nt EEJs is for each transcript, where n is the number of exon in each transcript. By doing so, we collected 60-nt EEJs for 21 species. The refFlat versions, number of EEJs and scientific names of these 21 species are available in Additional file 1.\nCollection of exon-exon junctions. By our definition, the EEJs can be continuous or discrete. The former represent known alternative splicing products. The latter, however, represent novel alternative splicing isoforms.", "NGS technologies have produced millions of reads, some of which may occur with high frequency. Such high-occurrence reads cause redundancy problems, and should be solved first. Therefore, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve this redundancy problem. NRP identifies unique sequence reads from input data, assigned a new ID and tabulates the occurrence frequency (copy number) of each unique read. After NRP processing, non-redundant un-mappable reads may be mapped back to viral genomes or EEJs by UMARS. In the studies involved in mapping sequence reads back to genomes, 100% identity is usually demanded [12,13]. Because viral genomes usually have higher mutation rates than eukaryotic ones, we allowed one nucleotide variation, including mismatch and gap, when mapping back to viral genomes or to EEJ sequences. The mapping procedures in this study were done with blast [15].", "After the mapping procedure, the viral genomic loci mapped by sequence reads are considered as candidate miRNAs. These genomic loci and their flanking sequences were extracted, followed by alignment using miRNA identification pipeline [27]. For each candidate miRNA, the pipeline first calculated the values of ten features, which serve as discrimination indices in a Support Vector Machine (SVM) algorithm. Then, the SVM was used as a classifier to classify candidate miRNAs into positive or negative sets.", "In this study, we used sequence reads from L2 library (Additional file 2) to scan the EEJs of novel alternative splicing isoforms, followed by experimental validation of the detected EEJs in 23 human tissues. Bellow we described how to prepare cDNAs from these tissues. Human tissue poly(A) RNAs (5μg) or total RNAs (40μg) purchased from Clontech (Clontech, Palo Alto, CA) were reverse-transcribed by Transcriptor reverse transcriptase (Roche Applied Science), primed by oligo (dT)15 according to the supplier's instructions. After the reverse transcription reaction, the mixtures were phenol-extracted once, followed by chloroform extraction. Excess primers were removed by applying the mixtures to Chroma Spin-200 (Clontech) gel filtration column. The purified cDNAs were properly diluted and subjected to Polymerase Chain Reaction (PCR) as the amplification templates. In this study, cDNAs from 23 tissues were investigated and they were labeled as follow: M: DNA marker, 1: blood, 2: bone marrow, 3: brain, 4: colon, 5: heart, 6: kidney, 7: liver, 8: lung, 9: ovary, 10: pancreas, 11: placenta, 12: skeletal muscle, 13: small intestine, 14: stomach, 15: testis, 16: whole fetus, 17: breast tumor, 18: cervix tumor, 19: colon tumor, 20: kidney tumor, 21: lung tumor, 22: ovary tumor, 23: gastric tumor, 24: PCR no-template control.", "The dtetcted EEJs were verified by PCR amplification, followed by capillary sequencing confirmation. Primer pair sequences were picked from each couple of the “discrete” EEJ-spanning exons and were listed in Additional file 3. PCR components include mainly 1mM dNTP, 1μM primer separately, 0.1U Takara Taq DNA polymerase (Takara) per 10μL reaction volume, and the diluted cDNA. The thermal reaction was set at 94°C for 3 minutes, 40 cycles (GAPDH was run specifically for 30 cycles) of denaturing at 94°C for 20 seconds, annealing at 58°C for 30 seconds, and extension at 72°C for 30 seconds, finally at 72°C for 10 minutes. PCR products were separated by 3% NuSieve (Lonza, Rockland, ME) conventional TAE-Agarose gel, and visualized through the ultraviolet light source. The detected and the estimated target size regions of the gel were cut-out and the nucleic-acid contents were purified by Viogene Gel Purification reagents. Minor bands eluted were further subjected to additional 30 PCR cycles with the same pair of primer. The amplified nucleic acid fragments were directly sequenced by ABI 3730xl DNA Analyzer (Applied Biosystems).", "[SUBTITLE] UMARS pipeline and interface [SUBSECTION] The main concept of UMARS is to provide a user-friendly solution for biologists to retrieve valuable information from discarded NGS reads that could not be mapped back to a reference genome. We have initially produced two major datasets for interrogation, a virus genome sequence collection and an animal exon-exon junction sequence collection. Since many NGS studies intend to study miRNAs, we also provided an additional portal for a miRNA discovery pipeline. The overall UMARS schema is illustrated in Fig. 2. The first step of UMARS deals with the redundancy problem of NGS reads. For convenience and efficiency in data processing and network traffic, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve the problem. The reads uploaded to UMARS must be processed by NRP in advance or be presented in the format of NRP output. NRP accepts the files containing unmapped read sequences in the FASTA format. An NRP standalone program can be downloaded from the UMARS website. Next, the uploaded non-redundant NGS reads could be further processed by either UMARS:EEJ or UMARS:Vir.\nFlowchart of UMARS service. The UMARS service can be divided into two subservices: for discovering viral genomic regions (UMARS:Vir); and discovering novel alternative splicing exon-exon junctions (UMARS:EEJ). Un-mappable reads must be processed by NRP to solving redundancy problem before uploaded to UMARS.\nThe purpose of UMARS:EEJ is to identify novel alternative splicing exon-exon junctions (EEJs) from un-mappable reads. The sequences of all possible EEJs of 21 species were collected in advance. In UMARS:EEJ, uploaded reads are mapped to EEJs. To avoid random sequence matches, besides our mapping criteria (see Materials and methods), a mapping match must overlap both exons for at least five nucleotides, not skewing too much to either exon. Following the mapping procedure, UMARS tabulats detected EEJs and their expression levels. The detected EEJs are reported as either continuous or discrete EEJs. Continuous EEJs represent known mRNA transcripts. However, discrete EEJs could represent novel splicing isoforms.\nThe purpose of UMARS:Vir is to identify possible virus genomic regions from un-mappable NGS reads. In UMARS:Vir, the uploaded NGS reads are mapped to all 3,602 known virus genomes. Following the mapping procedure, UMARS tabulates detected virus species and their expression levels. The detected viral genomic regions may locate at intergenic, protein-coding gene, pre-miRNA regions and so on according to the annotations of RefSeq 40 and miRBase 15. Such information of genomic annotation is also provided by the UMARS service. Several viruses are reported to encode viral miRNAs, regulating expression of host genes and playing important roles in host immune misfunctions [28-30]. Therefore, UMARS:Vir may further have the option to detect viral miRNAs by an additional miRNA identification pipeline from viral intergenic genomic regions.\nThe UMARS service is freely available to the academic community. Users may access UMARS via http://musk.ibms.sinica.edu.tw/UMARS/. The non-redundant reads uploaded into UMARS must not exceed 10 MB in size. Such size limitation has nothing to do with pipeline performance but reduces the network load. As shown in Fig. 3a, in UMARS:Vir, users can select the host category (animals, plants, fungi, protozoan or bacteria) of their expected virus. In UMARS:EEJ, users can select the species corresponding to the EEJs (Fig. 3b). The UMARS results will be sent back to the users via e-mail.\nInterface of UMARS service. UMARS:Vir and UMARS:EEJ have individual parameters. Users must adjust the parameters according to their specific data source to obtain optimal results. In UMARS:Vir service, users must specify the value, Standard or Loose, of “Parameter Criteria” parameter. Specifying Standard outputs only the viruses with CN >= 100 and RN >= 10 (see Table 1); while specifying Loose outputs all virus with CN >= 1 and RNA >= 1. This is an empirical criterion to reduce random match.\nSummary of viruses detected from L1 un-mappable reads.\nCN denotes copy number of total reads mappable to the virus; RN denotes the number of distinct genomic regions mapped by reads. The values in parentheses denote the corresponding values when no mismatch was allowed in the mapping procedure. Because their high similarity, most of the reads of type 1 EBV were also mapped back to type 2 EBV.\nThe main concept of UMARS is to provide a user-friendly solution for biologists to retrieve valuable information from discarded NGS reads that could not be mapped back to a reference genome. We have initially produced two major datasets for interrogation, a virus genome sequence collection and an animal exon-exon junction sequence collection. Since many NGS studies intend to study miRNAs, we also provided an additional portal for a miRNA discovery pipeline. The overall UMARS schema is illustrated in Fig. 2. The first step of UMARS deals with the redundancy problem of NGS reads. For convenience and efficiency in data processing and network traffic, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve the problem. The reads uploaded to UMARS must be processed by NRP in advance or be presented in the format of NRP output. NRP accepts the files containing unmapped read sequences in the FASTA format. An NRP standalone program can be downloaded from the UMARS website. Next, the uploaded non-redundant NGS reads could be further processed by either UMARS:EEJ or UMARS:Vir.\nFlowchart of UMARS service. The UMARS service can be divided into two subservices: for discovering viral genomic regions (UMARS:Vir); and discovering novel alternative splicing exon-exon junctions (UMARS:EEJ). Un-mappable reads must be processed by NRP to solving redundancy problem before uploaded to UMARS.\nThe purpose of UMARS:EEJ is to identify novel alternative splicing exon-exon junctions (EEJs) from un-mappable reads. The sequences of all possible EEJs of 21 species were collected in advance. In UMARS:EEJ, uploaded reads are mapped to EEJs. To avoid random sequence matches, besides our mapping criteria (see Materials and methods), a mapping match must overlap both exons for at least five nucleotides, not skewing too much to either exon. Following the mapping procedure, UMARS tabulats detected EEJs and their expression levels. The detected EEJs are reported as either continuous or discrete EEJs. Continuous EEJs represent known mRNA transcripts. However, discrete EEJs could represent novel splicing isoforms.\nThe purpose of UMARS:Vir is to identify possible virus genomic regions from un-mappable NGS reads. In UMARS:Vir, the uploaded NGS reads are mapped to all 3,602 known virus genomes. Following the mapping procedure, UMARS tabulates detected virus species and their expression levels. The detected viral genomic regions may locate at intergenic, protein-coding gene, pre-miRNA regions and so on according to the annotations of RefSeq 40 and miRBase 15. Such information of genomic annotation is also provided by the UMARS service. Several viruses are reported to encode viral miRNAs, regulating expression of host genes and playing important roles in host immune misfunctions [28-30]. Therefore, UMARS:Vir may further have the option to detect viral miRNAs by an additional miRNA identification pipeline from viral intergenic genomic regions.\nThe UMARS service is freely available to the academic community. Users may access UMARS via http://musk.ibms.sinica.edu.tw/UMARS/. The non-redundant reads uploaded into UMARS must not exceed 10 MB in size. Such size limitation has nothing to do with pipeline performance but reduces the network load. As shown in Fig. 3a, in UMARS:Vir, users can select the host category (animals, plants, fungi, protozoan or bacteria) of their expected virus. In UMARS:EEJ, users can select the species corresponding to the EEJs (Fig. 3b). The UMARS results will be sent back to the users via e-mail.\nInterface of UMARS service. UMARS:Vir and UMARS:EEJ have individual parameters. Users must adjust the parameters according to their specific data source to obtain optimal results. In UMARS:Vir service, users must specify the value, Standard or Loose, of “Parameter Criteria” parameter. Specifying Standard outputs only the viruses with CN >= 100 and RN >= 10 (see Table 1); while specifying Loose outputs all virus with CN >= 1 and RNA >= 1. This is an empirical criterion to reduce random match.\nSummary of viruses detected from L1 un-mappable reads.\nCN denotes copy number of total reads mappable to the virus; RN denotes the number of distinct genomic regions mapped by reads. The values in parentheses denote the corresponding values when no mismatch was allowed in the mapping procedure. Because their high similarity, most of the reads of type 1 EBV were also mapped back to type 2 EBV.\n[SUBTITLE] Case study and demonstration of UMARS:Vir [SUBSECTION] To demonstrate the utility of UMARS, we have analyzed NGS reads using UMARS:Vir. In the first case, we investigated the un-mappable reads from the human NPC cells (L1 library) infected with Epstein-Barr virus (EBV, also named human herpes virus 4 type 1). We examined whether the expected EBV genome could be detected by UMARS:Vir. As a result, eight viruses were detected under our mapping criteria. As shown in Table 1, the expected EBV matches dominated over other un-expected viruses in terms of expression level, which shows that UMARS:Vir can be used to detect infections by a specific virus from un-mappable reads. Besides EBV, there were seven un-expected viruses detected, most of which infect primates, and all of them belong to the herpes virus family.\nAs demonstrated in Table 1, 105,325 reads were mapped back to EBV under our mapping criteria. These EBV reads spanned the EBV genome at 629 genomic regions. The genomic contents, intergenic or intragenic, and detailed information about these regions are listed in Additional file 4. As shown in Table 2 and Additional file 4, 30.1% of these regions are located at the inter-genic regions; 60.1% at the EBV pre-miRNA regions; and only 9.1% at the protein-coding regions. As well, 21 out of 98 protein-coding genes and 22 (excluding mir-BHRF1-1, mir-BHRF1-2, mir-BHRF1-3) out of 25 pre-miRNAs were detected, respectively. This miRNA dominance phenomenon was observed because the L1 sample RNA was extracted with a small RNA kit rather than with a transcriptome kit.\nEBV genomic regions mapped by reads.\nThe regions mapped by reads were classified into three categories according to the annotation of RefSeq 40 and miRBase 15. pre-miRNA regions dominated over other regions in terms of copy number and region number.\nBecause of the strong sequencing intensity, additional mature miRNAs from the same precursors, including isomiRs at the same arm and minor forms of mature miRNA at the opposite arm, are usually detected from NGS reads [12,13,31]. After arranging miRNA reads in order within their corresponding pre-miRNAs, we observed many isomiRs at all of the 22 detected pre-miRNAs (Additional file 5). Compared with EBV pre-miRNAs in miRBase 15, the reference mature miRNAs do not always represent the most abundant reads. In addition, according to miRBase 15 annotation, mir-BART12, mir-BART16 and mir-BART22 encode mature miRNAs only at their 3’, 5’ and 3’ arm respectively. However, we detected additional mature miRNAs at the 5’ arm of mir-BART12, the 3’ arm of mir-BART16 and the 5’ arm of mir-BART22. Moreover, the 5’ arm of mir-BART12 and the 3’ arm of mir-BART16 encode more reads than the original arms. This result is similar to that in Wheeler’s report [14] and should be noted in future data updates of the miRBase.\nTo demonstrate the utility of UMARS, we have analyzed NGS reads using UMARS:Vir. In the first case, we investigated the un-mappable reads from the human NPC cells (L1 library) infected with Epstein-Barr virus (EBV, also named human herpes virus 4 type 1). We examined whether the expected EBV genome could be detected by UMARS:Vir. As a result, eight viruses were detected under our mapping criteria. As shown in Table 1, the expected EBV matches dominated over other un-expected viruses in terms of expression level, which shows that UMARS:Vir can be used to detect infections by a specific virus from un-mappable reads. Besides EBV, there were seven un-expected viruses detected, most of which infect primates, and all of them belong to the herpes virus family.\nAs demonstrated in Table 1, 105,325 reads were mapped back to EBV under our mapping criteria. These EBV reads spanned the EBV genome at 629 genomic regions. The genomic contents, intergenic or intragenic, and detailed information about these regions are listed in Additional file 4. As shown in Table 2 and Additional file 4, 30.1% of these regions are located at the inter-genic regions; 60.1% at the EBV pre-miRNA regions; and only 9.1% at the protein-coding regions. As well, 21 out of 98 protein-coding genes and 22 (excluding mir-BHRF1-1, mir-BHRF1-2, mir-BHRF1-3) out of 25 pre-miRNAs were detected, respectively. This miRNA dominance phenomenon was observed because the L1 sample RNA was extracted with a small RNA kit rather than with a transcriptome kit.\nEBV genomic regions mapped by reads.\nThe regions mapped by reads were classified into three categories according to the annotation of RefSeq 40 and miRBase 15. pre-miRNA regions dominated over other regions in terms of copy number and region number.\nBecause of the strong sequencing intensity, additional mature miRNAs from the same precursors, including isomiRs at the same arm and minor forms of mature miRNA at the opposite arm, are usually detected from NGS reads [12,13,31]. After arranging miRNA reads in order within their corresponding pre-miRNAs, we observed many isomiRs at all of the 22 detected pre-miRNAs (Additional file 5). Compared with EBV pre-miRNAs in miRBase 15, the reference mature miRNAs do not always represent the most abundant reads. In addition, according to miRBase 15 annotation, mir-BART12, mir-BART16 and mir-BART22 encode mature miRNAs only at their 3’, 5’ and 3’ arm respectively. However, we detected additional mature miRNAs at the 5’ arm of mir-BART12, the 3’ arm of mir-BART16 and the 5’ arm of mir-BART22. Moreover, the 5’ arm of mir-BART12 and the 3’ arm of mir-BART16 encode more reads than the original arms. This result is similar to that in Wheeler’s report [14] and should be noted in future data updates of the miRBase.\n[SUBTITLE] Case study and demonstration of UMARS:EEJ [SUBSECTION] In addition to virus infections, splicing exon-exon junctions of gene transcripts may also contribute to un-mappable reads. We, therefore, had the un-mappable reads from human H23 cells (L2 library) processed by UMARS:EEJ. As a result, 35,915 un-mappable reads were mapped back to 4,254 EEJs, extracted from 2,738 transcripts (Table 3). On further examination, 70.2% of the detected EEJs were continuous ones, which indicated they were the major mRNA transcription forms.\nSummary of EEJs detected from L2 un-mappable reads.\nThe digits in parenthesis denote the values when no mismatch was allowed in the mapping procedure. Continuous EEJ dominated over discrete ones in terms of read copy number and EST abundance.\nHowever, there were 1,269 discrete EEJs detected by UMARS:EEJ from the L2 library. We randomly selected five detected EEJs for PCR validation and three of them were successfully validated. Referring to the UCSC Genome Browser, some discrete EEJs had EST evidence, although these splicing isoforms were not deposited in RefSeq. As shown in Fig. 4a, four ESTs, including BU934587, D80989, D52190 and D53867, matched the detected EEJ from NM_007108 (4:2-4), which was assembled from the exon-end fragments of exon 2 and exon 4. We further designed a PCR experiment in which the forward and reverse primers were located at exon 2 and exon 4, respectively (see Materials and methods). As a result, we succeeded in detecting the expected fragment (d transcript) of the alternative transcript in most of the 23 tissues (Fig. 4b and Fig. 4c), which demonstrated that the EEJ, NM_007108 (4:2-4), occurred almost universally. The sequencing result also proved the authenticity of this discrete transcript (Fig. 4d). Another two detected EEJs, NM_021019(7:2-4) and NM_178580(13:10-13), were also successfully validated (Additional file 6 and Additional file 7).\nPCR experimental validation of the detected EEJ, NM_007108 (4:2-4). The marker lane is labeled with M and is presented as 50bp ladder. The central bright band in the marker lane is equal to 350 bp. (a) UCSC Genome Browser shows that four ESTs match the detected EEJ. (b, c) The PCR result showed that the expected EEJ (d transcript) can be experimentally detected. c and d denotes continuous and discrete transcripts, respectively. (d) The sequencing result provided the authenticity of the detected EEJs.\nTable 4 demonstrates the detailed output sample of UMARS:EEJ alignment. The first value “19:6-8” in the “EEJ pattern” column demonstrates that NM_001009566 has 19 exons and the detected EEJ was assembled from the exon-end fragments of exon 6 and exon 8, leading to a discrete EEJ. The first value “14 + 5” in the “Mapping pattern” column demonstrates that the read was split into 14-nt and 5-nt fragments; the former mapped to exon 6 and the latter to exon 8. CN and MM denote the value of the expression level and the number of variations. Because one-nucleotide variation was allowed, three unique reads were mapped back to the same position of the same EEJ. NM_014944 (18:5-7) was also detected owing to alternative splicing from the same gene. All results of our L2 library case are listed in Additional file 8.\nDetected EEJs from CLSTN1.\nNM_001009566 and NM_014944 are splicing isoforms of CLSTN1. The EEJ (19:6-8) from NM_001009566 is identical to the EEJ (18:5-7) from NM_014944.\nMYL6 (myosin light polypeptide 6) encodes a myosin alkali light chain and is associated with cell migration [32]. It was also reported that fibroblasts promote the growth of breast tumor cells by enhancing the expression of several genes, including MYL6 [33]. In this study, c alternative transcript and d transcript of MYL6 have similar expression levels in most normal tissues (Additional file 6). However, the d transcript dominates over c alternative transcript in most tumor tissues, including breast tumor (17th lane in Additional file 6a). It is possible that these alternative splicing isoforms function differently with each other and are associated with tumor genesis.\nIn addition to virus infections, splicing exon-exon junctions of gene transcripts may also contribute to un-mappable reads. We, therefore, had the un-mappable reads from human H23 cells (L2 library) processed by UMARS:EEJ. As a result, 35,915 un-mappable reads were mapped back to 4,254 EEJs, extracted from 2,738 transcripts (Table 3). On further examination, 70.2% of the detected EEJs were continuous ones, which indicated they were the major mRNA transcription forms.\nSummary of EEJs detected from L2 un-mappable reads.\nThe digits in parenthesis denote the values when no mismatch was allowed in the mapping procedure. Continuous EEJ dominated over discrete ones in terms of read copy number and EST abundance.\nHowever, there were 1,269 discrete EEJs detected by UMARS:EEJ from the L2 library. We randomly selected five detected EEJs for PCR validation and three of them were successfully validated. Referring to the UCSC Genome Browser, some discrete EEJs had EST evidence, although these splicing isoforms were not deposited in RefSeq. As shown in Fig. 4a, four ESTs, including BU934587, D80989, D52190 and D53867, matched the detected EEJ from NM_007108 (4:2-4), which was assembled from the exon-end fragments of exon 2 and exon 4. We further designed a PCR experiment in which the forward and reverse primers were located at exon 2 and exon 4, respectively (see Materials and methods). As a result, we succeeded in detecting the expected fragment (d transcript) of the alternative transcript in most of the 23 tissues (Fig. 4b and Fig. 4c), which demonstrated that the EEJ, NM_007108 (4:2-4), occurred almost universally. The sequencing result also proved the authenticity of this discrete transcript (Fig. 4d). Another two detected EEJs, NM_021019(7:2-4) and NM_178580(13:10-13), were also successfully validated (Additional file 6 and Additional file 7).\nPCR experimental validation of the detected EEJ, NM_007108 (4:2-4). The marker lane is labeled with M and is presented as 50bp ladder. The central bright band in the marker lane is equal to 350 bp. (a) UCSC Genome Browser shows that four ESTs match the detected EEJ. (b, c) The PCR result showed that the expected EEJ (d transcript) can be experimentally detected. c and d denotes continuous and discrete transcripts, respectively. (d) The sequencing result provided the authenticity of the detected EEJs.\nTable 4 demonstrates the detailed output sample of UMARS:EEJ alignment. The first value “19:6-8” in the “EEJ pattern” column demonstrates that NM_001009566 has 19 exons and the detected EEJ was assembled from the exon-end fragments of exon 6 and exon 8, leading to a discrete EEJ. The first value “14 + 5” in the “Mapping pattern” column demonstrates that the read was split into 14-nt and 5-nt fragments; the former mapped to exon 6 and the latter to exon 8. CN and MM denote the value of the expression level and the number of variations. Because one-nucleotide variation was allowed, three unique reads were mapped back to the same position of the same EEJ. NM_014944 (18:5-7) was also detected owing to alternative splicing from the same gene. All results of our L2 library case are listed in Additional file 8.\nDetected EEJs from CLSTN1.\nNM_001009566 and NM_014944 are splicing isoforms of CLSTN1. The EEJ (19:6-8) from NM_001009566 is identical to the EEJ (18:5-7) from NM_014944.\nMYL6 (myosin light polypeptide 6) encodes a myosin alkali light chain and is associated with cell migration [32]. It was also reported that fibroblasts promote the growth of breast tumor cells by enhancing the expression of several genes, including MYL6 [33]. In this study, c alternative transcript and d transcript of MYL6 have similar expression levels in most normal tissues (Additional file 6). However, the d transcript dominates over c alternative transcript in most tumor tissues, including breast tumor (17th lane in Additional file 6a). It is possible that these alternative splicing isoforms function differently with each other and are associated with tumor genesis.", "The main concept of UMARS is to provide a user-friendly solution for biologists to retrieve valuable information from discarded NGS reads that could not be mapped back to a reference genome. We have initially produced two major datasets for interrogation, a virus genome sequence collection and an animal exon-exon junction sequence collection. Since many NGS studies intend to study miRNAs, we also provided an additional portal for a miRNA discovery pipeline. The overall UMARS schema is illustrated in Fig. 2. The first step of UMARS deals with the redundancy problem of NGS reads. For convenience and efficiency in data processing and network traffic, we developed an in-house tool, called Non-redundant Reads Producer (NRP), to solve the problem. The reads uploaded to UMARS must be processed by NRP in advance or be presented in the format of NRP output. NRP accepts the files containing unmapped read sequences in the FASTA format. An NRP standalone program can be downloaded from the UMARS website. Next, the uploaded non-redundant NGS reads could be further processed by either UMARS:EEJ or UMARS:Vir.\nFlowchart of UMARS service. The UMARS service can be divided into two subservices: for discovering viral genomic regions (UMARS:Vir); and discovering novel alternative splicing exon-exon junctions (UMARS:EEJ). Un-mappable reads must be processed by NRP to solving redundancy problem before uploaded to UMARS.\nThe purpose of UMARS:EEJ is to identify novel alternative splicing exon-exon junctions (EEJs) from un-mappable reads. The sequences of all possible EEJs of 21 species were collected in advance. In UMARS:EEJ, uploaded reads are mapped to EEJs. To avoid random sequence matches, besides our mapping criteria (see Materials and methods), a mapping match must overlap both exons for at least five nucleotides, not skewing too much to either exon. Following the mapping procedure, UMARS tabulats detected EEJs and their expression levels. The detected EEJs are reported as either continuous or discrete EEJs. Continuous EEJs represent known mRNA transcripts. However, discrete EEJs could represent novel splicing isoforms.\nThe purpose of UMARS:Vir is to identify possible virus genomic regions from un-mappable NGS reads. In UMARS:Vir, the uploaded NGS reads are mapped to all 3,602 known virus genomes. Following the mapping procedure, UMARS tabulates detected virus species and their expression levels. The detected viral genomic regions may locate at intergenic, protein-coding gene, pre-miRNA regions and so on according to the annotations of RefSeq 40 and miRBase 15. Such information of genomic annotation is also provided by the UMARS service. Several viruses are reported to encode viral miRNAs, regulating expression of host genes and playing important roles in host immune misfunctions [28-30]. Therefore, UMARS:Vir may further have the option to detect viral miRNAs by an additional miRNA identification pipeline from viral intergenic genomic regions.\nThe UMARS service is freely available to the academic community. Users may access UMARS via http://musk.ibms.sinica.edu.tw/UMARS/. The non-redundant reads uploaded into UMARS must not exceed 10 MB in size. Such size limitation has nothing to do with pipeline performance but reduces the network load. As shown in Fig. 3a, in UMARS:Vir, users can select the host category (animals, plants, fungi, protozoan or bacteria) of their expected virus. In UMARS:EEJ, users can select the species corresponding to the EEJs (Fig. 3b). The UMARS results will be sent back to the users via e-mail.\nInterface of UMARS service. UMARS:Vir and UMARS:EEJ have individual parameters. Users must adjust the parameters according to their specific data source to obtain optimal results. In UMARS:Vir service, users must specify the value, Standard or Loose, of “Parameter Criteria” parameter. Specifying Standard outputs only the viruses with CN >= 100 and RN >= 10 (see Table 1); while specifying Loose outputs all virus with CN >= 1 and RNA >= 1. This is an empirical criterion to reduce random match.\nSummary of viruses detected from L1 un-mappable reads.\nCN denotes copy number of total reads mappable to the virus; RN denotes the number of distinct genomic regions mapped by reads. The values in parentheses denote the corresponding values when no mismatch was allowed in the mapping procedure. Because their high similarity, most of the reads of type 1 EBV were also mapped back to type 2 EBV.", "To demonstrate the utility of UMARS, we have analyzed NGS reads using UMARS:Vir. In the first case, we investigated the un-mappable reads from the human NPC cells (L1 library) infected with Epstein-Barr virus (EBV, also named human herpes virus 4 type 1). We examined whether the expected EBV genome could be detected by UMARS:Vir. As a result, eight viruses were detected under our mapping criteria. As shown in Table 1, the expected EBV matches dominated over other un-expected viruses in terms of expression level, which shows that UMARS:Vir can be used to detect infections by a specific virus from un-mappable reads. Besides EBV, there were seven un-expected viruses detected, most of which infect primates, and all of them belong to the herpes virus family.\nAs demonstrated in Table 1, 105,325 reads were mapped back to EBV under our mapping criteria. These EBV reads spanned the EBV genome at 629 genomic regions. The genomic contents, intergenic or intragenic, and detailed information about these regions are listed in Additional file 4. As shown in Table 2 and Additional file 4, 30.1% of these regions are located at the inter-genic regions; 60.1% at the EBV pre-miRNA regions; and only 9.1% at the protein-coding regions. As well, 21 out of 98 protein-coding genes and 22 (excluding mir-BHRF1-1, mir-BHRF1-2, mir-BHRF1-3) out of 25 pre-miRNAs were detected, respectively. This miRNA dominance phenomenon was observed because the L1 sample RNA was extracted with a small RNA kit rather than with a transcriptome kit.\nEBV genomic regions mapped by reads.\nThe regions mapped by reads were classified into three categories according to the annotation of RefSeq 40 and miRBase 15. pre-miRNA regions dominated over other regions in terms of copy number and region number.\nBecause of the strong sequencing intensity, additional mature miRNAs from the same precursors, including isomiRs at the same arm and minor forms of mature miRNA at the opposite arm, are usually detected from NGS reads [12,13,31]. After arranging miRNA reads in order within their corresponding pre-miRNAs, we observed many isomiRs at all of the 22 detected pre-miRNAs (Additional file 5). Compared with EBV pre-miRNAs in miRBase 15, the reference mature miRNAs do not always represent the most abundant reads. In addition, according to miRBase 15 annotation, mir-BART12, mir-BART16 and mir-BART22 encode mature miRNAs only at their 3’, 5’ and 3’ arm respectively. However, we detected additional mature miRNAs at the 5’ arm of mir-BART12, the 3’ arm of mir-BART16 and the 5’ arm of mir-BART22. Moreover, the 5’ arm of mir-BART12 and the 3’ arm of mir-BART16 encode more reads than the original arms. This result is similar to that in Wheeler’s report [14] and should be noted in future data updates of the miRBase.", "In addition to virus infections, splicing exon-exon junctions of gene transcripts may also contribute to un-mappable reads. We, therefore, had the un-mappable reads from human H23 cells (L2 library) processed by UMARS:EEJ. As a result, 35,915 un-mappable reads were mapped back to 4,254 EEJs, extracted from 2,738 transcripts (Table 3). On further examination, 70.2% of the detected EEJs were continuous ones, which indicated they were the major mRNA transcription forms.\nSummary of EEJs detected from L2 un-mappable reads.\nThe digits in parenthesis denote the values when no mismatch was allowed in the mapping procedure. Continuous EEJ dominated over discrete ones in terms of read copy number and EST abundance.\nHowever, there were 1,269 discrete EEJs detected by UMARS:EEJ from the L2 library. We randomly selected five detected EEJs for PCR validation and three of them were successfully validated. Referring to the UCSC Genome Browser, some discrete EEJs had EST evidence, although these splicing isoforms were not deposited in RefSeq. As shown in Fig. 4a, four ESTs, including BU934587, D80989, D52190 and D53867, matched the detected EEJ from NM_007108 (4:2-4), which was assembled from the exon-end fragments of exon 2 and exon 4. We further designed a PCR experiment in which the forward and reverse primers were located at exon 2 and exon 4, respectively (see Materials and methods). As a result, we succeeded in detecting the expected fragment (d transcript) of the alternative transcript in most of the 23 tissues (Fig. 4b and Fig. 4c), which demonstrated that the EEJ, NM_007108 (4:2-4), occurred almost universally. The sequencing result also proved the authenticity of this discrete transcript (Fig. 4d). Another two detected EEJs, NM_021019(7:2-4) and NM_178580(13:10-13), were also successfully validated (Additional file 6 and Additional file 7).\nPCR experimental validation of the detected EEJ, NM_007108 (4:2-4). The marker lane is labeled with M and is presented as 50bp ladder. The central bright band in the marker lane is equal to 350 bp. (a) UCSC Genome Browser shows that four ESTs match the detected EEJ. (b, c) The PCR result showed that the expected EEJ (d transcript) can be experimentally detected. c and d denotes continuous and discrete transcripts, respectively. (d) The sequencing result provided the authenticity of the detected EEJs.\nTable 4 demonstrates the detailed output sample of UMARS:EEJ alignment. The first value “19:6-8” in the “EEJ pattern” column demonstrates that NM_001009566 has 19 exons and the detected EEJ was assembled from the exon-end fragments of exon 6 and exon 8, leading to a discrete EEJ. The first value “14 + 5” in the “Mapping pattern” column demonstrates that the read was split into 14-nt and 5-nt fragments; the former mapped to exon 6 and the latter to exon 8. CN and MM denote the value of the expression level and the number of variations. Because one-nucleotide variation was allowed, three unique reads were mapped back to the same position of the same EEJ. NM_014944 (18:5-7) was also detected owing to alternative splicing from the same gene. All results of our L2 library case are listed in Additional file 8.\nDetected EEJs from CLSTN1.\nNM_001009566 and NM_014944 are splicing isoforms of CLSTN1. The EEJ (19:6-8) from NM_001009566 is identical to the EEJ (18:5-7) from NM_014944.\nMYL6 (myosin light polypeptide 6) encodes a myosin alkali light chain and is associated with cell migration [32]. It was also reported that fibroblasts promote the growth of breast tumor cells by enhancing the expression of several genes, including MYL6 [33]. In this study, c alternative transcript and d transcript of MYL6 have similar expression levels in most normal tissues (Additional file 6). However, the d transcript dominates over c alternative transcript in most tumor tissues, including breast tumor (17th lane in Additional file 6a). It is possible that these alternative splicing isoforms function differently with each other and are associated with tumor genesis.", "With the rapid increase of sequencing data, UMARS can detect more and more un-expected splicing isoforms which may provide us insights deeper into gene functions and relations to disease. Although NGS technology has been considered a powerful sequencing tool in biological research, large-scale studies, such as those using microarrays, seem to produce un-expected data unavoidably. Such un-expected data could be background noise, and should be eliminated for data accuracy. In this study, we have shown that some un-mappable reads are not caused by sequencing errors. They can originate from viral infection or transcript splicing. Our UMARS pipeline provides another way to examine and recycle the un-mappable reads that are commonly discarded as garbage. Although we have proposed two possible sources for generating un-mappable redas, a fraction of un-mappable reads still failed to be detected by UMARS. More effort should be expended in investigating the biological relevance and possible sources of un-mappable reads.", "SCL and WCC executed this study and wrote the draft of this manuscript. WCC and CNH were responsible for Construction of UMARS interface. CHL did the jobs of cDNA preparation and PCR experiments. YSJ and HCC provided the SOLiD read data. KWT and CHC participated in discussion on research design. WCL supervised the study and edited the manuscript.", "The authors declare that they have no competing interests.", "The refFlat version, genome version, number of transcript, number of exon-exon junction and scientific names of the 21 species.\nClick here for file\nThe sequencing platform, RNA source species, RNA source tissue of these libraries.\nClick here for file\nPrimer sequences involved in this study.\nClick here for file\nThe mapped reads and their corresponding genomic loci. In the Region info column, “-” denoted intergenic regions without known gene annotation. SN. denoted serial numbers of each mapping match.\nClick here for file\nThe detected isomiRs of EBV pre-miRNAs.\nClick here for file\nPCR experimental validation of the detected EEJ, NM_021019 (7:2-4). The description of the marker lane and the abbreviations are the same with the ones of Fig. 4c. The forward and reverse primers were located at exon 2 and exon 4, respectively. (a) The PCR result showed that the expected EEJ (d transcript) can be experimentally detected. Besides, the major c and the minor d transcripts, the c alternative transcript (with 108 bp fewer than c transcript) was also detected. Both of the detected c alternative and d transcript have EST evidences. (b) The sequencing result provided the authenticity of the detected EEJs.\nClick here for file\nPCR experimental validation of the detected EEJ, NM_178580 (13:10-13). The description of the marker lane and the abbreviations are the same with the ones of Fig. 4c. The forward and reverse primers were located at exon 10 and exon 13, respectively. (a) The PCR result showed that the expected EEJ (d2 transcript) can be experimentally detected. A c alternative transcript (with 56 bp more than c transcript) was also detected. In addition the d2 transcript (with exon 11 and 12 skiping), we also detected a d1 transcript (with exon 12 skipping, not originally detected by UMARS). (b) The sequencing result provided the authenticity of the detected EEJs.\nClick here for file\nThe information of all detected discrete EEJs. Values in this table are tab separated and can be opened with Excel.\nClick here for file" ]

[ null, "materials|methods", null, null, null, null, null, null, null, null, null, null, "conclusions", null, null, "supplementary-material" ]

[]

[ "Acromioclavicular Joint", "Evidence-Based Medicine", "Humans", "Joint Dislocations", "Physical Therapy Modalities", "Trauma Severity Indices" ]

[ "Study eligibility", "Search strategy", "Study identification", "Data extraction", "Methodological appraisal", "Outcomes of interest", "Data analysis", "Search strategy results", "Methodological quality", "Study characteristics", "Meta-analysis", "Sensitivity analysis" ]

[ "To be eligible for inclusion in the systematic review, studies had to compare operative to non-operative management following an acute, closed grade III acromioclavicular dislocation. Studies had to report at least one outcome of interest (see below). All randomised controlled trials (RCTs) and non-randomised controlled trials (nRCT) were included.\nAll included studies reported that all patients recruited gave informed consent prior to being included. All studies were authorized by a local ethical committee, and performed in accordance with the ethical standards of the 1964 Declaration of Helsinki, as revised in 2000.", "The electronic databases: MEDLINE, Embase, Cinahl, Ahmed, Cochrane library and Scopus were searched from their inception to 1st May 2010 in accordance to PRISMA guidelines [10]. A secondary search was conducted reviewing unpublished literature databases including: Greynet, SIGLE, National Technological Information Service, British Library Integrated catalogue, Current Controlled Trials and the Cochrane Central Register of Controlled Trials.\nIn order not to omit any important papers, a broad search was initially undertaken using the MeSH terms and Boolean operators (“acromi$” OR “acromioclavicular”) AND (“injur$” OR “disrupt$” OR “dislocat” OR “subluxat$” OR “ruptur$”) AND (“operat$” OR “surg”) AND (“conservat$” OR “non-surg$” OR “immobilis$” OR “rehabilit$” OR “physical therapy” OR “physiotherapy”).\nThe reference lists of all potentially eligible studies were reviewed. Finally, the corresponding authors of all eligible studies were contacted and asked to review the search results to identify any studies which may have been initially missed.", "Two reviewers (TS, CH) independently screened the titles and/or abstracts of all identified citations against the inclusion and exclusion criteria. The full texts of all potentially eligible studies were obtained. These were then reviewed against the eligibility criteria before inclusion in the review.", "One reviewer extracted all the data onto a pre-defined database (CH). This was then independently verified by a second reviewer for accuracy (RC). Data collected included patients’ characteristics, study design, interventions, follow-up periods and relevant outcomes.", "Study methodological assessment was evaluated using the PEDro score. This is an eleven-item critical appraisal tool which assesses documentation of eligibility, subject allocation and randomisation, subject assessment and blinding, subject follow-up, data assessment and analysis. This has previously been demonstrated to be a reliable and valid scoring system [11, 12]. The critical appraisal was conducted by one reviewer (CH), and independently verified by a second reviewer (RC).", "The primary outcome was the Constant score [13]. Secondary outcomes included: duration of sick leave, strength, pain, cosmetic outcome, implant failure, infection rate, throwing ability, loss of reduction of anatomical position, ossification of the coracoclavicular ligament, range of motion, and the incidence of acromioclavicular joint osteoarthritis (OA).", "An assessment of study heterogeneity was made by observing for population or interventional differences between the studies from the data extraction tables. Secondly, statistical heterogeneity was evaluated using the chi2 (χ2) test and I2 statistics. For outcomes when I2 andχ2 were less than 20% or P < 0.05, a fixed-effects model was adopted. When these assumptions were not met, a random-effects model was adopted. A meta-analysis was conducted where appropriate to pool outcomes. For dichotomous outcomes, the effects measure was the risk difference (RD). For continuous outcome measures, the effect measure was mean difference (MD) or standardised mean difference (Std MD). In each case, a P < 0.05 was considered statistically significant, and 95% confidence intervals (CI) were calculated.\nThe principal analysis was to compare outcomes between operative and non-operative management of acromioclavicular joint grade III dislocations. A secondary analysis included a sensitivity analysis to compare outcomes for RCTs only. Publication and small study bias was assessed using a funnel plot. All meta-analyses were performed using the Review Manager software (RevMan Version 5.0; Nordic Cochrane Centre, Copenhagen, Denmark) and the Mantel–Haenszel method [14].", "A total of 724 citations were identified (Fig. 1). Twenty-four were identified as potentially relevant. On second review, thirteen were deemed not appropriate, whilst one study reported the outcomes of the same cohort in two publications [15, 16]. The most recent version of this paper was included in the review [15]. Four studies did not clearly define the grade of acromioclavicular displacement [17–20]. To minimise review heterogeneity, these studies were excluded, leaving six eligible studies. All were retrospective case series. The funnel plot of infection rate indicated mild evidence of small study exclusion and publication bias (Fig. 2).Fig. 1PRISMA chart illustrating the results of the search strategyFig. 2Funnel plot illustrating publication bias using the cosmetic results outcome measure\nPRISMA chart illustrating the results of the search strategy\nFunnel plot illustrating publication bias using the cosmetic results outcome measure", "The findings of the PEDro critical appraisal indicated that the methodological quality of the current evidence base was poor (Table 1). Although all studies clearly defined their study participants, only two studies demonstrated baseline comparability between the operative and non-operative groups [21]. Furthermore, no study randomised their patients to the allocated intervention. No study based their sample size on a power calculation. Whilst it may have been impractical to blind subjects or clinicians to treatment allocation, no study blinded their assessors during the investigations. Although subject drop-out was more than 85% in all but two studies, no study analysed their results by intention-to-treat principles, or adjusted their results to estimate this missing data. Nonetheless, all clearly described their results and appropriately used descriptive and inferential statistical tests to analyse their cohorts.Table 1PEDro scoreCalvo et al. [8]Fremerey et al. [15]Galpin et al. [26]Gstettner et al. [22]Press et al. [21]Taft et al. [9]Eligibility criteria111111Random allocation000000Concealed allocation000000Baseline comparability111100Blind subject000000Blind clinician000000Blind assessor000100Adequate follow-up (≥85%)100001Intention-to treat analysis000000Between-group analysis111111Point estimates and variability111111Total score5445341: criterion satisfied; 0: criterion not satisfied\nPEDro score\n1: criterion satisfied; 0: criterion not satisfied", "In total, 380 patients were included in the review (Table 2). The operative management cohort consisted of 195 shoulders, 125 males and 15 females with a mean age of 24.4 [standard deviation (SD) = 4.5] years. The non-operative group consisted of 185 shoulders, 96 males and 13 females with a mean age of 27.8 (SD = 6.1) years. One study did not document the cohort age or gender, and therefore total numbers for age and gender are incomplete [9]. Five studies solely evaluated outcomes in patients with Grade III Rockwood injuries. One study’s cohort consisted of 78% grade III injuries, and 22% grade V injuries [15]. Given this high proportion, and since this study provided some outcomes based on grade of injury separately, this study was included in the review.Table 2Study characteristicsStudyStudySampleAge (years)Gender (m/f)Gd of ACJ dis.Surgical mgmtNon-surgical mgmtMean follow up (years)OpNOpOpNOpOpNOpCalvo et al. [8]Retro3211403527/511/0Type IIIPhemister techniqueSling—type not specified. Physiotherapy—type not specifiedOp: 123 mNOp: 41 mFremerey et al. [15]Retro514633.735.948/339/777Gd III20Gd VPDSPhysiotherapy exercisesOp: 6.1NOp: 6.5Galpin et al. [26]Retro1621293716/017/3Gd IIIBosworthSling—type not specified. Strength and ROM exercisesOp: 35.0 mNOp: 33.7 mGstettner et al. [22]Retro282237.236.225/320/2Gd IIIHook plateSling—type not specified. Physiotherapy—type not specified34 mPress et al. [21]Retro161030.749.612/49/1Gd IIIWeaver–Dunn procedureSling—type not specified. Rehabilitation—not specified32.3 mTaft et al. [9]Retro5275N/SN/SN/SN/SGd IIIBosworth screw or 1–3 Steinmann pin fixationSling, Kenny–Howard splint, taping or cast. Mobilisation exercisesOp: 10.8Nop: 9.5av average, Co conservative management, Comp complete, Dis disruption, exs exercises, Gd grade, gp group, immob immobilised, inj injury, K wire Kirschner wire, Lig ligament, m months, Mgmt management, ND Not documented, Op operative management, physio physiotherapy, Pros Prospective, rec recreational, recon reconstructed, Ret retrospective, RTA road traffic accident, Sed sedentary, sh shoulder, wks weeks\nStudy characteristics\nOp: 123 m\nNOp: 41 m\n77\nGd III\n20\nGd V\nOp: 6.1\nNOp: 6.5\nOp: 35.0 m\nNOp: 33.7 m\nOp: 10.8\nNop: 9.5\nav average, Co conservative management, Comp complete, Dis disruption, exs exercises, Gd grade, gp group, immob immobilised, inj injury, K wire Kirschner wire, Lig ligament, m months, Mgmt management, ND Not documented, Op operative management, physio physiotherapy, Pros Prospective, rec recreational, recon reconstructed, Ret retrospective, RTA road traffic accident, Sed sedentary, sh shoulder, wks weeks\nThe operative procedures were clearly described in all papers. Five studies included fixation using Kirschner wire or screw fixation methods, whilst one study used hook plates as the form of fixation [22]. All studies reported repair of coracoclavicular and acromioclavicular ligaments using sutures. The non-operative management adopted was poorly described. All subjects were immobilised using a sling of some description. However, Taft et al. [9] reported immobilising patients using a taping technique or cast, but did not specify the method of application. Immobilisation varied between the studies from 2 weeks [8] to 4 weeks [9]. The remaining papers reported immobilising patients until pain and symptoms had resolved. Following this, subjects commenced range of motion and/or strength rehabilitation programmes, but this was not described in detail in the studies. The follow-up period ranged from 32 months [21] to 10.8 years [9].", "The results of the meta-analysis are shown in Table 3. The primary outcome of this study was the Constant score. This revealed that there was a significantly better functional outcome following operative compared to non-operative management of grade III acromioclavicular separation (MD = 9.70; 95% CI: 1.00, 18.40; P = 0.03; Fig. 3). However, this is based on the complete data from one study [22]. There was no statistically significant difference between the interventions in respect to strength, pain, throwing ability, loss of anatomical reduction, ossification of the coracoclavicular ligament or acromioclavicular joint osteoarthritis (P > 0.05). There were significantly poorer cosmetic results following non-operative management (RD = 0.64; 95% CI: 1.09, 0.19; P < 0.0001; Fig. 4). The results also suggested that there was a significantly greater duration of sick leave following operative management compared to non-operative management (MD = 3.3; 95% CI: 2.10, 4.50; P < 0.001; Fig. 5). Although a relatively low incidence, unsurprisingly, the infection rate was significantly higher in the operative compared to the non-operative group (RD = 0.05; 95% CI: 0.01, 0.09; Fig. 6).Fig. 3Forest plot illustrating constant scoreFig. 4Forest plot illustrating cosmetic outcomeFig. 5Forest plot illustrating duration of sick leaveFig. 6Forest plot illustrating infection rateTable 3Results of the meta-analysisOutcomeStudiesEffect estimateP-valueHeterogeneity\nI\n2\nChi2 (P value)Duration of sick leave23.30 (2.10, 4.50)<0.0001NENEConstant score29.70 (1.00, −18.50)0.03NENEThrowing ability3−0.00 (−0.15, 0.15)0.9800.57Strength (≥90% normal)2−0.01 (−0.12, 0.11)0.9000.82Strength (≤70% normal)20.35 (0.04, 3.51)0.3700.88No pain20.90 (0.33, 2.41)0.8300.60Severe pain2−0.00 (−0.06, 0.06)0.9500.97Poor cosmetic outcome4−0.79 (−0.92, −0.66)<0.0001520.10Tenderness over the acromioclavicular joint20.08 (−0.23, 0.40)0.61750.05Implant failure20.04 (−0.05, 0.13)0.4200.91Infection50.05 (0.01, 0.09)0.0300.66Loss of anatomical reduction20.50 (−1.07, 0.52)0.5098<0.0001Ossification of the coracoclavicular ligament20.17 (−0.32, 0.66)0.50820.02Acromioclavicular joint osteoarthritis30.12 (−0.21, 0.46)0.46890.001NE not estimable\nForest plot illustrating constant score\nForest plot illustrating cosmetic outcome\nForest plot illustrating duration of sick leave\nForest plot illustrating infection rate\nResults of the meta-analysis\nNE not estimable\nOne study assessed the effect of range of motion. Fremerey et al. [15] reported no substantial difference between the interventions, with two patients demonstrating a loss of abduction and external rotation following operative management compared to one patient following non-operative rehabilitation (P > 0.05). Finally, Press et al. [21] reported loss of reduction from the anatomical position. They documented that two patients following operative management presented with loss of reduction, compared to no cases following non-operative management.", "It was not possible to undertake a sensitivity analysis since none of the studies included in the meta-analysis were randomised controlled trials." ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Materials and methods", "Study eligibility", "Search strategy", "Study identification", "Data extraction", "Methodological appraisal", "Outcomes of interest", "Data analysis", "Results", "Search strategy results", "Methodological quality", "Study characteristics", "Meta-analysis", "Sensitivity analysis", "Discussion" ]

[ "Rockwood’s classification of acromioclavicular dislocation is based on the degree and direction of clavicular displacement [1]. Grades I and II are benign and are widely regarded as best managed conservatively [2, 3]. There is a general consensus that type V and VI lesions should be treated operatively [2, 4]. However, there remains controversy over the optimal management strategy for grade III and IV injuries [4–7]. Grade III is classified as a superior displacement of the lateral end of the clavicle of one clavicular diameter or 1 cm on the anteroposterior radiograph, whilst grade IV is described as a separation of the acromioclavicular joint with the distal clavicle displaced posterior into the trapezial fascia [6, 7]. In both grades the acromioclavicular and coracoclavicular ligaments are torn.\nAdvocates of non-operative treatment suggest that patients often regain excellent clinical results and painless shoulder function, although for some there is the potential for chronic instability and pain [8, 9]. Alternatively, operative treatment strategies are able to address these shortcomings, but occasionally compromise shoulder function [2, 8].\nGiven this degree of equipoise, the purpose of this study was to compare the clinical outcomes of patients managed operatively and non-operatively following grade III acromioclavicular dislocation.", "[SUBTITLE] Study eligibility [SUBSECTION] To be eligible for inclusion in the systematic review, studies had to compare operative to non-operative management following an acute, closed grade III acromioclavicular dislocation. Studies had to report at least one outcome of interest (see below). All randomised controlled trials (RCTs) and non-randomised controlled trials (nRCT) were included.\nAll included studies reported that all patients recruited gave informed consent prior to being included. All studies were authorized by a local ethical committee, and performed in accordance with the ethical standards of the 1964 Declaration of Helsinki, as revised in 2000.\nTo be eligible for inclusion in the systematic review, studies had to compare operative to non-operative management following an acute, closed grade III acromioclavicular dislocation. Studies had to report at least one outcome of interest (see below). All randomised controlled trials (RCTs) and non-randomised controlled trials (nRCT) were included.\nAll included studies reported that all patients recruited gave informed consent prior to being included. All studies were authorized by a local ethical committee, and performed in accordance with the ethical standards of the 1964 Declaration of Helsinki, as revised in 2000.\n[SUBTITLE] Search strategy [SUBSECTION] The electronic databases: MEDLINE, Embase, Cinahl, Ahmed, Cochrane library and Scopus were searched from their inception to 1st May 2010 in accordance to PRISMA guidelines [10]. A secondary search was conducted reviewing unpublished literature databases including: Greynet, SIGLE, National Technological Information Service, British Library Integrated catalogue, Current Controlled Trials and the Cochrane Central Register of Controlled Trials.\nIn order not to omit any important papers, a broad search was initially undertaken using the MeSH terms and Boolean operators (“acromi$” OR “acromioclavicular”) AND (“injur$” OR “disrupt$” OR “dislocat” OR “subluxat$” OR “ruptur$”) AND (“operat$” OR “surg”) AND (“conservat$” OR “non-surg$” OR “immobilis$” OR “rehabilit$” OR “physical therapy” OR “physiotherapy”).\nThe reference lists of all potentially eligible studies were reviewed. Finally, the corresponding authors of all eligible studies were contacted and asked to review the search results to identify any studies which may have been initially missed.\nThe electronic databases: MEDLINE, Embase, Cinahl, Ahmed, Cochrane library and Scopus were searched from their inception to 1st May 2010 in accordance to PRISMA guidelines [10]. A secondary search was conducted reviewing unpublished literature databases including: Greynet, SIGLE, National Technological Information Service, British Library Integrated catalogue, Current Controlled Trials and the Cochrane Central Register of Controlled Trials.\nIn order not to omit any important papers, a broad search was initially undertaken using the MeSH terms and Boolean operators (“acromi$” OR “acromioclavicular”) AND (“injur$” OR “disrupt$” OR “dislocat” OR “subluxat$” OR “ruptur$”) AND (“operat$” OR “surg”) AND (“conservat$” OR “non-surg$” OR “immobilis$” OR “rehabilit$” OR “physical therapy” OR “physiotherapy”).\nThe reference lists of all potentially eligible studies were reviewed. Finally, the corresponding authors of all eligible studies were contacted and asked to review the search results to identify any studies which may have been initially missed.\n[SUBTITLE] Study identification [SUBSECTION] Two reviewers (TS, CH) independently screened the titles and/or abstracts of all identified citations against the inclusion and exclusion criteria. The full texts of all potentially eligible studies were obtained. These were then reviewed against the eligibility criteria before inclusion in the review.\nTwo reviewers (TS, CH) independently screened the titles and/or abstracts of all identified citations against the inclusion and exclusion criteria. The full texts of all potentially eligible studies were obtained. These were then reviewed against the eligibility criteria before inclusion in the review.\n[SUBTITLE] Data extraction [SUBSECTION] One reviewer extracted all the data onto a pre-defined database (CH). This was then independently verified by a second reviewer for accuracy (RC). Data collected included patients’ characteristics, study design, interventions, follow-up periods and relevant outcomes.\nOne reviewer extracted all the data onto a pre-defined database (CH). This was then independently verified by a second reviewer for accuracy (RC). Data collected included patients’ characteristics, study design, interventions, follow-up periods and relevant outcomes.\n[SUBTITLE] Methodological appraisal [SUBSECTION] Study methodological assessment was evaluated using the PEDro score. This is an eleven-item critical appraisal tool which assesses documentation of eligibility, subject allocation and randomisation, subject assessment and blinding, subject follow-up, data assessment and analysis. This has previously been demonstrated to be a reliable and valid scoring system [11, 12]. The critical appraisal was conducted by one reviewer (CH), and independently verified by a second reviewer (RC).\nStudy methodological assessment was evaluated using the PEDro score. This is an eleven-item critical appraisal tool which assesses documentation of eligibility, subject allocation and randomisation, subject assessment and blinding, subject follow-up, data assessment and analysis. This has previously been demonstrated to be a reliable and valid scoring system [11, 12]. The critical appraisal was conducted by one reviewer (CH), and independently verified by a second reviewer (RC).\n[SUBTITLE] Outcomes of interest [SUBSECTION] The primary outcome was the Constant score [13]. Secondary outcomes included: duration of sick leave, strength, pain, cosmetic outcome, implant failure, infection rate, throwing ability, loss of reduction of anatomical position, ossification of the coracoclavicular ligament, range of motion, and the incidence of acromioclavicular joint osteoarthritis (OA).\nThe primary outcome was the Constant score [13]. Secondary outcomes included: duration of sick leave, strength, pain, cosmetic outcome, implant failure, infection rate, throwing ability, loss of reduction of anatomical position, ossification of the coracoclavicular ligament, range of motion, and the incidence of acromioclavicular joint osteoarthritis (OA).\n[SUBTITLE] Data analysis [SUBSECTION] An assessment of study heterogeneity was made by observing for population or interventional differences between the studies from the data extraction tables. Secondly, statistical heterogeneity was evaluated using the chi2 (χ2) test and I2 statistics. For outcomes when I2 andχ2 were less than 20% or P < 0.05, a fixed-effects model was adopted. When these assumptions were not met, a random-effects model was adopted. A meta-analysis was conducted where appropriate to pool outcomes. For dichotomous outcomes, the effects measure was the risk difference (RD). For continuous outcome measures, the effect measure was mean difference (MD) or standardised mean difference (Std MD). In each case, a P < 0.05 was considered statistically significant, and 95% confidence intervals (CI) were calculated.\nThe principal analysis was to compare outcomes between operative and non-operative management of acromioclavicular joint grade III dislocations. A secondary analysis included a sensitivity analysis to compare outcomes for RCTs only. Publication and small study bias was assessed using a funnel plot. All meta-analyses were performed using the Review Manager software (RevMan Version 5.0; Nordic Cochrane Centre, Copenhagen, Denmark) and the Mantel–Haenszel method [14].\nAn assessment of study heterogeneity was made by observing for population or interventional differences between the studies from the data extraction tables. Secondly, statistical heterogeneity was evaluated using the chi2 (χ2) test and I2 statistics. For outcomes when I2 andχ2 were less than 20% or P < 0.05, a fixed-effects model was adopted. When these assumptions were not met, a random-effects model was adopted. A meta-analysis was conducted where appropriate to pool outcomes. For dichotomous outcomes, the effects measure was the risk difference (RD). For continuous outcome measures, the effect measure was mean difference (MD) or standardised mean difference (Std MD). In each case, a P < 0.05 was considered statistically significant, and 95% confidence intervals (CI) were calculated.\nThe principal analysis was to compare outcomes between operative and non-operative management of acromioclavicular joint grade III dislocations. A secondary analysis included a sensitivity analysis to compare outcomes for RCTs only. Publication and small study bias was assessed using a funnel plot. All meta-analyses were performed using the Review Manager software (RevMan Version 5.0; Nordic Cochrane Centre, Copenhagen, Denmark) and the Mantel–Haenszel method [14].", "To be eligible for inclusion in the systematic review, studies had to compare operative to non-operative management following an acute, closed grade III acromioclavicular dislocation. Studies had to report at least one outcome of interest (see below). All randomised controlled trials (RCTs) and non-randomised controlled trials (nRCT) were included.\nAll included studies reported that all patients recruited gave informed consent prior to being included. All studies were authorized by a local ethical committee, and performed in accordance with the ethical standards of the 1964 Declaration of Helsinki, as revised in 2000.", "The electronic databases: MEDLINE, Embase, Cinahl, Ahmed, Cochrane library and Scopus were searched from their inception to 1st May 2010 in accordance to PRISMA guidelines [10]. A secondary search was conducted reviewing unpublished literature databases including: Greynet, SIGLE, National Technological Information Service, British Library Integrated catalogue, Current Controlled Trials and the Cochrane Central Register of Controlled Trials.\nIn order not to omit any important papers, a broad search was initially undertaken using the MeSH terms and Boolean operators (“acromi$” OR “acromioclavicular”) AND (“injur$” OR “disrupt$” OR “dislocat” OR “subluxat$” OR “ruptur$”) AND (“operat$” OR “surg”) AND (“conservat$” OR “non-surg$” OR “immobilis$” OR “rehabilit$” OR “physical therapy” OR “physiotherapy”).\nThe reference lists of all potentially eligible studies were reviewed. Finally, the corresponding authors of all eligible studies were contacted and asked to review the search results to identify any studies which may have been initially missed.", "Two reviewers (TS, CH) independently screened the titles and/or abstracts of all identified citations against the inclusion and exclusion criteria. The full texts of all potentially eligible studies were obtained. These were then reviewed against the eligibility criteria before inclusion in the review.", "One reviewer extracted all the data onto a pre-defined database (CH). This was then independently verified by a second reviewer for accuracy (RC). Data collected included patients’ characteristics, study design, interventions, follow-up periods and relevant outcomes.", "Study methodological assessment was evaluated using the PEDro score. This is an eleven-item critical appraisal tool which assesses documentation of eligibility, subject allocation and randomisation, subject assessment and blinding, subject follow-up, data assessment and analysis. This has previously been demonstrated to be a reliable and valid scoring system [11, 12]. The critical appraisal was conducted by one reviewer (CH), and independently verified by a second reviewer (RC).", "The primary outcome was the Constant score [13]. Secondary outcomes included: duration of sick leave, strength, pain, cosmetic outcome, implant failure, infection rate, throwing ability, loss of reduction of anatomical position, ossification of the coracoclavicular ligament, range of motion, and the incidence of acromioclavicular joint osteoarthritis (OA).", "An assessment of study heterogeneity was made by observing for population or interventional differences between the studies from the data extraction tables. Secondly, statistical heterogeneity was evaluated using the chi2 (χ2) test and I2 statistics. For outcomes when I2 andχ2 were less than 20% or P < 0.05, a fixed-effects model was adopted. When these assumptions were not met, a random-effects model was adopted. A meta-analysis was conducted where appropriate to pool outcomes. For dichotomous outcomes, the effects measure was the risk difference (RD). For continuous outcome measures, the effect measure was mean difference (MD) or standardised mean difference (Std MD). In each case, a P < 0.05 was considered statistically significant, and 95% confidence intervals (CI) were calculated.\nThe principal analysis was to compare outcomes between operative and non-operative management of acromioclavicular joint grade III dislocations. A secondary analysis included a sensitivity analysis to compare outcomes for RCTs only. Publication and small study bias was assessed using a funnel plot. All meta-analyses were performed using the Review Manager software (RevMan Version 5.0; Nordic Cochrane Centre, Copenhagen, Denmark) and the Mantel–Haenszel method [14].", "[SUBTITLE] Search strategy results [SUBSECTION] A total of 724 citations were identified (Fig. 1). Twenty-four were identified as potentially relevant. On second review, thirteen were deemed not appropriate, whilst one study reported the outcomes of the same cohort in two publications [15, 16]. The most recent version of this paper was included in the review [15]. Four studies did not clearly define the grade of acromioclavicular displacement [17–20]. To minimise review heterogeneity, these studies were excluded, leaving six eligible studies. All were retrospective case series. The funnel plot of infection rate indicated mild evidence of small study exclusion and publication bias (Fig. 2).Fig. 1PRISMA chart illustrating the results of the search strategyFig. 2Funnel plot illustrating publication bias using the cosmetic results outcome measure\nPRISMA chart illustrating the results of the search strategy\nFunnel plot illustrating publication bias using the cosmetic results outcome measure\nA total of 724 citations were identified (Fig. 1). Twenty-four were identified as potentially relevant. On second review, thirteen were deemed not appropriate, whilst one study reported the outcomes of the same cohort in two publications [15, 16]. The most recent version of this paper was included in the review [15]. Four studies did not clearly define the grade of acromioclavicular displacement [17–20]. To minimise review heterogeneity, these studies were excluded, leaving six eligible studies. All were retrospective case series. The funnel plot of infection rate indicated mild evidence of small study exclusion and publication bias (Fig. 2).Fig. 1PRISMA chart illustrating the results of the search strategyFig. 2Funnel plot illustrating publication bias using the cosmetic results outcome measure\nPRISMA chart illustrating the results of the search strategy\nFunnel plot illustrating publication bias using the cosmetic results outcome measure\n[SUBTITLE] Methodological quality [SUBSECTION] The findings of the PEDro critical appraisal indicated that the methodological quality of the current evidence base was poor (Table 1). Although all studies clearly defined their study participants, only two studies demonstrated baseline comparability between the operative and non-operative groups [21]. Furthermore, no study randomised their patients to the allocated intervention. No study based their sample size on a power calculation. Whilst it may have been impractical to blind subjects or clinicians to treatment allocation, no study blinded their assessors during the investigations. Although subject drop-out was more than 85% in all but two studies, no study analysed their results by intention-to-treat principles, or adjusted their results to estimate this missing data. Nonetheless, all clearly described their results and appropriately used descriptive and inferential statistical tests to analyse their cohorts.Table 1PEDro scoreCalvo et al. [8]Fremerey et al. [15]Galpin et al. [26]Gstettner et al. [22]Press et al. [21]Taft et al. [9]Eligibility criteria111111Random allocation000000Concealed allocation000000Baseline comparability111100Blind subject000000Blind clinician000000Blind assessor000100Adequate follow-up (≥85%)100001Intention-to treat analysis000000Between-group analysis111111Point estimates and variability111111Total score5445341: criterion satisfied; 0: criterion not satisfied\nPEDro score\n1: criterion satisfied; 0: criterion not satisfied\nThe findings of the PEDro critical appraisal indicated that the methodological quality of the current evidence base was poor (Table 1). Although all studies clearly defined their study participants, only two studies demonstrated baseline comparability between the operative and non-operative groups [21]. Furthermore, no study randomised their patients to the allocated intervention. No study based their sample size on a power calculation. Whilst it may have been impractical to blind subjects or clinicians to treatment allocation, no study blinded their assessors during the investigations. Although subject drop-out was more than 85% in all but two studies, no study analysed their results by intention-to-treat principles, or adjusted their results to estimate this missing data. Nonetheless, all clearly described their results and appropriately used descriptive and inferential statistical tests to analyse their cohorts.Table 1PEDro scoreCalvo et al. [8]Fremerey et al. [15]Galpin et al. [26]Gstettner et al. [22]Press et al. [21]Taft et al. [9]Eligibility criteria111111Random allocation000000Concealed allocation000000Baseline comparability111100Blind subject000000Blind clinician000000Blind assessor000100Adequate follow-up (≥85%)100001Intention-to treat analysis000000Between-group analysis111111Point estimates and variability111111Total score5445341: criterion satisfied; 0: criterion not satisfied\nPEDro score\n1: criterion satisfied; 0: criterion not satisfied\n[SUBTITLE] Study characteristics [SUBSECTION] In total, 380 patients were included in the review (Table 2). The operative management cohort consisted of 195 shoulders, 125 males and 15 females with a mean age of 24.4 [standard deviation (SD) = 4.5] years. The non-operative group consisted of 185 shoulders, 96 males and 13 females with a mean age of 27.8 (SD = 6.1) years. One study did not document the cohort age or gender, and therefore total numbers for age and gender are incomplete [9]. Five studies solely evaluated outcomes in patients with Grade III Rockwood injuries. One study’s cohort consisted of 78% grade III injuries, and 22% grade V injuries [15]. Given this high proportion, and since this study provided some outcomes based on grade of injury separately, this study was included in the review.Table 2Study characteristicsStudyStudySampleAge (years)Gender (m/f)Gd of ACJ dis.Surgical mgmtNon-surgical mgmtMean follow up (years)OpNOpOpNOpOpNOpCalvo et al. [8]Retro3211403527/511/0Type IIIPhemister techniqueSling—type not specified. Physiotherapy—type not specifiedOp: 123 mNOp: 41 mFremerey et al. [15]Retro514633.735.948/339/777Gd III20Gd VPDSPhysiotherapy exercisesOp: 6.1NOp: 6.5Galpin et al. [26]Retro1621293716/017/3Gd IIIBosworthSling—type not specified. Strength and ROM exercisesOp: 35.0 mNOp: 33.7 mGstettner et al. [22]Retro282237.236.225/320/2Gd IIIHook plateSling—type not specified. Physiotherapy—type not specified34 mPress et al. [21]Retro161030.749.612/49/1Gd IIIWeaver–Dunn procedureSling—type not specified. Rehabilitation—not specified32.3 mTaft et al. [9]Retro5275N/SN/SN/SN/SGd IIIBosworth screw or 1–3 Steinmann pin fixationSling, Kenny–Howard splint, taping or cast. Mobilisation exercisesOp: 10.8Nop: 9.5av average, Co conservative management, Comp complete, Dis disruption, exs exercises, Gd grade, gp group, immob immobilised, inj injury, K wire Kirschner wire, Lig ligament, m months, Mgmt management, ND Not documented, Op operative management, physio physiotherapy, Pros Prospective, rec recreational, recon reconstructed, Ret retrospective, RTA road traffic accident, Sed sedentary, sh shoulder, wks weeks\nStudy characteristics\nOp: 123 m\nNOp: 41 m\n77\nGd III\n20\nGd V\nOp: 6.1\nNOp: 6.5\nOp: 35.0 m\nNOp: 33.7 m\nOp: 10.8\nNop: 9.5\nav average, Co conservative management, Comp complete, Dis disruption, exs exercises, Gd grade, gp group, immob immobilised, inj injury, K wire Kirschner wire, Lig ligament, m months, Mgmt management, ND Not documented, Op operative management, physio physiotherapy, Pros Prospective, rec recreational, recon reconstructed, Ret retrospective, RTA road traffic accident, Sed sedentary, sh shoulder, wks weeks\nThe operative procedures were clearly described in all papers. Five studies included fixation using Kirschner wire or screw fixation methods, whilst one study used hook plates as the form of fixation [22]. All studies reported repair of coracoclavicular and acromioclavicular ligaments using sutures. The non-operative management adopted was poorly described. All subjects were immobilised using a sling of some description. However, Taft et al. [9] reported immobilising patients using a taping technique or cast, but did not specify the method of application. Immobilisation varied between the studies from 2 weeks [8] to 4 weeks [9]. The remaining papers reported immobilising patients until pain and symptoms had resolved. Following this, subjects commenced range of motion and/or strength rehabilitation programmes, but this was not described in detail in the studies. The follow-up period ranged from 32 months [21] to 10.8 years [9].\nIn total, 380 patients were included in the review (Table 2). The operative management cohort consisted of 195 shoulders, 125 males and 15 females with a mean age of 24.4 [standard deviation (SD) = 4.5] years. The non-operative group consisted of 185 shoulders, 96 males and 13 females with a mean age of 27.8 (SD = 6.1) years. One study did not document the cohort age or gender, and therefore total numbers for age and gender are incomplete [9]. Five studies solely evaluated outcomes in patients with Grade III Rockwood injuries. One study’s cohort consisted of 78% grade III injuries, and 22% grade V injuries [15]. Given this high proportion, and since this study provided some outcomes based on grade of injury separately, this study was included in the review.Table 2Study characteristicsStudyStudySampleAge (years)Gender (m/f)Gd of ACJ dis.Surgical mgmtNon-surgical mgmtMean follow up (years)OpNOpOpNOpOpNOpCalvo et al. [8]Retro3211403527/511/0Type IIIPhemister techniqueSling—type not specified. Physiotherapy—type not specifiedOp: 123 mNOp: 41 mFremerey et al. [15]Retro514633.735.948/339/777Gd III20Gd VPDSPhysiotherapy exercisesOp: 6.1NOp: 6.5Galpin et al. [26]Retro1621293716/017/3Gd IIIBosworthSling—type not specified. Strength and ROM exercisesOp: 35.0 mNOp: 33.7 mGstettner et al. [22]Retro282237.236.225/320/2Gd IIIHook plateSling—type not specified. Physiotherapy—type not specified34 mPress et al. [21]Retro161030.749.612/49/1Gd IIIWeaver–Dunn procedureSling—type not specified. Rehabilitation—not specified32.3 mTaft et al. [9]Retro5275N/SN/SN/SN/SGd IIIBosworth screw or 1–3 Steinmann pin fixationSling, Kenny–Howard splint, taping or cast. Mobilisation exercisesOp: 10.8Nop: 9.5av average, Co conservative management, Comp complete, Dis disruption, exs exercises, Gd grade, gp group, immob immobilised, inj injury, K wire Kirschner wire, Lig ligament, m months, Mgmt management, ND Not documented, Op operative management, physio physiotherapy, Pros Prospective, rec recreational, recon reconstructed, Ret retrospective, RTA road traffic accident, Sed sedentary, sh shoulder, wks weeks\nStudy characteristics\nOp: 123 m\nNOp: 41 m\n77\nGd III\n20\nGd V\nOp: 6.1\nNOp: 6.5\nOp: 35.0 m\nNOp: 33.7 m\nOp: 10.8\nNop: 9.5\nav average, Co conservative management, Comp complete, Dis disruption, exs exercises, Gd grade, gp group, immob immobilised, inj injury, K wire Kirschner wire, Lig ligament, m months, Mgmt management, ND Not documented, Op operative management, physio physiotherapy, Pros Prospective, rec recreational, recon reconstructed, Ret retrospective, RTA road traffic accident, Sed sedentary, sh shoulder, wks weeks\nThe operative procedures were clearly described in all papers. Five studies included fixation using Kirschner wire or screw fixation methods, whilst one study used hook plates as the form of fixation [22]. All studies reported repair of coracoclavicular and acromioclavicular ligaments using sutures. The non-operative management adopted was poorly described. All subjects were immobilised using a sling of some description. However, Taft et al. [9] reported immobilising patients using a taping technique or cast, but did not specify the method of application. Immobilisation varied between the studies from 2 weeks [8] to 4 weeks [9]. The remaining papers reported immobilising patients until pain and symptoms had resolved. Following this, subjects commenced range of motion and/or strength rehabilitation programmes, but this was not described in detail in the studies. The follow-up period ranged from 32 months [21] to 10.8 years [9].\n[SUBTITLE] Meta-analysis [SUBSECTION] The results of the meta-analysis are shown in Table 3. The primary outcome of this study was the Constant score. This revealed that there was a significantly better functional outcome following operative compared to non-operative management of grade III acromioclavicular separation (MD = 9.70; 95% CI: 1.00, 18.40; P = 0.03; Fig. 3). However, this is based on the complete data from one study [22]. There was no statistically significant difference between the interventions in respect to strength, pain, throwing ability, loss of anatomical reduction, ossification of the coracoclavicular ligament or acromioclavicular joint osteoarthritis (P > 0.05). There were significantly poorer cosmetic results following non-operative management (RD = 0.64; 95% CI: 1.09, 0.19; P < 0.0001; Fig. 4). The results also suggested that there was a significantly greater duration of sick leave following operative management compared to non-operative management (MD = 3.3; 95% CI: 2.10, 4.50; P < 0.001; Fig. 5). Although a relatively low incidence, unsurprisingly, the infection rate was significantly higher in the operative compared to the non-operative group (RD = 0.05; 95% CI: 0.01, 0.09; Fig. 6).Fig. 3Forest plot illustrating constant scoreFig. 4Forest plot illustrating cosmetic outcomeFig. 5Forest plot illustrating duration of sick leaveFig. 6Forest plot illustrating infection rateTable 3Results of the meta-analysisOutcomeStudiesEffect estimateP-valueHeterogeneity\nI\n2\nChi2 (P value)Duration of sick leave23.30 (2.10, 4.50)<0.0001NENEConstant score29.70 (1.00, −18.50)0.03NENEThrowing ability3−0.00 (−0.15, 0.15)0.9800.57Strength (≥90% normal)2−0.01 (−0.12, 0.11)0.9000.82Strength (≤70% normal)20.35 (0.04, 3.51)0.3700.88No pain20.90 (0.33, 2.41)0.8300.60Severe pain2−0.00 (−0.06, 0.06)0.9500.97Poor cosmetic outcome4−0.79 (−0.92, −0.66)<0.0001520.10Tenderness over the acromioclavicular joint20.08 (−0.23, 0.40)0.61750.05Implant failure20.04 (−0.05, 0.13)0.4200.91Infection50.05 (0.01, 0.09)0.0300.66Loss of anatomical reduction20.50 (−1.07, 0.52)0.5098<0.0001Ossification of the coracoclavicular ligament20.17 (−0.32, 0.66)0.50820.02Acromioclavicular joint osteoarthritis30.12 (−0.21, 0.46)0.46890.001NE not estimable\nForest plot illustrating constant score\nForest plot illustrating cosmetic outcome\nForest plot illustrating duration of sick leave\nForest plot illustrating infection rate\nResults of the meta-analysis\nNE not estimable\nOne study assessed the effect of range of motion. Fremerey et al. [15] reported no substantial difference between the interventions, with two patients demonstrating a loss of abduction and external rotation following operative management compared to one patient following non-operative rehabilitation (P > 0.05). Finally, Press et al. [21] reported loss of reduction from the anatomical position. They documented that two patients following operative management presented with loss of reduction, compared to no cases following non-operative management.\nThe results of the meta-analysis are shown in Table 3. The primary outcome of this study was the Constant score. This revealed that there was a significantly better functional outcome following operative compared to non-operative management of grade III acromioclavicular separation (MD = 9.70; 95% CI: 1.00, 18.40; P = 0.03; Fig. 3). However, this is based on the complete data from one study [22]. There was no statistically significant difference between the interventions in respect to strength, pain, throwing ability, loss of anatomical reduction, ossification of the coracoclavicular ligament or acromioclavicular joint osteoarthritis (P > 0.05). There were significantly poorer cosmetic results following non-operative management (RD = 0.64; 95% CI: 1.09, 0.19; P < 0.0001; Fig. 4). The results also suggested that there was a significantly greater duration of sick leave following operative management compared to non-operative management (MD = 3.3; 95% CI: 2.10, 4.50; P < 0.001; Fig. 5). Although a relatively low incidence, unsurprisingly, the infection rate was significantly higher in the operative compared to the non-operative group (RD = 0.05; 95% CI: 0.01, 0.09; Fig. 6).Fig. 3Forest plot illustrating constant scoreFig. 4Forest plot illustrating cosmetic outcomeFig. 5Forest plot illustrating duration of sick leaveFig. 6Forest plot illustrating infection rateTable 3Results of the meta-analysisOutcomeStudiesEffect estimateP-valueHeterogeneity\nI\n2\nChi2 (P value)Duration of sick leave23.30 (2.10, 4.50)<0.0001NENEConstant score29.70 (1.00, −18.50)0.03NENEThrowing ability3−0.00 (−0.15, 0.15)0.9800.57Strength (≥90% normal)2−0.01 (−0.12, 0.11)0.9000.82Strength (≤70% normal)20.35 (0.04, 3.51)0.3700.88No pain20.90 (0.33, 2.41)0.8300.60Severe pain2−0.00 (−0.06, 0.06)0.9500.97Poor cosmetic outcome4−0.79 (−0.92, −0.66)<0.0001520.10Tenderness over the acromioclavicular joint20.08 (−0.23, 0.40)0.61750.05Implant failure20.04 (−0.05, 0.13)0.4200.91Infection50.05 (0.01, 0.09)0.0300.66Loss of anatomical reduction20.50 (−1.07, 0.52)0.5098<0.0001Ossification of the coracoclavicular ligament20.17 (−0.32, 0.66)0.50820.02Acromioclavicular joint osteoarthritis30.12 (−0.21, 0.46)0.46890.001NE not estimable\nForest plot illustrating constant score\nForest plot illustrating cosmetic outcome\nForest plot illustrating duration of sick leave\nForest plot illustrating infection rate\nResults of the meta-analysis\nNE not estimable\nOne study assessed the effect of range of motion. Fremerey et al. [15] reported no substantial difference between the interventions, with two patients demonstrating a loss of abduction and external rotation following operative management compared to one patient following non-operative rehabilitation (P > 0.05). Finally, Press et al. [21] reported loss of reduction from the anatomical position. They documented that two patients following operative management presented with loss of reduction, compared to no cases following non-operative management.\n[SUBTITLE] Sensitivity analysis [SUBSECTION] It was not possible to undertake a sensitivity analysis since none of the studies included in the meta-analysis were randomised controlled trials.\nIt was not possible to undertake a sensitivity analysis since none of the studies included in the meta-analysis were randomised controlled trials.", "A total of 724 citations were identified (Fig. 1). Twenty-four were identified as potentially relevant. On second review, thirteen were deemed not appropriate, whilst one study reported the outcomes of the same cohort in two publications [15, 16]. The most recent version of this paper was included in the review [15]. Four studies did not clearly define the grade of acromioclavicular displacement [17–20]. To minimise review heterogeneity, these studies were excluded, leaving six eligible studies. All were retrospective case series. The funnel plot of infection rate indicated mild evidence of small study exclusion and publication bias (Fig. 2).Fig. 1PRISMA chart illustrating the results of the search strategyFig. 2Funnel plot illustrating publication bias using the cosmetic results outcome measure\nPRISMA chart illustrating the results of the search strategy\nFunnel plot illustrating publication bias using the cosmetic results outcome measure", "The findings of the PEDro critical appraisal indicated that the methodological quality of the current evidence base was poor (Table 1). Although all studies clearly defined their study participants, only two studies demonstrated baseline comparability between the operative and non-operative groups [21]. Furthermore, no study randomised their patients to the allocated intervention. No study based their sample size on a power calculation. Whilst it may have been impractical to blind subjects or clinicians to treatment allocation, no study blinded their assessors during the investigations. Although subject drop-out was more than 85% in all but two studies, no study analysed their results by intention-to-treat principles, or adjusted their results to estimate this missing data. Nonetheless, all clearly described their results and appropriately used descriptive and inferential statistical tests to analyse their cohorts.Table 1PEDro scoreCalvo et al. [8]Fremerey et al. [15]Galpin et al. [26]Gstettner et al. [22]Press et al. [21]Taft et al. [9]Eligibility criteria111111Random allocation000000Concealed allocation000000Baseline comparability111100Blind subject000000Blind clinician000000Blind assessor000100Adequate follow-up (≥85%)100001Intention-to treat analysis000000Between-group analysis111111Point estimates and variability111111Total score5445341: criterion satisfied; 0: criterion not satisfied\nPEDro score\n1: criterion satisfied; 0: criterion not satisfied", "In total, 380 patients were included in the review (Table 2). The operative management cohort consisted of 195 shoulders, 125 males and 15 females with a mean age of 24.4 [standard deviation (SD) = 4.5] years. The non-operative group consisted of 185 shoulders, 96 males and 13 females with a mean age of 27.8 (SD = 6.1) years. One study did not document the cohort age or gender, and therefore total numbers for age and gender are incomplete [9]. Five studies solely evaluated outcomes in patients with Grade III Rockwood injuries. One study’s cohort consisted of 78% grade III injuries, and 22% grade V injuries [15]. Given this high proportion, and since this study provided some outcomes based on grade of injury separately, this study was included in the review.Table 2Study characteristicsStudyStudySampleAge (years)Gender (m/f)Gd of ACJ dis.Surgical mgmtNon-surgical mgmtMean follow up (years)OpNOpOpNOpOpNOpCalvo et al. [8]Retro3211403527/511/0Type IIIPhemister techniqueSling—type not specified. Physiotherapy—type not specifiedOp: 123 mNOp: 41 mFremerey et al. [15]Retro514633.735.948/339/777Gd III20Gd VPDSPhysiotherapy exercisesOp: 6.1NOp: 6.5Galpin et al. [26]Retro1621293716/017/3Gd IIIBosworthSling—type not specified. Strength and ROM exercisesOp: 35.0 mNOp: 33.7 mGstettner et al. [22]Retro282237.236.225/320/2Gd IIIHook plateSling—type not specified. Physiotherapy—type not specified34 mPress et al. [21]Retro161030.749.612/49/1Gd IIIWeaver–Dunn procedureSling—type not specified. Rehabilitation—not specified32.3 mTaft et al. [9]Retro5275N/SN/SN/SN/SGd IIIBosworth screw or 1–3 Steinmann pin fixationSling, Kenny–Howard splint, taping or cast. Mobilisation exercisesOp: 10.8Nop: 9.5av average, Co conservative management, Comp complete, Dis disruption, exs exercises, Gd grade, gp group, immob immobilised, inj injury, K wire Kirschner wire, Lig ligament, m months, Mgmt management, ND Not documented, Op operative management, physio physiotherapy, Pros Prospective, rec recreational, recon reconstructed, Ret retrospective, RTA road traffic accident, Sed sedentary, sh shoulder, wks weeks\nStudy characteristics\nOp: 123 m\nNOp: 41 m\n77\nGd III\n20\nGd V\nOp: 6.1\nNOp: 6.5\nOp: 35.0 m\nNOp: 33.7 m\nOp: 10.8\nNop: 9.5\nav average, Co conservative management, Comp complete, Dis disruption, exs exercises, Gd grade, gp group, immob immobilised, inj injury, K wire Kirschner wire, Lig ligament, m months, Mgmt management, ND Not documented, Op operative management, physio physiotherapy, Pros Prospective, rec recreational, recon reconstructed, Ret retrospective, RTA road traffic accident, Sed sedentary, sh shoulder, wks weeks\nThe operative procedures were clearly described in all papers. Five studies included fixation using Kirschner wire or screw fixation methods, whilst one study used hook plates as the form of fixation [22]. All studies reported repair of coracoclavicular and acromioclavicular ligaments using sutures. The non-operative management adopted was poorly described. All subjects were immobilised using a sling of some description. However, Taft et al. [9] reported immobilising patients using a taping technique or cast, but did not specify the method of application. Immobilisation varied between the studies from 2 weeks [8] to 4 weeks [9]. The remaining papers reported immobilising patients until pain and symptoms had resolved. Following this, subjects commenced range of motion and/or strength rehabilitation programmes, but this was not described in detail in the studies. The follow-up period ranged from 32 months [21] to 10.8 years [9].", "The results of the meta-analysis are shown in Table 3. The primary outcome of this study was the Constant score. This revealed that there was a significantly better functional outcome following operative compared to non-operative management of grade III acromioclavicular separation (MD = 9.70; 95% CI: 1.00, 18.40; P = 0.03; Fig. 3). However, this is based on the complete data from one study [22]. There was no statistically significant difference between the interventions in respect to strength, pain, throwing ability, loss of anatomical reduction, ossification of the coracoclavicular ligament or acromioclavicular joint osteoarthritis (P > 0.05). There were significantly poorer cosmetic results following non-operative management (RD = 0.64; 95% CI: 1.09, 0.19; P < 0.0001; Fig. 4). The results also suggested that there was a significantly greater duration of sick leave following operative management compared to non-operative management (MD = 3.3; 95% CI: 2.10, 4.50; P < 0.001; Fig. 5). Although a relatively low incidence, unsurprisingly, the infection rate was significantly higher in the operative compared to the non-operative group (RD = 0.05; 95% CI: 0.01, 0.09; Fig. 6).Fig. 3Forest plot illustrating constant scoreFig. 4Forest plot illustrating cosmetic outcomeFig. 5Forest plot illustrating duration of sick leaveFig. 6Forest plot illustrating infection rateTable 3Results of the meta-analysisOutcomeStudiesEffect estimateP-valueHeterogeneity\nI\n2\nChi2 (P value)Duration of sick leave23.30 (2.10, 4.50)<0.0001NENEConstant score29.70 (1.00, −18.50)0.03NENEThrowing ability3−0.00 (−0.15, 0.15)0.9800.57Strength (≥90% normal)2−0.01 (−0.12, 0.11)0.9000.82Strength (≤70% normal)20.35 (0.04, 3.51)0.3700.88No pain20.90 (0.33, 2.41)0.8300.60Severe pain2−0.00 (−0.06, 0.06)0.9500.97Poor cosmetic outcome4−0.79 (−0.92, −0.66)<0.0001520.10Tenderness over the acromioclavicular joint20.08 (−0.23, 0.40)0.61750.05Implant failure20.04 (−0.05, 0.13)0.4200.91Infection50.05 (0.01, 0.09)0.0300.66Loss of anatomical reduction20.50 (−1.07, 0.52)0.5098<0.0001Ossification of the coracoclavicular ligament20.17 (−0.32, 0.66)0.50820.02Acromioclavicular joint osteoarthritis30.12 (−0.21, 0.46)0.46890.001NE not estimable\nForest plot illustrating constant score\nForest plot illustrating cosmetic outcome\nForest plot illustrating duration of sick leave\nForest plot illustrating infection rate\nResults of the meta-analysis\nNE not estimable\nOne study assessed the effect of range of motion. Fremerey et al. [15] reported no substantial difference between the interventions, with two patients demonstrating a loss of abduction and external rotation following operative management compared to one patient following non-operative rehabilitation (P > 0.05). Finally, Press et al. [21] reported loss of reduction from the anatomical position. They documented that two patients following operative management presented with loss of reduction, compared to no cases following non-operative management.", "It was not possible to undertake a sensitivity analysis since none of the studies included in the meta-analysis were randomised controlled trials.", "The principal finding of this study was that, for the majority of outcomes, there was no statistically significant difference in clinical or radiological outcomes between operative and non-operative management for this patient group. Nonetheless, there was some evidence to suggest that operative management provided a significantly better Constant score compared to non-operative following grade III acromioclavicular dislocation, but this was based on the results from a single study. Non-operative management was associated with significantly poorer cosmetic outcome but less sick leave compared to operative management (P < 0.001).\nThe current evidence base presented with a number of methodological limitations, including not randomising patients to group allocation, permitting allocation bias [23], and not blinding assessors to subject groups, therefore increasing the risk of assessment bias [24]. Finally, the studies did not base their sample sizes on power calculations, increasing the risk of a type II statistical error due to an insufficient sample size [25]. Accordingly, future robust, well-designed RCTs are required to improve the currently poor evidence base in order to determine the optimal management strategy for patients following grade III acromioclavicular separation.\nA previous meta-analysis by Philips et al. [6] ultimately advised against surgical treatment following grade III acromioclavicular separation. This review differed to the previous review as it specifically included only those studies with cohorts of predominantly grade III acromicoclavicular separation. Furthermore, with the advantage of time, we have been able to include a number of studies which have recently been published on this topic. Whilst there is agreement with some of Philips et al.’s [6] conclusions, this study concludes, with some reservations, that there is little difference in the outcome of operative and non-operative management for patients following grade III acromioclavicular separation, with the exception that non-operative management provides cosmetically poorer outcomes. A more recent paper [22] has shown that maintenance of reduction is possible with the operative group having a statistically better outcome than the non-operative group. As operative techniques improve, there may be a paradigm shift from the historically poor results of fixation with K-wires.\nThe mechanism of injury appeared similar among the studies, with a combination of sporting, accidental and occupation trauma as the associated factor. Few studies distinguished whether upper limb dominance was a factor in outcome. This may have a been particularly important confounding variable for functional-based outcomes and return to sports measures, where those with a dominant limb injury may present with poorer outcomes—particularly during early review—compared to non-dominant limb injury. A further confounding factor which may have affected outcome was time from injury to surgery. Rolf et al. [2] reported that those patients who had an acute acromioclavicular reconstruction after trauma reported significantly better functional outcomes and patient satisfaction rates as well as lower complication rates compared to patients with delayed reconstruction. Whilst the four studies reported that all operations were acute, the duration from injury to surgical reconstruction was not clearly stated in the papers of Galphin et al. [26] or Taft et al. [9]. Finally, to the study’s credit, the follow-up period of the evidence base was reasonable, providing some evidence for detecting late failures and longer-term outcomes.\nThe results of this meta-analysis indicate that there was no significant difference in respect to maintenance of anatomical reduction between operative and non-operative management of grade III acromioclavicular separation (P = 0.15). Calvo et al. [8] acknowledged that complete reduction may not necessarily be a pre-requisite for optimal functional outcome [6, 8, 27]. They suggested that the rationale of surgical reconstruction to achieve anatomic alignment for full functional recovery may not always be achieved following grade III acromioclavicular separation [8]. Thus, anatomical reduction alone cannot justify operative intervention. However, the method of assessing anatomical alignment was unclear from the included studies. Previous authors have argued that only by assessing the acromioclavicular joint with stress radiography can anatomical position be determined [28]. Accordingly, future study is recommended to determine the optimal method of radiographic evaluation of acromioclavicular displacement following operative and non-operative management strategies.\nThe current meta-analysis suggests that there was no difference in the incidence of OA or ossification of the coraclavicular ligament between the two management strategies. Authors such as Calvo et al. [8] have suggested that the incidence of OA changes may be related to the surgical manipulation and inability to maintain reduction, whilst ossification of the coracoclavicular ligaments has been associated with the manipulation of ligament tissue when attempting to repair it [8, 29]. Fremerey et al. [15] and Taft et al. [9] suggested that post-traumatic OA in surgically managed patients is related to the unphysiological contact of traumatised joint surface and subsequent joint cartilage injury.\nSeveral authors have suggested that surgical reconstruction should be advocated for those patients who have physically demanding occupations or sporting interests. However, since the mean age of each study’s cohort was under 28 years, and the mechanism of injury was largely sporting or occupationally related, there was little evidence to substantiate this claim based on clinical outcomes. Furthermore, since this study suggested that duration of sick leave was significantly higher following non-operative procedures, and that there was no significant difference in strength outcomes, then non-operative management may be seen as superior to manage this patient group. For those patients who carry heavy weights on their shoulders, such as soldiers carrying rucksacks, operative intervention may be indicated to prevent anatomical deformities from affecting return to normal activities.\nThe literature poorly described the non-operative management strategies used. Historically, various straps, harnesses, casting techniques and traction methods have been used as part of closed reduction [30–34]. Currently, there appears greater support for the use of internal rotation slings. Since non-operative management strategies were not clearly defined, it remains unclear as to whether there was a variation in these strategies between the studies. Furthermore, it also remains unclear as to whether clinical outcomes are affected by the type of rehabilitation programme adopted, immobilisation method or period of immobilisation.\nAs Gstettner et al. [22] acknowledged, the disadvantage of all operative strategies is the risk of complications. This was mirrored by our study, which demonstrated a significantly higher risk of infection following surgical management compared to non-operative treatment (P = 0.03). However, the incidence of infection was relatively low following acromioclavicular surgery. There has been a paradigm shift in clinical practice. Earlier studies adopted Phemister fixation methods. This developed into a consensus of using Bosworth screw and then later Hook plate fixation methods [35–38]. Currently, TightRope fixation methods and biodegradable slings have been introduced [39, 40]. Whilst clinical differences between operative and non-operative strategies have evaluated previous surgical interventions, the comparison to biodegradable sling fixation is yet to be evaluated using a large, well-designed RCT.\nFinally, no study compared cost-effectiveness with a formal economical evaluation. Since the meta-analysis indicated that whilst patients reported a shorter duration of sick leave following non-operative management, and that higher costs of hospitalisation, the operative procedure and prolonged rehabilitation are associated with this strategy, there initially appears to be greater support from an economic perspective for adopting a non-operative management strategy for this patient group. Formal health economical assessment is therefore imperative to assess the differences in this and clinical outcomes when developing the evidence base with well-designed, sufficiently powerful RCTs.\nTo conclude, based on the current evidence base, operative management of grade III acromioclavicular dislocations results in a better cosmetic outcome (P < 0.0001) but a greater duration of sick leave (P < 0.001) compared to non-operative management. There was no difference between the two interventions in terms of strength, pain and throwing ability (P > 0.05)." ]

[ "introduction", "materials|methods", null, null, null, null, null, null, null, "results", null, null, null, null, null, "discussion" ]

[ "Acromioclavicular", "Dislocation", "ACJT", "Rockwood type", "Systematic review" ]

[ "Bone and Bones", "Image Processing, Computer-Assisted", "Magnetic Resonance Imaging", "Radiation Dosage", "Tomography, X-Ray Computed" ]

[ "Search Strategy and Criteria", "Fundamentals of CT Imaging", "Micro-CT", "MDCT", "HR-pQCT" ]

[ "To determine the relevant articles, we used the PubMed (PM) and Google Scholar (GS) search engines. The following list of search phrases was used, with the number of results reported parenthetically: “CT” AND “trabecular bone microarchitecture” (PM: 14; GS: 549); “microCT iliac crest biopsy” (PM: 17; GS: 122); “microct” AND “cortical bone porosity” (PM: 19; GS: 74); “XtremeCT” (PM: 8; GS: 139); “HR-pQCT” (PM: 68; GS: 247); “MDCT” AND “bone structure” (PM: 7; GS: 111); “SR-μCT” AND “bone structure” (PM: 8; GS: 82). From these, we selected those articles we believed most relevant.", "CT is a three-dimensional radiographic imaging technique. The image formation process begins with the acquisition of sequential radiographic projections captured over a range of angular positions around the object of interest. The cross-sectional field of view is reconstructed using established computational techniques based on radon projection theory [50]. Similar to simple radiography, the reconstructed image’s intensity values represent the local radiographic attenuation: a material property related to the object’s electron density (atomic number and mass density). The contrast between soft and mineralized tissue in CT is high, due to the relative electron-dense inorganic component (calcium hydroxyapatite) of the bone matrix [8]. Since the logarithm of the measured absorption scales linearly with the length of material the beam has penetrated, simultaneous quantitative measurements of bone density are possible. Calibration of grayscale linear attenuation to BMD is accomplished by imaging reference phantoms containing objects with known hydroxyapatite concentrations [21, 49].\nThese principles capture high-resolution images of bone across a range of structural scales. Several classes of CT devices are presently used for high-resolution imaging of trabecular microarchitecture and cortical ultrastructure (Table 1). Techniques with spatial resolution between 1 and 100 μm are referred to as micro-CT and may replace tedious serial staining procedures required by histomorphometric analysis of thin sections and offer the possibility of longitudinal in vivo investigations in small animals, such as mice and rats. Many early micro-CT approaches used synchrotron radiation (SR) [59], which is still the method of choice for ultrahigh-resolution applications. The use of desktop laboratory scanners equipped with xray tubes is much more convenient than performing an experiment at one of the few synchrotron facilities available worldwide. Initial and ongoing university-based research in the past decade has led to the development of a variety of commercial xray tube-based micro-CT scanners (Table 1).Table 1Descriptive summary of high-resolution CT technologies availableModalityReferencesPrimary manufacturersSkeletal sitesField of view size (mm)Voxel size (µm)Effective doseTypical scan timeMicro-CT\n55, 129, 135, 151\nGE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland)Siemens (New York, NY)SkyScan (Kontich, Belgium)Xradia (Pleasanton, CA)SpecimensBiopsies(ex vivo)2–1000.3–100 (isotropic)NA30 minutes to 3 hoursMDCT/fp-vCT\n5, 41, 43, 44, 69, 70, 73, 85, 118, 124\nGE Heathcare (Waukesha, WI)Philips (Amsterdam, The Netherlands)Siemens (New York, NY)Toshiba Corp (Tokyo, Japan)Specimens (ex vivo)SpineFemurForearm (in vivo)100–250156–300 (in plane)300–500 (slice thickness)0.1–5 mSv< 30 secondsHR-pQCT\n14, 82, 130\nScanco Medical AG (Brüttisellen, Switzerland)Specimens (ex vivo)Distal radiusDistal tibia(in vivo)12641–123(isotropic)3–4 µSv3 minutesNA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.\n\nDescriptive summary of high-resolution CT technologies available\nGE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland)\nSiemens (New York, NY)\nSkyScan (Kontich, Belgium)\nXradia (Pleasanton, CA)\nSpecimens\nBiopsies\n(ex vivo)\nGE Heathcare (Waukesha, WI)\nPhilips (Amsterdam, The Netherlands)\nSiemens (New York, NY)\nToshiba Corp (Tokyo, Japan)\nSpecimens (ex vivo)\nSpine\nFemur\nForearm (in vivo)\n156–300 (in plane)\n300–500 (slice thickness)\nSpecimens (ex vivo)\nDistal radius\nDistal tibia(in vivo)\nNA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.\nApplication of standard whole-body MDCT to imaging trabecular bone in the central and peripheral skeleton was investigated at several research institutes [5, 43, 58, 70, 73]. Compared to standard two-dimensional QCT, volumetric MDCT imaging studies use higher-dose acquisition protocols (3 mSv versus 0.06–0.3 mSv) with a higher in plane resolution (200–300 μm versus 500–1000 μm) and smaller slice thickness and spacing (500 μm versus 1–10 mm). Recently, a standard MDCT gantry was combined with two-dimensional flat panel detector technology to provide rapid continuous acquisitions at high isotropic spatial resolution [60, 124]. In the last 5 years, a high-resolution, limited-field-of-view CT device became commercially available for dedicated imaging of bone structure in the peripheral skeleton [14, 80, 82, 97]. The HR-pQCT imaging system consists of a microfocus xray source and high-resolution charge-coupled device (CCD) detector that can produce tomographic images with a nominal resolution as high as 41 μm for a 12.6-mm field of view.", "Micro-CT has achieved widespread use in the laboratory for rapid, nondestructive imaging of bone specimens [47, 51, 108] and noninvasive imaging in animal models [54, 148]. The pervasive use of this technology at many research institutes invested in bone science research has widely eclipsed traditional histomorphometry for evaluating bone microarchitecture. However, micro-CT does not provide direct information on cellular function and remodeling activity, which continues to be the domain of bone histology.\nConventional laboratory micro-CT typically uses a cone-beam, polychromatic xray source, which produces photons spanning a broad range of energies. In contrast, SR micro-CT, only available at a limited number of particle accelerator facilities worldwide, is typically performed using a parallel, monochromatic beam [72, 83, 127, 141]. While the first commercial micro-CT scanner consisted of a single-row 512-pixel detector [126], modern scanners employ areal CCD detectors up to 11 megapixels and are capable of acquiring projection data for more than 1000 slices simultaneously [129, 135, 151]. Dedicated specimen ex vivo scanners are typically designed with the specimen oriented on a high-resolution motorized stage for translation and rotation. In this scenario, the source and detector remain in a fixed position during the scan, while the specimen rotates in the field of view. Conversely, preclinical micro-CT systems designed for in vivo imaging of small animals utilize a fixed gantry with the xray source and detector rotating and translating about the field of view [54, 147].\nAnalogous to clinical QCT, the grayscale attenuation values of reconstructed micro-CT images can be converted to hydroxyapatite concentration. Various calibration procedures based on idealized phantoms or theoretical calculations were established to derive relations between attenuation and BMD [21, 72, 102, 107, 110, 113]. However, micro-CT imaging is subject to xray scattering (SR and polychromatic micro-CT) and beam-hardening effects (polychromatic micro-CT) that can introduce nonnegligible error in the depiction of mineralization that is spatially variant and dependent on the geometry and composition of the object imaged [48, 79, 102]. Methods to minimize beam-hardening effects in conventional micro-CT include xray filtration to block “soft” low-energy xrays [102] and empirical corrections based on phantom measurements [21]. Using these procedures, apparent BMD and tissue level mineral density can be accurately measured (r2 = 0.78–0.99) in specimens of similar size and composition [21, 79].\nMorphometric indices analogous to classical histomorphometry can be calculated from micro-CT images of trabecular and cortical bone. Comparison of structural parameters of specimens scanned with these systems and mechanical testing suggest the amount of bone and the architecture of trabecular bone contribute to mechanical strength [57]. Advanced image-processing methodologies are used to quantify trabecular bone microarchitecture beyond measures of bone volume fraction (BV/TV). Specifically, direct three-dimensional measures of mean distances and measures of structural heterogeneity are used to characterize trabeculae and marrow spaces [64, 65] (Fig. 1). The degree of anisotropy, a measure of the degree of structural orientation of the trabecular network, can be calculated from the principal structural directions calculated by the mean intercept length techniques [63] and is highly related to the directional dependence of bone’s biomechanical properties [115]. A measure of the structural connectedness was also adapted to bone, based on the Euler number [114]. The shape of the trabecular structure is characterized using the structure model index (SMI), an index of surface convexity that estimates the degree to which the structure consists of rod-like or plate-like elements [66]. Furthermore, several groups developed algorithms to decompose the trabecular structure to independently quantify the volume and scale of rod-like and plate-like elements [92, 117, 139].Fig. 1A–BThree-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].\n\nThree-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].\nThe ultrastructure of cortical bone is an important determinant of bone strength [128, 131], critical in fracture initiation and propagation [146], and known to change with age [37], disease [7], and therapy [11]. Volumetric and morphologic characterization of the cortical ultrastructure has predominantly focused on Haversian and Volkmann canal network of cadaveric femoral neck specimens [13, 32, 36, 38, 153]. Resolution improvements heralding the evolution of nano-CT (CT with submicron resolution) recently paved the way for a complete evaluation of cortical bone ultrastructure, including the distribution of osteocyte lacunae [131, 146].\nWhile volumetric density and microarchitecture information provide improved fracture risk prediction and some explanation for treatment efficacy, more direct estimates of bone mechanical strength that inherently account for geometry, microarchitecture, and even composition are the ultimate goal for improving fracture risk prediction and management of osteoporosis. Computational modeling approaches were introduced to take advantage of the detailed information in high-resolution images of bone. Finite element analysis (FEA) is a common computational tool in engineering fields, critical to design and failure analysis. Applied to high-resolution images of bone, the apparent biomechanical properties (stiffness, elastic modulus) of a biologically complex microstructure are computed by decomposing the structure into small cubic elements (the voxels) with assumed mechanical properties [109, 143] (Fig. 2). Numerous studies have utilized micro-FEA techniques to investigate the micromechanics of bone strength, failure, and relation to bone microarchitecture [81, 142].Fig. 2A micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.\n\nA micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.\nAs an ex vivo imaging modality, the application of micro-CT to clinical research is limited to examining small bone biopsy specimens. There are numerous data over the last 20 years characterizing age, gender, and anatomic differences in cadaveric specimens of bone [13, 45, 52, 64, 67, 68, 94]. Clinical applications generally preclude access to the most relevant sites, such as the spine and proximal femur. Accordingly, a minimally invasive bone biopsy is typically acquired from the iliac crest [75]. Trabecular microarchitecture at the iliac crest reflects vertebral fracture status [56, 74] and changes after the onset of menopause [1]. Quantification of bone structure from iliac crest biopsies is also an important end point in longitudinal drug efficacy studies of parathyroid hormone [33, 53, 76, 121], strontium ranelate [3], and various bisphosphonates [11, 12, 46, 112, 120, 122, 123]. Borah et al. [11] recently reported major ultrastructural changes to the cortical bone of the iliac crest after 5 years of treatment with risedronate. Despite the resolution advantages of in vitro imaging studies, the invasiveness of the procedure, inherent variability in specimen collection [28, 29], and the limited correlation to bone quality at clinically relevant sites for fragility fracture (proximal femur, lumbar spine, distal radius) [35] are drawbacks to the clinical application of micro-CT.", "MDCT is a clinical CT technique available in most diagnostic imaging departments, using scanners from a number of manufacturers (Table 1). Therefore, a dedicated scanner is not required. The spatial resolution of this technique is limited, with an in-plane resolution ranging from 150 to 300 μm and a minimum slice thickness of around 300 μm. These spatial resolutions are above trabecular dimensions, and imaging of individual trabeculae is subject to considerable partial-volume effects. However, given the larger size of intertrabecular spaces, trabecular bone parameters obtained with this technique are observed to correlate moderately well with those determined by contact radiography and micro-CT of bone specimens (r = 0.53–0.70) [71, 91], as well as micro-CT (r = 0.44–0.99) [43, 44, 69, 70, 85, 118]. The advantage of the MDCT technique is that central regions of the skeleton critically relevant to osteoporosis and fracture risk assessment, such as the spine [70, 73, 85] and proximal femur [5, 43], can be visualized. However, to achieve adequate spatial resolution and image quality, the required radiation exposure is substantial, offsetting the technique’s applicability in clinical, routine, and scientific studies (Table 1). High-resolution CT protocols are typically associated with an effective dose of approximately 3 mSv (1.5 years of natural background radiation), several orders of magnitude greater than standard DXA or HR-pQCT (4–13 μSv) and an order of magnitude higher than standard two-dimensional QCT (0.06–0.3 mSv) [41].\nWhile early MDCT systems used for bone structure imaging typically consisted of four to 16 rows of detector elements [73], modern systems have 64 to 320 rows [69]. Most recently, high-resolution areal CCD detectors have been combined with a standard clinical CT gantry to provide substantial improvements in scan time and resolution [124]. In each case, the tomographic acquisition is performed with the subject lying supine on the scan table within the gantry. The xray source and detector ensemble continuously rotate about the field of view while the gantry translates along the rotational axis, effectively producing a helical series of projections [78]. Simultaneous calibration of Hounsfield units to mineral density is typically accomplished by placement of a solid hydroxyapatite phantom below the patient [49].\nThe analysis of trabecular microarchitecture from MDCT and flat-panel volumetric CT image data primarily involves the application of traditional histomorphometry [5, 17, 43, 70, 118], where BV/TV, Tb.N, Tb.Th, and Tb.Sp are calculated two-dimensionally using plate model assumptions [116]. In contrast, Ito et al. [73] used direct three-dimensional measures of trabecular dimensions, connectivity, and SMI in a MDCT study of the lumbar spine, finding a strong correlation (r = 0.98) between BV/TV measured by micro-CT and MDCT. Other specialized measures of trabecular dimensions using the fuzzy distance transform (which does not require a threshold binarization process) have been proposed by Krebs et al. [85].\nTo date, in vivo human studies using MDCT to assess bone structure are limited due to radiation dose concerns. Ito et al. [73] demonstrated SMI and BV/TV measured from MDCT images of the lumbar spine provided superior fracture discrimination to areal BMD by DXA. Graeff et al. [58] showed teriparatide treatment effects are better monitored by architectural parameters of the spine obtained through MDCT than by BMD (Fig. 3). In a cross-sectional cohort study of adolescent girls with and without anorexia nervosa, Bredella et al. [17] observed diminished trabecular microarchitecture at the distal radius in subjects with anorexia nervosa compared to controls, despite no differences in lumbar areal BMD. In a companion study, Lawson et al. [89] observed the abnormal trabecular microarchitecture in these patients are predicted by IGF-1, leptin, and androgen levels, with positive correlations (r = 0.32–0.72) to BV/TV, Tb.Th, and Tb.N.Fig. 3A–DIn vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.\n\nIn vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.", "A dedicated extremity imaging system designed for trabecular-scale imaging is currently available from a single manufacturer (XtremeCT; Scanco Medical AG, Brüttisellen, Switzerland). This device has the advantage of a higher signal-to-noise ratio and spatial resolution (nominal isotropic voxel dimension of 82 μm) when compared to MDCT. Furthermore, the radiation dose is lower when compared to whole-body CT and does not involve critical, radiosensitive organs in skeletally mature adults. There are several disadvantages to this technology. It is limited to peripheral skeletal sites and provides no direct insight into bone quality in the lumbar spine or proximal femur, common sites for osteoporotic fragility fractures. Additionally, there are currently a limited number of devices installed globally, which are primarily located at major research institutions with few available in clinical radiology departments.\nIn HR-pQCT, a standard protocol recommended by the manufacturer is utilized for most studies [14, 82]. The patient’s forearm or ankle is immobilized in a carbon fiber cast fixed within the gantry of the scanner. A single scout projection image of the distal radius or tibia is acquired to define the tomographic scan region. This scout image is acquired at an AP orientation at the wrist and at an oblique (45°) anterolateral-posteromedial orientation at the ankle. This tomographic region spans 9.02 mm in length (110 slices) and is localized to a fixed offset proximal from either the radial or tibial midjoint line and extends proximally. The offset is 9.5 mm in the radius and 22.5 mm in the tibia (Fig. 4). This method does not account for differences in bone length and may be a confounding source of variability in cross-sectional studies [16]. In the radius, the default axial scan location partially includes the most common site for fracture and location, where the bone microstructure is most strongly correlated to experimental strength of the forearm under a simulated falling load [106].Fig. 4A–BScout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.\n\nScout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.\nThere are several different protocol modifications for developmental studies in children and adolescents [26, 84] to account for patient size and age-related changes in bone length and to avoid radiation exposure to the epiphyseal growth plate and inclusion of provisional mineralized tissues from this region. In a cross-sectional study of age- and gender-related differences in the microarchitecture of the distal forearm of adolescents, Kirmani et al. [84] used a fixed offset (1 mm) with respect to the proximal extent of the distal epiphyseal growth plate of the radius. In contrast, Burrows et al. [26] selected a region offset of 8% of the total tibial length proximal to the tibial endplate. While there are a number of studies underway investigating other scan locations in adults, including more proximal sites dominated by cortical bone, the internal configuration of the XtremeCT gantry prohibits the positioning of true diaphyseal sites in the radius or tibia within the limited 15-cm longitudinal range of the source-detector ensemble.\nThe reconstructed images are analyzed using a standard protocol provided by the manufacturer. The operator chaperones a semiautomated contouring process to identify the periosteal boundary and segment the cortical and trabecular compartments [42]. The trabecular bone structure is extracted using an edge enhancement and threshold procedure [88]. While the importance of threshold selection for morphometric analysis is well described for other CT devices [61], few studies to date investigate threshold effects or propose new methods for HR-pQCT [23, 42]. The default compartmental and structural segmentation provides the basis for the subsequent densitometric, morphometric, and biomechanical analyses. For subjects with very thin or highly porous cortical bone, this segmentation procedure may fail to capture the cortical structure [42, 80]; therefore, more sophisticated autocontouring techniques that operate on the fine-structure segmentation have been proposed [18, 23]. For the XtremeCT, reproducibility of densitometric measures is very high (coefficient of variation < 1%), while biomechanical and morphometric measures typically have a coefficient of variation of 4% to 5% [14, 80, 82, 100, 105].\nThe linear attenuation values of the tomographic images are converted to hydroxyapatite mineral densities using a beam-hardening correction and phantom calibration procedure previously described for the ex vivo micro-CT system [21]. Based on this calibration, volumetric BMD is determined independently for cortical and trabecular bone compartments using the segmentation process described previously. HR-pQCT images are used to derive surrogate measures of areal BMD in the ultradistal radius [22]. This technique is associated with a high level of agreement (r2 > 0.8) with multiple clinical DXA devices.\nUnlike MDCT, which has a large slice thickness relative to the in-plane resolution, the high isotropic resolution of HR-pQCT (82 μm) permits direct three-dimensional assessment of intertrabecular distances. These measures were validated against micro-CT gold standards [23, 93, 97]. From the binary image of the extracted trabecular structure, three-dimensional distance transformation techniques are used to calculated trabecular number [65]. While the intertrabecular distances are large compared to the voxel dimension, the average trabecular thickness (100–150 μm) is only one to two voxels wide. Accordingly, direct measures of thickness and bone volume are complicated by considerable partial-volume effects. In the standard analysis protocol, BV/TV is derived from the trabecular volumetric BMD, assuming a fixed mineralization of 1200 mg hydroxyapatite per cm3 for compact bone (BV/TV = Tb.BMD/1200). Trabecular BMD is determined by calculating the mean mineral density of the full medullary compartment. From the direct measure of Tb.N and the densitometrically derived BV/TV, Tb.Th and Tb.Sp are derived using standard stereologic relations, assuming plate model geometry [87, 116].\nThere are several potential concerns with this approach. First, Sekhon et al. [134] documented substantial errors in the measurement of trabecular BMD related to biologically relevant variations in cortical thickness and the magnitude of trabecular BMD itself. This may be related to xray scatter effects and residual beam-hardening artifacts. These errors are primarily a concern for cross-sectional studies, when cortical thickness and trabecular BMD may span a broad range. It is less of a concern in longitudinal studies, where the percent of change as the primary end point, such as age-, disease-, and therapy-related changes in cortical thickness and trabecular BMD, is comparatively small. Second, the assumption of a fixed-matrix mineralization is inconsistent with the established action of many common antifracture therapeutics [10]. Changes in matrix mineral density is expected to cause an increase in BMD, irrespective of bone volume changes, and result in an overestimation of BV/TV and propagate error to the derivative measures of Tb.Th and Tb.Sp, confounding any actual therapy-related effect on trabecular bone volume and structure.\nSeveral studies investigate other measures of bone microarchitecture and topology from HR-pQCT images, including connectivity, SMI, and anisotropy; however, there is mixed evidence of their reliability at in vivo resolutions [23, 93, 97, 136]. Recently, more sophisticated approaches to cortical bone segmentation have been proposed [18] that allow direct three-dimensional assessment of cortical thickness and quantification of cortical ultrastructure (Fig. 5), including intracortical porosity and canal diameter [24, 111]Fig. 5A–D(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.\n\n(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.\nThe ability of HR-pQCT to resolve the trabecular microarchitecture and a level of the cortical ultrastructure lends itself to calculating direct estimates of bone strength by voxel-based micro-FEA. For HR-pQCT scans of the distal radius and tibia, this is typically used to estimate strength under uniaxial compression, which approximates to the common loading condition for Colles’ fracture of the radius [106] and normal gaited loading for the tibia. This technique was validated against both higher-resolution models (based on micro-CT images) and empirical measures of strength [93, 99]. The application of micro-FEA is primarily performed assuming homogeneous material properties [15, 24, 40, 93, 99, 105], although it is also possible to produce models with material properties scaled according to mineralization [95, 99]. In addition to whole-bone mechanics, micro-FEA can be used to determine the relative load distribution between cortical and trabecular compartments [98] and estimate mechanical implications of specific structural features, such as the resolvable cortical ultrastructure [24].\nFor clinical investigations into longitudinal changes in HR-pQCT-derived measures of bone quality, it is critical baseline and followup scans be matched, since bone structure and geometry vary substantially along the axial direction [16]. Operator positioning for followup scans are aided by visual reference to the positioning of the baseline scan. Furthermore, postprocess image registration is performed to ensure comparable regions of interest are used for the image analysis. The manufacturer provides software that matches slices based on the periosteal cross-sectional area and limits the analyzed region to the slices common in baseline and followup [87]. Alternatively, MacNeil and Boyd [100] demonstrated three-dimensional image registration techniques can provide improved short- and medium-term reproducibility when compared to the default slice-matching approach. This approach may be more appropriate in longitudinal studies (long-term studies, anabolic therapy trials) where periosteal apposition would confound registration based strictly on cross-sectional area. As discussed earlier, the challenge in obtaining meaningful results in longitudinal studies in children or adolescents experiencing rapid growth is not trivial and requires careful consideration of standardized procedures for scan positioning and analysis [26, 101].\nThere is a growing body of literature featuring HR-pQCT assessment of bone quality. The first cross-sectional studies by Boutroy et al. [14] and Khosla et al. [82] reported gender-specific, age-related differences in trabecular bone microarchitecture. Several centers have observed age-related differences in micro-FEA estimates of bone strength in normative cross-sectional cohorts [24, 40, 96]. Furthermore, Burghardt et al. [24] and Macdonald et al. [96] demonstrated the ability of HR-pQCT to detect dramatic age-related differences in cortical porosity in females using new techniques for the analysis of cortical ultrastructure [19]. A microstructural basis for ethnicity-related differences in bone strength between East Asian and white women was reported in two studies [149, 150]. Sornay-Rendu et al. [137] suggested cortical and trabecular morphology provided additional fracture discrimination independent of areal BMD in osteopenic women. In the same cohort, Boutroy et al. [15] showed micro-FEA mechanical measures provided additional discriminatory power between osteopenic women with and without distal radius fractures. While the initial focus was predominantly related to fracture discrimination in postmenopausal osteopenia and osteoporosis [15, 103, 104, 137, 138, 140, 144, 145], a number of studies have utilized HR-pQCT to investigate developmental changes in bone quality and fracture risk [27, 34, 84], as well as secondary causes of bone loss [4, 20, 31, 62, 90].\nMost recently, data from the first HR-pQCT single- and multicenter longitudinal trials were published. In a multicenter, head-to-head, randomized, placebo-controlled trial of denosumab (a RANKL inhibitor) and alendronate (a bisphosphonate), Seeman et al. [133] reported more pronounced antiresorptive efficacy with denosumab than alendronate. In particular, cortical thickness was preserved or improved at the radius and tibia with either treatment, while cortical bone loss progressed in the control group. Burghardt et al. [25] reported similar cortical bone changes and additionally showed a preservation of compressive bone strength by micro-FEA after 24 months of alendronate treatment. Longitudinal microarchitectural changes also occurred with strontium ranelate [125] and teriparatide [95]." ]

[ null, null, null, null, null ]

[ "Introduction", "Search Strategy and Criteria", "Fundamentals of CT Imaging", "Micro-CT", "MDCT", "HR-pQCT", "Discussion" ]

[ "The skeleton is composed of cortical and trabecular bone, both contributing to bone strength and the resistance of bone to fracture. The strength of bone and risk of fracture are important outcomes in the study of growth and peak bone accrual, aging, postmenopausal bone loss, cancer-related bone loss, and for conditions such as diabetes, osteogenesis imperfecta, osteoarthritis, rheumatoid arthritis, and others. Typically, clinical assessment of skeletal health is based on measures of bone mineral density (BMD), usually obtained using dual-energy xray absorptiometry (DXA), a two-dimensional, projection-based radiographic technique that measures integral BMD of both cortical and trabecular bone (areal BMD). In addition to DXA, three-dimensional quantitative CT (QCT) is used to assess BMD. This three-dimensional technique measures volumetric BMD and permits characterization of bone geometry and density as elements of fracture risk. Furthermore, QCT can examine cortical and trabecular bone independently.\nBMD only explains about 70% to 75% of the variance in strength [2], while the remaining variance is due to the cumulative and synergistic effect of factors such as bone macro- and microarchitecture, tissue composition, and microdamage [30, 132]. In a multicenter fracture intervention trial, the antifracture efficacy of all drugs tested was only partially explained by their effects on BMD [9]. In this context, and specifically in osteoporosis, the concept of bone quality emerged [132]. Bone quality represents different properties of bone to researchers and could encompass one or all of the factors mentioned. Trabecular and cortical bone play important roles in the prediction of bone strength and are affected by age, gender, and metabolic conditions and have varying responses to therapy. Sites containing predominantly trabecular bone, such as the hip, spine, and wrist, are most frequently associated with increased fracture risk [39]. Traditionally, trabecular microarchitecture is assessed from bone biopsies by two-dimensional histomorphometry [116], applying stereologic principals to measure trabecular bone volume fraction (the ratio of mineralized bone volume to total volume [BV/TV]), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), and trabecular number (Tb.N). On the other hand, cortical bone constitutes about 80% of total bone mass. In previous studies, cortical thinning [119] and increased cortical porosity [13] were important factors in the assessment of osteoporosis and bone strength. Recently, focus increased on the ultrastructure of cortical bone that can be attributed to resorption spaces, merging of Haversian canals, and clustering of osteons [6, 77]. Since the ultrastructure of cortical bone has a major impact on its mechanical properties [131, 152], characterizing cortical ultrastructure is also important in the context of bone strength and prediction of fracture risk. With the advent of improved three-dimensional imaging techniques, such as micro-CT [51, 126], high-resolution peripheral QCT (HR-pQCT) [14, 82], and multidetector CT (MDCT) [73], it is possible to perform in vitro and in vivo imaging of bone across different structural scales from the whole bone to the ultrastructural level.\nIn a recent publication for the radiographic imaging communities, we reviewed three-dimensional techniques for assessing bone structure in osteoporosis [86]. We described MRI, image processing, and CT, concluding these modalities have the potential to play an important role in imaging three-dimensional trabecular microarchitecture in osteoporosis. In this review, we focus on the development and application of high-resolution CT for quantifying cortical and trabecular bone structure covering specific clinical applications of interest to the orthopaedic research community.\nWe reviewed (1) the fundamentals of high-resolution CT for evaluating trabecular microarchitecture and cortical ultrastructure, (2) use of micro-CT for studying bone specimens ex vivo, (3) use of MDCT for in vivo human imaging studies, and (4) recent studies using HR-pQCT to characterize bone structure in the context of skeletal disease, particularly its ability to discriminate between subjects with and without fractures and monitor longitudinal response to therapeutic intervention.\n[SUBTITLE] Search Strategy and Criteria [SUBSECTION] To determine the relevant articles, we used the PubMed (PM) and Google Scholar (GS) search engines. The following list of search phrases was used, with the number of results reported parenthetically: “CT” AND “trabecular bone microarchitecture” (PM: 14; GS: 549); “microCT iliac crest biopsy” (PM: 17; GS: 122); “microct” AND “cortical bone porosity” (PM: 19; GS: 74); “XtremeCT” (PM: 8; GS: 139); “HR-pQCT” (PM: 68; GS: 247); “MDCT” AND “bone structure” (PM: 7; GS: 111); “SR-μCT” AND “bone structure” (PM: 8; GS: 82). From these, we selected those articles we believed most relevant.\nTo determine the relevant articles, we used the PubMed (PM) and Google Scholar (GS) search engines. The following list of search phrases was used, with the number of results reported parenthetically: “CT” AND “trabecular bone microarchitecture” (PM: 14; GS: 549); “microCT iliac crest biopsy” (PM: 17; GS: 122); “microct” AND “cortical bone porosity” (PM: 19; GS: 74); “XtremeCT” (PM: 8; GS: 139); “HR-pQCT” (PM: 68; GS: 247); “MDCT” AND “bone structure” (PM: 7; GS: 111); “SR-μCT” AND “bone structure” (PM: 8; GS: 82). From these, we selected those articles we believed most relevant.\n[SUBTITLE] Fundamentals of CT Imaging [SUBSECTION] CT is a three-dimensional radiographic imaging technique. The image formation process begins with the acquisition of sequential radiographic projections captured over a range of angular positions around the object of interest. The cross-sectional field of view is reconstructed using established computational techniques based on radon projection theory [50]. Similar to simple radiography, the reconstructed image’s intensity values represent the local radiographic attenuation: a material property related to the object’s electron density (atomic number and mass density). The contrast between soft and mineralized tissue in CT is high, due to the relative electron-dense inorganic component (calcium hydroxyapatite) of the bone matrix [8]. Since the logarithm of the measured absorption scales linearly with the length of material the beam has penetrated, simultaneous quantitative measurements of bone density are possible. Calibration of grayscale linear attenuation to BMD is accomplished by imaging reference phantoms containing objects with known hydroxyapatite concentrations [21, 49].\nThese principles capture high-resolution images of bone across a range of structural scales. Several classes of CT devices are presently used for high-resolution imaging of trabecular microarchitecture and cortical ultrastructure (Table 1). Techniques with spatial resolution between 1 and 100 μm are referred to as micro-CT and may replace tedious serial staining procedures required by histomorphometric analysis of thin sections and offer the possibility of longitudinal in vivo investigations in small animals, such as mice and rats. Many early micro-CT approaches used synchrotron radiation (SR) [59], which is still the method of choice for ultrahigh-resolution applications. The use of desktop laboratory scanners equipped with xray tubes is much more convenient than performing an experiment at one of the few synchrotron facilities available worldwide. Initial and ongoing university-based research in the past decade has led to the development of a variety of commercial xray tube-based micro-CT scanners (Table 1).Table 1Descriptive summary of high-resolution CT technologies availableModalityReferencesPrimary manufacturersSkeletal sitesField of view size (mm)Voxel size (µm)Effective doseTypical scan timeMicro-CT\n55, 129, 135, 151\nGE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland)Siemens (New York, NY)SkyScan (Kontich, Belgium)Xradia (Pleasanton, CA)SpecimensBiopsies(ex vivo)2–1000.3–100 (isotropic)NA30 minutes to 3 hoursMDCT/fp-vCT\n5, 41, 43, 44, 69, 70, 73, 85, 118, 124\nGE Heathcare (Waukesha, WI)Philips (Amsterdam, The Netherlands)Siemens (New York, NY)Toshiba Corp (Tokyo, Japan)Specimens (ex vivo)SpineFemurForearm (in vivo)100–250156–300 (in plane)300–500 (slice thickness)0.1–5 mSv< 30 secondsHR-pQCT\n14, 82, 130\nScanco Medical AG (Brüttisellen, Switzerland)Specimens (ex vivo)Distal radiusDistal tibia(in vivo)12641–123(isotropic)3–4 µSv3 minutesNA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.\n\nDescriptive summary of high-resolution CT technologies available\nGE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland)\nSiemens (New York, NY)\nSkyScan (Kontich, Belgium)\nXradia (Pleasanton, CA)\nSpecimens\nBiopsies\n(ex vivo)\nGE Heathcare (Waukesha, WI)\nPhilips (Amsterdam, The Netherlands)\nSiemens (New York, NY)\nToshiba Corp (Tokyo, Japan)\nSpecimens (ex vivo)\nSpine\nFemur\nForearm (in vivo)\n156–300 (in plane)\n300–500 (slice thickness)\nSpecimens (ex vivo)\nDistal radius\nDistal tibia(in vivo)\nNA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.\nApplication of standard whole-body MDCT to imaging trabecular bone in the central and peripheral skeleton was investigated at several research institutes [5, 43, 58, 70, 73]. Compared to standard two-dimensional QCT, volumetric MDCT imaging studies use higher-dose acquisition protocols (3 mSv versus 0.06–0.3 mSv) with a higher in plane resolution (200–300 μm versus 500–1000 μm) and smaller slice thickness and spacing (500 μm versus 1–10 mm). Recently, a standard MDCT gantry was combined with two-dimensional flat panel detector technology to provide rapid continuous acquisitions at high isotropic spatial resolution [60, 124]. In the last 5 years, a high-resolution, limited-field-of-view CT device became commercially available for dedicated imaging of bone structure in the peripheral skeleton [14, 80, 82, 97]. The HR-pQCT imaging system consists of a microfocus xray source and high-resolution charge-coupled device (CCD) detector that can produce tomographic images with a nominal resolution as high as 41 μm for a 12.6-mm field of view.\nCT is a three-dimensional radiographic imaging technique. The image formation process begins with the acquisition of sequential radiographic projections captured over a range of angular positions around the object of interest. The cross-sectional field of view is reconstructed using established computational techniques based on radon projection theory [50]. Similar to simple radiography, the reconstructed image’s intensity values represent the local radiographic attenuation: a material property related to the object’s electron density (atomic number and mass density). The contrast between soft and mineralized tissue in CT is high, due to the relative electron-dense inorganic component (calcium hydroxyapatite) of the bone matrix [8]. Since the logarithm of the measured absorption scales linearly with the length of material the beam has penetrated, simultaneous quantitative measurements of bone density are possible. Calibration of grayscale linear attenuation to BMD is accomplished by imaging reference phantoms containing objects with known hydroxyapatite concentrations [21, 49].\nThese principles capture high-resolution images of bone across a range of structural scales. Several classes of CT devices are presently used for high-resolution imaging of trabecular microarchitecture and cortical ultrastructure (Table 1). Techniques with spatial resolution between 1 and 100 μm are referred to as micro-CT and may replace tedious serial staining procedures required by histomorphometric analysis of thin sections and offer the possibility of longitudinal in vivo investigations in small animals, such as mice and rats. Many early micro-CT approaches used synchrotron radiation (SR) [59], which is still the method of choice for ultrahigh-resolution applications. The use of desktop laboratory scanners equipped with xray tubes is much more convenient than performing an experiment at one of the few synchrotron facilities available worldwide. Initial and ongoing university-based research in the past decade has led to the development of a variety of commercial xray tube-based micro-CT scanners (Table 1).Table 1Descriptive summary of high-resolution CT technologies availableModalityReferencesPrimary manufacturersSkeletal sitesField of view size (mm)Voxel size (µm)Effective doseTypical scan timeMicro-CT\n55, 129, 135, 151\nGE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland)Siemens (New York, NY)SkyScan (Kontich, Belgium)Xradia (Pleasanton, CA)SpecimensBiopsies(ex vivo)2–1000.3–100 (isotropic)NA30 minutes to 3 hoursMDCT/fp-vCT\n5, 41, 43, 44, 69, 70, 73, 85, 118, 124\nGE Heathcare (Waukesha, WI)Philips (Amsterdam, The Netherlands)Siemens (New York, NY)Toshiba Corp (Tokyo, Japan)Specimens (ex vivo)SpineFemurForearm (in vivo)100–250156–300 (in plane)300–500 (slice thickness)0.1–5 mSv< 30 secondsHR-pQCT\n14, 82, 130\nScanco Medical AG (Brüttisellen, Switzerland)Specimens (ex vivo)Distal radiusDistal tibia(in vivo)12641–123(isotropic)3–4 µSv3 minutesNA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.\n\nDescriptive summary of high-resolution CT technologies available\nGE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland)\nSiemens (New York, NY)\nSkyScan (Kontich, Belgium)\nXradia (Pleasanton, CA)\nSpecimens\nBiopsies\n(ex vivo)\nGE Heathcare (Waukesha, WI)\nPhilips (Amsterdam, The Netherlands)\nSiemens (New York, NY)\nToshiba Corp (Tokyo, Japan)\nSpecimens (ex vivo)\nSpine\nFemur\nForearm (in vivo)\n156–300 (in plane)\n300–500 (slice thickness)\nSpecimens (ex vivo)\nDistal radius\nDistal tibia(in vivo)\nNA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.\nApplication of standard whole-body MDCT to imaging trabecular bone in the central and peripheral skeleton was investigated at several research institutes [5, 43, 58, 70, 73]. Compared to standard two-dimensional QCT, volumetric MDCT imaging studies use higher-dose acquisition protocols (3 mSv versus 0.06–0.3 mSv) with a higher in plane resolution (200–300 μm versus 500–1000 μm) and smaller slice thickness and spacing (500 μm versus 1–10 mm). Recently, a standard MDCT gantry was combined with two-dimensional flat panel detector technology to provide rapid continuous acquisitions at high isotropic spatial resolution [60, 124]. In the last 5 years, a high-resolution, limited-field-of-view CT device became commercially available for dedicated imaging of bone structure in the peripheral skeleton [14, 80, 82, 97]. The HR-pQCT imaging system consists of a microfocus xray source and high-resolution charge-coupled device (CCD) detector that can produce tomographic images with a nominal resolution as high as 41 μm for a 12.6-mm field of view.\n[SUBTITLE] Micro-CT [SUBSECTION] Micro-CT has achieved widespread use in the laboratory for rapid, nondestructive imaging of bone specimens [47, 51, 108] and noninvasive imaging in animal models [54, 148]. The pervasive use of this technology at many research institutes invested in bone science research has widely eclipsed traditional histomorphometry for evaluating bone microarchitecture. However, micro-CT does not provide direct information on cellular function and remodeling activity, which continues to be the domain of bone histology.\nConventional laboratory micro-CT typically uses a cone-beam, polychromatic xray source, which produces photons spanning a broad range of energies. In contrast, SR micro-CT, only available at a limited number of particle accelerator facilities worldwide, is typically performed using a parallel, monochromatic beam [72, 83, 127, 141]. While the first commercial micro-CT scanner consisted of a single-row 512-pixel detector [126], modern scanners employ areal CCD detectors up to 11 megapixels and are capable of acquiring projection data for more than 1000 slices simultaneously [129, 135, 151]. Dedicated specimen ex vivo scanners are typically designed with the specimen oriented on a high-resolution motorized stage for translation and rotation. In this scenario, the source and detector remain in a fixed position during the scan, while the specimen rotates in the field of view. Conversely, preclinical micro-CT systems designed for in vivo imaging of small animals utilize a fixed gantry with the xray source and detector rotating and translating about the field of view [54, 147].\nAnalogous to clinical QCT, the grayscale attenuation values of reconstructed micro-CT images can be converted to hydroxyapatite concentration. Various calibration procedures based on idealized phantoms or theoretical calculations were established to derive relations between attenuation and BMD [21, 72, 102, 107, 110, 113]. However, micro-CT imaging is subject to xray scattering (SR and polychromatic micro-CT) and beam-hardening effects (polychromatic micro-CT) that can introduce nonnegligible error in the depiction of mineralization that is spatially variant and dependent on the geometry and composition of the object imaged [48, 79, 102]. Methods to minimize beam-hardening effects in conventional micro-CT include xray filtration to block “soft” low-energy xrays [102] and empirical corrections based on phantom measurements [21]. Using these procedures, apparent BMD and tissue level mineral density can be accurately measured (r2 = 0.78–0.99) in specimens of similar size and composition [21, 79].\nMorphometric indices analogous to classical histomorphometry can be calculated from micro-CT images of trabecular and cortical bone. Comparison of structural parameters of specimens scanned with these systems and mechanical testing suggest the amount of bone and the architecture of trabecular bone contribute to mechanical strength [57]. Advanced image-processing methodologies are used to quantify trabecular bone microarchitecture beyond measures of bone volume fraction (BV/TV). Specifically, direct three-dimensional measures of mean distances and measures of structural heterogeneity are used to characterize trabeculae and marrow spaces [64, 65] (Fig. 1). The degree of anisotropy, a measure of the degree of structural orientation of the trabecular network, can be calculated from the principal structural directions calculated by the mean intercept length techniques [63] and is highly related to the directional dependence of bone’s biomechanical properties [115]. A measure of the structural connectedness was also adapted to bone, based on the Euler number [114]. The shape of the trabecular structure is characterized using the structure model index (SMI), an index of surface convexity that estimates the degree to which the structure consists of rod-like or plate-like elements [66]. Furthermore, several groups developed algorithms to decompose the trabecular structure to independently quantify the volume and scale of rod-like and plate-like elements [92, 117, 139].Fig. 1A–BThree-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].\n\nThree-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].\nThe ultrastructure of cortical bone is an important determinant of bone strength [128, 131], critical in fracture initiation and propagation [146], and known to change with age [37], disease [7], and therapy [11]. Volumetric and morphologic characterization of the cortical ultrastructure has predominantly focused on Haversian and Volkmann canal network of cadaveric femoral neck specimens [13, 32, 36, 38, 153]. Resolution improvements heralding the evolution of nano-CT (CT with submicron resolution) recently paved the way for a complete evaluation of cortical bone ultrastructure, including the distribution of osteocyte lacunae [131, 146].\nWhile volumetric density and microarchitecture information provide improved fracture risk prediction and some explanation for treatment efficacy, more direct estimates of bone mechanical strength that inherently account for geometry, microarchitecture, and even composition are the ultimate goal for improving fracture risk prediction and management of osteoporosis. Computational modeling approaches were introduced to take advantage of the detailed information in high-resolution images of bone. Finite element analysis (FEA) is a common computational tool in engineering fields, critical to design and failure analysis. Applied to high-resolution images of bone, the apparent biomechanical properties (stiffness, elastic modulus) of a biologically complex microstructure are computed by decomposing the structure into small cubic elements (the voxels) with assumed mechanical properties [109, 143] (Fig. 2). Numerous studies have utilized micro-FEA techniques to investigate the micromechanics of bone strength, failure, and relation to bone microarchitecture [81, 142].Fig. 2A micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.\n\nA micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.\nAs an ex vivo imaging modality, the application of micro-CT to clinical research is limited to examining small bone biopsy specimens. There are numerous data over the last 20 years characterizing age, gender, and anatomic differences in cadaveric specimens of bone [13, 45, 52, 64, 67, 68, 94]. Clinical applications generally preclude access to the most relevant sites, such as the spine and proximal femur. Accordingly, a minimally invasive bone biopsy is typically acquired from the iliac crest [75]. Trabecular microarchitecture at the iliac crest reflects vertebral fracture status [56, 74] and changes after the onset of menopause [1]. Quantification of bone structure from iliac crest biopsies is also an important end point in longitudinal drug efficacy studies of parathyroid hormone [33, 53, 76, 121], strontium ranelate [3], and various bisphosphonates [11, 12, 46, 112, 120, 122, 123]. Borah et al. [11] recently reported major ultrastructural changes to the cortical bone of the iliac crest after 5 years of treatment with risedronate. Despite the resolution advantages of in vitro imaging studies, the invasiveness of the procedure, inherent variability in specimen collection [28, 29], and the limited correlation to bone quality at clinically relevant sites for fragility fracture (proximal femur, lumbar spine, distal radius) [35] are drawbacks to the clinical application of micro-CT.\nMicro-CT has achieved widespread use in the laboratory for rapid, nondestructive imaging of bone specimens [47, 51, 108] and noninvasive imaging in animal models [54, 148]. The pervasive use of this technology at many research institutes invested in bone science research has widely eclipsed traditional histomorphometry for evaluating bone microarchitecture. However, micro-CT does not provide direct information on cellular function and remodeling activity, which continues to be the domain of bone histology.\nConventional laboratory micro-CT typically uses a cone-beam, polychromatic xray source, which produces photons spanning a broad range of energies. In contrast, SR micro-CT, only available at a limited number of particle accelerator facilities worldwide, is typically performed using a parallel, monochromatic beam [72, 83, 127, 141]. While the first commercial micro-CT scanner consisted of a single-row 512-pixel detector [126], modern scanners employ areal CCD detectors up to 11 megapixels and are capable of acquiring projection data for more than 1000 slices simultaneously [129, 135, 151]. Dedicated specimen ex vivo scanners are typically designed with the specimen oriented on a high-resolution motorized stage for translation and rotation. In this scenario, the source and detector remain in a fixed position during the scan, while the specimen rotates in the field of view. Conversely, preclinical micro-CT systems designed for in vivo imaging of small animals utilize a fixed gantry with the xray source and detector rotating and translating about the field of view [54, 147].\nAnalogous to clinical QCT, the grayscale attenuation values of reconstructed micro-CT images can be converted to hydroxyapatite concentration. Various calibration procedures based on idealized phantoms or theoretical calculations were established to derive relations between attenuation and BMD [21, 72, 102, 107, 110, 113]. However, micro-CT imaging is subject to xray scattering (SR and polychromatic micro-CT) and beam-hardening effects (polychromatic micro-CT) that can introduce nonnegligible error in the depiction of mineralization that is spatially variant and dependent on the geometry and composition of the object imaged [48, 79, 102]. Methods to minimize beam-hardening effects in conventional micro-CT include xray filtration to block “soft” low-energy xrays [102] and empirical corrections based on phantom measurements [21]. Using these procedures, apparent BMD and tissue level mineral density can be accurately measured (r2 = 0.78–0.99) in specimens of similar size and composition [21, 79].\nMorphometric indices analogous to classical histomorphometry can be calculated from micro-CT images of trabecular and cortical bone. Comparison of structural parameters of specimens scanned with these systems and mechanical testing suggest the amount of bone and the architecture of trabecular bone contribute to mechanical strength [57]. Advanced image-processing methodologies are used to quantify trabecular bone microarchitecture beyond measures of bone volume fraction (BV/TV). Specifically, direct three-dimensional measures of mean distances and measures of structural heterogeneity are used to characterize trabeculae and marrow spaces [64, 65] (Fig. 1). The degree of anisotropy, a measure of the degree of structural orientation of the trabecular network, can be calculated from the principal structural directions calculated by the mean intercept length techniques [63] and is highly related to the directional dependence of bone’s biomechanical properties [115]. A measure of the structural connectedness was also adapted to bone, based on the Euler number [114]. The shape of the trabecular structure is characterized using the structure model index (SMI), an index of surface convexity that estimates the degree to which the structure consists of rod-like or plate-like elements [66]. Furthermore, several groups developed algorithms to decompose the trabecular structure to independently quantify the volume and scale of rod-like and plate-like elements [92, 117, 139].Fig. 1A–BThree-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].\n\nThree-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].\nThe ultrastructure of cortical bone is an important determinant of bone strength [128, 131], critical in fracture initiation and propagation [146], and known to change with age [37], disease [7], and therapy [11]. Volumetric and morphologic characterization of the cortical ultrastructure has predominantly focused on Haversian and Volkmann canal network of cadaveric femoral neck specimens [13, 32, 36, 38, 153]. Resolution improvements heralding the evolution of nano-CT (CT with submicron resolution) recently paved the way for a complete evaluation of cortical bone ultrastructure, including the distribution of osteocyte lacunae [131, 146].\nWhile volumetric density and microarchitecture information provide improved fracture risk prediction and some explanation for treatment efficacy, more direct estimates of bone mechanical strength that inherently account for geometry, microarchitecture, and even composition are the ultimate goal for improving fracture risk prediction and management of osteoporosis. Computational modeling approaches were introduced to take advantage of the detailed information in high-resolution images of bone. Finite element analysis (FEA) is a common computational tool in engineering fields, critical to design and failure analysis. Applied to high-resolution images of bone, the apparent biomechanical properties (stiffness, elastic modulus) of a biologically complex microstructure are computed by decomposing the structure into small cubic elements (the voxels) with assumed mechanical properties [109, 143] (Fig. 2). Numerous studies have utilized micro-FEA techniques to investigate the micromechanics of bone strength, failure, and relation to bone microarchitecture [81, 142].Fig. 2A micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.\n\nA micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.\nAs an ex vivo imaging modality, the application of micro-CT to clinical research is limited to examining small bone biopsy specimens. There are numerous data over the last 20 years characterizing age, gender, and anatomic differences in cadaveric specimens of bone [13, 45, 52, 64, 67, 68, 94]. Clinical applications generally preclude access to the most relevant sites, such as the spine and proximal femur. Accordingly, a minimally invasive bone biopsy is typically acquired from the iliac crest [75]. Trabecular microarchitecture at the iliac crest reflects vertebral fracture status [56, 74] and changes after the onset of menopause [1]. Quantification of bone structure from iliac crest biopsies is also an important end point in longitudinal drug efficacy studies of parathyroid hormone [33, 53, 76, 121], strontium ranelate [3], and various bisphosphonates [11, 12, 46, 112, 120, 122, 123]. Borah et al. [11] recently reported major ultrastructural changes to the cortical bone of the iliac crest after 5 years of treatment with risedronate. Despite the resolution advantages of in vitro imaging studies, the invasiveness of the procedure, inherent variability in specimen collection [28, 29], and the limited correlation to bone quality at clinically relevant sites for fragility fracture (proximal femur, lumbar spine, distal radius) [35] are drawbacks to the clinical application of micro-CT.\n[SUBTITLE] MDCT [SUBSECTION] MDCT is a clinical CT technique available in most diagnostic imaging departments, using scanners from a number of manufacturers (Table 1). Therefore, a dedicated scanner is not required. The spatial resolution of this technique is limited, with an in-plane resolution ranging from 150 to 300 μm and a minimum slice thickness of around 300 μm. These spatial resolutions are above trabecular dimensions, and imaging of individual trabeculae is subject to considerable partial-volume effects. However, given the larger size of intertrabecular spaces, trabecular bone parameters obtained with this technique are observed to correlate moderately well with those determined by contact radiography and micro-CT of bone specimens (r = 0.53–0.70) [71, 91], as well as micro-CT (r = 0.44–0.99) [43, 44, 69, 70, 85, 118]. The advantage of the MDCT technique is that central regions of the skeleton critically relevant to osteoporosis and fracture risk assessment, such as the spine [70, 73, 85] and proximal femur [5, 43], can be visualized. However, to achieve adequate spatial resolution and image quality, the required radiation exposure is substantial, offsetting the technique’s applicability in clinical, routine, and scientific studies (Table 1). High-resolution CT protocols are typically associated with an effective dose of approximately 3 mSv (1.5 years of natural background radiation), several orders of magnitude greater than standard DXA or HR-pQCT (4–13 μSv) and an order of magnitude higher than standard two-dimensional QCT (0.06–0.3 mSv) [41].\nWhile early MDCT systems used for bone structure imaging typically consisted of four to 16 rows of detector elements [73], modern systems have 64 to 320 rows [69]. Most recently, high-resolution areal CCD detectors have been combined with a standard clinical CT gantry to provide substantial improvements in scan time and resolution [124]. In each case, the tomographic acquisition is performed with the subject lying supine on the scan table within the gantry. The xray source and detector ensemble continuously rotate about the field of view while the gantry translates along the rotational axis, effectively producing a helical series of projections [78]. Simultaneous calibration of Hounsfield units to mineral density is typically accomplished by placement of a solid hydroxyapatite phantom below the patient [49].\nThe analysis of trabecular microarchitecture from MDCT and flat-panel volumetric CT image data primarily involves the application of traditional histomorphometry [5, 17, 43, 70, 118], where BV/TV, Tb.N, Tb.Th, and Tb.Sp are calculated two-dimensionally using plate model assumptions [116]. In contrast, Ito et al. [73] used direct three-dimensional measures of trabecular dimensions, connectivity, and SMI in a MDCT study of the lumbar spine, finding a strong correlation (r = 0.98) between BV/TV measured by micro-CT and MDCT. Other specialized measures of trabecular dimensions using the fuzzy distance transform (which does not require a threshold binarization process) have been proposed by Krebs et al. [85].\nTo date, in vivo human studies using MDCT to assess bone structure are limited due to radiation dose concerns. Ito et al. [73] demonstrated SMI and BV/TV measured from MDCT images of the lumbar spine provided superior fracture discrimination to areal BMD by DXA. Graeff et al. [58] showed teriparatide treatment effects are better monitored by architectural parameters of the spine obtained through MDCT than by BMD (Fig. 3). In a cross-sectional cohort study of adolescent girls with and without anorexia nervosa, Bredella et al. [17] observed diminished trabecular microarchitecture at the distal radius in subjects with anorexia nervosa compared to controls, despite no differences in lumbar areal BMD. In a companion study, Lawson et al. [89] observed the abnormal trabecular microarchitecture in these patients are predicted by IGF-1, leptin, and androgen levels, with positive correlations (r = 0.32–0.72) to BV/TV, Tb.Th, and Tb.N.Fig. 3A–DIn vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.\n\nIn vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.\nMDCT is a clinical CT technique available in most diagnostic imaging departments, using scanners from a number of manufacturers (Table 1). Therefore, a dedicated scanner is not required. The spatial resolution of this technique is limited, with an in-plane resolution ranging from 150 to 300 μm and a minimum slice thickness of around 300 μm. These spatial resolutions are above trabecular dimensions, and imaging of individual trabeculae is subject to considerable partial-volume effects. However, given the larger size of intertrabecular spaces, trabecular bone parameters obtained with this technique are observed to correlate moderately well with those determined by contact radiography and micro-CT of bone specimens (r = 0.53–0.70) [71, 91], as well as micro-CT (r = 0.44–0.99) [43, 44, 69, 70, 85, 118]. The advantage of the MDCT technique is that central regions of the skeleton critically relevant to osteoporosis and fracture risk assessment, such as the spine [70, 73, 85] and proximal femur [5, 43], can be visualized. However, to achieve adequate spatial resolution and image quality, the required radiation exposure is substantial, offsetting the technique’s applicability in clinical, routine, and scientific studies (Table 1). High-resolution CT protocols are typically associated with an effective dose of approximately 3 mSv (1.5 years of natural background radiation), several orders of magnitude greater than standard DXA or HR-pQCT (4–13 μSv) and an order of magnitude higher than standard two-dimensional QCT (0.06–0.3 mSv) [41].\nWhile early MDCT systems used for bone structure imaging typically consisted of four to 16 rows of detector elements [73], modern systems have 64 to 320 rows [69]. Most recently, high-resolution areal CCD detectors have been combined with a standard clinical CT gantry to provide substantial improvements in scan time and resolution [124]. In each case, the tomographic acquisition is performed with the subject lying supine on the scan table within the gantry. The xray source and detector ensemble continuously rotate about the field of view while the gantry translates along the rotational axis, effectively producing a helical series of projections [78]. Simultaneous calibration of Hounsfield units to mineral density is typically accomplished by placement of a solid hydroxyapatite phantom below the patient [49].\nThe analysis of trabecular microarchitecture from MDCT and flat-panel volumetric CT image data primarily involves the application of traditional histomorphometry [5, 17, 43, 70, 118], where BV/TV, Tb.N, Tb.Th, and Tb.Sp are calculated two-dimensionally using plate model assumptions [116]. In contrast, Ito et al. [73] used direct three-dimensional measures of trabecular dimensions, connectivity, and SMI in a MDCT study of the lumbar spine, finding a strong correlation (r = 0.98) between BV/TV measured by micro-CT and MDCT. Other specialized measures of trabecular dimensions using the fuzzy distance transform (which does not require a threshold binarization process) have been proposed by Krebs et al. [85].\nTo date, in vivo human studies using MDCT to assess bone structure are limited due to radiation dose concerns. Ito et al. [73] demonstrated SMI and BV/TV measured from MDCT images of the lumbar spine provided superior fracture discrimination to areal BMD by DXA. Graeff et al. [58] showed teriparatide treatment effects are better monitored by architectural parameters of the spine obtained through MDCT than by BMD (Fig. 3). In a cross-sectional cohort study of adolescent girls with and without anorexia nervosa, Bredella et al. [17] observed diminished trabecular microarchitecture at the distal radius in subjects with anorexia nervosa compared to controls, despite no differences in lumbar areal BMD. In a companion study, Lawson et al. [89] observed the abnormal trabecular microarchitecture in these patients are predicted by IGF-1, leptin, and androgen levels, with positive correlations (r = 0.32–0.72) to BV/TV, Tb.Th, and Tb.N.Fig. 3A–DIn vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.\n\nIn vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.\n[SUBTITLE] HR-pQCT [SUBSECTION] A dedicated extremity imaging system designed for trabecular-scale imaging is currently available from a single manufacturer (XtremeCT; Scanco Medical AG, Brüttisellen, Switzerland). This device has the advantage of a higher signal-to-noise ratio and spatial resolution (nominal isotropic voxel dimension of 82 μm) when compared to MDCT. Furthermore, the radiation dose is lower when compared to whole-body CT and does not involve critical, radiosensitive organs in skeletally mature adults. There are several disadvantages to this technology. It is limited to peripheral skeletal sites and provides no direct insight into bone quality in the lumbar spine or proximal femur, common sites for osteoporotic fragility fractures. Additionally, there are currently a limited number of devices installed globally, which are primarily located at major research institutions with few available in clinical radiology departments.\nIn HR-pQCT, a standard protocol recommended by the manufacturer is utilized for most studies [14, 82]. The patient’s forearm or ankle is immobilized in a carbon fiber cast fixed within the gantry of the scanner. A single scout projection image of the distal radius or tibia is acquired to define the tomographic scan region. This scout image is acquired at an AP orientation at the wrist and at an oblique (45°) anterolateral-posteromedial orientation at the ankle. This tomographic region spans 9.02 mm in length (110 slices) and is localized to a fixed offset proximal from either the radial or tibial midjoint line and extends proximally. The offset is 9.5 mm in the radius and 22.5 mm in the tibia (Fig. 4). This method does not account for differences in bone length and may be a confounding source of variability in cross-sectional studies [16]. In the radius, the default axial scan location partially includes the most common site for fracture and location, where the bone microstructure is most strongly correlated to experimental strength of the forearm under a simulated falling load [106].Fig. 4A–BScout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.\n\nScout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.\nThere are several different protocol modifications for developmental studies in children and adolescents [26, 84] to account for patient size and age-related changes in bone length and to avoid radiation exposure to the epiphyseal growth plate and inclusion of provisional mineralized tissues from this region. In a cross-sectional study of age- and gender-related differences in the microarchitecture of the distal forearm of adolescents, Kirmani et al. [84] used a fixed offset (1 mm) with respect to the proximal extent of the distal epiphyseal growth plate of the radius. In contrast, Burrows et al. [26] selected a region offset of 8% of the total tibial length proximal to the tibial endplate. While there are a number of studies underway investigating other scan locations in adults, including more proximal sites dominated by cortical bone, the internal configuration of the XtremeCT gantry prohibits the positioning of true diaphyseal sites in the radius or tibia within the limited 15-cm longitudinal range of the source-detector ensemble.\nThe reconstructed images are analyzed using a standard protocol provided by the manufacturer. The operator chaperones a semiautomated contouring process to identify the periosteal boundary and segment the cortical and trabecular compartments [42]. The trabecular bone structure is extracted using an edge enhancement and threshold procedure [88]. While the importance of threshold selection for morphometric analysis is well described for other CT devices [61], few studies to date investigate threshold effects or propose new methods for HR-pQCT [23, 42]. The default compartmental and structural segmentation provides the basis for the subsequent densitometric, morphometric, and biomechanical analyses. For subjects with very thin or highly porous cortical bone, this segmentation procedure may fail to capture the cortical structure [42, 80]; therefore, more sophisticated autocontouring techniques that operate on the fine-structure segmentation have been proposed [18, 23]. For the XtremeCT, reproducibility of densitometric measures is very high (coefficient of variation < 1%), while biomechanical and morphometric measures typically have a coefficient of variation of 4% to 5% [14, 80, 82, 100, 105].\nThe linear attenuation values of the tomographic images are converted to hydroxyapatite mineral densities using a beam-hardening correction and phantom calibration procedure previously described for the ex vivo micro-CT system [21]. Based on this calibration, volumetric BMD is determined independently for cortical and trabecular bone compartments using the segmentation process described previously. HR-pQCT images are used to derive surrogate measures of areal BMD in the ultradistal radius [22]. This technique is associated with a high level of agreement (r2 > 0.8) with multiple clinical DXA devices.\nUnlike MDCT, which has a large slice thickness relative to the in-plane resolution, the high isotropic resolution of HR-pQCT (82 μm) permits direct three-dimensional assessment of intertrabecular distances. These measures were validated against micro-CT gold standards [23, 93, 97]. From the binary image of the extracted trabecular structure, three-dimensional distance transformation techniques are used to calculated trabecular number [65]. While the intertrabecular distances are large compared to the voxel dimension, the average trabecular thickness (100–150 μm) is only one to two voxels wide. Accordingly, direct measures of thickness and bone volume are complicated by considerable partial-volume effects. In the standard analysis protocol, BV/TV is derived from the trabecular volumetric BMD, assuming a fixed mineralization of 1200 mg hydroxyapatite per cm3 for compact bone (BV/TV = Tb.BMD/1200). Trabecular BMD is determined by calculating the mean mineral density of the full medullary compartment. From the direct measure of Tb.N and the densitometrically derived BV/TV, Tb.Th and Tb.Sp are derived using standard stereologic relations, assuming plate model geometry [87, 116].\nThere are several potential concerns with this approach. First, Sekhon et al. [134] documented substantial errors in the measurement of trabecular BMD related to biologically relevant variations in cortical thickness and the magnitude of trabecular BMD itself. This may be related to xray scatter effects and residual beam-hardening artifacts. These errors are primarily a concern for cross-sectional studies, when cortical thickness and trabecular BMD may span a broad range. It is less of a concern in longitudinal studies, where the percent of change as the primary end point, such as age-, disease-, and therapy-related changes in cortical thickness and trabecular BMD, is comparatively small. Second, the assumption of a fixed-matrix mineralization is inconsistent with the established action of many common antifracture therapeutics [10]. Changes in matrix mineral density is expected to cause an increase in BMD, irrespective of bone volume changes, and result in an overestimation of BV/TV and propagate error to the derivative measures of Tb.Th and Tb.Sp, confounding any actual therapy-related effect on trabecular bone volume and structure.\nSeveral studies investigate other measures of bone microarchitecture and topology from HR-pQCT images, including connectivity, SMI, and anisotropy; however, there is mixed evidence of their reliability at in vivo resolutions [23, 93, 97, 136]. Recently, more sophisticated approaches to cortical bone segmentation have been proposed [18] that allow direct three-dimensional assessment of cortical thickness and quantification of cortical ultrastructure (Fig. 5), including intracortical porosity and canal diameter [24, 111]Fig. 5A–D(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.\n\n(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.\nThe ability of HR-pQCT to resolve the trabecular microarchitecture and a level of the cortical ultrastructure lends itself to calculating direct estimates of bone strength by voxel-based micro-FEA. For HR-pQCT scans of the distal radius and tibia, this is typically used to estimate strength under uniaxial compression, which approximates to the common loading condition for Colles’ fracture of the radius [106] and normal gaited loading for the tibia. This technique was validated against both higher-resolution models (based on micro-CT images) and empirical measures of strength [93, 99]. The application of micro-FEA is primarily performed assuming homogeneous material properties [15, 24, 40, 93, 99, 105], although it is also possible to produce models with material properties scaled according to mineralization [95, 99]. In addition to whole-bone mechanics, micro-FEA can be used to determine the relative load distribution between cortical and trabecular compartments [98] and estimate mechanical implications of specific structural features, such as the resolvable cortical ultrastructure [24].\nFor clinical investigations into longitudinal changes in HR-pQCT-derived measures of bone quality, it is critical baseline and followup scans be matched, since bone structure and geometry vary substantially along the axial direction [16]. Operator positioning for followup scans are aided by visual reference to the positioning of the baseline scan. Furthermore, postprocess image registration is performed to ensure comparable regions of interest are used for the image analysis. The manufacturer provides software that matches slices based on the periosteal cross-sectional area and limits the analyzed region to the slices common in baseline and followup [87]. Alternatively, MacNeil and Boyd [100] demonstrated three-dimensional image registration techniques can provide improved short- and medium-term reproducibility when compared to the default slice-matching approach. This approach may be more appropriate in longitudinal studies (long-term studies, anabolic therapy trials) where periosteal apposition would confound registration based strictly on cross-sectional area. As discussed earlier, the challenge in obtaining meaningful results in longitudinal studies in children or adolescents experiencing rapid growth is not trivial and requires careful consideration of standardized procedures for scan positioning and analysis [26, 101].\nThere is a growing body of literature featuring HR-pQCT assessment of bone quality. The first cross-sectional studies by Boutroy et al. [14] and Khosla et al. [82] reported gender-specific, age-related differences in trabecular bone microarchitecture. Several centers have observed age-related differences in micro-FEA estimates of bone strength in normative cross-sectional cohorts [24, 40, 96]. Furthermore, Burghardt et al. [24] and Macdonald et al. [96] demonstrated the ability of HR-pQCT to detect dramatic age-related differences in cortical porosity in females using new techniques for the analysis of cortical ultrastructure [19]. A microstructural basis for ethnicity-related differences in bone strength between East Asian and white women was reported in two studies [149, 150]. Sornay-Rendu et al. [137] suggested cortical and trabecular morphology provided additional fracture discrimination independent of areal BMD in osteopenic women. In the same cohort, Boutroy et al. [15] showed micro-FEA mechanical measures provided additional discriminatory power between osteopenic women with and without distal radius fractures. While the initial focus was predominantly related to fracture discrimination in postmenopausal osteopenia and osteoporosis [15, 103, 104, 137, 138, 140, 144, 145], a number of studies have utilized HR-pQCT to investigate developmental changes in bone quality and fracture risk [27, 34, 84], as well as secondary causes of bone loss [4, 20, 31, 62, 90].\nMost recently, data from the first HR-pQCT single- and multicenter longitudinal trials were published. In a multicenter, head-to-head, randomized, placebo-controlled trial of denosumab (a RANKL inhibitor) and alendronate (a bisphosphonate), Seeman et al. [133] reported more pronounced antiresorptive efficacy with denosumab than alendronate. In particular, cortical thickness was preserved or improved at the radius and tibia with either treatment, while cortical bone loss progressed in the control group. Burghardt et al. [25] reported similar cortical bone changes and additionally showed a preservation of compressive bone strength by micro-FEA after 24 months of alendronate treatment. Longitudinal microarchitectural changes also occurred with strontium ranelate [125] and teriparatide [95].\nA dedicated extremity imaging system designed for trabecular-scale imaging is currently available from a single manufacturer (XtremeCT; Scanco Medical AG, Brüttisellen, Switzerland). This device has the advantage of a higher signal-to-noise ratio and spatial resolution (nominal isotropic voxel dimension of 82 μm) when compared to MDCT. Furthermore, the radiation dose is lower when compared to whole-body CT and does not involve critical, radiosensitive organs in skeletally mature adults. There are several disadvantages to this technology. It is limited to peripheral skeletal sites and provides no direct insight into bone quality in the lumbar spine or proximal femur, common sites for osteoporotic fragility fractures. Additionally, there are currently a limited number of devices installed globally, which are primarily located at major research institutions with few available in clinical radiology departments.\nIn HR-pQCT, a standard protocol recommended by the manufacturer is utilized for most studies [14, 82]. The patient’s forearm or ankle is immobilized in a carbon fiber cast fixed within the gantry of the scanner. A single scout projection image of the distal radius or tibia is acquired to define the tomographic scan region. This scout image is acquired at an AP orientation at the wrist and at an oblique (45°) anterolateral-posteromedial orientation at the ankle. This tomographic region spans 9.02 mm in length (110 slices) and is localized to a fixed offset proximal from either the radial or tibial midjoint line and extends proximally. The offset is 9.5 mm in the radius and 22.5 mm in the tibia (Fig. 4). This method does not account for differences in bone length and may be a confounding source of variability in cross-sectional studies [16]. In the radius, the default axial scan location partially includes the most common site for fracture and location, where the bone microstructure is most strongly correlated to experimental strength of the forearm under a simulated falling load [106].Fig. 4A–BScout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.\n\nScout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.\nThere are several different protocol modifications for developmental studies in children and adolescents [26, 84] to account for patient size and age-related changes in bone length and to avoid radiation exposure to the epiphyseal growth plate and inclusion of provisional mineralized tissues from this region. In a cross-sectional study of age- and gender-related differences in the microarchitecture of the distal forearm of adolescents, Kirmani et al. [84] used a fixed offset (1 mm) with respect to the proximal extent of the distal epiphyseal growth plate of the radius. In contrast, Burrows et al. [26] selected a region offset of 8% of the total tibial length proximal to the tibial endplate. While there are a number of studies underway investigating other scan locations in adults, including more proximal sites dominated by cortical bone, the internal configuration of the XtremeCT gantry prohibits the positioning of true diaphyseal sites in the radius or tibia within the limited 15-cm longitudinal range of the source-detector ensemble.\nThe reconstructed images are analyzed using a standard protocol provided by the manufacturer. The operator chaperones a semiautomated contouring process to identify the periosteal boundary and segment the cortical and trabecular compartments [42]. The trabecular bone structure is extracted using an edge enhancement and threshold procedure [88]. While the importance of threshold selection for morphometric analysis is well described for other CT devices [61], few studies to date investigate threshold effects or propose new methods for HR-pQCT [23, 42]. The default compartmental and structural segmentation provides the basis for the subsequent densitometric, morphometric, and biomechanical analyses. For subjects with very thin or highly porous cortical bone, this segmentation procedure may fail to capture the cortical structure [42, 80]; therefore, more sophisticated autocontouring techniques that operate on the fine-structure segmentation have been proposed [18, 23]. For the XtremeCT, reproducibility of densitometric measures is very high (coefficient of variation < 1%), while biomechanical and morphometric measures typically have a coefficient of variation of 4% to 5% [14, 80, 82, 100, 105].\nThe linear attenuation values of the tomographic images are converted to hydroxyapatite mineral densities using a beam-hardening correction and phantom calibration procedure previously described for the ex vivo micro-CT system [21]. Based on this calibration, volumetric BMD is determined independently for cortical and trabecular bone compartments using the segmentation process described previously. HR-pQCT images are used to derive surrogate measures of areal BMD in the ultradistal radius [22]. This technique is associated with a high level of agreement (r2 > 0.8) with multiple clinical DXA devices.\nUnlike MDCT, which has a large slice thickness relative to the in-plane resolution, the high isotropic resolution of HR-pQCT (82 μm) permits direct three-dimensional assessment of intertrabecular distances. These measures were validated against micro-CT gold standards [23, 93, 97]. From the binary image of the extracted trabecular structure, three-dimensional distance transformation techniques are used to calculated trabecular number [65]. While the intertrabecular distances are large compared to the voxel dimension, the average trabecular thickness (100–150 μm) is only one to two voxels wide. Accordingly, direct measures of thickness and bone volume are complicated by considerable partial-volume effects. In the standard analysis protocol, BV/TV is derived from the trabecular volumetric BMD, assuming a fixed mineralization of 1200 mg hydroxyapatite per cm3 for compact bone (BV/TV = Tb.BMD/1200). Trabecular BMD is determined by calculating the mean mineral density of the full medullary compartment. From the direct measure of Tb.N and the densitometrically derived BV/TV, Tb.Th and Tb.Sp are derived using standard stereologic relations, assuming plate model geometry [87, 116].\nThere are several potential concerns with this approach. First, Sekhon et al. [134] documented substantial errors in the measurement of trabecular BMD related to biologically relevant variations in cortical thickness and the magnitude of trabecular BMD itself. This may be related to xray scatter effects and residual beam-hardening artifacts. These errors are primarily a concern for cross-sectional studies, when cortical thickness and trabecular BMD may span a broad range. It is less of a concern in longitudinal studies, where the percent of change as the primary end point, such as age-, disease-, and therapy-related changes in cortical thickness and trabecular BMD, is comparatively small. Second, the assumption of a fixed-matrix mineralization is inconsistent with the established action of many common antifracture therapeutics [10]. Changes in matrix mineral density is expected to cause an increase in BMD, irrespective of bone volume changes, and result in an overestimation of BV/TV and propagate error to the derivative measures of Tb.Th and Tb.Sp, confounding any actual therapy-related effect on trabecular bone volume and structure.\nSeveral studies investigate other measures of bone microarchitecture and topology from HR-pQCT images, including connectivity, SMI, and anisotropy; however, there is mixed evidence of their reliability at in vivo resolutions [23, 93, 97, 136]. Recently, more sophisticated approaches to cortical bone segmentation have been proposed [18] that allow direct three-dimensional assessment of cortical thickness and quantification of cortical ultrastructure (Fig. 5), including intracortical porosity and canal diameter [24, 111]Fig. 5A–D(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.\n\n(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.\nThe ability of HR-pQCT to resolve the trabecular microarchitecture and a level of the cortical ultrastructure lends itself to calculating direct estimates of bone strength by voxel-based micro-FEA. For HR-pQCT scans of the distal radius and tibia, this is typically used to estimate strength under uniaxial compression, which approximates to the common loading condition for Colles’ fracture of the radius [106] and normal gaited loading for the tibia. This technique was validated against both higher-resolution models (based on micro-CT images) and empirical measures of strength [93, 99]. The application of micro-FEA is primarily performed assuming homogeneous material properties [15, 24, 40, 93, 99, 105], although it is also possible to produce models with material properties scaled according to mineralization [95, 99]. In addition to whole-bone mechanics, micro-FEA can be used to determine the relative load distribution between cortical and trabecular compartments [98] and estimate mechanical implications of specific structural features, such as the resolvable cortical ultrastructure [24].\nFor clinical investigations into longitudinal changes in HR-pQCT-derived measures of bone quality, it is critical baseline and followup scans be matched, since bone structure and geometry vary substantially along the axial direction [16]. Operator positioning for followup scans are aided by visual reference to the positioning of the baseline scan. Furthermore, postprocess image registration is performed to ensure comparable regions of interest are used for the image analysis. The manufacturer provides software that matches slices based on the periosteal cross-sectional area and limits the analyzed region to the slices common in baseline and followup [87]. Alternatively, MacNeil and Boyd [100] demonstrated three-dimensional image registration techniques can provide improved short- and medium-term reproducibility when compared to the default slice-matching approach. This approach may be more appropriate in longitudinal studies (long-term studies, anabolic therapy trials) where periosteal apposition would confound registration based strictly on cross-sectional area. As discussed earlier, the challenge in obtaining meaningful results in longitudinal studies in children or adolescents experiencing rapid growth is not trivial and requires careful consideration of standardized procedures for scan positioning and analysis [26, 101].\nThere is a growing body of literature featuring HR-pQCT assessment of bone quality. The first cross-sectional studies by Boutroy et al. [14] and Khosla et al. [82] reported gender-specific, age-related differences in trabecular bone microarchitecture. Several centers have observed age-related differences in micro-FEA estimates of bone strength in normative cross-sectional cohorts [24, 40, 96]. Furthermore, Burghardt et al. [24] and Macdonald et al. [96] demonstrated the ability of HR-pQCT to detect dramatic age-related differences in cortical porosity in females using new techniques for the analysis of cortical ultrastructure [19]. A microstructural basis for ethnicity-related differences in bone strength between East Asian and white women was reported in two studies [149, 150]. Sornay-Rendu et al. [137] suggested cortical and trabecular morphology provided additional fracture discrimination independent of areal BMD in osteopenic women. In the same cohort, Boutroy et al. [15] showed micro-FEA mechanical measures provided additional discriminatory power between osteopenic women with and without distal radius fractures. While the initial focus was predominantly related to fracture discrimination in postmenopausal osteopenia and osteoporosis [15, 103, 104, 137, 138, 140, 144, 145], a number of studies have utilized HR-pQCT to investigate developmental changes in bone quality and fracture risk [27, 34, 84], as well as secondary causes of bone loss [4, 20, 31, 62, 90].\nMost recently, data from the first HR-pQCT single- and multicenter longitudinal trials were published. In a multicenter, head-to-head, randomized, placebo-controlled trial of denosumab (a RANKL inhibitor) and alendronate (a bisphosphonate), Seeman et al. [133] reported more pronounced antiresorptive efficacy with denosumab than alendronate. In particular, cortical thickness was preserved or improved at the radius and tibia with either treatment, while cortical bone loss progressed in the control group. Burghardt et al. [25] reported similar cortical bone changes and additionally showed a preservation of compressive bone strength by micro-FEA after 24 months of alendronate treatment. Longitudinal microarchitectural changes also occurred with strontium ranelate [125] and teriparatide [95].", "To determine the relevant articles, we used the PubMed (PM) and Google Scholar (GS) search engines. The following list of search phrases was used, with the number of results reported parenthetically: “CT” AND “trabecular bone microarchitecture” (PM: 14; GS: 549); “microCT iliac crest biopsy” (PM: 17; GS: 122); “microct” AND “cortical bone porosity” (PM: 19; GS: 74); “XtremeCT” (PM: 8; GS: 139); “HR-pQCT” (PM: 68; GS: 247); “MDCT” AND “bone structure” (PM: 7; GS: 111); “SR-μCT” AND “bone structure” (PM: 8; GS: 82). From these, we selected those articles we believed most relevant.", "CT is a three-dimensional radiographic imaging technique. The image formation process begins with the acquisition of sequential radiographic projections captured over a range of angular positions around the object of interest. The cross-sectional field of view is reconstructed using established computational techniques based on radon projection theory [50]. Similar to simple radiography, the reconstructed image’s intensity values represent the local radiographic attenuation: a material property related to the object’s electron density (atomic number and mass density). The contrast between soft and mineralized tissue in CT is high, due to the relative electron-dense inorganic component (calcium hydroxyapatite) of the bone matrix [8]. Since the logarithm of the measured absorption scales linearly with the length of material the beam has penetrated, simultaneous quantitative measurements of bone density are possible. Calibration of grayscale linear attenuation to BMD is accomplished by imaging reference phantoms containing objects with known hydroxyapatite concentrations [21, 49].\nThese principles capture high-resolution images of bone across a range of structural scales. Several classes of CT devices are presently used for high-resolution imaging of trabecular microarchitecture and cortical ultrastructure (Table 1). Techniques with spatial resolution between 1 and 100 μm are referred to as micro-CT and may replace tedious serial staining procedures required by histomorphometric analysis of thin sections and offer the possibility of longitudinal in vivo investigations in small animals, such as mice and rats. Many early micro-CT approaches used synchrotron radiation (SR) [59], which is still the method of choice for ultrahigh-resolution applications. The use of desktop laboratory scanners equipped with xray tubes is much more convenient than performing an experiment at one of the few synchrotron facilities available worldwide. Initial and ongoing university-based research in the past decade has led to the development of a variety of commercial xray tube-based micro-CT scanners (Table 1).Table 1Descriptive summary of high-resolution CT technologies availableModalityReferencesPrimary manufacturersSkeletal sitesField of view size (mm)Voxel size (µm)Effective doseTypical scan timeMicro-CT\n55, 129, 135, 151\nGE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland)Siemens (New York, NY)SkyScan (Kontich, Belgium)Xradia (Pleasanton, CA)SpecimensBiopsies(ex vivo)2–1000.3–100 (isotropic)NA30 minutes to 3 hoursMDCT/fp-vCT\n5, 41, 43, 44, 69, 70, 73, 85, 118, 124\nGE Heathcare (Waukesha, WI)Philips (Amsterdam, The Netherlands)Siemens (New York, NY)Toshiba Corp (Tokyo, Japan)Specimens (ex vivo)SpineFemurForearm (in vivo)100–250156–300 (in plane)300–500 (slice thickness)0.1–5 mSv< 30 secondsHR-pQCT\n14, 82, 130\nScanco Medical AG (Brüttisellen, Switzerland)Specimens (ex vivo)Distal radiusDistal tibia(in vivo)12641–123(isotropic)3–4 µSv3 minutesNA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.\n\nDescriptive summary of high-resolution CT technologies available\nGE Heathcare (Waukesha, WI) Scanco Medical AG (Brüttisellen, Switzerland)\nSiemens (New York, NY)\nSkyScan (Kontich, Belgium)\nXradia (Pleasanton, CA)\nSpecimens\nBiopsies\n(ex vivo)\nGE Heathcare (Waukesha, WI)\nPhilips (Amsterdam, The Netherlands)\nSiemens (New York, NY)\nToshiba Corp (Tokyo, Japan)\nSpecimens (ex vivo)\nSpine\nFemur\nForearm (in vivo)\n156–300 (in plane)\n300–500 (slice thickness)\nSpecimens (ex vivo)\nDistal radius\nDistal tibia(in vivo)\nNA = not applicable; MDCT = multidetector CT; fp-vCT = flat-panel volumetric CT; HR-pQCT = high-resolution quantitative CT.\nApplication of standard whole-body MDCT to imaging trabecular bone in the central and peripheral skeleton was investigated at several research institutes [5, 43, 58, 70, 73]. Compared to standard two-dimensional QCT, volumetric MDCT imaging studies use higher-dose acquisition protocols (3 mSv versus 0.06–0.3 mSv) with a higher in plane resolution (200–300 μm versus 500–1000 μm) and smaller slice thickness and spacing (500 μm versus 1–10 mm). Recently, a standard MDCT gantry was combined with two-dimensional flat panel detector technology to provide rapid continuous acquisitions at high isotropic spatial resolution [60, 124]. In the last 5 years, a high-resolution, limited-field-of-view CT device became commercially available for dedicated imaging of bone structure in the peripheral skeleton [14, 80, 82, 97]. The HR-pQCT imaging system consists of a microfocus xray source and high-resolution charge-coupled device (CCD) detector that can produce tomographic images with a nominal resolution as high as 41 μm for a 12.6-mm field of view.", "Micro-CT has achieved widespread use in the laboratory for rapid, nondestructive imaging of bone specimens [47, 51, 108] and noninvasive imaging in animal models [54, 148]. The pervasive use of this technology at many research institutes invested in bone science research has widely eclipsed traditional histomorphometry for evaluating bone microarchitecture. However, micro-CT does not provide direct information on cellular function and remodeling activity, which continues to be the domain of bone histology.\nConventional laboratory micro-CT typically uses a cone-beam, polychromatic xray source, which produces photons spanning a broad range of energies. In contrast, SR micro-CT, only available at a limited number of particle accelerator facilities worldwide, is typically performed using a parallel, monochromatic beam [72, 83, 127, 141]. While the first commercial micro-CT scanner consisted of a single-row 512-pixel detector [126], modern scanners employ areal CCD detectors up to 11 megapixels and are capable of acquiring projection data for more than 1000 slices simultaneously [129, 135, 151]. Dedicated specimen ex vivo scanners are typically designed with the specimen oriented on a high-resolution motorized stage for translation and rotation. In this scenario, the source and detector remain in a fixed position during the scan, while the specimen rotates in the field of view. Conversely, preclinical micro-CT systems designed for in vivo imaging of small animals utilize a fixed gantry with the xray source and detector rotating and translating about the field of view [54, 147].\nAnalogous to clinical QCT, the grayscale attenuation values of reconstructed micro-CT images can be converted to hydroxyapatite concentration. Various calibration procedures based on idealized phantoms or theoretical calculations were established to derive relations between attenuation and BMD [21, 72, 102, 107, 110, 113]. However, micro-CT imaging is subject to xray scattering (SR and polychromatic micro-CT) and beam-hardening effects (polychromatic micro-CT) that can introduce nonnegligible error in the depiction of mineralization that is spatially variant and dependent on the geometry and composition of the object imaged [48, 79, 102]. Methods to minimize beam-hardening effects in conventional micro-CT include xray filtration to block “soft” low-energy xrays [102] and empirical corrections based on phantom measurements [21]. Using these procedures, apparent BMD and tissue level mineral density can be accurately measured (r2 = 0.78–0.99) in specimens of similar size and composition [21, 79].\nMorphometric indices analogous to classical histomorphometry can be calculated from micro-CT images of trabecular and cortical bone. Comparison of structural parameters of specimens scanned with these systems and mechanical testing suggest the amount of bone and the architecture of trabecular bone contribute to mechanical strength [57]. Advanced image-processing methodologies are used to quantify trabecular bone microarchitecture beyond measures of bone volume fraction (BV/TV). Specifically, direct three-dimensional measures of mean distances and measures of structural heterogeneity are used to characterize trabeculae and marrow spaces [64, 65] (Fig. 1). The degree of anisotropy, a measure of the degree of structural orientation of the trabecular network, can be calculated from the principal structural directions calculated by the mean intercept length techniques [63] and is highly related to the directional dependence of bone’s biomechanical properties [115]. A measure of the structural connectedness was also adapted to bone, based on the Euler number [114]. The shape of the trabecular structure is characterized using the structure model index (SMI), an index of surface convexity that estimates the degree to which the structure consists of rod-like or plate-like elements [66]. Furthermore, several groups developed algorithms to decompose the trabecular structure to independently quantify the volume and scale of rod-like and plate-like elements [92, 117, 139].Fig. 1A–BThree-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].\n\nThree-dimensional micro-CT images (16-μm isotropic voxel size) show (A) the spatial distribution of thickness and (B) the spatial distribution of the diameters of the intertrabecular marrow space in two specimens of human trabecular bone calculated by the direct three-dimensional distance transformation method of Hildebrand and Ruegsegger [65].\nThe ultrastructure of cortical bone is an important determinant of bone strength [128, 131], critical in fracture initiation and propagation [146], and known to change with age [37], disease [7], and therapy [11]. Volumetric and morphologic characterization of the cortical ultrastructure has predominantly focused on Haversian and Volkmann canal network of cadaveric femoral neck specimens [13, 32, 36, 38, 153]. Resolution improvements heralding the evolution of nano-CT (CT with submicron resolution) recently paved the way for a complete evaluation of cortical bone ultrastructure, including the distribution of osteocyte lacunae [131, 146].\nWhile volumetric density and microarchitecture information provide improved fracture risk prediction and some explanation for treatment efficacy, more direct estimates of bone mechanical strength that inherently account for geometry, microarchitecture, and even composition are the ultimate goal for improving fracture risk prediction and management of osteoporosis. Computational modeling approaches were introduced to take advantage of the detailed information in high-resolution images of bone. Finite element analysis (FEA) is a common computational tool in engineering fields, critical to design and failure analysis. Applied to high-resolution images of bone, the apparent biomechanical properties (stiffness, elastic modulus) of a biologically complex microstructure are computed by decomposing the structure into small cubic elements (the voxels) with assumed mechanical properties [109, 143] (Fig. 2). Numerous studies have utilized micro-FEA techniques to investigate the micromechanics of bone strength, failure, and relation to bone microarchitecture [81, 142].Fig. 2A micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.\n\nA micro-CT image shows the distribution of stresses in a human vertebral trabecular bone specimen under a simulated 1% compressive strain by micro-FEA. Red areas correspond to the locations of highest stress and blue to the areas of low stress. FEA = finite element analysis.\nAs an ex vivo imaging modality, the application of micro-CT to clinical research is limited to examining small bone biopsy specimens. There are numerous data over the last 20 years characterizing age, gender, and anatomic differences in cadaveric specimens of bone [13, 45, 52, 64, 67, 68, 94]. Clinical applications generally preclude access to the most relevant sites, such as the spine and proximal femur. Accordingly, a minimally invasive bone biopsy is typically acquired from the iliac crest [75]. Trabecular microarchitecture at the iliac crest reflects vertebral fracture status [56, 74] and changes after the onset of menopause [1]. Quantification of bone structure from iliac crest biopsies is also an important end point in longitudinal drug efficacy studies of parathyroid hormone [33, 53, 76, 121], strontium ranelate [3], and various bisphosphonates [11, 12, 46, 112, 120, 122, 123]. Borah et al. [11] recently reported major ultrastructural changes to the cortical bone of the iliac crest after 5 years of treatment with risedronate. Despite the resolution advantages of in vitro imaging studies, the invasiveness of the procedure, inherent variability in specimen collection [28, 29], and the limited correlation to bone quality at clinically relevant sites for fragility fracture (proximal femur, lumbar spine, distal radius) [35] are drawbacks to the clinical application of micro-CT.", "MDCT is a clinical CT technique available in most diagnostic imaging departments, using scanners from a number of manufacturers (Table 1). Therefore, a dedicated scanner is not required. The spatial resolution of this technique is limited, with an in-plane resolution ranging from 150 to 300 μm and a minimum slice thickness of around 300 μm. These spatial resolutions are above trabecular dimensions, and imaging of individual trabeculae is subject to considerable partial-volume effects. However, given the larger size of intertrabecular spaces, trabecular bone parameters obtained with this technique are observed to correlate moderately well with those determined by contact radiography and micro-CT of bone specimens (r = 0.53–0.70) [71, 91], as well as micro-CT (r = 0.44–0.99) [43, 44, 69, 70, 85, 118]. The advantage of the MDCT technique is that central regions of the skeleton critically relevant to osteoporosis and fracture risk assessment, such as the spine [70, 73, 85] and proximal femur [5, 43], can be visualized. However, to achieve adequate spatial resolution and image quality, the required radiation exposure is substantial, offsetting the technique’s applicability in clinical, routine, and scientific studies (Table 1). High-resolution CT protocols are typically associated with an effective dose of approximately 3 mSv (1.5 years of natural background radiation), several orders of magnitude greater than standard DXA or HR-pQCT (4–13 μSv) and an order of magnitude higher than standard two-dimensional QCT (0.06–0.3 mSv) [41].\nWhile early MDCT systems used for bone structure imaging typically consisted of four to 16 rows of detector elements [73], modern systems have 64 to 320 rows [69]. Most recently, high-resolution areal CCD detectors have been combined with a standard clinical CT gantry to provide substantial improvements in scan time and resolution [124]. In each case, the tomographic acquisition is performed with the subject lying supine on the scan table within the gantry. The xray source and detector ensemble continuously rotate about the field of view while the gantry translates along the rotational axis, effectively producing a helical series of projections [78]. Simultaneous calibration of Hounsfield units to mineral density is typically accomplished by placement of a solid hydroxyapatite phantom below the patient [49].\nThe analysis of trabecular microarchitecture from MDCT and flat-panel volumetric CT image data primarily involves the application of traditional histomorphometry [5, 17, 43, 70, 118], where BV/TV, Tb.N, Tb.Th, and Tb.Sp are calculated two-dimensionally using plate model assumptions [116]. In contrast, Ito et al. [73] used direct three-dimensional measures of trabecular dimensions, connectivity, and SMI in a MDCT study of the lumbar spine, finding a strong correlation (r = 0.98) between BV/TV measured by micro-CT and MDCT. Other specialized measures of trabecular dimensions using the fuzzy distance transform (which does not require a threshold binarization process) have been proposed by Krebs et al. [85].\nTo date, in vivo human studies using MDCT to assess bone structure are limited due to radiation dose concerns. Ito et al. [73] demonstrated SMI and BV/TV measured from MDCT images of the lumbar spine provided superior fracture discrimination to areal BMD by DXA. Graeff et al. [58] showed teriparatide treatment effects are better monitored by architectural parameters of the spine obtained through MDCT than by BMD (Fig. 3). In a cross-sectional cohort study of adolescent girls with and without anorexia nervosa, Bredella et al. [17] observed diminished trabecular microarchitecture at the distal radius in subjects with anorexia nervosa compared to controls, despite no differences in lumbar areal BMD. In a companion study, Lawson et al. [89] observed the abnormal trabecular microarchitecture in these patients are predicted by IGF-1, leptin, and androgen levels, with positive correlations (r = 0.32–0.72) to BV/TV, Tb.Th, and Tb.N.Fig. 3A–DIn vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.\n\nIn vivo MDCT images of the vertebral body show three-dimensional reconstructions (A) pre- and (B) postteriparatide therapy for 6 months, (C) 12 months, and (D) 24 months. Images provided by and printed with permission of Claus C. Glüer, Medizinische Physik, Klinik für Diagnostische Radiologie, Universitätsklinikum Schleswig Holstein–Campus Kiel, Kiel, Germany. MDCT = multidetector CT.", "A dedicated extremity imaging system designed for trabecular-scale imaging is currently available from a single manufacturer (XtremeCT; Scanco Medical AG, Brüttisellen, Switzerland). This device has the advantage of a higher signal-to-noise ratio and spatial resolution (nominal isotropic voxel dimension of 82 μm) when compared to MDCT. Furthermore, the radiation dose is lower when compared to whole-body CT and does not involve critical, radiosensitive organs in skeletally mature adults. There are several disadvantages to this technology. It is limited to peripheral skeletal sites and provides no direct insight into bone quality in the lumbar spine or proximal femur, common sites for osteoporotic fragility fractures. Additionally, there are currently a limited number of devices installed globally, which are primarily located at major research institutions with few available in clinical radiology departments.\nIn HR-pQCT, a standard protocol recommended by the manufacturer is utilized for most studies [14, 82]. The patient’s forearm or ankle is immobilized in a carbon fiber cast fixed within the gantry of the scanner. A single scout projection image of the distal radius or tibia is acquired to define the tomographic scan region. This scout image is acquired at an AP orientation at the wrist and at an oblique (45°) anterolateral-posteromedial orientation at the ankle. This tomographic region spans 9.02 mm in length (110 slices) and is localized to a fixed offset proximal from either the radial or tibial midjoint line and extends proximally. The offset is 9.5 mm in the radius and 22.5 mm in the tibia (Fig. 4). This method does not account for differences in bone length and may be a confounding source of variability in cross-sectional studies [16]. In the radius, the default axial scan location partially includes the most common site for fracture and location, where the bone microstructure is most strongly correlated to experimental strength of the forearm under a simulated falling load [106].Fig. 4A–BScout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.\n\nScout acquisition is used to define the HR-pQCT scan region for (A) the distal radius and (B) the distal tibia. The solid green region corresponds to the imaging location and consists of 110 slices spanning 9.02 mm longitudinally. In the radius the scan region is fixed 9.5 mm proximal from the midjoint line, while in the tibia the scan region is 22.5 mm proximal from the tibial plafond. HR-pQCT = high-resolution peripheral quantitative CT.\nThere are several different protocol modifications for developmental studies in children and adolescents [26, 84] to account for patient size and age-related changes in bone length and to avoid radiation exposure to the epiphyseal growth plate and inclusion of provisional mineralized tissues from this region. In a cross-sectional study of age- and gender-related differences in the microarchitecture of the distal forearm of adolescents, Kirmani et al. [84] used a fixed offset (1 mm) with respect to the proximal extent of the distal epiphyseal growth plate of the radius. In contrast, Burrows et al. [26] selected a region offset of 8% of the total tibial length proximal to the tibial endplate. While there are a number of studies underway investigating other scan locations in adults, including more proximal sites dominated by cortical bone, the internal configuration of the XtremeCT gantry prohibits the positioning of true diaphyseal sites in the radius or tibia within the limited 15-cm longitudinal range of the source-detector ensemble.\nThe reconstructed images are analyzed using a standard protocol provided by the manufacturer. The operator chaperones a semiautomated contouring process to identify the periosteal boundary and segment the cortical and trabecular compartments [42]. The trabecular bone structure is extracted using an edge enhancement and threshold procedure [88]. While the importance of threshold selection for morphometric analysis is well described for other CT devices [61], few studies to date investigate threshold effects or propose new methods for HR-pQCT [23, 42]. The default compartmental and structural segmentation provides the basis for the subsequent densitometric, morphometric, and biomechanical analyses. For subjects with very thin or highly porous cortical bone, this segmentation procedure may fail to capture the cortical structure [42, 80]; therefore, more sophisticated autocontouring techniques that operate on the fine-structure segmentation have been proposed [18, 23]. For the XtremeCT, reproducibility of densitometric measures is very high (coefficient of variation < 1%), while biomechanical and morphometric measures typically have a coefficient of variation of 4% to 5% [14, 80, 82, 100, 105].\nThe linear attenuation values of the tomographic images are converted to hydroxyapatite mineral densities using a beam-hardening correction and phantom calibration procedure previously described for the ex vivo micro-CT system [21]. Based on this calibration, volumetric BMD is determined independently for cortical and trabecular bone compartments using the segmentation process described previously. HR-pQCT images are used to derive surrogate measures of areal BMD in the ultradistal radius [22]. This technique is associated with a high level of agreement (r2 > 0.8) with multiple clinical DXA devices.\nUnlike MDCT, which has a large slice thickness relative to the in-plane resolution, the high isotropic resolution of HR-pQCT (82 μm) permits direct three-dimensional assessment of intertrabecular distances. These measures were validated against micro-CT gold standards [23, 93, 97]. From the binary image of the extracted trabecular structure, three-dimensional distance transformation techniques are used to calculated trabecular number [65]. While the intertrabecular distances are large compared to the voxel dimension, the average trabecular thickness (100–150 μm) is only one to two voxels wide. Accordingly, direct measures of thickness and bone volume are complicated by considerable partial-volume effects. In the standard analysis protocol, BV/TV is derived from the trabecular volumetric BMD, assuming a fixed mineralization of 1200 mg hydroxyapatite per cm3 for compact bone (BV/TV = Tb.BMD/1200). Trabecular BMD is determined by calculating the mean mineral density of the full medullary compartment. From the direct measure of Tb.N and the densitometrically derived BV/TV, Tb.Th and Tb.Sp are derived using standard stereologic relations, assuming plate model geometry [87, 116].\nThere are several potential concerns with this approach. First, Sekhon et al. [134] documented substantial errors in the measurement of trabecular BMD related to biologically relevant variations in cortical thickness and the magnitude of trabecular BMD itself. This may be related to xray scatter effects and residual beam-hardening artifacts. These errors are primarily a concern for cross-sectional studies, when cortical thickness and trabecular BMD may span a broad range. It is less of a concern in longitudinal studies, where the percent of change as the primary end point, such as age-, disease-, and therapy-related changes in cortical thickness and trabecular BMD, is comparatively small. Second, the assumption of a fixed-matrix mineralization is inconsistent with the established action of many common antifracture therapeutics [10]. Changes in matrix mineral density is expected to cause an increase in BMD, irrespective of bone volume changes, and result in an overestimation of BV/TV and propagate error to the derivative measures of Tb.Th and Tb.Sp, confounding any actual therapy-related effect on trabecular bone volume and structure.\nSeveral studies investigate other measures of bone microarchitecture and topology from HR-pQCT images, including connectivity, SMI, and anisotropy; however, there is mixed evidence of their reliability at in vivo resolutions [23, 93, 97, 136]. Recently, more sophisticated approaches to cortical bone segmentation have been proposed [18] that allow direct three-dimensional assessment of cortical thickness and quantification of cortical ultrastructure (Fig. 5), including intracortical porosity and canal diameter [24, 111]Fig. 5A–D(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.\n\n(A, B) Cross-sectional HR-pQCT images through the distal radius show two individuals with identical areal BMD by DXA at the ultradistal radius but substantial differences in trabecular and cortical structure. (C, D) Three-dimensional renderings of the cortical and trabecular bone compartments and intracortical porosity (highlighted in red) were segmented using software described by Burghardt et al. [19]. HR-pQCT = high-resolution peripheral quantitative CT; BMD = bone mineral density; DXA = dual-energy xray absorptiometry.\nThe ability of HR-pQCT to resolve the trabecular microarchitecture and a level of the cortical ultrastructure lends itself to calculating direct estimates of bone strength by voxel-based micro-FEA. For HR-pQCT scans of the distal radius and tibia, this is typically used to estimate strength under uniaxial compression, which approximates to the common loading condition for Colles’ fracture of the radius [106] and normal gaited loading for the tibia. This technique was validated against both higher-resolution models (based on micro-CT images) and empirical measures of strength [93, 99]. The application of micro-FEA is primarily performed assuming homogeneous material properties [15, 24, 40, 93, 99, 105], although it is also possible to produce models with material properties scaled according to mineralization [95, 99]. In addition to whole-bone mechanics, micro-FEA can be used to determine the relative load distribution between cortical and trabecular compartments [98] and estimate mechanical implications of specific structural features, such as the resolvable cortical ultrastructure [24].\nFor clinical investigations into longitudinal changes in HR-pQCT-derived measures of bone quality, it is critical baseline and followup scans be matched, since bone structure and geometry vary substantially along the axial direction [16]. Operator positioning for followup scans are aided by visual reference to the positioning of the baseline scan. Furthermore, postprocess image registration is performed to ensure comparable regions of interest are used for the image analysis. The manufacturer provides software that matches slices based on the periosteal cross-sectional area and limits the analyzed region to the slices common in baseline and followup [87]. Alternatively, MacNeil and Boyd [100] demonstrated three-dimensional image registration techniques can provide improved short- and medium-term reproducibility when compared to the default slice-matching approach. This approach may be more appropriate in longitudinal studies (long-term studies, anabolic therapy trials) where periosteal apposition would confound registration based strictly on cross-sectional area. As discussed earlier, the challenge in obtaining meaningful results in longitudinal studies in children or adolescents experiencing rapid growth is not trivial and requires careful consideration of standardized procedures for scan positioning and analysis [26, 101].\nThere is a growing body of literature featuring HR-pQCT assessment of bone quality. The first cross-sectional studies by Boutroy et al. [14] and Khosla et al. [82] reported gender-specific, age-related differences in trabecular bone microarchitecture. Several centers have observed age-related differences in micro-FEA estimates of bone strength in normative cross-sectional cohorts [24, 40, 96]. Furthermore, Burghardt et al. [24] and Macdonald et al. [96] demonstrated the ability of HR-pQCT to detect dramatic age-related differences in cortical porosity in females using new techniques for the analysis of cortical ultrastructure [19]. A microstructural basis for ethnicity-related differences in bone strength between East Asian and white women was reported in two studies [149, 150]. Sornay-Rendu et al. [137] suggested cortical and trabecular morphology provided additional fracture discrimination independent of areal BMD in osteopenic women. In the same cohort, Boutroy et al. [15] showed micro-FEA mechanical measures provided additional discriminatory power between osteopenic women with and without distal radius fractures. While the initial focus was predominantly related to fracture discrimination in postmenopausal osteopenia and osteoporosis [15, 103, 104, 137, 138, 140, 144, 145], a number of studies have utilized HR-pQCT to investigate developmental changes in bone quality and fracture risk [27, 34, 84], as well as secondary causes of bone loss [4, 20, 31, 62, 90].\nMost recently, data from the first HR-pQCT single- and multicenter longitudinal trials were published. In a multicenter, head-to-head, randomized, placebo-controlled trial of denosumab (a RANKL inhibitor) and alendronate (a bisphosphonate), Seeman et al. [133] reported more pronounced antiresorptive efficacy with denosumab than alendronate. In particular, cortical thickness was preserved or improved at the radius and tibia with either treatment, while cortical bone loss progressed in the control group. Burghardt et al. [25] reported similar cortical bone changes and additionally showed a preservation of compressive bone strength by micro-FEA after 24 months of alendronate treatment. Longitudinal microarchitectural changes also occurred with strontium ranelate [125] and teriparatide [95].", "With the advent of new imaging equipment, image-processing methods, and mathematical modeling techniques, the importance of trabecular microarchitecture and cortical ultrastructure in the context of bone strength and fracture risk has received considerable attention. This review provided the essential background of high-resolution CT techniques now in use for clinical skeletal research and summarized the important clinical applications of this technology reported to date.\nIn focusing on the clinical applications of CT imaging, we did not directly cover a number of important applications of this technology in basic skeletal research. In particular, we did not review (1) numerous ex vivo cadaveric studies conducted to investigate structure-function relationships in cortical and trabecular bone; (2) preclinical imaging studies in animal models of disease, genetic intervention, and environmental effects on skeletal health; and (3) analysis of microarchitecture using texture analysis. Furthermore, it is important to recognize other imaging modalities, such as high-resolution projection radiography, MRI, and phase-contrast xray imaging, may also be used to assess bone microarchitecture. This review did not cover the substantial body of literature on these subjects.\nEx vivo micro-CT is the longest-standing technology used for nondestructive three-dimensional imaging of bone microstructure. Micro-CT scanners resolve the cortical and trabecular structure of human bone specimens and are used to evaluate aspects of bone quality in the context of skeletal disease and therapeutic interventions. However, their clinical utility is limited to small biopsy samples, typically taken from the iliac crest.\nTranslation of these techniques to in vivo imaging modalities, such as the MDCT and HR-pQCT, are subject to technical limitations related to image quality, radiation dose considerations, and subject motion. They provide images at resolutions approximately equal to Tb.Th (HR-pQCT) or intertrabecular distances (MDCT). Furthermore, while MDCT has the advantage of being able to image central skeletal sites, such as the spine and proximal femur, the images represent trabecular texture more than a true visualization of the individual trabecular structure. Similarly, HR-pQCT may not resolve the finest trabeculae or the complete scale of cortical ultrastructure features. Despite the challenges and limitations facing current in vivo CT imaging technologies, morphologic and biomechanical indices determined from these techniques correlate well with micro-CT [73, 97].\nFinally, we demonstrated trabecular microarchitecture and cortical ultrastructure measured in vivo show age-, gender-, and race-dependent differences and provide improved fracture discrimination. Early longitudinal HR-pQCT observations suggest it is possible to detect structural changes induced by treatment. Ongoing studies and new results from therapeutic trials will provide a clearer indication as to whether trabecular microarchitecture and cortical ultrastructure measured using these in vivo methods will play a role in further understanding the affect of aging, disease, and interventions on skeletal health.\nTo date, there are no prospective fracture trials or large therapeutic trials used to draw conclusions regarding the role of in vivo assessment of trabecular microarchitecture and cortical ultrastructure in predicting bone strength and fracture status. These studies are warranted, and it would greatly enhance the field to establish a new set of diagnostic biomarkers that complement or improve upon areal BMD measured by DXA. The primary obstacles to achieving this goal are the substantial cost of large-scale human studies, the limited dissemination of the technology to clinical centers, the minimal standardization of image acquisition and image-processing protocols, and a lack of crosscalibration and quality control procedures (including standardized subject motion detection) for reliable data pooling in multicenter trials. Efforts to overcome the calibration issues, standardization, and multicenter comparative studies are needed to lay the groundwork for future large-scale multicenter trials and prospective studies." ]

[ "introduction", null, null, null, null, null, "discussion" ]

[]

[ "Amyotrophic Lateral Sclerosis", "Animals", "DNA-Binding Proteins", "Disease Models, Animal", "Disease Progression", "Humans", "Lectins", "Lymphocytes", "Mice", "Mice, Inbred C57BL", "Mice, Transgenic", "Microglia", "Nerve Degeneration", "RNA, Messenger", "Spinal Cord", "Superoxide Dismutase", "Survival Rate", "beta-N-Acetylhexosaminidases" ]

[ "Background", "Animals", "Symptomatic analysis and hanging-wire test", "Histochemistry and quantification of microglia and astrocytes", "Western blots", "Quantitative RT-PCR", "Statistical Analyses", "Delayed clinical onset in mutant SOD1 transgenic mice deficient in RAG2", "Increased microglial activation at the early stage in mSOD1 mice without mature lymphocytes", "Increased mRNA expression of Ym1, a neuroprotective factor, in mSOD1/RAG2-/- mice at the early stage", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Amyotrophic lateral sclerosis (ALS) is characterized by a progressive degeneration of motor neurons in brain and the spinal cord, resulting in muscle weakness. Patients eventually become paralyzed and approximately 50% die within 3 years of onset of symptoms, usually as the result of respiratory failure [1]. Although the precise mechanisms of ALS remain unclear, approximately 2% of patients with ALS have dominant mutations in the Cu/Zn superoxide dismutase 1 (SOD1) gene [2]. Transgenic mice overexpressing the mutant human SOD1 gene (mSOD1 mice) develop progressive motor neuron degeneration that resembles ALS and therefore these mice serve as an appropriate animal model for the disease [3].\nAlthough ALS is characterized by motor neuron degeneration, activation of microglia and astrocytes and infiltration of T lymphocytes are significant pathological hallmarks in the spinal cord lesions of ALS patients and mSOD1 mice, and a role for these cells in the pathogenesis of ALS has been suggested [4-6]. Recent experiments in mSOD1 mice suggest that neurons do not die alone, but rather that the process is non-cell-autonomous and depends on the active participation of non-neuronal cells, such as microglia, astrocytes, and T cells [7-9].\nMicroglia, resident immune effector cells in the central nervous system (CNS), display functional plasticity during activation, which involves changes in cell number, morphology, surface receptors, and production of growth factors and cytokines [10]. T-cell-derived cytokines play critical roles in the control of the microglial phenotype. For example, classically activated microglia (M1 microglia) differentiate in response to granulocyte macrophage colony-stimulating factor (GM-CSF) and are primed by interferon gamma (IFN-γ), one of the most important cytokines produced by T helper 1 (Th1) cells, in the presence of lipopolysaccharide (LPS) [10,11]. M1 microglia secrete increased proinflammatory cytokines, superoxide radicals, nitric oxide (·NO), and reduced neurotrophic factors, which promote neuronal death [12]. In contrast, representative T helper 2 (Th2) cytokines, such as interleukin 4 (IL-4) and interleukin 13 (IL-13), can convert microglia, primed by macrophage colony-stimulating factor (M-CSF), to an alternatively activated M2 phenotype [12]. M2 microglia are also characterized by increased expressions of arginase 1 (Arg1), resistin-like alpha (Retnla), and chitinase 3-like 3 (Ym1), which play important roles in tissue repair and remodeling [10]. However, the precise roles of crosstalk between T cells and microglia in the pathology of ALS remain unknown.\nIn this study, we established mSOD1 mice lacking recombination-activating gene 2 (mSOD1/RAG2-/-), an animal model for inherited ALS that lacks mature lymphocytes, and compared their phenotype and microglial characteristics with that of mutant human SOD1 transgenic mice (mSOD1/RAG2+/+). The clinical onset of mSOD1/RAG2-/- mice was significantly delayed compared to the control group. Consistent with this, increased numbers of activated microglia/macrophages and the expression of Ym1, a molecule with matrix reorganization and wound-healing effects [13,14], were observed at the early stage of the disease in mSOD1/RAG2-/- mice compared to mSOD1/RAG2+/+ mice. These results suggest an important role for lymphocytes interacting with microglia in the early phase of neurodegeneration in ALS.", "All mice were housed in microisolator cages within a modified pathogen-free barrier facility at the Animal Resource Center for Infectious Diseases, Research Institute for Microbial Diseases, Osaka University. All animals had free access to food and water ad libitum, and all of the experimental procedures followed our institutional guidelines.\nTransgenic mice overexpressing the familial ALS-associated G93A SOD1 mutation (harboring a single amino acid substitution of glycine to alanine at codon 93) (mSOD1 mice) were obtained from The Jackson Laboratory (Bar Harbor, ME) (strain designated: B6SJL-TgN(SOD1-G93A)1Gurd/J) and were backcrossed with C57BL/6 mice for at least 10 generations. Transgenic progeny were identified by a polymerase chain reaction (PCR) of genomic DNA using specific primers for exon 4 of the human SOD1 gene: ex4Pla 5'-CATCAGCCCTAATCCATCTGA-3' and ex4P2a 5'-TGGATCTTAGAATTCGCGAC-3'. PCR conditions included 35 cycles at 94°C for 30 s, 60°C for 30 s and 72°C for 1 min. The PCR products were resolved on a 1.0% agarose/ethidium bromide gel and photographed under UV illumination.\nRAG2-/- mice on a C57BL/6 background were obtained from The Jackson Laboratory. RAG2-/- mice were initially bred with mSOD1 mice. The presence or absence of the RAG2 gene was determined by PCR using 250 ng of tail DNA. The following primers were: Rag A, 5'-TAAAAGACCTATTCACAATCAAAAATGTCC-3'; and Rag B, 5'-TCAATCGTGTTGTCCCCTAGAGAGACAAGG-3'. PCR products for the different mice include: a 1100-bp band for homozygote mice, 1100- and 917-bp bands for heterozygote mice, and a 917-bp band for wild-type. PCR was conducted with 35 cycles at 94°C for 30 s, 60°C for 30 s and 72°C for 1 min. The PCR products were resolved by 1.0% agarose/ethidium bromide gel electrophoresis and photographed under UV illumination.", "Symptomatic analysis and weight measurements of male mice were conducted every 3-4 days starting at day 64 (n = 13 for mSOD1/RAG2+/+ mice, n = 11 for mSOD1/RAG2-/- mice). Animals were scored for motor symptoms using a 4-point scoring system as follows: 4 points, normal (no sign of motor dysfunction); 3 points, hind limb tremors were evident when suspended by the tail; 2 points, gait abnormalities were present; 1 point, dragging of at least one hind limb; and 0 points, inability to right self within 30 s. Onset was defined retrospectively as the earliest time when the mice showed symptoms (i.e., score < 4).\nFor the hanging-wire test, each male mouse was placed on a wire lid of a conventional housing cage. The lid was shaken gently to prompt the mouse to hold onto the grid before the lid was swiftly turned upside down. The latency until the mouse let go with at least both hind limbs was timed. Each mouse was allowed up to three attempts to hold on to the inverted lid for an arbitrary maximum of 90 s and the longest latency was recorded.", "Mice were anesthetized with an overdose of pentobarbital (60 mg/kg i.p.) and transcardially perfused with ice-cold 4% paraformaldehyde. Spinal cords were dissected out, postfixed in the same fixative for 4 h, placed overnight in 30% sucrose, and embedded in O.C.T. Compound (Sakura Finetek, Tokyo, Japan). Sections were frozen in liquid nitrogen and the blocks were stored at -80°C until use. Ten-micrometer-thick transverse sections of spinal cords were prepared using a Leica 3050 S cryostat (Thermo Shandon, Inc., Pittsburgh, PA) and collected onto Superfrost slides (Matsunami Glass Industries, Ltd., Osaka, Japan).\nFor lectin staining, sections were permeabilized with 0.2% Tween-20 in 0.05 M Tris-buffered saline (TBST, pH 7.2) for 10 min. They were then incubated with FITC-conjugated lectin (Sigma, St. Louis, MO) diluted 1:750 in phosphate-buffered saline (PBS) overnight at 4°C. Sections were washed 3 times in 0.2% TBST for 5 min, and mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA). The fluorescence-labeled sections were examined using a LSM 510 confocal microscope (Zeiss, Thornwood, NY).\nFor immunohistochemistry, sections were permeabilized with 0.2% TBST for 10 min and pretreated with 10% normal goat serum (Jackson ImmunoResearch Laboratories, West Grove, PA) for 1 h at RT to block non-specific IgG binding. Sections were then incubated with rabbit anti-mouse glial fibrillary acidic protein (GFAP) antibody (Ab) (1:1000) (Dako, Carpinteria, CA) overnight at 4°C. As a negative control, the primary Ab was omitted during the reaction. After rinsing with 0.2% TBST 3 times, sections were incubated with Cy5-conjugated F(ab') 2 fragment donkey anti-rabbit IgG (1:500 in PBS) for 3 h at RT. Sections were then washed 3 times in 0.2% TBST for 5 min and mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories). The fluorescence-labeled sections were examined using an LSM 510 confocal microscope (Zeiss).\nFor quantification of microglia and astrocytes, we analyzed the density of lectin- or GFAP-positive cells, respectively, in every third sample section of lumbar horizontal spinal cord sections. A total of 25 sections were analyzed and cell densities were counted in 4 randomly selected fields per section (a total of 100 fields).", "Mice were anesthetized with an overdose of pentobarbital (60 mg/kg i.p.) and transcardially perfused with ice-cold PBS. Dissected spinal cords were homogenized with protease inhibitors and loaded onto 10% SDS-PAGE gels. Blots were incubated with an antibody against Iba-1 (Wako, Richmond, VA) and visualized by chemiluminescence.", "RNA was isolated from homogenized flash-frozen mice spinal cords using TRIzol (GIBCO/BRL, Grand Island, NY) and purified using RNeasy (QIAGEN, Valencia, CA) according to the manufacturer's recommendations. The concentrations were determined spectrophotometrically. cDNA was generated from 100 ng of each RNA sample digested with RNase-free DNase I (QIAGEN) and SuperScript VILO cDNA Synthesis Kit (Invitrogen, Carlsbad, CA) according to the manufacturer's recommendations. The synthesized cDNA was amplified with SYBR Premix Ex Taq II (Takara Bio Inc., Shiga, Japan) and the primer set, which is listed below, using the StepOnePlus real-time PCR system (Applied BioSystems, Foster City, CA). Primer sets used in this study were: glyceraldehyde-3-phosphate dehydrogenase (GAPDH): 5'-TGTGTCCGTCGTGGATCTGA-3' and 5'-TTGCTGTTGAAGTCGCAGGAG-3'; nitric oxide synthase 2 (NOS2): 5'-GCAGAGATTGGAGGCCTTGTG-3' and 5'-GGGTTGTTGCTGAACTTCCAGTC-3'; IFN-γ: 5'-CGGCACAGTCATTGAAAGCCTA-3' and 5'-GTTGCTGATGGCCTGATTGTC-3'; Ym1: 5'-TTTGATGGCCTCAACCTGGA-3' and 5'-AGTGAGTAGCAGCCTTGGAATGTC-3'; Retnla: 5'-TCAGCAATCCCATGGCGTATAA-3' and 5'-TCATCAGTATTCACTGGGACCATCA-3'; glial cell line derived neurotrophic factor (GDNF): 5'-CCCACGTTTCGCATGGTTC-3' and 5'-TGGGCAGCTGAGGTTGTCAC-3'; solute carrier family 1 (glial high affinity glutamate transporter), member 3 (SLC1A3): 5'-CTTGTCTCATCCATTGGCCTCA-3' and 5'-CCATTCCCAATAAGTGTCCTGTCTG-3'; Mannose receptor, C type 1 (MRC1): 5'-TCGAGACTGCTGCTGAGTCCA-3' and 5'-AGACAGGATTGTCGTTCAACCAAAG-3'; M-CSF: 5'-GAACAGCCTGTCCCATCCATC-3' and 5'-TGAGGCCAGCTCAGTGCAA-3'; GM-CSF: 5'-AAGGGCGCCTTGAACATGAC-3' and 5'-AAATCCGCATAGGTGGTAACTTGTG-3'; interferon gamma receptor 1 (IFN-γR): 5'-TGACGGGAGCACCTGTTACAC-3' and 5'-TTTCGACCGTATGTTTCGTATGTAG-3'; IL-4: 5'-TCTCGAATGTACCAGGAGCCATATC-3' and 5'-AGCACCTTGGAAGCCCTACAGA-3'; interleukin 4 receptor, alpha (IL-4R): 5'-TGAGCCCGTGGTATCTGCAA-3' and 5'-CCAGATCCCTGTGCCTCAAAC-3'; IL-13: 5'-CAATTGCAATGCCATCTACAGGAC-3' and 5'-CGAAACAGTTGCTTTGTGTAGCTGA-3'; interleukin 13 receptor alpha 1 (IL-13R): 5'-CAGTCTTGCAGCATGGGAACA-3' and 5'-TGAGTCCCTAAGGCCTGGAGATTAC-3'. A standard thermal cycling program was used for all PCRs: 95°C for 60 s, 40 cycles of 95°C for 5 s, and 60°C for 30 s. The relative expression levels of each mRNA were calculated using the ΔΔCt method, normalizing to GAPDH and relative to the control samples.", "Differences in onset and survival (Figure 1B and 1E) were computed using Kaplan-Meier survival statistics (generalized Wilcoxon test). Data in Figure 1A, C and 1D were analyzed using a two-way repeated ANOVA. All other data were analyzed using an unpaired two-tailed Student's t test using Excel software (Microsoft, Dedmond, WA). Data are expressed as mean ± SEM. p < 0.05 was considered statistically significant. Results\nDelayed clinical onset in male mutant SOD1 transgenic mice deficient in the RAG2 gene. (A) Disease progression was significantly delayed in mSOD1/RAG2-/- mice (filled circles; n = 11) compared with mSOD1/RAG2+/+ mice (open circles; n = 13). (B) Kaplan Meier curve for symptomatic onset in mSOD1/RAG2+/+ mice (n = 13) and mSOD1/RAG2-/- mice (n = 11). Mice were evaluated for signs of motor deficiency with the following 4-point scoring system: 4 points, normal (no sign of motor dysfunction); 3 points, hind limb tremors were evident when suspended by the tail; 2 points, gait abnormalities were present; 1 point, dragging of at least one hind limb; 0 points, inability to right self within 30 s. We defined the point when the clinical score became 3 as being clinical onset. The mSOD1/RAG2-/- mice exhibited the disease phenotype at a later stage than the mSOD1/RAG2+/+ mice. (C) Time course of motor performance in the hanging-wire test. mSOD1/RAG2-/- mice (n = 11) performed significantly better than mSOD1/RAG2+/+ mice (n = 13). (D) Changes in mean body weight of mSOD1/RAG2-/- mice (n = 11) and mSOD1/RAG2+/+ mice (n = 13). (E) Kaplan-Meier survival curve. Absence of T and B cells did not affect the survival of mSOD1 mice (mSOD1/RAG2+/+ mice; n = 13, mSOD1/RAG2-/- mice; n = 11) (p = 0.13). Data are expressed as the mean ± SEM. Statistical analysis was done using the generalized Wilcoxon test for Kaplan-Meier curves (B,E) and a two-way repeated multivariate ANOVA (A, C, D) during 12 to 16 weeks of age. *:p < 0.05, #:p < 0.005.", "To investigate the role of lymphocytes in ALS, mSOD1 mice were crossed with RAG2-/- mice to generate mSOD1/RAG2-/- mice that lack functional T- and B-lymphocytes. Although there were no differences in relative copy number of mSOD1 (data not shown), delayed clinical onset was observed in mSOD1/RAG2-/- mice when compared with mSOD1/RAG2+/+ mice (p = 0.0033) (Figure 1A and 1B) or mSOD1/RAG2+/- mice (mSOD1/RAG2+/- mice: 94 ± 4 days (n = 8); mSOD1/RAG2-/- mice: 118 ± 3 days (n = 11), p = 0.0007, generalized Wilcoxon test). Next, neuromuscular dysfunction in mSOD1 mice was evaluated with the hanging-wire test twice a week. Consistent with the clinical onset and scores, genetic depletion of lymphocytes significantly improved motor function of mSOD1/RAG2-/- mice during the early stage (p = 0.0043) (Figure 1C). As neurodegeneration in mSOD1 mice is accompanied by loss of body weight, we also measured the body weight of these mice. We found that mSOD1/RAG2-/- mice had increased weight gain compared to mSOD1/RAG2+/+ mice, suggesting alleviated neurodegeneration in mSOD1/RAG2-/- mice (p = 0.0457) (Figure 1D). mSOD1/RAG2+/+ mice had normal numbers of white blood cells and normal ratios of T and B lymphocytes (white blood cell count: mSOD1tg 8067 ± 1324/ul, control 10867 ± 1629/ul, percentage of CD3+ cells: mSOD1tg 32.00 ± 1.21%, control 31.18 ± 1.23%, percentage of B220+ cells: mSOD1tg 55.97 ± 2.19%, control 57.85 ± 1.52%, n = 4 for each group). Although the absence of mature lymphocytes delayed onset of the disease, it did not affect the survival of mSOD1 mice (Figure 1E) (p = 0.13). Furthermore, survival times for mSOD1/RAG2+/- mice were comparable to those of mSOD1/RAG2-/- mice (average survival time; mSOD1/RAG2+/- mice: 142 ± 3 days (n = 8); mSOD1/RAG2-/- mice: 142 ± 3 days (n = 11), p = 0.74, generalized Wilcoxon test). Collectively, these results suggest that lymphocytes may contribute to acceleration of neurodegeneration in mSOD1 mice at the early stage.", "Neurodegeneration in ALS is accompanied by microglial-mediated neuroinflammation, which greatly affects disease progression [12]. Persistent activation of microglia is achieved by the sustained production of various cytokines induced by T lymphocytes [10]. To assess whether the absence of lymphocytes affects activation of microglia/macrophages in mSOD1 mice, spinal cord sections were stained with lectin. Lumbar spinal cord sections from early stage (70-90 days of age) mSOD1/RAG2-/- mice showed significantly increased numbers of lectin-positive microglia/macrophages compared with mSOD1/RAG2+/+ mice. However, there was no significant difference in the number of these cells between mSOD1/RAG2-/- mice and mSOD1/RAG2+/+ mice at the end stage of disease (130-150 days of age) (Figure 2A and 2B). Consistent with these results, western blot analysis showed an intense band for Iba-1 at the early stage of disease in spinal cord of mSOD1/RAG2-/- mice (Figure 2C). However, this striking difference disappeared at the end stage.\nSignificantly increased lectin-positive cells at the early stage of disease. (A) Lumbar sections of spinal cord from mSOD1/RAG2-/- and mSOD1/RAG2+/+ mice at early (70-90 days of age) and end (130-150 days of age) stage of disease were stained for lectin (microglia/macrophage). Scale bar = 50 μm. (B) Significantly increased numbers of lectin-positive microglia/macrophages are observed at the early stage of the disease in spinal cord sections of mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice. There was no significant difference in the number of lectin-positive cells at the end stage of disease. Significance was evaluated by using a generalized Wilcoxon test. **: p < 0.01. (C) Western blot analysis of Iba-1 shows markedly increased expression in lumbar spinal cord of mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice at early stage. Data presented in A-C is representative of three animals.\nAstrocytes are the most abundant non-neuronal cells in the CNS and their important roles in neurodegenerative disease have been suggested [8,9]. To assess the relevance of astrocytes, GFAP-immunoreactive cells were evaluated in lumbar spinal cords. At both early and end stages, there were no differences in intensity of GFAP immunoreactivity (IR) between mSOD1/RAG2-/- mice and control mice (Figure 3A). Also, both groups had comparable numbers of GFAP positive cells (Figure 3B). Collectively, these results suggest that mature lymphocytes may play a role in the inhibition of microglial activation but not of astrocyte activation.\nThe number of GFAP-positive cells in spinal cord does not differ between mSOD1/RAG2+/+ and mSOD1/RAG2-/- mice. (A) Lumbar sections of spinal cord from mSOD1/RAG2-/- mice and control mice stained with anti-GFAP antibody. Scale bar = 50 μm. (B) The number of GFAP-positive astrocytes in lumbar cord of mSOD1/RAG2-/- mice does not significantly differ from that of control mSOD1 mice at early (70-90 days of age) and end (130-150 days of age) stages. Data presented in A-B is representative of three animals.", "Activation of microglia in the presence of GM-CSF and IFN-γ results in a classically activated phenotype (M1 microglia) characterized by the production of ·NO [11]. The primary function of M1 microglia is thought to be activation of inflammation and tissue injury, whereas microglia activated alternatively (M2 microglia) by M-CSF, IL-4 and IL-13 express mannose receptor, MHC class II and Ym1, which leads to resolution of inflammation and promotion of wound healing [10,11]. To determine if delayed onset in mSOD1/RAG2-/- mice is attributable to increased numbers of M2 microglia, we evaluated the mRNA expression of M2 microglial target genes in spinal cord of mSOD1/RAG2-/- mice and control mice. Quantitative RT-PCR demonstrated that mRNA expression of Ym1, which is induced by activated M2 microglia/macrophages, was significantly elevated in mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice at early stage but not at end stage disease. The expression of IL-4R and IL-13 mRNA tended to be elevated in the mSOD1/RAG2-/- mice at the early stage, although this change was not significant (Figure 4A). In contrast, levels of M1 microglia-related molecules such as GM-CSF, IFN-γ, IFN-γ R and NOS2 were similar between mSOD1/RAG2-/- mice and control mice both at early and end stages (Figure 4B). It is worthy to note that considerable expression of GM-CSF, which can induce M1 microglia/macrophages, was observed at the end stage of disease in both groups. There were no significant differences in the expression of mRNA for neurotrophic factors such as GDNF and SLC1A3 between the two groups (Figure 4C). These results suggest that M2 microglia are more dominant than M1 microglia in the spinal cord of mSOD1/RAG2-/- mice at the early stage.\nExpression of Ym1 mRNA is significantly increased in the spinal cord of mSOD1/RAG2-/- mice. Quantitative RT-PCR analyses of spinal cord of mSOD1/RAG2-/- mice and mSOD1/RAG2+/+ mice at early (70-90 days of age) and end (130-150 days of age) stages were performed (n = 3-4 in each group). (A) Relative mRNA expression of M-CSF, IL-4, IL-4R, IL-13, IL-13R, Retnla, and Ym1, which are related to M2 macrophages, is shown. Ym1 expression levels are significantly increased in mSOD1-RAG2-/- mice at the early stage of disease (p = 0.039). (B) Relative mRNA expression of GM-CSF, IFN-γ, IFN-γ R, and NOS2, which are related to M1 macrophages, is shown. M1 macrophage-related molecules were comparable between mSOD1/RAG2-/- mice and control mice. (C) Relative mRNA expression of GDNF and SLC1A3 is shown. Expression levels of GDNF and SLC1A3 did not differ between mSOD1/RAG2-/- mice and control mice at early and end stages. Significance was evaluated using the student's t test. *: p < 0.05.", "In this study, we demonstrate delayed clinical onset in mSOD1/RAG2-/- mice compared to mSOD1/RAG2+/+ mice, with increased number of activated microglia in the spinal cord. Microgliosis in mSOD1/RAG2-/- mice was accompanied by increased Ym1 mRNA expression, one of the hallmarks of M2 microglia/macrophages. However, the differences in weakness, microgliosis, and Ym1 expression between SOD1/RAG2-/- and mSOD1/RAG2+/+ mice disappeared in the later stage of disease. This suggests that this affect on M2 microglia/macrophages, as the result of lymphocyte depletion, may be associated with delayed clinical onset in mSOD1/RAG2-/- mice. As a previous study has shown that B lymphocytes do not have a major role in neurodegeneration in G93A-SOD1 mice [15], T lymphocytes rather than B lymphocytes are thought to be responsible for the increased M2 activation of microglia. Although few T lymphocytes have been shown to infiltrate spinal cord in mSOD1 mice in the early stage, T cells in the meninges may have some effects on neurons or glial cells without entering the parenchyma [16-18]. Increased GM-CSF at the late stage of disease, as shown in Figure 4B, might counteract the neuroprotective effect of M2 microglia observed at the early phase.\nIn contrast to our finding of a deleterious effect of lymphocytes in ALS, previous studies with mSOD1/RAG2-/- mice have indicated a protective rather than deleterious role of lymphocytes in mSOD1 mice [6,16,19]. A possible explanation for these discrepancies is differences in transgene copy number in the mice used by us and other research groups. The mutant G93A transgene is reported to undergo a background level of copy loss due to meiotic rearrangement of the transgene array [20,21]. Since mSOD1 has been suggested to activate innate immunity downstream of the toll-like receptor (TLR) 2, TLR4 and CD14 [22], a difference in transgene copy number may cause different levels of microglial activation. Another possibility is environmental factors. Mice maintained under germ-free conditions develop significantly attenuated symptoms of experimental autoimmune encephalomyelitis (EAE) compared with conventionally colonized mice, and microbiota are reported to have an effect on the T-helper/Treg axis outside of the gut [23]. Thus, different microbiota of mice maintained in different facilities could change the characteristics of lymphocytes and other immune cells in the CNS, thus leading to the discrepancy between our study and previous reports.\nPiecing together our results and previous studies suggests that lymphocytes might have biphasic effects on neurons in this animal model of ALS. That is, lymphocytes may have a deleterious effect in the early phase of neurodegeneration by suppressing M2 microglia, whereas they may exert neuroprotective effects in the later stage of the disease, thus leading to the shortened survival of mSOD1/RAG2-/- mice reported in a previous study [6]. Transgene copy number of our mice and the SPF condition of our facilities might be responsible for the apparent difference seen in the early stage of neurodegeneration between the two groups, while those of other facilities appear to affect the end stage.\nThe present findings suggest that mature lymphocytes act to produce a deterioration of the ALS phenotype in mSOD1 mice at the early stage of the disease by modulating the trophic balance of microglia. If the course of ALS is so modulated, then disease progression could be delayed by expanding populations of M2 microglia/macrophages, thus producing neuroprotective effects.", "This study reports delayed onset of ALS phenotype in mSOD1/RAG2-/- mice compared to controls. We observed an increased number of lectin-positive microglia/macrophages at the early stage of disease in mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice. In addition, quantitative RT-PCR revealed increased expression of mRNA for M2-microglial-related genes in mSOD1/RAG2-/- mice. In conclusion, these results indicate that peripheral lymphocytes act to produce a deterioration of the ALS phenotype in mSOD1 transgenic mice at the early stage of disease by inhibiting M2 microglia/macrophage skewing.", "The authors declare that they have no competing interests.", "ST designed the experiments, analyzed the data, and wrote the manuscript.\nTO, TY and YN participated in the design of the study and helped draft the manuscript. TS contributed analysis of data as well as provided editing of the manuscript. HK and SS conceived of the study, and participated in its design and conduct and helped draft the manuscript. All authors have read and approved the final version of the manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Animals", "Symptomatic analysis and hanging-wire test", "Histochemistry and quantification of microglia and astrocytes", "Western blots", "Quantitative RT-PCR", "Statistical Analyses", "Delayed clinical onset in mutant SOD1 transgenic mice deficient in RAG2", "Increased microglial activation at the early stage in mSOD1 mice without mature lymphocytes", "Increased mRNA expression of Ym1, a neuroprotective factor, in mSOD1/RAG2-/- mice at the early stage", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Amyotrophic lateral sclerosis (ALS) is characterized by a progressive degeneration of motor neurons in brain and the spinal cord, resulting in muscle weakness. Patients eventually become paralyzed and approximately 50% die within 3 years of onset of symptoms, usually as the result of respiratory failure [1]. Although the precise mechanisms of ALS remain unclear, approximately 2% of patients with ALS have dominant mutations in the Cu/Zn superoxide dismutase 1 (SOD1) gene [2]. Transgenic mice overexpressing the mutant human SOD1 gene (mSOD1 mice) develop progressive motor neuron degeneration that resembles ALS and therefore these mice serve as an appropriate animal model for the disease [3].\nAlthough ALS is characterized by motor neuron degeneration, activation of microglia and astrocytes and infiltration of T lymphocytes are significant pathological hallmarks in the spinal cord lesions of ALS patients and mSOD1 mice, and a role for these cells in the pathogenesis of ALS has been suggested [4-6]. Recent experiments in mSOD1 mice suggest that neurons do not die alone, but rather that the process is non-cell-autonomous and depends on the active participation of non-neuronal cells, such as microglia, astrocytes, and T cells [7-9].\nMicroglia, resident immune effector cells in the central nervous system (CNS), display functional plasticity during activation, which involves changes in cell number, morphology, surface receptors, and production of growth factors and cytokines [10]. T-cell-derived cytokines play critical roles in the control of the microglial phenotype. For example, classically activated microglia (M1 microglia) differentiate in response to granulocyte macrophage colony-stimulating factor (GM-CSF) and are primed by interferon gamma (IFN-γ), one of the most important cytokines produced by T helper 1 (Th1) cells, in the presence of lipopolysaccharide (LPS) [10,11]. M1 microglia secrete increased proinflammatory cytokines, superoxide radicals, nitric oxide (·NO), and reduced neurotrophic factors, which promote neuronal death [12]. In contrast, representative T helper 2 (Th2) cytokines, such as interleukin 4 (IL-4) and interleukin 13 (IL-13), can convert microglia, primed by macrophage colony-stimulating factor (M-CSF), to an alternatively activated M2 phenotype [12]. M2 microglia are also characterized by increased expressions of arginase 1 (Arg1), resistin-like alpha (Retnla), and chitinase 3-like 3 (Ym1), which play important roles in tissue repair and remodeling [10]. However, the precise roles of crosstalk between T cells and microglia in the pathology of ALS remain unknown.\nIn this study, we established mSOD1 mice lacking recombination-activating gene 2 (mSOD1/RAG2-/-), an animal model for inherited ALS that lacks mature lymphocytes, and compared their phenotype and microglial characteristics with that of mutant human SOD1 transgenic mice (mSOD1/RAG2+/+). The clinical onset of mSOD1/RAG2-/- mice was significantly delayed compared to the control group. Consistent with this, increased numbers of activated microglia/macrophages and the expression of Ym1, a molecule with matrix reorganization and wound-healing effects [13,14], were observed at the early stage of the disease in mSOD1/RAG2-/- mice compared to mSOD1/RAG2+/+ mice. These results suggest an important role for lymphocytes interacting with microglia in the early phase of neurodegeneration in ALS.", "[SUBTITLE] Animals [SUBSECTION] All mice were housed in microisolator cages within a modified pathogen-free barrier facility at the Animal Resource Center for Infectious Diseases, Research Institute for Microbial Diseases, Osaka University. All animals had free access to food and water ad libitum, and all of the experimental procedures followed our institutional guidelines.\nTransgenic mice overexpressing the familial ALS-associated G93A SOD1 mutation (harboring a single amino acid substitution of glycine to alanine at codon 93) (mSOD1 mice) were obtained from The Jackson Laboratory (Bar Harbor, ME) (strain designated: B6SJL-TgN(SOD1-G93A)1Gurd/J) and were backcrossed with C57BL/6 mice for at least 10 generations. Transgenic progeny were identified by a polymerase chain reaction (PCR) of genomic DNA using specific primers for exon 4 of the human SOD1 gene: ex4Pla 5'-CATCAGCCCTAATCCATCTGA-3' and ex4P2a 5'-TGGATCTTAGAATTCGCGAC-3'. PCR conditions included 35 cycles at 94°C for 30 s, 60°C for 30 s and 72°C for 1 min. The PCR products were resolved on a 1.0% agarose/ethidium bromide gel and photographed under UV illumination.\nRAG2-/- mice on a C57BL/6 background were obtained from The Jackson Laboratory. RAG2-/- mice were initially bred with mSOD1 mice. The presence or absence of the RAG2 gene was determined by PCR using 250 ng of tail DNA. The following primers were: Rag A, 5'-TAAAAGACCTATTCACAATCAAAAATGTCC-3'; and Rag B, 5'-TCAATCGTGTTGTCCCCTAGAGAGACAAGG-3'. PCR products for the different mice include: a 1100-bp band for homozygote mice, 1100- and 917-bp bands for heterozygote mice, and a 917-bp band for wild-type. PCR was conducted with 35 cycles at 94°C for 30 s, 60°C for 30 s and 72°C for 1 min. The PCR products were resolved by 1.0% agarose/ethidium bromide gel electrophoresis and photographed under UV illumination.\nAll mice were housed in microisolator cages within a modified pathogen-free barrier facility at the Animal Resource Center for Infectious Diseases, Research Institute for Microbial Diseases, Osaka University. All animals had free access to food and water ad libitum, and all of the experimental procedures followed our institutional guidelines.\nTransgenic mice overexpressing the familial ALS-associated G93A SOD1 mutation (harboring a single amino acid substitution of glycine to alanine at codon 93) (mSOD1 mice) were obtained from The Jackson Laboratory (Bar Harbor, ME) (strain designated: B6SJL-TgN(SOD1-G93A)1Gurd/J) and were backcrossed with C57BL/6 mice for at least 10 generations. Transgenic progeny were identified by a polymerase chain reaction (PCR) of genomic DNA using specific primers for exon 4 of the human SOD1 gene: ex4Pla 5'-CATCAGCCCTAATCCATCTGA-3' and ex4P2a 5'-TGGATCTTAGAATTCGCGAC-3'. PCR conditions included 35 cycles at 94°C for 30 s, 60°C for 30 s and 72°C for 1 min. The PCR products were resolved on a 1.0% agarose/ethidium bromide gel and photographed under UV illumination.\nRAG2-/- mice on a C57BL/6 background were obtained from The Jackson Laboratory. RAG2-/- mice were initially bred with mSOD1 mice. The presence or absence of the RAG2 gene was determined by PCR using 250 ng of tail DNA. The following primers were: Rag A, 5'-TAAAAGACCTATTCACAATCAAAAATGTCC-3'; and Rag B, 5'-TCAATCGTGTTGTCCCCTAGAGAGACAAGG-3'. PCR products for the different mice include: a 1100-bp band for homozygote mice, 1100- and 917-bp bands for heterozygote mice, and a 917-bp band for wild-type. PCR was conducted with 35 cycles at 94°C for 30 s, 60°C for 30 s and 72°C for 1 min. The PCR products were resolved by 1.0% agarose/ethidium bromide gel electrophoresis and photographed under UV illumination.\n[SUBTITLE] Symptomatic analysis and hanging-wire test [SUBSECTION] Symptomatic analysis and weight measurements of male mice were conducted every 3-4 days starting at day 64 (n = 13 for mSOD1/RAG2+/+ mice, n = 11 for mSOD1/RAG2-/- mice). Animals were scored for motor symptoms using a 4-point scoring system as follows: 4 points, normal (no sign of motor dysfunction); 3 points, hind limb tremors were evident when suspended by the tail; 2 points, gait abnormalities were present; 1 point, dragging of at least one hind limb; and 0 points, inability to right self within 30 s. Onset was defined retrospectively as the earliest time when the mice showed symptoms (i.e., score < 4).\nFor the hanging-wire test, each male mouse was placed on a wire lid of a conventional housing cage. The lid was shaken gently to prompt the mouse to hold onto the grid before the lid was swiftly turned upside down. The latency until the mouse let go with at least both hind limbs was timed. Each mouse was allowed up to three attempts to hold on to the inverted lid for an arbitrary maximum of 90 s and the longest latency was recorded.\nSymptomatic analysis and weight measurements of male mice were conducted every 3-4 days starting at day 64 (n = 13 for mSOD1/RAG2+/+ mice, n = 11 for mSOD1/RAG2-/- mice). Animals were scored for motor symptoms using a 4-point scoring system as follows: 4 points, normal (no sign of motor dysfunction); 3 points, hind limb tremors were evident when suspended by the tail; 2 points, gait abnormalities were present; 1 point, dragging of at least one hind limb; and 0 points, inability to right self within 30 s. Onset was defined retrospectively as the earliest time when the mice showed symptoms (i.e., score < 4).\nFor the hanging-wire test, each male mouse was placed on a wire lid of a conventional housing cage. The lid was shaken gently to prompt the mouse to hold onto the grid before the lid was swiftly turned upside down. The latency until the mouse let go with at least both hind limbs was timed. Each mouse was allowed up to three attempts to hold on to the inverted lid for an arbitrary maximum of 90 s and the longest latency was recorded.\n[SUBTITLE] Histochemistry and quantification of microglia and astrocytes [SUBSECTION] Mice were anesthetized with an overdose of pentobarbital (60 mg/kg i.p.) and transcardially perfused with ice-cold 4% paraformaldehyde. Spinal cords were dissected out, postfixed in the same fixative for 4 h, placed overnight in 30% sucrose, and embedded in O.C.T. Compound (Sakura Finetek, Tokyo, Japan). Sections were frozen in liquid nitrogen and the blocks were stored at -80°C until use. Ten-micrometer-thick transverse sections of spinal cords were prepared using a Leica 3050 S cryostat (Thermo Shandon, Inc., Pittsburgh, PA) and collected onto Superfrost slides (Matsunami Glass Industries, Ltd., Osaka, Japan).\nFor lectin staining, sections were permeabilized with 0.2% Tween-20 in 0.05 M Tris-buffered saline (TBST, pH 7.2) for 10 min. They were then incubated with FITC-conjugated lectin (Sigma, St. Louis, MO) diluted 1:750 in phosphate-buffered saline (PBS) overnight at 4°C. Sections were washed 3 times in 0.2% TBST for 5 min, and mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA). The fluorescence-labeled sections were examined using a LSM 510 confocal microscope (Zeiss, Thornwood, NY).\nFor immunohistochemistry, sections were permeabilized with 0.2% TBST for 10 min and pretreated with 10% normal goat serum (Jackson ImmunoResearch Laboratories, West Grove, PA) for 1 h at RT to block non-specific IgG binding. Sections were then incubated with rabbit anti-mouse glial fibrillary acidic protein (GFAP) antibody (Ab) (1:1000) (Dako, Carpinteria, CA) overnight at 4°C. As a negative control, the primary Ab was omitted during the reaction. After rinsing with 0.2% TBST 3 times, sections were incubated with Cy5-conjugated F(ab') 2 fragment donkey anti-rabbit IgG (1:500 in PBS) for 3 h at RT. Sections were then washed 3 times in 0.2% TBST for 5 min and mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories). The fluorescence-labeled sections were examined using an LSM 510 confocal microscope (Zeiss).\nFor quantification of microglia and astrocytes, we analyzed the density of lectin- or GFAP-positive cells, respectively, in every third sample section of lumbar horizontal spinal cord sections. A total of 25 sections were analyzed and cell densities were counted in 4 randomly selected fields per section (a total of 100 fields).\nMice were anesthetized with an overdose of pentobarbital (60 mg/kg i.p.) and transcardially perfused with ice-cold 4% paraformaldehyde. Spinal cords were dissected out, postfixed in the same fixative for 4 h, placed overnight in 30% sucrose, and embedded in O.C.T. Compound (Sakura Finetek, Tokyo, Japan). Sections were frozen in liquid nitrogen and the blocks were stored at -80°C until use. Ten-micrometer-thick transverse sections of spinal cords were prepared using a Leica 3050 S cryostat (Thermo Shandon, Inc., Pittsburgh, PA) and collected onto Superfrost slides (Matsunami Glass Industries, Ltd., Osaka, Japan).\nFor lectin staining, sections were permeabilized with 0.2% Tween-20 in 0.05 M Tris-buffered saline (TBST, pH 7.2) for 10 min. They were then incubated with FITC-conjugated lectin (Sigma, St. Louis, MO) diluted 1:750 in phosphate-buffered saline (PBS) overnight at 4°C. Sections were washed 3 times in 0.2% TBST for 5 min, and mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA). The fluorescence-labeled sections were examined using a LSM 510 confocal microscope (Zeiss, Thornwood, NY).\nFor immunohistochemistry, sections were permeabilized with 0.2% TBST for 10 min and pretreated with 10% normal goat serum (Jackson ImmunoResearch Laboratories, West Grove, PA) for 1 h at RT to block non-specific IgG binding. Sections were then incubated with rabbit anti-mouse glial fibrillary acidic protein (GFAP) antibody (Ab) (1:1000) (Dako, Carpinteria, CA) overnight at 4°C. As a negative control, the primary Ab was omitted during the reaction. After rinsing with 0.2% TBST 3 times, sections were incubated with Cy5-conjugated F(ab') 2 fragment donkey anti-rabbit IgG (1:500 in PBS) for 3 h at RT. Sections were then washed 3 times in 0.2% TBST for 5 min and mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories). The fluorescence-labeled sections were examined using an LSM 510 confocal microscope (Zeiss).\nFor quantification of microglia and astrocytes, we analyzed the density of lectin- or GFAP-positive cells, respectively, in every third sample section of lumbar horizontal spinal cord sections. A total of 25 sections were analyzed and cell densities were counted in 4 randomly selected fields per section (a total of 100 fields).\n[SUBTITLE] Western blots [SUBSECTION] Mice were anesthetized with an overdose of pentobarbital (60 mg/kg i.p.) and transcardially perfused with ice-cold PBS. Dissected spinal cords were homogenized with protease inhibitors and loaded onto 10% SDS-PAGE gels. Blots were incubated with an antibody against Iba-1 (Wako, Richmond, VA) and visualized by chemiluminescence.\nMice were anesthetized with an overdose of pentobarbital (60 mg/kg i.p.) and transcardially perfused with ice-cold PBS. Dissected spinal cords were homogenized with protease inhibitors and loaded onto 10% SDS-PAGE gels. Blots were incubated with an antibody against Iba-1 (Wako, Richmond, VA) and visualized by chemiluminescence.\n[SUBTITLE] Quantitative RT-PCR [SUBSECTION] RNA was isolated from homogenized flash-frozen mice spinal cords using TRIzol (GIBCO/BRL, Grand Island, NY) and purified using RNeasy (QIAGEN, Valencia, CA) according to the manufacturer's recommendations. The concentrations were determined spectrophotometrically. cDNA was generated from 100 ng of each RNA sample digested with RNase-free DNase I (QIAGEN) and SuperScript VILO cDNA Synthesis Kit (Invitrogen, Carlsbad, CA) according to the manufacturer's recommendations. The synthesized cDNA was amplified with SYBR Premix Ex Taq II (Takara Bio Inc., Shiga, Japan) and the primer set, which is listed below, using the StepOnePlus real-time PCR system (Applied BioSystems, Foster City, CA). Primer sets used in this study were: glyceraldehyde-3-phosphate dehydrogenase (GAPDH): 5'-TGTGTCCGTCGTGGATCTGA-3' and 5'-TTGCTGTTGAAGTCGCAGGAG-3'; nitric oxide synthase 2 (NOS2): 5'-GCAGAGATTGGAGGCCTTGTG-3' and 5'-GGGTTGTTGCTGAACTTCCAGTC-3'; IFN-γ: 5'-CGGCACAGTCATTGAAAGCCTA-3' and 5'-GTTGCTGATGGCCTGATTGTC-3'; Ym1: 5'-TTTGATGGCCTCAACCTGGA-3' and 5'-AGTGAGTAGCAGCCTTGGAATGTC-3'; Retnla: 5'-TCAGCAATCCCATGGCGTATAA-3' and 5'-TCATCAGTATTCACTGGGACCATCA-3'; glial cell line derived neurotrophic factor (GDNF): 5'-CCCACGTTTCGCATGGTTC-3' and 5'-TGGGCAGCTGAGGTTGTCAC-3'; solute carrier family 1 (glial high affinity glutamate transporter), member 3 (SLC1A3): 5'-CTTGTCTCATCCATTGGCCTCA-3' and 5'-CCATTCCCAATAAGTGTCCTGTCTG-3'; Mannose receptor, C type 1 (MRC1): 5'-TCGAGACTGCTGCTGAGTCCA-3' and 5'-AGACAGGATTGTCGTTCAACCAAAG-3'; M-CSF: 5'-GAACAGCCTGTCCCATCCATC-3' and 5'-TGAGGCCAGCTCAGTGCAA-3'; GM-CSF: 5'-AAGGGCGCCTTGAACATGAC-3' and 5'-AAATCCGCATAGGTGGTAACTTGTG-3'; interferon gamma receptor 1 (IFN-γR): 5'-TGACGGGAGCACCTGTTACAC-3' and 5'-TTTCGACCGTATGTTTCGTATGTAG-3'; IL-4: 5'-TCTCGAATGTACCAGGAGCCATATC-3' and 5'-AGCACCTTGGAAGCCCTACAGA-3'; interleukin 4 receptor, alpha (IL-4R): 5'-TGAGCCCGTGGTATCTGCAA-3' and 5'-CCAGATCCCTGTGCCTCAAAC-3'; IL-13: 5'-CAATTGCAATGCCATCTACAGGAC-3' and 5'-CGAAACAGTTGCTTTGTGTAGCTGA-3'; interleukin 13 receptor alpha 1 (IL-13R): 5'-CAGTCTTGCAGCATGGGAACA-3' and 5'-TGAGTCCCTAAGGCCTGGAGATTAC-3'. A standard thermal cycling program was used for all PCRs: 95°C for 60 s, 40 cycles of 95°C for 5 s, and 60°C for 30 s. The relative expression levels of each mRNA were calculated using the ΔΔCt method, normalizing to GAPDH and relative to the control samples.\nRNA was isolated from homogenized flash-frozen mice spinal cords using TRIzol (GIBCO/BRL, Grand Island, NY) and purified using RNeasy (QIAGEN, Valencia, CA) according to the manufacturer's recommendations. The concentrations were determined spectrophotometrically. cDNA was generated from 100 ng of each RNA sample digested with RNase-free DNase I (QIAGEN) and SuperScript VILO cDNA Synthesis Kit (Invitrogen, Carlsbad, CA) according to the manufacturer's recommendations. The synthesized cDNA was amplified with SYBR Premix Ex Taq II (Takara Bio Inc., Shiga, Japan) and the primer set, which is listed below, using the StepOnePlus real-time PCR system (Applied BioSystems, Foster City, CA). Primer sets used in this study were: glyceraldehyde-3-phosphate dehydrogenase (GAPDH): 5'-TGTGTCCGTCGTGGATCTGA-3' and 5'-TTGCTGTTGAAGTCGCAGGAG-3'; nitric oxide synthase 2 (NOS2): 5'-GCAGAGATTGGAGGCCTTGTG-3' and 5'-GGGTTGTTGCTGAACTTCCAGTC-3'; IFN-γ: 5'-CGGCACAGTCATTGAAAGCCTA-3' and 5'-GTTGCTGATGGCCTGATTGTC-3'; Ym1: 5'-TTTGATGGCCTCAACCTGGA-3' and 5'-AGTGAGTAGCAGCCTTGGAATGTC-3'; Retnla: 5'-TCAGCAATCCCATGGCGTATAA-3' and 5'-TCATCAGTATTCACTGGGACCATCA-3'; glial cell line derived neurotrophic factor (GDNF): 5'-CCCACGTTTCGCATGGTTC-3' and 5'-TGGGCAGCTGAGGTTGTCAC-3'; solute carrier family 1 (glial high affinity glutamate transporter), member 3 (SLC1A3): 5'-CTTGTCTCATCCATTGGCCTCA-3' and 5'-CCATTCCCAATAAGTGTCCTGTCTG-3'; Mannose receptor, C type 1 (MRC1): 5'-TCGAGACTGCTGCTGAGTCCA-3' and 5'-AGACAGGATTGTCGTTCAACCAAAG-3'; M-CSF: 5'-GAACAGCCTGTCCCATCCATC-3' and 5'-TGAGGCCAGCTCAGTGCAA-3'; GM-CSF: 5'-AAGGGCGCCTTGAACATGAC-3' and 5'-AAATCCGCATAGGTGGTAACTTGTG-3'; interferon gamma receptor 1 (IFN-γR): 5'-TGACGGGAGCACCTGTTACAC-3' and 5'-TTTCGACCGTATGTTTCGTATGTAG-3'; IL-4: 5'-TCTCGAATGTACCAGGAGCCATATC-3' and 5'-AGCACCTTGGAAGCCCTACAGA-3'; interleukin 4 receptor, alpha (IL-4R): 5'-TGAGCCCGTGGTATCTGCAA-3' and 5'-CCAGATCCCTGTGCCTCAAAC-3'; IL-13: 5'-CAATTGCAATGCCATCTACAGGAC-3' and 5'-CGAAACAGTTGCTTTGTGTAGCTGA-3'; interleukin 13 receptor alpha 1 (IL-13R): 5'-CAGTCTTGCAGCATGGGAACA-3' and 5'-TGAGTCCCTAAGGCCTGGAGATTAC-3'. A standard thermal cycling program was used for all PCRs: 95°C for 60 s, 40 cycles of 95°C for 5 s, and 60°C for 30 s. The relative expression levels of each mRNA were calculated using the ΔΔCt method, normalizing to GAPDH and relative to the control samples.\n[SUBTITLE] Statistical Analyses [SUBSECTION] Differences in onset and survival (Figure 1B and 1E) were computed using Kaplan-Meier survival statistics (generalized Wilcoxon test). Data in Figure 1A, C and 1D were analyzed using a two-way repeated ANOVA. All other data were analyzed using an unpaired two-tailed Student's t test using Excel software (Microsoft, Dedmond, WA). Data are expressed as mean ± SEM. p < 0.05 was considered statistically significant. Results\nDelayed clinical onset in male mutant SOD1 transgenic mice deficient in the RAG2 gene. (A) Disease progression was significantly delayed in mSOD1/RAG2-/- mice (filled circles; n = 11) compared with mSOD1/RAG2+/+ mice (open circles; n = 13). (B) Kaplan Meier curve for symptomatic onset in mSOD1/RAG2+/+ mice (n = 13) and mSOD1/RAG2-/- mice (n = 11). Mice were evaluated for signs of motor deficiency with the following 4-point scoring system: 4 points, normal (no sign of motor dysfunction); 3 points, hind limb tremors were evident when suspended by the tail; 2 points, gait abnormalities were present; 1 point, dragging of at least one hind limb; 0 points, inability to right self within 30 s. We defined the point when the clinical score became 3 as being clinical onset. The mSOD1/RAG2-/- mice exhibited the disease phenotype at a later stage than the mSOD1/RAG2+/+ mice. (C) Time course of motor performance in the hanging-wire test. mSOD1/RAG2-/- mice (n = 11) performed significantly better than mSOD1/RAG2+/+ mice (n = 13). (D) Changes in mean body weight of mSOD1/RAG2-/- mice (n = 11) and mSOD1/RAG2+/+ mice (n = 13). (E) Kaplan-Meier survival curve. Absence of T and B cells did not affect the survival of mSOD1 mice (mSOD1/RAG2+/+ mice; n = 13, mSOD1/RAG2-/- mice; n = 11) (p = 0.13). Data are expressed as the mean ± SEM. Statistical analysis was done using the generalized Wilcoxon test for Kaplan-Meier curves (B,E) and a two-way repeated multivariate ANOVA (A, C, D) during 12 to 16 weeks of age. *:p < 0.05, #:p < 0.005.\nDifferences in onset and survival (Figure 1B and 1E) were computed using Kaplan-Meier survival statistics (generalized Wilcoxon test). Data in Figure 1A, C and 1D were analyzed using a two-way repeated ANOVA. All other data were analyzed using an unpaired two-tailed Student's t test using Excel software (Microsoft, Dedmond, WA). Data are expressed as mean ± SEM. p < 0.05 was considered statistically significant. Results\nDelayed clinical onset in male mutant SOD1 transgenic mice deficient in the RAG2 gene. (A) Disease progression was significantly delayed in mSOD1/RAG2-/- mice (filled circles; n = 11) compared with mSOD1/RAG2+/+ mice (open circles; n = 13). (B) Kaplan Meier curve for symptomatic onset in mSOD1/RAG2+/+ mice (n = 13) and mSOD1/RAG2-/- mice (n = 11). Mice were evaluated for signs of motor deficiency with the following 4-point scoring system: 4 points, normal (no sign of motor dysfunction); 3 points, hind limb tremors were evident when suspended by the tail; 2 points, gait abnormalities were present; 1 point, dragging of at least one hind limb; 0 points, inability to right self within 30 s. We defined the point when the clinical score became 3 as being clinical onset. The mSOD1/RAG2-/- mice exhibited the disease phenotype at a later stage than the mSOD1/RAG2+/+ mice. (C) Time course of motor performance in the hanging-wire test. mSOD1/RAG2-/- mice (n = 11) performed significantly better than mSOD1/RAG2+/+ mice (n = 13). (D) Changes in mean body weight of mSOD1/RAG2-/- mice (n = 11) and mSOD1/RAG2+/+ mice (n = 13). (E) Kaplan-Meier survival curve. Absence of T and B cells did not affect the survival of mSOD1 mice (mSOD1/RAG2+/+ mice; n = 13, mSOD1/RAG2-/- mice; n = 11) (p = 0.13). Data are expressed as the mean ± SEM. Statistical analysis was done using the generalized Wilcoxon test for Kaplan-Meier curves (B,E) and a two-way repeated multivariate ANOVA (A, C, D) during 12 to 16 weeks of age. *:p < 0.05, #:p < 0.005.\n[SUBTITLE] Delayed clinical onset in mutant SOD1 transgenic mice deficient in RAG2 [SUBSECTION] To investigate the role of lymphocytes in ALS, mSOD1 mice were crossed with RAG2-/- mice to generate mSOD1/RAG2-/- mice that lack functional T- and B-lymphocytes. Although there were no differences in relative copy number of mSOD1 (data not shown), delayed clinical onset was observed in mSOD1/RAG2-/- mice when compared with mSOD1/RAG2+/+ mice (p = 0.0033) (Figure 1A and 1B) or mSOD1/RAG2+/- mice (mSOD1/RAG2+/- mice: 94 ± 4 days (n = 8); mSOD1/RAG2-/- mice: 118 ± 3 days (n = 11), p = 0.0007, generalized Wilcoxon test). Next, neuromuscular dysfunction in mSOD1 mice was evaluated with the hanging-wire test twice a week. Consistent with the clinical onset and scores, genetic depletion of lymphocytes significantly improved motor function of mSOD1/RAG2-/- mice during the early stage (p = 0.0043) (Figure 1C). As neurodegeneration in mSOD1 mice is accompanied by loss of body weight, we also measured the body weight of these mice. We found that mSOD1/RAG2-/- mice had increased weight gain compared to mSOD1/RAG2+/+ mice, suggesting alleviated neurodegeneration in mSOD1/RAG2-/- mice (p = 0.0457) (Figure 1D). mSOD1/RAG2+/+ mice had normal numbers of white blood cells and normal ratios of T and B lymphocytes (white blood cell count: mSOD1tg 8067 ± 1324/ul, control 10867 ± 1629/ul, percentage of CD3+ cells: mSOD1tg 32.00 ± 1.21%, control 31.18 ± 1.23%, percentage of B220+ cells: mSOD1tg 55.97 ± 2.19%, control 57.85 ± 1.52%, n = 4 for each group). Although the absence of mature lymphocytes delayed onset of the disease, it did not affect the survival of mSOD1 mice (Figure 1E) (p = 0.13). Furthermore, survival times for mSOD1/RAG2+/- mice were comparable to those of mSOD1/RAG2-/- mice (average survival time; mSOD1/RAG2+/- mice: 142 ± 3 days (n = 8); mSOD1/RAG2-/- mice: 142 ± 3 days (n = 11), p = 0.74, generalized Wilcoxon test). Collectively, these results suggest that lymphocytes may contribute to acceleration of neurodegeneration in mSOD1 mice at the early stage.\nTo investigate the role of lymphocytes in ALS, mSOD1 mice were crossed with RAG2-/- mice to generate mSOD1/RAG2-/- mice that lack functional T- and B-lymphocytes. Although there were no differences in relative copy number of mSOD1 (data not shown), delayed clinical onset was observed in mSOD1/RAG2-/- mice when compared with mSOD1/RAG2+/+ mice (p = 0.0033) (Figure 1A and 1B) or mSOD1/RAG2+/- mice (mSOD1/RAG2+/- mice: 94 ± 4 days (n = 8); mSOD1/RAG2-/- mice: 118 ± 3 days (n = 11), p = 0.0007, generalized Wilcoxon test). Next, neuromuscular dysfunction in mSOD1 mice was evaluated with the hanging-wire test twice a week. Consistent with the clinical onset and scores, genetic depletion of lymphocytes significantly improved motor function of mSOD1/RAG2-/- mice during the early stage (p = 0.0043) (Figure 1C). As neurodegeneration in mSOD1 mice is accompanied by loss of body weight, we also measured the body weight of these mice. We found that mSOD1/RAG2-/- mice had increased weight gain compared to mSOD1/RAG2+/+ mice, suggesting alleviated neurodegeneration in mSOD1/RAG2-/- mice (p = 0.0457) (Figure 1D). mSOD1/RAG2+/+ mice had normal numbers of white blood cells and normal ratios of T and B lymphocytes (white blood cell count: mSOD1tg 8067 ± 1324/ul, control 10867 ± 1629/ul, percentage of CD3+ cells: mSOD1tg 32.00 ± 1.21%, control 31.18 ± 1.23%, percentage of B220+ cells: mSOD1tg 55.97 ± 2.19%, control 57.85 ± 1.52%, n = 4 for each group). Although the absence of mature lymphocytes delayed onset of the disease, it did not affect the survival of mSOD1 mice (Figure 1E) (p = 0.13). Furthermore, survival times for mSOD1/RAG2+/- mice were comparable to those of mSOD1/RAG2-/- mice (average survival time; mSOD1/RAG2+/- mice: 142 ± 3 days (n = 8); mSOD1/RAG2-/- mice: 142 ± 3 days (n = 11), p = 0.74, generalized Wilcoxon test). Collectively, these results suggest that lymphocytes may contribute to acceleration of neurodegeneration in mSOD1 mice at the early stage.\n[SUBTITLE] Increased microglial activation at the early stage in mSOD1 mice without mature lymphocytes [SUBSECTION] Neurodegeneration in ALS is accompanied by microglial-mediated neuroinflammation, which greatly affects disease progression [12]. Persistent activation of microglia is achieved by the sustained production of various cytokines induced by T lymphocytes [10]. To assess whether the absence of lymphocytes affects activation of microglia/macrophages in mSOD1 mice, spinal cord sections were stained with lectin. Lumbar spinal cord sections from early stage (70-90 days of age) mSOD1/RAG2-/- mice showed significantly increased numbers of lectin-positive microglia/macrophages compared with mSOD1/RAG2+/+ mice. However, there was no significant difference in the number of these cells between mSOD1/RAG2-/- mice and mSOD1/RAG2+/+ mice at the end stage of disease (130-150 days of age) (Figure 2A and 2B). Consistent with these results, western blot analysis showed an intense band for Iba-1 at the early stage of disease in spinal cord of mSOD1/RAG2-/- mice (Figure 2C). However, this striking difference disappeared at the end stage.\nSignificantly increased lectin-positive cells at the early stage of disease. (A) Lumbar sections of spinal cord from mSOD1/RAG2-/- and mSOD1/RAG2+/+ mice at early (70-90 days of age) and end (130-150 days of age) stage of disease were stained for lectin (microglia/macrophage). Scale bar = 50 μm. (B) Significantly increased numbers of lectin-positive microglia/macrophages are observed at the early stage of the disease in spinal cord sections of mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice. There was no significant difference in the number of lectin-positive cells at the end stage of disease. Significance was evaluated by using a generalized Wilcoxon test. **: p < 0.01. (C) Western blot analysis of Iba-1 shows markedly increased expression in lumbar spinal cord of mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice at early stage. Data presented in A-C is representative of three animals.\nAstrocytes are the most abundant non-neuronal cells in the CNS and their important roles in neurodegenerative disease have been suggested [8,9]. To assess the relevance of astrocytes, GFAP-immunoreactive cells were evaluated in lumbar spinal cords. At both early and end stages, there were no differences in intensity of GFAP immunoreactivity (IR) between mSOD1/RAG2-/- mice and control mice (Figure 3A). Also, both groups had comparable numbers of GFAP positive cells (Figure 3B). Collectively, these results suggest that mature lymphocytes may play a role in the inhibition of microglial activation but not of astrocyte activation.\nThe number of GFAP-positive cells in spinal cord does not differ between mSOD1/RAG2+/+ and mSOD1/RAG2-/- mice. (A) Lumbar sections of spinal cord from mSOD1/RAG2-/- mice and control mice stained with anti-GFAP antibody. Scale bar = 50 μm. (B) The number of GFAP-positive astrocytes in lumbar cord of mSOD1/RAG2-/- mice does not significantly differ from that of control mSOD1 mice at early (70-90 days of age) and end (130-150 days of age) stages. Data presented in A-B is representative of three animals.\nNeurodegeneration in ALS is accompanied by microglial-mediated neuroinflammation, which greatly affects disease progression [12]. Persistent activation of microglia is achieved by the sustained production of various cytokines induced by T lymphocytes [10]. To assess whether the absence of lymphocytes affects activation of microglia/macrophages in mSOD1 mice, spinal cord sections were stained with lectin. Lumbar spinal cord sections from early stage (70-90 days of age) mSOD1/RAG2-/- mice showed significantly increased numbers of lectin-positive microglia/macrophages compared with mSOD1/RAG2+/+ mice. However, there was no significant difference in the number of these cells between mSOD1/RAG2-/- mice and mSOD1/RAG2+/+ mice at the end stage of disease (130-150 days of age) (Figure 2A and 2B). Consistent with these results, western blot analysis showed an intense band for Iba-1 at the early stage of disease in spinal cord of mSOD1/RAG2-/- mice (Figure 2C). However, this striking difference disappeared at the end stage.\nSignificantly increased lectin-positive cells at the early stage of disease. (A) Lumbar sections of spinal cord from mSOD1/RAG2-/- and mSOD1/RAG2+/+ mice at early (70-90 days of age) and end (130-150 days of age) stage of disease were stained for lectin (microglia/macrophage). Scale bar = 50 μm. (B) Significantly increased numbers of lectin-positive microglia/macrophages are observed at the early stage of the disease in spinal cord sections of mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice. There was no significant difference in the number of lectin-positive cells at the end stage of disease. Significance was evaluated by using a generalized Wilcoxon test. **: p < 0.01. (C) Western blot analysis of Iba-1 shows markedly increased expression in lumbar spinal cord of mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice at early stage. Data presented in A-C is representative of three animals.\nAstrocytes are the most abundant non-neuronal cells in the CNS and their important roles in neurodegenerative disease have been suggested [8,9]. To assess the relevance of astrocytes, GFAP-immunoreactive cells were evaluated in lumbar spinal cords. At both early and end stages, there were no differences in intensity of GFAP immunoreactivity (IR) between mSOD1/RAG2-/- mice and control mice (Figure 3A). Also, both groups had comparable numbers of GFAP positive cells (Figure 3B). Collectively, these results suggest that mature lymphocytes may play a role in the inhibition of microglial activation but not of astrocyte activation.\nThe number of GFAP-positive cells in spinal cord does not differ between mSOD1/RAG2+/+ and mSOD1/RAG2-/- mice. (A) Lumbar sections of spinal cord from mSOD1/RAG2-/- mice and control mice stained with anti-GFAP antibody. Scale bar = 50 μm. (B) The number of GFAP-positive astrocytes in lumbar cord of mSOD1/RAG2-/- mice does not significantly differ from that of control mSOD1 mice at early (70-90 days of age) and end (130-150 days of age) stages. Data presented in A-B is representative of three animals.\n[SUBTITLE] Increased mRNA expression of Ym1, a neuroprotective factor, in mSOD1/RAG2-/- mice at the early stage [SUBSECTION] Activation of microglia in the presence of GM-CSF and IFN-γ results in a classically activated phenotype (M1 microglia) characterized by the production of ·NO [11]. The primary function of M1 microglia is thought to be activation of inflammation and tissue injury, whereas microglia activated alternatively (M2 microglia) by M-CSF, IL-4 and IL-13 express mannose receptor, MHC class II and Ym1, which leads to resolution of inflammation and promotion of wound healing [10,11]. To determine if delayed onset in mSOD1/RAG2-/- mice is attributable to increased numbers of M2 microglia, we evaluated the mRNA expression of M2 microglial target genes in spinal cord of mSOD1/RAG2-/- mice and control mice. Quantitative RT-PCR demonstrated that mRNA expression of Ym1, which is induced by activated M2 microglia/macrophages, was significantly elevated in mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice at early stage but not at end stage disease. The expression of IL-4R and IL-13 mRNA tended to be elevated in the mSOD1/RAG2-/- mice at the early stage, although this change was not significant (Figure 4A). In contrast, levels of M1 microglia-related molecules such as GM-CSF, IFN-γ, IFN-γ R and NOS2 were similar between mSOD1/RAG2-/- mice and control mice both at early and end stages (Figure 4B). It is worthy to note that considerable expression of GM-CSF, which can induce M1 microglia/macrophages, was observed at the end stage of disease in both groups. There were no significant differences in the expression of mRNA for neurotrophic factors such as GDNF and SLC1A3 between the two groups (Figure 4C). These results suggest that M2 microglia are more dominant than M1 microglia in the spinal cord of mSOD1/RAG2-/- mice at the early stage.\nExpression of Ym1 mRNA is significantly increased in the spinal cord of mSOD1/RAG2-/- mice. Quantitative RT-PCR analyses of spinal cord of mSOD1/RAG2-/- mice and mSOD1/RAG2+/+ mice at early (70-90 days of age) and end (130-150 days of age) stages were performed (n = 3-4 in each group). (A) Relative mRNA expression of M-CSF, IL-4, IL-4R, IL-13, IL-13R, Retnla, and Ym1, which are related to M2 macrophages, is shown. Ym1 expression levels are significantly increased in mSOD1-RAG2-/- mice at the early stage of disease (p = 0.039). (B) Relative mRNA expression of GM-CSF, IFN-γ, IFN-γ R, and NOS2, which are related to M1 macrophages, is shown. M1 macrophage-related molecules were comparable between mSOD1/RAG2-/- mice and control mice. (C) Relative mRNA expression of GDNF and SLC1A3 is shown. Expression levels of GDNF and SLC1A3 did not differ between mSOD1/RAG2-/- mice and control mice at early and end stages. Significance was evaluated using the student's t test. *: p < 0.05.\nActivation of microglia in the presence of GM-CSF and IFN-γ results in a classically activated phenotype (M1 microglia) characterized by the production of ·NO [11]. The primary function of M1 microglia is thought to be activation of inflammation and tissue injury, whereas microglia activated alternatively (M2 microglia) by M-CSF, IL-4 and IL-13 express mannose receptor, MHC class II and Ym1, which leads to resolution of inflammation and promotion of wound healing [10,11]. To determine if delayed onset in mSOD1/RAG2-/- mice is attributable to increased numbers of M2 microglia, we evaluated the mRNA expression of M2 microglial target genes in spinal cord of mSOD1/RAG2-/- mice and control mice. Quantitative RT-PCR demonstrated that mRNA expression of Ym1, which is induced by activated M2 microglia/macrophages, was significantly elevated in mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice at early stage but not at end stage disease. The expression of IL-4R and IL-13 mRNA tended to be elevated in the mSOD1/RAG2-/- mice at the early stage, although this change was not significant (Figure 4A). In contrast, levels of M1 microglia-related molecules such as GM-CSF, IFN-γ, IFN-γ R and NOS2 were similar between mSOD1/RAG2-/- mice and control mice both at early and end stages (Figure 4B). It is worthy to note that considerable expression of GM-CSF, which can induce M1 microglia/macrophages, was observed at the end stage of disease in both groups. There were no significant differences in the expression of mRNA for neurotrophic factors such as GDNF and SLC1A3 between the two groups (Figure 4C). These results suggest that M2 microglia are more dominant than M1 microglia in the spinal cord of mSOD1/RAG2-/- mice at the early stage.\nExpression of Ym1 mRNA is significantly increased in the spinal cord of mSOD1/RAG2-/- mice. Quantitative RT-PCR analyses of spinal cord of mSOD1/RAG2-/- mice and mSOD1/RAG2+/+ mice at early (70-90 days of age) and end (130-150 days of age) stages were performed (n = 3-4 in each group). (A) Relative mRNA expression of M-CSF, IL-4, IL-4R, IL-13, IL-13R, Retnla, and Ym1, which are related to M2 macrophages, is shown. Ym1 expression levels are significantly increased in mSOD1-RAG2-/- mice at the early stage of disease (p = 0.039). (B) Relative mRNA expression of GM-CSF, IFN-γ, IFN-γ R, and NOS2, which are related to M1 macrophages, is shown. M1 macrophage-related molecules were comparable between mSOD1/RAG2-/- mice and control mice. (C) Relative mRNA expression of GDNF and SLC1A3 is shown. Expression levels of GDNF and SLC1A3 did not differ between mSOD1/RAG2-/- mice and control mice at early and end stages. Significance was evaluated using the student's t test. *: p < 0.05.", "All mice were housed in microisolator cages within a modified pathogen-free barrier facility at the Animal Resource Center for Infectious Diseases, Research Institute for Microbial Diseases, Osaka University. All animals had free access to food and water ad libitum, and all of the experimental procedures followed our institutional guidelines.\nTransgenic mice overexpressing the familial ALS-associated G93A SOD1 mutation (harboring a single amino acid substitution of glycine to alanine at codon 93) (mSOD1 mice) were obtained from The Jackson Laboratory (Bar Harbor, ME) (strain designated: B6SJL-TgN(SOD1-G93A)1Gurd/J) and were backcrossed with C57BL/6 mice for at least 10 generations. Transgenic progeny were identified by a polymerase chain reaction (PCR) of genomic DNA using specific primers for exon 4 of the human SOD1 gene: ex4Pla 5'-CATCAGCCCTAATCCATCTGA-3' and ex4P2a 5'-TGGATCTTAGAATTCGCGAC-3'. PCR conditions included 35 cycles at 94°C for 30 s, 60°C for 30 s and 72°C for 1 min. The PCR products were resolved on a 1.0% agarose/ethidium bromide gel and photographed under UV illumination.\nRAG2-/- mice on a C57BL/6 background were obtained from The Jackson Laboratory. RAG2-/- mice were initially bred with mSOD1 mice. The presence or absence of the RAG2 gene was determined by PCR using 250 ng of tail DNA. The following primers were: Rag A, 5'-TAAAAGACCTATTCACAATCAAAAATGTCC-3'; and Rag B, 5'-TCAATCGTGTTGTCCCCTAGAGAGACAAGG-3'. PCR products for the different mice include: a 1100-bp band for homozygote mice, 1100- and 917-bp bands for heterozygote mice, and a 917-bp band for wild-type. PCR was conducted with 35 cycles at 94°C for 30 s, 60°C for 30 s and 72°C for 1 min. The PCR products were resolved by 1.0% agarose/ethidium bromide gel electrophoresis and photographed under UV illumination.", "Symptomatic analysis and weight measurements of male mice were conducted every 3-4 days starting at day 64 (n = 13 for mSOD1/RAG2+/+ mice, n = 11 for mSOD1/RAG2-/- mice). Animals were scored for motor symptoms using a 4-point scoring system as follows: 4 points, normal (no sign of motor dysfunction); 3 points, hind limb tremors were evident when suspended by the tail; 2 points, gait abnormalities were present; 1 point, dragging of at least one hind limb; and 0 points, inability to right self within 30 s. Onset was defined retrospectively as the earliest time when the mice showed symptoms (i.e., score < 4).\nFor the hanging-wire test, each male mouse was placed on a wire lid of a conventional housing cage. The lid was shaken gently to prompt the mouse to hold onto the grid before the lid was swiftly turned upside down. The latency until the mouse let go with at least both hind limbs was timed. Each mouse was allowed up to three attempts to hold on to the inverted lid for an arbitrary maximum of 90 s and the longest latency was recorded.", "Mice were anesthetized with an overdose of pentobarbital (60 mg/kg i.p.) and transcardially perfused with ice-cold 4% paraformaldehyde. Spinal cords were dissected out, postfixed in the same fixative for 4 h, placed overnight in 30% sucrose, and embedded in O.C.T. Compound (Sakura Finetek, Tokyo, Japan). Sections were frozen in liquid nitrogen and the blocks were stored at -80°C until use. Ten-micrometer-thick transverse sections of spinal cords were prepared using a Leica 3050 S cryostat (Thermo Shandon, Inc., Pittsburgh, PA) and collected onto Superfrost slides (Matsunami Glass Industries, Ltd., Osaka, Japan).\nFor lectin staining, sections were permeabilized with 0.2% Tween-20 in 0.05 M Tris-buffered saline (TBST, pH 7.2) for 10 min. They were then incubated with FITC-conjugated lectin (Sigma, St. Louis, MO) diluted 1:750 in phosphate-buffered saline (PBS) overnight at 4°C. Sections were washed 3 times in 0.2% TBST for 5 min, and mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA). The fluorescence-labeled sections were examined using a LSM 510 confocal microscope (Zeiss, Thornwood, NY).\nFor immunohistochemistry, sections were permeabilized with 0.2% TBST for 10 min and pretreated with 10% normal goat serum (Jackson ImmunoResearch Laboratories, West Grove, PA) for 1 h at RT to block non-specific IgG binding. Sections were then incubated with rabbit anti-mouse glial fibrillary acidic protein (GFAP) antibody (Ab) (1:1000) (Dako, Carpinteria, CA) overnight at 4°C. As a negative control, the primary Ab was omitted during the reaction. After rinsing with 0.2% TBST 3 times, sections were incubated with Cy5-conjugated F(ab') 2 fragment donkey anti-rabbit IgG (1:500 in PBS) for 3 h at RT. Sections were then washed 3 times in 0.2% TBST for 5 min and mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories). The fluorescence-labeled sections were examined using an LSM 510 confocal microscope (Zeiss).\nFor quantification of microglia and astrocytes, we analyzed the density of lectin- or GFAP-positive cells, respectively, in every third sample section of lumbar horizontal spinal cord sections. A total of 25 sections were analyzed and cell densities were counted in 4 randomly selected fields per section (a total of 100 fields).", "Mice were anesthetized with an overdose of pentobarbital (60 mg/kg i.p.) and transcardially perfused with ice-cold PBS. Dissected spinal cords were homogenized with protease inhibitors and loaded onto 10% SDS-PAGE gels. Blots were incubated with an antibody against Iba-1 (Wako, Richmond, VA) and visualized by chemiluminescence.", "RNA was isolated from homogenized flash-frozen mice spinal cords using TRIzol (GIBCO/BRL, Grand Island, NY) and purified using RNeasy (QIAGEN, Valencia, CA) according to the manufacturer's recommendations. The concentrations were determined spectrophotometrically. cDNA was generated from 100 ng of each RNA sample digested with RNase-free DNase I (QIAGEN) and SuperScript VILO cDNA Synthesis Kit (Invitrogen, Carlsbad, CA) according to the manufacturer's recommendations. The synthesized cDNA was amplified with SYBR Premix Ex Taq II (Takara Bio Inc., Shiga, Japan) and the primer set, which is listed below, using the StepOnePlus real-time PCR system (Applied BioSystems, Foster City, CA). Primer sets used in this study were: glyceraldehyde-3-phosphate dehydrogenase (GAPDH): 5'-TGTGTCCGTCGTGGATCTGA-3' and 5'-TTGCTGTTGAAGTCGCAGGAG-3'; nitric oxide synthase 2 (NOS2): 5'-GCAGAGATTGGAGGCCTTGTG-3' and 5'-GGGTTGTTGCTGAACTTCCAGTC-3'; IFN-γ: 5'-CGGCACAGTCATTGAAAGCCTA-3' and 5'-GTTGCTGATGGCCTGATTGTC-3'; Ym1: 5'-TTTGATGGCCTCAACCTGGA-3' and 5'-AGTGAGTAGCAGCCTTGGAATGTC-3'; Retnla: 5'-TCAGCAATCCCATGGCGTATAA-3' and 5'-TCATCAGTATTCACTGGGACCATCA-3'; glial cell line derived neurotrophic factor (GDNF): 5'-CCCACGTTTCGCATGGTTC-3' and 5'-TGGGCAGCTGAGGTTGTCAC-3'; solute carrier family 1 (glial high affinity glutamate transporter), member 3 (SLC1A3): 5'-CTTGTCTCATCCATTGGCCTCA-3' and 5'-CCATTCCCAATAAGTGTCCTGTCTG-3'; Mannose receptor, C type 1 (MRC1): 5'-TCGAGACTGCTGCTGAGTCCA-3' and 5'-AGACAGGATTGTCGTTCAACCAAAG-3'; M-CSF: 5'-GAACAGCCTGTCCCATCCATC-3' and 5'-TGAGGCCAGCTCAGTGCAA-3'; GM-CSF: 5'-AAGGGCGCCTTGAACATGAC-3' and 5'-AAATCCGCATAGGTGGTAACTTGTG-3'; interferon gamma receptor 1 (IFN-γR): 5'-TGACGGGAGCACCTGTTACAC-3' and 5'-TTTCGACCGTATGTTTCGTATGTAG-3'; IL-4: 5'-TCTCGAATGTACCAGGAGCCATATC-3' and 5'-AGCACCTTGGAAGCCCTACAGA-3'; interleukin 4 receptor, alpha (IL-4R): 5'-TGAGCCCGTGGTATCTGCAA-3' and 5'-CCAGATCCCTGTGCCTCAAAC-3'; IL-13: 5'-CAATTGCAATGCCATCTACAGGAC-3' and 5'-CGAAACAGTTGCTTTGTGTAGCTGA-3'; interleukin 13 receptor alpha 1 (IL-13R): 5'-CAGTCTTGCAGCATGGGAACA-3' and 5'-TGAGTCCCTAAGGCCTGGAGATTAC-3'. A standard thermal cycling program was used for all PCRs: 95°C for 60 s, 40 cycles of 95°C for 5 s, and 60°C for 30 s. The relative expression levels of each mRNA were calculated using the ΔΔCt method, normalizing to GAPDH and relative to the control samples.", "Differences in onset and survival (Figure 1B and 1E) were computed using Kaplan-Meier survival statistics (generalized Wilcoxon test). Data in Figure 1A, C and 1D were analyzed using a two-way repeated ANOVA. All other data were analyzed using an unpaired two-tailed Student's t test using Excel software (Microsoft, Dedmond, WA). Data are expressed as mean ± SEM. p < 0.05 was considered statistically significant. Results\nDelayed clinical onset in male mutant SOD1 transgenic mice deficient in the RAG2 gene. (A) Disease progression was significantly delayed in mSOD1/RAG2-/- mice (filled circles; n = 11) compared with mSOD1/RAG2+/+ mice (open circles; n = 13). (B) Kaplan Meier curve for symptomatic onset in mSOD1/RAG2+/+ mice (n = 13) and mSOD1/RAG2-/- mice (n = 11). Mice were evaluated for signs of motor deficiency with the following 4-point scoring system: 4 points, normal (no sign of motor dysfunction); 3 points, hind limb tremors were evident when suspended by the tail; 2 points, gait abnormalities were present; 1 point, dragging of at least one hind limb; 0 points, inability to right self within 30 s. We defined the point when the clinical score became 3 as being clinical onset. The mSOD1/RAG2-/- mice exhibited the disease phenotype at a later stage than the mSOD1/RAG2+/+ mice. (C) Time course of motor performance in the hanging-wire test. mSOD1/RAG2-/- mice (n = 11) performed significantly better than mSOD1/RAG2+/+ mice (n = 13). (D) Changes in mean body weight of mSOD1/RAG2-/- mice (n = 11) and mSOD1/RAG2+/+ mice (n = 13). (E) Kaplan-Meier survival curve. Absence of T and B cells did not affect the survival of mSOD1 mice (mSOD1/RAG2+/+ mice; n = 13, mSOD1/RAG2-/- mice; n = 11) (p = 0.13). Data are expressed as the mean ± SEM. Statistical analysis was done using the generalized Wilcoxon test for Kaplan-Meier curves (B,E) and a two-way repeated multivariate ANOVA (A, C, D) during 12 to 16 weeks of age. *:p < 0.05, #:p < 0.005.", "To investigate the role of lymphocytes in ALS, mSOD1 mice were crossed with RAG2-/- mice to generate mSOD1/RAG2-/- mice that lack functional T- and B-lymphocytes. Although there were no differences in relative copy number of mSOD1 (data not shown), delayed clinical onset was observed in mSOD1/RAG2-/- mice when compared with mSOD1/RAG2+/+ mice (p = 0.0033) (Figure 1A and 1B) or mSOD1/RAG2+/- mice (mSOD1/RAG2+/- mice: 94 ± 4 days (n = 8); mSOD1/RAG2-/- mice: 118 ± 3 days (n = 11), p = 0.0007, generalized Wilcoxon test). Next, neuromuscular dysfunction in mSOD1 mice was evaluated with the hanging-wire test twice a week. Consistent with the clinical onset and scores, genetic depletion of lymphocytes significantly improved motor function of mSOD1/RAG2-/- mice during the early stage (p = 0.0043) (Figure 1C). As neurodegeneration in mSOD1 mice is accompanied by loss of body weight, we also measured the body weight of these mice. We found that mSOD1/RAG2-/- mice had increased weight gain compared to mSOD1/RAG2+/+ mice, suggesting alleviated neurodegeneration in mSOD1/RAG2-/- mice (p = 0.0457) (Figure 1D). mSOD1/RAG2+/+ mice had normal numbers of white blood cells and normal ratios of T and B lymphocytes (white blood cell count: mSOD1tg 8067 ± 1324/ul, control 10867 ± 1629/ul, percentage of CD3+ cells: mSOD1tg 32.00 ± 1.21%, control 31.18 ± 1.23%, percentage of B220+ cells: mSOD1tg 55.97 ± 2.19%, control 57.85 ± 1.52%, n = 4 for each group). Although the absence of mature lymphocytes delayed onset of the disease, it did not affect the survival of mSOD1 mice (Figure 1E) (p = 0.13). Furthermore, survival times for mSOD1/RAG2+/- mice were comparable to those of mSOD1/RAG2-/- mice (average survival time; mSOD1/RAG2+/- mice: 142 ± 3 days (n = 8); mSOD1/RAG2-/- mice: 142 ± 3 days (n = 11), p = 0.74, generalized Wilcoxon test). Collectively, these results suggest that lymphocytes may contribute to acceleration of neurodegeneration in mSOD1 mice at the early stage.", "Neurodegeneration in ALS is accompanied by microglial-mediated neuroinflammation, which greatly affects disease progression [12]. Persistent activation of microglia is achieved by the sustained production of various cytokines induced by T lymphocytes [10]. To assess whether the absence of lymphocytes affects activation of microglia/macrophages in mSOD1 mice, spinal cord sections were stained with lectin. Lumbar spinal cord sections from early stage (70-90 days of age) mSOD1/RAG2-/- mice showed significantly increased numbers of lectin-positive microglia/macrophages compared with mSOD1/RAG2+/+ mice. However, there was no significant difference in the number of these cells between mSOD1/RAG2-/- mice and mSOD1/RAG2+/+ mice at the end stage of disease (130-150 days of age) (Figure 2A and 2B). Consistent with these results, western blot analysis showed an intense band for Iba-1 at the early stage of disease in spinal cord of mSOD1/RAG2-/- mice (Figure 2C). However, this striking difference disappeared at the end stage.\nSignificantly increased lectin-positive cells at the early stage of disease. (A) Lumbar sections of spinal cord from mSOD1/RAG2-/- and mSOD1/RAG2+/+ mice at early (70-90 days of age) and end (130-150 days of age) stage of disease were stained for lectin (microglia/macrophage). Scale bar = 50 μm. (B) Significantly increased numbers of lectin-positive microglia/macrophages are observed at the early stage of the disease in spinal cord sections of mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice. There was no significant difference in the number of lectin-positive cells at the end stage of disease. Significance was evaluated by using a generalized Wilcoxon test. **: p < 0.01. (C) Western blot analysis of Iba-1 shows markedly increased expression in lumbar spinal cord of mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice at early stage. Data presented in A-C is representative of three animals.\nAstrocytes are the most abundant non-neuronal cells in the CNS and their important roles in neurodegenerative disease have been suggested [8,9]. To assess the relevance of astrocytes, GFAP-immunoreactive cells were evaluated in lumbar spinal cords. At both early and end stages, there were no differences in intensity of GFAP immunoreactivity (IR) between mSOD1/RAG2-/- mice and control mice (Figure 3A). Also, both groups had comparable numbers of GFAP positive cells (Figure 3B). Collectively, these results suggest that mature lymphocytes may play a role in the inhibition of microglial activation but not of astrocyte activation.\nThe number of GFAP-positive cells in spinal cord does not differ between mSOD1/RAG2+/+ and mSOD1/RAG2-/- mice. (A) Lumbar sections of spinal cord from mSOD1/RAG2-/- mice and control mice stained with anti-GFAP antibody. Scale bar = 50 μm. (B) The number of GFAP-positive astrocytes in lumbar cord of mSOD1/RAG2-/- mice does not significantly differ from that of control mSOD1 mice at early (70-90 days of age) and end (130-150 days of age) stages. Data presented in A-B is representative of three animals.", "Activation of microglia in the presence of GM-CSF and IFN-γ results in a classically activated phenotype (M1 microglia) characterized by the production of ·NO [11]. The primary function of M1 microglia is thought to be activation of inflammation and tissue injury, whereas microglia activated alternatively (M2 microglia) by M-CSF, IL-4 and IL-13 express mannose receptor, MHC class II and Ym1, which leads to resolution of inflammation and promotion of wound healing [10,11]. To determine if delayed onset in mSOD1/RAG2-/- mice is attributable to increased numbers of M2 microglia, we evaluated the mRNA expression of M2 microglial target genes in spinal cord of mSOD1/RAG2-/- mice and control mice. Quantitative RT-PCR demonstrated that mRNA expression of Ym1, which is induced by activated M2 microglia/macrophages, was significantly elevated in mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice at early stage but not at end stage disease. The expression of IL-4R and IL-13 mRNA tended to be elevated in the mSOD1/RAG2-/- mice at the early stage, although this change was not significant (Figure 4A). In contrast, levels of M1 microglia-related molecules such as GM-CSF, IFN-γ, IFN-γ R and NOS2 were similar between mSOD1/RAG2-/- mice and control mice both at early and end stages (Figure 4B). It is worthy to note that considerable expression of GM-CSF, which can induce M1 microglia/macrophages, was observed at the end stage of disease in both groups. There were no significant differences in the expression of mRNA for neurotrophic factors such as GDNF and SLC1A3 between the two groups (Figure 4C). These results suggest that M2 microglia are more dominant than M1 microglia in the spinal cord of mSOD1/RAG2-/- mice at the early stage.\nExpression of Ym1 mRNA is significantly increased in the spinal cord of mSOD1/RAG2-/- mice. Quantitative RT-PCR analyses of spinal cord of mSOD1/RAG2-/- mice and mSOD1/RAG2+/+ mice at early (70-90 days of age) and end (130-150 days of age) stages were performed (n = 3-4 in each group). (A) Relative mRNA expression of M-CSF, IL-4, IL-4R, IL-13, IL-13R, Retnla, and Ym1, which are related to M2 macrophages, is shown. Ym1 expression levels are significantly increased in mSOD1-RAG2-/- mice at the early stage of disease (p = 0.039). (B) Relative mRNA expression of GM-CSF, IFN-γ, IFN-γ R, and NOS2, which are related to M1 macrophages, is shown. M1 macrophage-related molecules were comparable between mSOD1/RAG2-/- mice and control mice. (C) Relative mRNA expression of GDNF and SLC1A3 is shown. Expression levels of GDNF and SLC1A3 did not differ between mSOD1/RAG2-/- mice and control mice at early and end stages. Significance was evaluated using the student's t test. *: p < 0.05.", "In this study, we demonstrate delayed clinical onset in mSOD1/RAG2-/- mice compared to mSOD1/RAG2+/+ mice, with increased number of activated microglia in the spinal cord. Microgliosis in mSOD1/RAG2-/- mice was accompanied by increased Ym1 mRNA expression, one of the hallmarks of M2 microglia/macrophages. However, the differences in weakness, microgliosis, and Ym1 expression between SOD1/RAG2-/- and mSOD1/RAG2+/+ mice disappeared in the later stage of disease. This suggests that this affect on M2 microglia/macrophages, as the result of lymphocyte depletion, may be associated with delayed clinical onset in mSOD1/RAG2-/- mice. As a previous study has shown that B lymphocytes do not have a major role in neurodegeneration in G93A-SOD1 mice [15], T lymphocytes rather than B lymphocytes are thought to be responsible for the increased M2 activation of microglia. Although few T lymphocytes have been shown to infiltrate spinal cord in mSOD1 mice in the early stage, T cells in the meninges may have some effects on neurons or glial cells without entering the parenchyma [16-18]. Increased GM-CSF at the late stage of disease, as shown in Figure 4B, might counteract the neuroprotective effect of M2 microglia observed at the early phase.\nIn contrast to our finding of a deleterious effect of lymphocytes in ALS, previous studies with mSOD1/RAG2-/- mice have indicated a protective rather than deleterious role of lymphocytes in mSOD1 mice [6,16,19]. A possible explanation for these discrepancies is differences in transgene copy number in the mice used by us and other research groups. The mutant G93A transgene is reported to undergo a background level of copy loss due to meiotic rearrangement of the transgene array [20,21]. Since mSOD1 has been suggested to activate innate immunity downstream of the toll-like receptor (TLR) 2, TLR4 and CD14 [22], a difference in transgene copy number may cause different levels of microglial activation. Another possibility is environmental factors. Mice maintained under germ-free conditions develop significantly attenuated symptoms of experimental autoimmune encephalomyelitis (EAE) compared with conventionally colonized mice, and microbiota are reported to have an effect on the T-helper/Treg axis outside of the gut [23]. Thus, different microbiota of mice maintained in different facilities could change the characteristics of lymphocytes and other immune cells in the CNS, thus leading to the discrepancy between our study and previous reports.\nPiecing together our results and previous studies suggests that lymphocytes might have biphasic effects on neurons in this animal model of ALS. That is, lymphocytes may have a deleterious effect in the early phase of neurodegeneration by suppressing M2 microglia, whereas they may exert neuroprotective effects in the later stage of the disease, thus leading to the shortened survival of mSOD1/RAG2-/- mice reported in a previous study [6]. Transgene copy number of our mice and the SPF condition of our facilities might be responsible for the apparent difference seen in the early stage of neurodegeneration between the two groups, while those of other facilities appear to affect the end stage.\nThe present findings suggest that mature lymphocytes act to produce a deterioration of the ALS phenotype in mSOD1 mice at the early stage of the disease by modulating the trophic balance of microglia. If the course of ALS is so modulated, then disease progression could be delayed by expanding populations of M2 microglia/macrophages, thus producing neuroprotective effects.", "This study reports delayed onset of ALS phenotype in mSOD1/RAG2-/- mice compared to controls. We observed an increased number of lectin-positive microglia/macrophages at the early stage of disease in mSOD1/RAG2-/- mice compared with mSOD1/RAG2+/+ mice. In addition, quantitative RT-PCR revealed increased expression of mRNA for M2-microglial-related genes in mSOD1/RAG2-/- mice. In conclusion, these results indicate that peripheral lymphocytes act to produce a deterioration of the ALS phenotype in mSOD1 transgenic mice at the early stage of disease by inhibiting M2 microglia/macrophage skewing.", "The authors declare that they have no competing interests.", "ST designed the experiments, analyzed the data, and wrote the manuscript.\nTO, TY and YN participated in the design of the study and helped draft the manuscript. TS contributed analysis of data as well as provided editing of the manuscript. HK and SS conceived of the study, and participated in its design and conduct and helped draft the manuscript. All authors have read and approved the final version of the manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Affective Symptoms", "Cluster Analysis", "Emotions", "Female", "Humans", "Male", "Personality", "Personality Inventory", "Psychiatric Status Rating Scales", "Severity of Illness Index", "Surveys and Questionnaires", "Young Adult" ]

[ "Background", "Participants", "Measurements", "Data analysis", "Results", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Alexithymia has been a familiar conception as \"no words for feeling\" in psychiatry and psychosomatic medicine since it was first termed by Sifneos [1]. Now its definition is more explicitly refined with five dominant features: (1) difficulty in identifying one's emotion; (2) difficulty in describing self feelings verbally;(3) a reduction or incapability to experience emotions;(4) an absence of tendencies to image one else's emotion, or an externally oriented cognitive style; and (5) poor capacity for fantasize or symbolic thought [2]. Alexithymia refers to a specific disturbance in emotional processing, especially reduced capabilities in verbalizing and realizing emotion. Longitudinal study also suggested that alexithymia was significantly associated with the severity of depression [3], anxiety [4] and schizophrenia [5]. The prevalence rate of alexithymia is significantly higher in patients with psychosomatic disorders, such as eating disorder[6], fibromyalgia syndrome[7] and low-back pain[8], than control groups.\nResearchers [9,10] found alexithymia overlaps with various dimensions like external locus of control and irrational beliefs, except impulsiveness, of the Five-Factor Model (FFM) of personality in an undergraduate student sample. It has been speculated that alexithymia is a cognitive state of externally oriented thinking with an emotional instability and unsecure performance in controlling stressful situation. However, alexithymia has also been criticized whether it is an affect-deficit disorder (state-orient) or a continuous personality variable (trait-orient). Tolmunen et al. [11] considered alexithymia as a stable personality trait in general. Their 11-year follow-up study also suggested that alexithymia might increase vulnerability to depressive symptoms[11]. Honkalampi [12] further demonstrated that depressive symptom might act as a mediator between alexithymia and psychiatric morbidity. Parker and Mattila used taxometric analysis to synthesize several studies about alexithymia in large sample pools including general population and psychotic patients [11,12]. These findings suggest that \"aleixhtymia is not a discrete affect deficit type of person but represents 'the lower tail' of an emotion processing ability that is continuously distributed in the general population\" [11].\nThe purpose of this study was to examine whether there might be subtypes of alexithymia characterized by different behavioural manifestations. In so doing, the current study adopted a cluster analytical approach to examine whether there were natural grouping of people characterized by different psychological features associating with alexithymia. Cluster analysis is a statistical procedure for determining cases can be placed into groups if they share properties in common, while the cases in different clusters are as dissimilar as possible. It was hypothesized that there were various subtypes of alexithymia characterized by different psychological features associating with alexithymia.\nIndividuals on various level of alexithymia would adopt different ways to express and regulate their emotion. The higher alexithymia groups would perform more serious level of depressive or anxious emotional status and more possible to adopt improper regulation strategy.", "1788 college students (freshmen and sophomore) were recruited from three regional universities in Guangzhou, south China. 1071 were males and 616 were females, aged 20.44 ± 1.40 years and 20.51 ± 1.39 years respectively, 101 individual did not mention their gender or age. Economic status was also recorded by a multiple-choice question in the checklist for monthly income per person a month. Among all subjects, 110 individuals did not mention their economic status.\nAll subjects were literarily informed the aim of current study was to examine about psychological status of Chinese youngsters and were voluntarily attended this study. All of them would receive a feedback on the assessment results via email. This study was approved by the ethics committee of the Sun Yat-Sen University.", "Alexithymia was assessed by the 20-item Toronto alexithymia scale (TAS-20) to assess the severity of alexithymia[13,14]. It is a 20-item self-report instrument rated on a 5-point Liker-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Total scores range from 20 to 100, with higher scores indicating higher level of alexithymia. Within the five criteria, the TAS-20 consists of 3 factors: difficulty identifying feelings (DIF); difficulty describing feelings (DDF); externally oriented cognitive style of thinking (EOT). The Chinese version has been shown with having the same factor structure of the original version and has been associated with good internal consistency [15], which was adopted in this study. The Cronbach's α coefficient of it was 0.83, the test-retest reliability coefficient was 0.87, the mean inter-item correlation coefficients ranged from 0.13 to 0.32, the correlate on coefficients of the three factors with the total scale score ranged from 0.72 to 0.82, the correlation coefficients among the three factors ranged from 0.29 to 0.54 [16].\nEmotion expression tendency was assessed by the Chinese version of the Emotional Expressivity Scale (EES) [17]. It is a 17-item self-report assessing the ability to express emotion rated on a 6-point Liker-type scale (1 = never true to 6 = always true). There were two factors in the Chinese version, namely the emotional suppression and emotional expression [18]. The Cronbach's alpha coefficient for the total scale showed a high internal consistency reliability of 0.816. Cronbach's alphas for the two factors were 0.84 and 0.78 respectively indicating adequate internal consistency [18]. Higher total score reflects a higher ability to express emotion, higher expressive factor score means higher intention to express, but lower suppressive factor score means higher inclination to control emotion.\nEmotion Regulation Questionnaire (ERQ) [19] was used to measure emotion regulation. The ERQ is a 10-item checklist capturing two commonly used emotion regulation strategies, i.e., reappraisal and suppression. Reappraisal refers to the use of methods changing the way of thinking about a potential emotional event, whereas suppression refers to the adoption of regulation to suppress when facing the same emotional event. Subjects were required to rate their respond to a 7-point Liker-type scale (1 = totally disagree to 7 = totally agree) on their usual ways of emotional regulation. The test-retest reliability and a coefficient of Chinese version of ERQ were 0.82 and 0.85 for reappraisal dimension, were 0.79 and 0.77 for suppression [20]. Higher score indicates a higher tendency to adopt such strategy. Reappraisal strategy was thought to be a more appropriate way to regulate emotion than suppression one.\nDepression was measured with the Beck depression inventory (BDI) [21,22]. It is a 21-item scale to assess depression problems with higher score representing more depression tendency. The current study adopted the Chinese version of BDI, which Cronbach's alpha coefficient was found to be 0.85 [23].\nAnxiety was assessed with the Chinese version of the state portion of State-Trait Anxiety inventory (STAI-T) [24]. This is a self-reported scale containing 20 items assessing level of anxious status rated on a 4-point Liker-type scale (1 = never to 4 = always). The Cronbach's alpha coefficient showed a high internal consistency reliability was 0.81. The higher score refers a more serious anxious state.", "The Statistical Package for Social Sciences (SPSS) 15.0 (SPSS Inc, Chicago, IL, USA) was used for all statistical analyses reported.\nIndependent sample t-tests were conducted to analyze the gender effect on the total scores of TAS-20. ANOVA was conducted to evaluate the potential effect of economic status upon the total scores of TAS-20, whereas correlation analyses were used to explore any association of education and age with the TAS-20 scores.\nCluster analyses were conducted in two phases. First, a hierarchical cluster analysis was conducted using the 3 factors: difficulty identifying feelings; difficulty describing feelings; externally oriented cognitive style of thinking scores of TAS-20 as the clustering variables and the between-group linkage method with a squared Euclidean distance measure to discriminate clusters. Second, the cluster solution was validated with analysis of variance (ANOVA) on scores of Emotional Expressivity Scale with its subscales, Emotion Regulation Questionnaire, Beck depression inventory and the state portion of State-Trait Anxiety inventory of the identified groups.", "No significant difference was found in the total mean TAS-20 scores between boys and girls (49.63; SD, 8.70 vs 48.96; SD, 8.60; p = 0.13). Age was significantly correlated the total mean TAS-20 (r = 0.05, p = 0.04), whereas there was no significant association between education and TAS-20 total score (r = 0.04, p = 0.11). No significant different in the total mean TAS-20 scores between economic status was found (F = 2.06, p = 0.08). Given the effect of age upon the TAS-20 score was negligible, it was not controlled for subsequent analyses between cluster comparisons.\nTable 1 shows there were four subtypes of alexithymia groups, namely extrovert-high alexithymia (EHA), general-high alexithymia (GHA), introvert-high alexithymia (IHA) and non-alexithymia (NA). The extrovert-high alexithymia (EHA) group was characterized a relative high in externally oriented cognitive style, regular scores in difficulty identifying feelings and difficulty describing feelings and contained most of the cases (77.3%). The general-high alexithymia (GHA) group was characterized by a high score of every factor of alexithymia. The introvert-high alexithymia (IHA) was characterized by a significant high score in difficulty identifying feelings and difficulty describing feelings, which referring to self emotional experience, but relative low score in externally oriented cognitive style of thinking scores. Finally, the non-alexithymia (NA) group was characterized a general low score of alexithymia problems.\nDescription of subtypes of alexithymia\nNote: TAS-20 = 20-itemToronto alexithymia scale, *P < 0.05, **P < 0.01 (two-detailed)\nThe four clusters did not differ significantly in terms of gender propotion (Chi-square = 0.84, df = 3, p = 0.84) and economic status (Chi-square = 16.18, df = 12, p = 0.17). However there showed significant difference in terms of age (t = 3.20, p = 0.02) and education (t = 3.85, p = 0.01).\nAn ANOVA showed that the four subtypes of alexithymia differed significantly in terms of emotional status, emotional expression and regulation as our estimation (Table 2). The general-high alexithymia (GHA) and introvert-high alexithymia (IHA) groups showed dominant higher level of depression and anxious than the extrovert-high alexithymia (EHA) and non-alexithymia (NA) groups. The GHA demonstrated significantly higher scores on suppressive factor in Emotional Expressivity Scale (EES) and Emotion Regulation Questionnaire (ERQ). The introvert-high alexithymia (IHA) group also demonstrated higher scores in suppressive tendency in expressing emotion but adopting more reappraisal strategies in regulating emotion than the GHA group. The non-alexithymia (NA) exhibited the highest will to express their emotions and to choose more reappraisal strategies to regulate their emotions, and was associated with the least depressive and anxiety problems. The extrovert-high alexithymia (EHA) were modest between NA and GHA groups.\nEmotional status, emotional expression and regulation traits among subtypes of alexithymia (ANOVA)\nNote: *P < 0.05, **P < 0.01 (two-detailed)", "The major findings of this study showed there were four subtypes of alexithymia and were consistent with previous studies. For example, Vorst and Bermond [25] suggested that there were two types of alexithymia characterized by the emotional and cognitive factors of the Bermond-Vorst Alexithymia Questionnaire (BVAQ)[26]. They proposed that Type I alexithymia is characterized by a low degree of conscious awareness of emotional arousal and a low degree of emotion accompanying cognitions; whereas Type II alexithymia is characterized by a normal or high degree of conscious awareness of emotional arousal together with a low degree of emotion accompanying cognitions. Our cluster analysis showed that there were 4 subtypes of participants associating with different degrees of alextithymia in the college students, namely the extroverted-high alexithymia (EHA), general-high alexithymia (GHA), introversive-high alexithymia (IHA) and non-alexithymia (NA). The GHA was characterized by a general low profile of emotional cognition including identifying, describing self emotion and external imagination. The general-high alexithymia (GHA) was similar to Type I alexithymia mentioned by Vorst. The introversive-high alexithymia (IHA) was dominant by low arousal of self emotional experience but normal ability of externally oriented thinking style, which was very similar to Type II. The extroverted-high alexithymia (EHA) is characterized by a normal range of self emotional arousal and a profile score of externally oriented thinking style. These features were very similar to those of Type II alexithymia.\nValidation of the cluster solution suggested that these subtypes of alexithymia were characterized by different emotional expression and regulation abilities. The general-high alexithymia (GHA) and introversive-high alexithymia (IHA) were characterized by poorer emotional regulation and expression with worse emotion status. In more details description, the extroverted-high alexithymia (EHA) seemed to be modest in emotion status, with emotion regulating more efficiently as compared to the general-high alexithymia (GHA). These results suggest the potential functional outcome of these different subtypes. Mattila [27] found that individuals with alexithymia showed significantly lower satisfaction to many dimensions of general life than individuals without alexithymia. Our current findings also showed that individuals with general-high alexithymia (GHA) and introversive-high alexithymia (IHA) tended to show less effective ways to regulate their emotion and might face more stress in their social life than other groups. These findings highlight the need for timely and appropriate psychological counseling for these individuals. The characteristics associated with the different clusters of individuals with alexithymia suggest that regulation ability of alexithymia may require different intervention regimes to protect or maintain their own emotional regulation and expression.\nIt should be noted that the EHA cluster includes the 77.3% of the sample. The EHA cluster shows alterations mostly on externally oriented cognitive style of thinking (EOT). Some studies showed difficulty identifying feelings (DIF) and difficulty describing feelings (DDF) had good internal reliability but not EOT [28]. Some researcher thought EOT dimension showed different development paths companying with DDI and DDF [29]. Moreover, it should also be cautious that the clusters we found in the current study may not be stable at all in time and/or it is an artefact due to the instruments that have been used by the current study. It should be necessary to reassess the students to evaluate validity. Without a longitudinal approach any speculation in the discussion should be shown just as an hypothesis to be confirmed. Our results indicated that most of the cases were clustered into extroverted-high alexithymia (EHA) and suggested alexithymia might be a general phenomenon of an emotion processing ability distributed in the general population. More investigations were needed to clarify the relation of alexithymia and personality.\nThe current study has several limitations. First, participants were recruited from a convenient sample pool was the main limitation of this study. Whether these cluster groups could discovered in more broad population needs future study adopted a more rigorous epidemiological approach to improve its representativeness. Second, the findings were based on subjective self-report measures. More rigorous methodologies adopting experimental designs or neurophysiological approaches such as ERP or imaging paradigms should be enforced in the near future in order to validate potential differential neural bases of these subtypes of alexithymia. Finally, the current cross-sectional design could not examine the stability of the cluster solutions across different time points. Future study should adopt a longitudinal design to test the stability of the cluster solution.", "The current findings suggest there were four subtypes of alexithymia characterized by different emotional regulation manifestations.", "ANOVA: analysis of variance; BDI: Beck Depression Inventory; BVAQ: Bermond-Vorst Alexithymia Questionnaire; DDF: difficulty describing feelings; DIF: difficulty identifying feelings; EES: Emotional Expressivity Scale; EHA: extrovert-high alexithymia; EOT: externally oriented cognitive style of thinking; ERP: Event-related Potential; ERQ: Emotion Regulation Questionnaire; FFM: Five-Factor Model; GHA: general-high alexithymia, IHA: introvert-high alexithymia; NA: non-alexithymia; SPSS: Statistical Package for Social Sciences; STAI-T: Trait portion of the State-Trait Anxiety inventory; TAS-20: 20-item Toronto alexithymia scale.", "The authors declare that they have no competing interests.", "JC designed the study, collected and analyzed data, and wrote the first draft of the manuscript. RCKC conceived and designed the study, and wrote the first draft of the paper. TX analyzed data and participated in writing up the manuscript. JJ participated in writing up the manuscript. All authors read and approved the final manuscript", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-244X/11/33/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Method", "Participants", "Measurements", "Data analysis", "Results", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Alexithymia has been a familiar conception as \"no words for feeling\" in psychiatry and psychosomatic medicine since it was first termed by Sifneos [1]. Now its definition is more explicitly refined with five dominant features: (1) difficulty in identifying one's emotion; (2) difficulty in describing self feelings verbally;(3) a reduction or incapability to experience emotions;(4) an absence of tendencies to image one else's emotion, or an externally oriented cognitive style; and (5) poor capacity for fantasize or symbolic thought [2]. Alexithymia refers to a specific disturbance in emotional processing, especially reduced capabilities in verbalizing and realizing emotion. Longitudinal study also suggested that alexithymia was significantly associated with the severity of depression [3], anxiety [4] and schizophrenia [5]. The prevalence rate of alexithymia is significantly higher in patients with psychosomatic disorders, such as eating disorder[6], fibromyalgia syndrome[7] and low-back pain[8], than control groups.\nResearchers [9,10] found alexithymia overlaps with various dimensions like external locus of control and irrational beliefs, except impulsiveness, of the Five-Factor Model (FFM) of personality in an undergraduate student sample. It has been speculated that alexithymia is a cognitive state of externally oriented thinking with an emotional instability and unsecure performance in controlling stressful situation. However, alexithymia has also been criticized whether it is an affect-deficit disorder (state-orient) or a continuous personality variable (trait-orient). Tolmunen et al. [11] considered alexithymia as a stable personality trait in general. Their 11-year follow-up study also suggested that alexithymia might increase vulnerability to depressive symptoms[11]. Honkalampi [12] further demonstrated that depressive symptom might act as a mediator between alexithymia and psychiatric morbidity. Parker and Mattila used taxometric analysis to synthesize several studies about alexithymia in large sample pools including general population and psychotic patients [11,12]. These findings suggest that \"aleixhtymia is not a discrete affect deficit type of person but represents 'the lower tail' of an emotion processing ability that is continuously distributed in the general population\" [11].\nThe purpose of this study was to examine whether there might be subtypes of alexithymia characterized by different behavioural manifestations. In so doing, the current study adopted a cluster analytical approach to examine whether there were natural grouping of people characterized by different psychological features associating with alexithymia. Cluster analysis is a statistical procedure for determining cases can be placed into groups if they share properties in common, while the cases in different clusters are as dissimilar as possible. It was hypothesized that there were various subtypes of alexithymia characterized by different psychological features associating with alexithymia.\nIndividuals on various level of alexithymia would adopt different ways to express and regulate their emotion. The higher alexithymia groups would perform more serious level of depressive or anxious emotional status and more possible to adopt improper regulation strategy.", "[SUBTITLE] Participants [SUBSECTION] 1788 college students (freshmen and sophomore) were recruited from three regional universities in Guangzhou, south China. 1071 were males and 616 were females, aged 20.44 ± 1.40 years and 20.51 ± 1.39 years respectively, 101 individual did not mention their gender or age. Economic status was also recorded by a multiple-choice question in the checklist for monthly income per person a month. Among all subjects, 110 individuals did not mention their economic status.\nAll subjects were literarily informed the aim of current study was to examine about psychological status of Chinese youngsters and were voluntarily attended this study. All of them would receive a feedback on the assessment results via email. This study was approved by the ethics committee of the Sun Yat-Sen University.\n1788 college students (freshmen and sophomore) were recruited from three regional universities in Guangzhou, south China. 1071 were males and 616 were females, aged 20.44 ± 1.40 years and 20.51 ± 1.39 years respectively, 101 individual did not mention their gender or age. Economic status was also recorded by a multiple-choice question in the checklist for monthly income per person a month. Among all subjects, 110 individuals did not mention their economic status.\nAll subjects were literarily informed the aim of current study was to examine about psychological status of Chinese youngsters and were voluntarily attended this study. All of them would receive a feedback on the assessment results via email. This study was approved by the ethics committee of the Sun Yat-Sen University.\n[SUBTITLE] Measurements [SUBSECTION] Alexithymia was assessed by the 20-item Toronto alexithymia scale (TAS-20) to assess the severity of alexithymia[13,14]. It is a 20-item self-report instrument rated on a 5-point Liker-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Total scores range from 20 to 100, with higher scores indicating higher level of alexithymia. Within the five criteria, the TAS-20 consists of 3 factors: difficulty identifying feelings (DIF); difficulty describing feelings (DDF); externally oriented cognitive style of thinking (EOT). The Chinese version has been shown with having the same factor structure of the original version and has been associated with good internal consistency [15], which was adopted in this study. The Cronbach's α coefficient of it was 0.83, the test-retest reliability coefficient was 0.87, the mean inter-item correlation coefficients ranged from 0.13 to 0.32, the correlate on coefficients of the three factors with the total scale score ranged from 0.72 to 0.82, the correlation coefficients among the three factors ranged from 0.29 to 0.54 [16].\nEmotion expression tendency was assessed by the Chinese version of the Emotional Expressivity Scale (EES) [17]. It is a 17-item self-report assessing the ability to express emotion rated on a 6-point Liker-type scale (1 = never true to 6 = always true). There were two factors in the Chinese version, namely the emotional suppression and emotional expression [18]. The Cronbach's alpha coefficient for the total scale showed a high internal consistency reliability of 0.816. Cronbach's alphas for the two factors were 0.84 and 0.78 respectively indicating adequate internal consistency [18]. Higher total score reflects a higher ability to express emotion, higher expressive factor score means higher intention to express, but lower suppressive factor score means higher inclination to control emotion.\nEmotion Regulation Questionnaire (ERQ) [19] was used to measure emotion regulation. The ERQ is a 10-item checklist capturing two commonly used emotion regulation strategies, i.e., reappraisal and suppression. Reappraisal refers to the use of methods changing the way of thinking about a potential emotional event, whereas suppression refers to the adoption of regulation to suppress when facing the same emotional event. Subjects were required to rate their respond to a 7-point Liker-type scale (1 = totally disagree to 7 = totally agree) on their usual ways of emotional regulation. The test-retest reliability and a coefficient of Chinese version of ERQ were 0.82 and 0.85 for reappraisal dimension, were 0.79 and 0.77 for suppression [20]. Higher score indicates a higher tendency to adopt such strategy. Reappraisal strategy was thought to be a more appropriate way to regulate emotion than suppression one.\nDepression was measured with the Beck depression inventory (BDI) [21,22]. It is a 21-item scale to assess depression problems with higher score representing more depression tendency. The current study adopted the Chinese version of BDI, which Cronbach's alpha coefficient was found to be 0.85 [23].\nAnxiety was assessed with the Chinese version of the state portion of State-Trait Anxiety inventory (STAI-T) [24]. This is a self-reported scale containing 20 items assessing level of anxious status rated on a 4-point Liker-type scale (1 = never to 4 = always). The Cronbach's alpha coefficient showed a high internal consistency reliability was 0.81. The higher score refers a more serious anxious state.\nAlexithymia was assessed by the 20-item Toronto alexithymia scale (TAS-20) to assess the severity of alexithymia[13,14]. It is a 20-item self-report instrument rated on a 5-point Liker-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Total scores range from 20 to 100, with higher scores indicating higher level of alexithymia. Within the five criteria, the TAS-20 consists of 3 factors: difficulty identifying feelings (DIF); difficulty describing feelings (DDF); externally oriented cognitive style of thinking (EOT). The Chinese version has been shown with having the same factor structure of the original version and has been associated with good internal consistency [15], which was adopted in this study. The Cronbach's α coefficient of it was 0.83, the test-retest reliability coefficient was 0.87, the mean inter-item correlation coefficients ranged from 0.13 to 0.32, the correlate on coefficients of the three factors with the total scale score ranged from 0.72 to 0.82, the correlation coefficients among the three factors ranged from 0.29 to 0.54 [16].\nEmotion expression tendency was assessed by the Chinese version of the Emotional Expressivity Scale (EES) [17]. It is a 17-item self-report assessing the ability to express emotion rated on a 6-point Liker-type scale (1 = never true to 6 = always true). There were two factors in the Chinese version, namely the emotional suppression and emotional expression [18]. The Cronbach's alpha coefficient for the total scale showed a high internal consistency reliability of 0.816. Cronbach's alphas for the two factors were 0.84 and 0.78 respectively indicating adequate internal consistency [18]. Higher total score reflects a higher ability to express emotion, higher expressive factor score means higher intention to express, but lower suppressive factor score means higher inclination to control emotion.\nEmotion Regulation Questionnaire (ERQ) [19] was used to measure emotion regulation. The ERQ is a 10-item checklist capturing two commonly used emotion regulation strategies, i.e., reappraisal and suppression. Reappraisal refers to the use of methods changing the way of thinking about a potential emotional event, whereas suppression refers to the adoption of regulation to suppress when facing the same emotional event. Subjects were required to rate their respond to a 7-point Liker-type scale (1 = totally disagree to 7 = totally agree) on their usual ways of emotional regulation. The test-retest reliability and a coefficient of Chinese version of ERQ were 0.82 and 0.85 for reappraisal dimension, were 0.79 and 0.77 for suppression [20]. Higher score indicates a higher tendency to adopt such strategy. Reappraisal strategy was thought to be a more appropriate way to regulate emotion than suppression one.\nDepression was measured with the Beck depression inventory (BDI) [21,22]. It is a 21-item scale to assess depression problems with higher score representing more depression tendency. The current study adopted the Chinese version of BDI, which Cronbach's alpha coefficient was found to be 0.85 [23].\nAnxiety was assessed with the Chinese version of the state portion of State-Trait Anxiety inventory (STAI-T) [24]. This is a self-reported scale containing 20 items assessing level of anxious status rated on a 4-point Liker-type scale (1 = never to 4 = always). The Cronbach's alpha coefficient showed a high internal consistency reliability was 0.81. The higher score refers a more serious anxious state.\n[SUBTITLE] Data analysis [SUBSECTION] The Statistical Package for Social Sciences (SPSS) 15.0 (SPSS Inc, Chicago, IL, USA) was used for all statistical analyses reported.\nIndependent sample t-tests were conducted to analyze the gender effect on the total scores of TAS-20. ANOVA was conducted to evaluate the potential effect of economic status upon the total scores of TAS-20, whereas correlation analyses were used to explore any association of education and age with the TAS-20 scores.\nCluster analyses were conducted in two phases. First, a hierarchical cluster analysis was conducted using the 3 factors: difficulty identifying feelings; difficulty describing feelings; externally oriented cognitive style of thinking scores of TAS-20 as the clustering variables and the between-group linkage method with a squared Euclidean distance measure to discriminate clusters. Second, the cluster solution was validated with analysis of variance (ANOVA) on scores of Emotional Expressivity Scale with its subscales, Emotion Regulation Questionnaire, Beck depression inventory and the state portion of State-Trait Anxiety inventory of the identified groups.\nThe Statistical Package for Social Sciences (SPSS) 15.0 (SPSS Inc, Chicago, IL, USA) was used for all statistical analyses reported.\nIndependent sample t-tests were conducted to analyze the gender effect on the total scores of TAS-20. ANOVA was conducted to evaluate the potential effect of economic status upon the total scores of TAS-20, whereas correlation analyses were used to explore any association of education and age with the TAS-20 scores.\nCluster analyses were conducted in two phases. First, a hierarchical cluster analysis was conducted using the 3 factors: difficulty identifying feelings; difficulty describing feelings; externally oriented cognitive style of thinking scores of TAS-20 as the clustering variables and the between-group linkage method with a squared Euclidean distance measure to discriminate clusters. Second, the cluster solution was validated with analysis of variance (ANOVA) on scores of Emotional Expressivity Scale with its subscales, Emotion Regulation Questionnaire, Beck depression inventory and the state portion of State-Trait Anxiety inventory of the identified groups.", "1788 college students (freshmen and sophomore) were recruited from three regional universities in Guangzhou, south China. 1071 were males and 616 were females, aged 20.44 ± 1.40 years and 20.51 ± 1.39 years respectively, 101 individual did not mention their gender or age. Economic status was also recorded by a multiple-choice question in the checklist for monthly income per person a month. Among all subjects, 110 individuals did not mention their economic status.\nAll subjects were literarily informed the aim of current study was to examine about psychological status of Chinese youngsters and were voluntarily attended this study. All of them would receive a feedback on the assessment results via email. This study was approved by the ethics committee of the Sun Yat-Sen University.", "Alexithymia was assessed by the 20-item Toronto alexithymia scale (TAS-20) to assess the severity of alexithymia[13,14]. It is a 20-item self-report instrument rated on a 5-point Liker-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Total scores range from 20 to 100, with higher scores indicating higher level of alexithymia. Within the five criteria, the TAS-20 consists of 3 factors: difficulty identifying feelings (DIF); difficulty describing feelings (DDF); externally oriented cognitive style of thinking (EOT). The Chinese version has been shown with having the same factor structure of the original version and has been associated with good internal consistency [15], which was adopted in this study. The Cronbach's α coefficient of it was 0.83, the test-retest reliability coefficient was 0.87, the mean inter-item correlation coefficients ranged from 0.13 to 0.32, the correlate on coefficients of the three factors with the total scale score ranged from 0.72 to 0.82, the correlation coefficients among the three factors ranged from 0.29 to 0.54 [16].\nEmotion expression tendency was assessed by the Chinese version of the Emotional Expressivity Scale (EES) [17]. It is a 17-item self-report assessing the ability to express emotion rated on a 6-point Liker-type scale (1 = never true to 6 = always true). There were two factors in the Chinese version, namely the emotional suppression and emotional expression [18]. The Cronbach's alpha coefficient for the total scale showed a high internal consistency reliability of 0.816. Cronbach's alphas for the two factors were 0.84 and 0.78 respectively indicating adequate internal consistency [18]. Higher total score reflects a higher ability to express emotion, higher expressive factor score means higher intention to express, but lower suppressive factor score means higher inclination to control emotion.\nEmotion Regulation Questionnaire (ERQ) [19] was used to measure emotion regulation. The ERQ is a 10-item checklist capturing two commonly used emotion regulation strategies, i.e., reappraisal and suppression. Reappraisal refers to the use of methods changing the way of thinking about a potential emotional event, whereas suppression refers to the adoption of regulation to suppress when facing the same emotional event. Subjects were required to rate their respond to a 7-point Liker-type scale (1 = totally disagree to 7 = totally agree) on their usual ways of emotional regulation. The test-retest reliability and a coefficient of Chinese version of ERQ were 0.82 and 0.85 for reappraisal dimension, were 0.79 and 0.77 for suppression [20]. Higher score indicates a higher tendency to adopt such strategy. Reappraisal strategy was thought to be a more appropriate way to regulate emotion than suppression one.\nDepression was measured with the Beck depression inventory (BDI) [21,22]. It is a 21-item scale to assess depression problems with higher score representing more depression tendency. The current study adopted the Chinese version of BDI, which Cronbach's alpha coefficient was found to be 0.85 [23].\nAnxiety was assessed with the Chinese version of the state portion of State-Trait Anxiety inventory (STAI-T) [24]. This is a self-reported scale containing 20 items assessing level of anxious status rated on a 4-point Liker-type scale (1 = never to 4 = always). The Cronbach's alpha coefficient showed a high internal consistency reliability was 0.81. The higher score refers a more serious anxious state.", "The Statistical Package for Social Sciences (SPSS) 15.0 (SPSS Inc, Chicago, IL, USA) was used for all statistical analyses reported.\nIndependent sample t-tests were conducted to analyze the gender effect on the total scores of TAS-20. ANOVA was conducted to evaluate the potential effect of economic status upon the total scores of TAS-20, whereas correlation analyses were used to explore any association of education and age with the TAS-20 scores.\nCluster analyses were conducted in two phases. First, a hierarchical cluster analysis was conducted using the 3 factors: difficulty identifying feelings; difficulty describing feelings; externally oriented cognitive style of thinking scores of TAS-20 as the clustering variables and the between-group linkage method with a squared Euclidean distance measure to discriminate clusters. Second, the cluster solution was validated with analysis of variance (ANOVA) on scores of Emotional Expressivity Scale with its subscales, Emotion Regulation Questionnaire, Beck depression inventory and the state portion of State-Trait Anxiety inventory of the identified groups.", "No significant difference was found in the total mean TAS-20 scores between boys and girls (49.63; SD, 8.70 vs 48.96; SD, 8.60; p = 0.13). Age was significantly correlated the total mean TAS-20 (r = 0.05, p = 0.04), whereas there was no significant association between education and TAS-20 total score (r = 0.04, p = 0.11). No significant different in the total mean TAS-20 scores between economic status was found (F = 2.06, p = 0.08). Given the effect of age upon the TAS-20 score was negligible, it was not controlled for subsequent analyses between cluster comparisons.\nTable 1 shows there were four subtypes of alexithymia groups, namely extrovert-high alexithymia (EHA), general-high alexithymia (GHA), introvert-high alexithymia (IHA) and non-alexithymia (NA). The extrovert-high alexithymia (EHA) group was characterized a relative high in externally oriented cognitive style, regular scores in difficulty identifying feelings and difficulty describing feelings and contained most of the cases (77.3%). The general-high alexithymia (GHA) group was characterized by a high score of every factor of alexithymia. The introvert-high alexithymia (IHA) was characterized by a significant high score in difficulty identifying feelings and difficulty describing feelings, which referring to self emotional experience, but relative low score in externally oriented cognitive style of thinking scores. Finally, the non-alexithymia (NA) group was characterized a general low score of alexithymia problems.\nDescription of subtypes of alexithymia\nNote: TAS-20 = 20-itemToronto alexithymia scale, *P < 0.05, **P < 0.01 (two-detailed)\nThe four clusters did not differ significantly in terms of gender propotion (Chi-square = 0.84, df = 3, p = 0.84) and economic status (Chi-square = 16.18, df = 12, p = 0.17). However there showed significant difference in terms of age (t = 3.20, p = 0.02) and education (t = 3.85, p = 0.01).\nAn ANOVA showed that the four subtypes of alexithymia differed significantly in terms of emotional status, emotional expression and regulation as our estimation (Table 2). The general-high alexithymia (GHA) and introvert-high alexithymia (IHA) groups showed dominant higher level of depression and anxious than the extrovert-high alexithymia (EHA) and non-alexithymia (NA) groups. The GHA demonstrated significantly higher scores on suppressive factor in Emotional Expressivity Scale (EES) and Emotion Regulation Questionnaire (ERQ). The introvert-high alexithymia (IHA) group also demonstrated higher scores in suppressive tendency in expressing emotion but adopting more reappraisal strategies in regulating emotion than the GHA group. The non-alexithymia (NA) exhibited the highest will to express their emotions and to choose more reappraisal strategies to regulate their emotions, and was associated with the least depressive and anxiety problems. The extrovert-high alexithymia (EHA) were modest between NA and GHA groups.\nEmotional status, emotional expression and regulation traits among subtypes of alexithymia (ANOVA)\nNote: *P < 0.05, **P < 0.01 (two-detailed)", "The major findings of this study showed there were four subtypes of alexithymia and were consistent with previous studies. For example, Vorst and Bermond [25] suggested that there were two types of alexithymia characterized by the emotional and cognitive factors of the Bermond-Vorst Alexithymia Questionnaire (BVAQ)[26]. They proposed that Type I alexithymia is characterized by a low degree of conscious awareness of emotional arousal and a low degree of emotion accompanying cognitions; whereas Type II alexithymia is characterized by a normal or high degree of conscious awareness of emotional arousal together with a low degree of emotion accompanying cognitions. Our cluster analysis showed that there were 4 subtypes of participants associating with different degrees of alextithymia in the college students, namely the extroverted-high alexithymia (EHA), general-high alexithymia (GHA), introversive-high alexithymia (IHA) and non-alexithymia (NA). The GHA was characterized by a general low profile of emotional cognition including identifying, describing self emotion and external imagination. The general-high alexithymia (GHA) was similar to Type I alexithymia mentioned by Vorst. The introversive-high alexithymia (IHA) was dominant by low arousal of self emotional experience but normal ability of externally oriented thinking style, which was very similar to Type II. The extroverted-high alexithymia (EHA) is characterized by a normal range of self emotional arousal and a profile score of externally oriented thinking style. These features were very similar to those of Type II alexithymia.\nValidation of the cluster solution suggested that these subtypes of alexithymia were characterized by different emotional expression and regulation abilities. The general-high alexithymia (GHA) and introversive-high alexithymia (IHA) were characterized by poorer emotional regulation and expression with worse emotion status. In more details description, the extroverted-high alexithymia (EHA) seemed to be modest in emotion status, with emotion regulating more efficiently as compared to the general-high alexithymia (GHA). These results suggest the potential functional outcome of these different subtypes. Mattila [27] found that individuals with alexithymia showed significantly lower satisfaction to many dimensions of general life than individuals without alexithymia. Our current findings also showed that individuals with general-high alexithymia (GHA) and introversive-high alexithymia (IHA) tended to show less effective ways to regulate their emotion and might face more stress in their social life than other groups. These findings highlight the need for timely and appropriate psychological counseling for these individuals. The characteristics associated with the different clusters of individuals with alexithymia suggest that regulation ability of alexithymia may require different intervention regimes to protect or maintain their own emotional regulation and expression.\nIt should be noted that the EHA cluster includes the 77.3% of the sample. The EHA cluster shows alterations mostly on externally oriented cognitive style of thinking (EOT). Some studies showed difficulty identifying feelings (DIF) and difficulty describing feelings (DDF) had good internal reliability but not EOT [28]. Some researcher thought EOT dimension showed different development paths companying with DDI and DDF [29]. Moreover, it should also be cautious that the clusters we found in the current study may not be stable at all in time and/or it is an artefact due to the instruments that have been used by the current study. It should be necessary to reassess the students to evaluate validity. Without a longitudinal approach any speculation in the discussion should be shown just as an hypothesis to be confirmed. Our results indicated that most of the cases were clustered into extroverted-high alexithymia (EHA) and suggested alexithymia might be a general phenomenon of an emotion processing ability distributed in the general population. More investigations were needed to clarify the relation of alexithymia and personality.\nThe current study has several limitations. First, participants were recruited from a convenient sample pool was the main limitation of this study. Whether these cluster groups could discovered in more broad population needs future study adopted a more rigorous epidemiological approach to improve its representativeness. Second, the findings were based on subjective self-report measures. More rigorous methodologies adopting experimental designs or neurophysiological approaches such as ERP or imaging paradigms should be enforced in the near future in order to validate potential differential neural bases of these subtypes of alexithymia. Finally, the current cross-sectional design could not examine the stability of the cluster solutions across different time points. Future study should adopt a longitudinal design to test the stability of the cluster solution.", "The current findings suggest there were four subtypes of alexithymia characterized by different emotional regulation manifestations.", "ANOVA: analysis of variance; BDI: Beck Depression Inventory; BVAQ: Bermond-Vorst Alexithymia Questionnaire; DDF: difficulty describing feelings; DIF: difficulty identifying feelings; EES: Emotional Expressivity Scale; EHA: extrovert-high alexithymia; EOT: externally oriented cognitive style of thinking; ERP: Event-related Potential; ERQ: Emotion Regulation Questionnaire; FFM: Five-Factor Model; GHA: general-high alexithymia, IHA: introvert-high alexithymia; NA: non-alexithymia; SPSS: Statistical Package for Social Sciences; STAI-T: Trait portion of the State-Trait Anxiety inventory; TAS-20: 20-item Toronto alexithymia scale.", "The authors declare that they have no competing interests.", "JC designed the study, collected and analyzed data, and wrote the first draft of the manuscript. RCKC conceived and designed the study, and wrote the first draft of the paper. TX analyzed data and participated in writing up the manuscript. JJ participated in writing up the manuscript. All authors read and approved the final manuscript", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-244X/11/33/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null ]

[]

[ "Aged", "Community-Acquired Infections", "Databases, Factual", "Electronic Health Records", "Endocarditis", "Female", "Humans", "Incidence", "Italy", "Male", "Middle Aged", "Staphylococcus" ]

[ "Background", "Results", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Over the last years epidemiological characteristics of infective endocarditis (IE) have been changing in industrialized countries as a result of advances in medical practice: decreasing prevalence of rheumatic heart disease as a predisposing condition, increased longevity, increasing number of patients who undergo invasive procedures [1-5]. Therefore, the emerging population at risk for IE consists of patients with health care-associated infections (acquired during hospitalization or following invasive procedures performed in other health-care settings) [6-8], elderly patients with valvular sclerosis, patients with valvular prostheses, and haemodialysis patients [1,5,9]. Most studies have recently shown a trend towards increasing incidence of Staphylococcus aureus endocarditis [1,10,11].\nMorbidity and mortality are still considerable [2,5,10]: the incidence of IE ranges within 3-10 episodes/100,000 person-year [2]; the in-hospital mortality rate of patients with IE varies from 9,6 to 26% [2]. Larger studies are usually multicentric surveys [1] with selected patients (e.g., only centers with cardiac surgery, with a high proportion of transferred patients) and limited follow-up (only in-hospital or short-term). On the other hand, few high quality population-based studies reviewed in 2007 [9] were heterogeneous as regards population size and demographic structure (male/female ratio ranging from 1.2/1 to 2.5/1, mean or median age ranging from 51 to 69 years), case and outcome definition (from inhospital mortality to 1 year mortality, with overall mortality ranging from 14% to 46% according to different definitions); crude incidence varied from 1.4 to 6.2 per 100,000.\nThe objective of our study was to provide population-based descriptive epidemiological data of IE in the Veneto Region (North-Eastern Italy) through linkage of electronic archives of hospital discharge records (HDR) and mortality records.", "After excluding prevalent cases, 1,863 residents in the Veneto Region were hospitalized for IE in the period 2000-2008. The number of incident IE increased from 562 in 2000-2002 to 700 in 2006-2008 (+25%), with a corresponding crude rate rising from 4.1 to 4.9 per 100,000 person-years (+17%; p = 0.003). Table 1 shows the distribution of demographic characteristics, risk factors and comorbidities in the selected population, and significant changes through 2000-2008. The male to female ratio was 1.7:1, and 60% of subjects were aged 65 and older with a significant increase across the study period. The median age (interquantile range) was 68 years (57-77) in the whole cohort, increasing from 66 years (54-74) in 2000-2002 to 70 years (58-78) in 2006-2008 (p < 0.001). Globally, the rate of IE in people aged 65 and older was 20.3 per 100,000 person-years in male and 10.9 per 100,000 in female subjects. About two out five subjects were hospitalized in the three months preceding the index admission; in 12% a diagnostic code for congestive heart failure was reported in the year before the onset of IE. Information retrieved both from the index and from prior hospitalizations showed that a relevant proportion of the cohort was affected by diabetes mellitus (16%), cancer (10%), and chronic renal failure (8%).\nDemographic and clinical characteristics of hospitalized subjects with infective endocarditis in 2000-2002, 2003-2005, 2006-2008, and p value of the Chi square test for linear trend across the study periods\nTable 2 shows the sparse microbiological data retrieved from diagnostic codes of HDR in the whole cohort (available in 502 subjects), and from blood isolates in the few hospitals participating to the microbiological archive (available in 106 subjects). Although raw (ICD9-CM codes can be hardly translated into a microbiological classification), limited, and related to different subsets of the cohort, the two sources of information are consistent in indicating that about 40% of IE were due to Staphylococci (mainly S. aureus), followed by Streptococci, Enterococci and Gram negatives.\nMicrobiological data retrieved from secondary diagnostic codes in discharge records (n = 502), and from blood culture samples (six hospitals, 2004-2006: n = 106)\nList of ICD-9-CM codes adopted:\n38.1 Staphylococcal septicemia; 38.10 Staphylococcal septicemia, unspecified; 38.11 Staphylococcus aureus septicaemia; 38.19 Other staphylococcal septicemia\nBacterial infection in conditions classified elsewhere and of unspecified site: 41.1 Staphylococcus; 41.10 Staphylococcus, unspecified; 41.11 Staphylococcus aureus; 41.19 Other Staphylococcus\n38.0 Streptococcal septicemia\nBacterial infection in conditions classified elsewhere and of unspecified site: 41.0 Streptococcus; 41.04 Group D [Enterococcus]\n38.4 Septicemia due to other gram-negative organisms\nBacterial infection in conditions classified elsewhere and of unspecified site: 41.3 K. pneumoniae; 41.4 E. coli; 41.5 H. influenzae; 41.6 P. mirabilis; 41.7 Pseudomonas\n38.2 Pneumococcal septicaemia; 38.3 Septicemia due to anaerobes\nBacterial infection in conditions classified elsewhere and of unspecified site: 41.2 Pneumococcus; 41.8 Other specified bacterial infections\n98.84 gonococcal endocarditis; 112.81 candidal endocarditis\nTable 3 shows surgical treatment and outcomes of IE. The median length of stay increased from 23 to 33 days when including hospital transfers, with a rising trend through the study period. If the analysis was restricted to survivors, the median stay was 24 and 35 days excluding or including hospital transfers, respectively. Furthermore, a substantial proportion of subjects surviving the first hospitalization were re-admitted (transfers excluded) with a diagnostic code of IE within 1 year. 23% of subjects underwent a cardiac valve procedure in the index admission or in the following year; such percentage heavily decreased with age (< 55 yrs = 36%; 55-64 yrs = 32%, 65-74 yrs = 23%; ≥75 yrs = 10%; p for trend <0.001). Among subjects submitted to cardiac valve procedures, 37% were treated in the index admission, 24% following an hospital transfer, and 38% in a subsequent readmission within 1 year. The inhospital mortality was 14.3% when excluding hospital transfers, with only a slight and non significant increase through study years; it rose to 18.5% when transfers were accounted for. The linkage with mortality data showed that a substantial proportion of deaths happened between 30 and 90 days from the admission for IE (Figure 1); an increasing time trend in overall mortality was observed at 90 and 365 days of follow-up (Table 3).\nHospitalization data and follow-up of subjects with infective endocarditis (IE) in 2000-2002, 2003-2005, 2006-2008, and p value of the test for linear trend across the study periods\n*data limited to 2006-2007\nSurvival curves of subjects with infective endocarditis according to age class.\nFigure 1 shows survival of subjects according to age: the survival curves of younger patients flatten after 60-90 days, while curves of older patients (especially those aged over 74) continue to drop until the end of follow-up probably due to several comorbidities.\nThe raise of incident cases observed among residents in the Veneto Region and the increase in 90-day fatality produced a growing mortality rate associated to IE, from 0.7 to 1.1 per 100,000 person-years (+65%; p < 0.001).\nTable 4 shows the association of demographic and clinical variables already reported in Table 1 with 90 days mortality; the risk of death increased with age and the Charlson comorbidity index, in subjects hospitalized in the previous three months (only at univariate analysis), in those with a previous diagnosis of heart failure, and in patients with a mention of cancer and chronic renal failure. In the subsets of the cohort with information on microbiology, infections due to Staphylococci (n = 210/502 according to HDR) or to S. aureus (n = 31/106 according to blood isolates) were significantly associated to mortality, with an estimated OR (95% CI) equal to 3.0 (1.8-5.0) and to 4.2 (1.5-11.4), respectively.\nVariables associated to 90-days mortality: odds ratio (OR) with 95% Confidence Interval (CI), at univariate analysis, and at stepwise logistic regression including (Model 1) or excluding (Model2) the Charlson index.", "Our study demonstrates an increasing trend in incidence and mortality for IE in the Veneto Region over the last decade. The main limit of the study is the lack of validation of IE tracked by ICD9-CM discharge codes: such disease misclassification could have led to the inclusion of false positive cases and the exclusion of false negatives. Discharge diagnoses were chosen based on a pilot investigation carried out in recent years in a single Veneto hospital by chart review of discharges with a broader extraction of diagnostic codes, including as well as 421.x, 98.84, 112.81, also 93.2 (syphilitic endocarditis), 391.1 (acute rheumatic endocarditis), 424.9x (endocarditis, valve unspecified), 966.61 (infection and inflammatory reaction due to cardiac device implant and graft). IE diagnoses were validated according to the modified Duke criteria [13], demonstrating a positive predictive value and a sensitivity for IE of the selection applied in the present study equal to 91/123 = 81% and 91/98 = 93%, respectively (Pellizzer, personnel communication). To what extent such findings are applicable to the entire region and study period remains uncertain, but disease misclassification does not probably account for such sharp time trends found in our study; moreover, the incidence rate (4.4 per 100000 person-years) is in the range reported in literature and in particular it's similar to incidences estimated by multicenter prospective surveys conducted in other regions of Northern Italy [14,15]. Furthermore, since the period analyzed to exclude prevalent cases increases over time (minimum = 365 days in 2000), possibly not all prevalent cases have been deleted from early study years, leading to a small underestimation of the real increase of IE incidence; this effect was tested using a constant wash-out time of 365 days through the study period.\nOur investigation is a large population-based survey spanning over several years that through record-linkage allows for a complete follow-up of IE cases as regards multiple outcomes: surgical interventions, hospital re-admissions, short-term and middle-term mortality. Descriptive data on incidence, surgical treatment, case-fatality obtained by the record-linkage system are within the range reported in the international literature. It must be remarked that in our analyses day 0 is the date of admission; as a consequence, health care-associated cases would have altered figures on length of stay and survival. However survival curves (except for the oldest age group) flatten after the first 60-90 days from admission; therefore 90-day mortality seems to be a good estimate of short term mortality associated to IE. Although recent studies report similar mortality rates in patients diagnosed and treated in tertiary care hospitals and those referred to tertiary centres from other hospitals [16,17], our data show that some usual outcome measures, such as simple in hospital mortality, could be inadequate since they miss a relevant burden of deaths occurring after discharge at home or in subsequent hospitalizations.\nOn the other hand, constraints in clinical data (risk factors, comorbidities, diagnostic work-out and medical therapy) and the almost complete lack of microbiological data derivable from HDR limit our interpretation. Although more accurate diagnostic techniques (such as a higher use of transesophageal echocardiography) can play a role in rising incidence rates [18,19], we cannot exclude that population at risk itself is increasing, considering the increasing share of elderly subjects among patients with IE.\nAccording to literature, we confirm a high mortality associated to IE, particularly in older ages and in subjects with comorbidities [2,10,20-22]; a novel finding is the poor prognosis in subjects with prior hospitalizations for heart failure. However, the increasing trend in mortality can only partially be explained by the growing number of affected elderly patients. It must be enlighten that some well identified predictors of poor prognosis such as some echocardiographic findings (size of vegetations, presence of periannular complications, severity of valve dysfunction at diagnosis) and the presence of foreign material [2,18,23-25] couldn't be deduced from HDR. Moreover, we couldn't get differences in treatment protocols over time and among different hospitals.\nIn particular, our survey lacks information on indications for surgery; we found that 23% of patients underwent surgical treatment, with a probability sharply decreasing with age. Recent multicentric studies reported surgery in about half of cases [1,26]. However, some studies included only centres with cardiac surgery units, and the percentage of surgical therapy decreased when restricting analysis to patients admitted directly to study sites [1]; moreover they dealt with a younger population (median age under 60). Among population-based studies, the share of patients submitted to surgery ranged from 12.8% to 40% [9,14,15], with the exception of a survey in France where 49% of subjects were surgically treated [27]. Although there are still some areas of debate and individualized factors must be considered, it's been advocated that early surgery could be worth even in patients with high operative risk [27]; in general, in all IE cases, early surgical consultation is recommended [2,26].\nThe high proportion of patients with previous hospitalization may explain the consistent - although limited - microbiological data, being S. aureus the leading microorganism isolated. We also found that infections caused by Staphylococci are significantly associated to mortality, a finding consistent with literature [1,3,4,11,14,28]. Whether an increasing number and severity of comorbidities in subjects developing IE or, as previously hypothized [1,4,11], an increasing share of IE due to Staphylococci, could have led to the observed raise in mortality remains uncertain and deserves further investigations on a clinical basis.", "The study demonstrates an increasing incidence and mortality for IE over the last decade. Some usual outcome measures (in hospital mortality) miss a relevant burden of deaths occurring after discharge. Analyses through electronic archives allow to draw a region-wide picture of IE, overcoming those referral biases that unavoidably affect single clinic or multicentric studies, and therefore represent a first fundamental step to detect critical issues related to infective endocarditis.", "The authors declare that they have no competing interests.", "GP and SP designed the study and revised the manuscript, ES and UF collected data and performed statistical analyses, UF and DB wrote the first draft of the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2334/11/48/prepub\n" ]

[ null, null, null, null, null, null, null ]

[ "Background", "Methods", "Results", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Over the last years epidemiological characteristics of infective endocarditis (IE) have been changing in industrialized countries as a result of advances in medical practice: decreasing prevalence of rheumatic heart disease as a predisposing condition, increased longevity, increasing number of patients who undergo invasive procedures [1-5]. Therefore, the emerging population at risk for IE consists of patients with health care-associated infections (acquired during hospitalization or following invasive procedures performed in other health-care settings) [6-8], elderly patients with valvular sclerosis, patients with valvular prostheses, and haemodialysis patients [1,5,9]. Most studies have recently shown a trend towards increasing incidence of Staphylococcus aureus endocarditis [1,10,11].\nMorbidity and mortality are still considerable [2,5,10]: the incidence of IE ranges within 3-10 episodes/100,000 person-year [2]; the in-hospital mortality rate of patients with IE varies from 9,6 to 26% [2]. Larger studies are usually multicentric surveys [1] with selected patients (e.g., only centers with cardiac surgery, with a high proportion of transferred patients) and limited follow-up (only in-hospital or short-term). On the other hand, few high quality population-based studies reviewed in 2007 [9] were heterogeneous as regards population size and demographic structure (male/female ratio ranging from 1.2/1 to 2.5/1, mean or median age ranging from 51 to 69 years), case and outcome definition (from inhospital mortality to 1 year mortality, with overall mortality ranging from 14% to 46% according to different definitions); crude incidence varied from 1.4 to 6.2 per 100,000.\nThe objective of our study was to provide population-based descriptive epidemiological data of IE in the Veneto Region (North-Eastern Italy) through linkage of electronic archives of hospital discharge records (HDR) and mortality records.", "The total population of the Veneto Region was 4,885,548 on January 1, 2009. In the Region there are about 65 hospitals (including public and private institutes) with an overall number of about 16,000 hospital beds for acute care; there are approximately 900,000 discharges from Veneto hospitals each year. One primary and up to five secondary discharge diagnoses, plus one primary and up to five secondary procedures are registered in the regional archive of HDRs, which includes all discharges from Veneto hospitals and all discharges of Veneto residents hospitalized outside the region. HDRs with a primary or secondary International Classification of Diseases, 9th Revision, Clinical Modification -versions 1997 and 2002- diagnosis code of IE (421.x = acute and sub-acute endocarditis; 98.84 = gonococcal endocarditis; 112.81 = candidal endocarditis) were extracted starting from 1 January 1999.\nIn order to identify a cohort of incident cases with adequate follow-up and to prevent double counting of the included subjects, the first hospitalization (day-case excluded) for IE in the years 2000-2008 was selected; patients already discharged in 1999 with a diagnosis of IE were removed (prevalent cases). Demographic (age, gender) and clinical information (Charlson comorbidity computed from discharge diagnoses [12]) were extracted from the selected HDR. For each subject, all hospitalizations in the year before the first admission for IE were traced and diagnostic and intervention codes analyzed in order to retrieve information on comorbidities and risk factors (cancer, diabetes mellitus, chronic renal failure, congestive heart failure, previous cardiac valve surgery). Furthermore, all hospitalizations in the year following the first admission for IE were selected in order to gain information on hospital transfers (zero or one day difference between discharge and subsequent admission for IE), overall in-hospital mortality, re-admissions, and cardiac valve surgeries. Finally, the cohort of hospitalized IE was linked to the regional mortality registry of the years 2000-2008 to estimate survival of patients; data were censored after a follow-up of 365 days.\nThe microbiologic aetiology of IE is rarely reported in HDR by means of diagnostic codes 38.x or 41.x. Additionally, six regional hospitals sent microbiological data to a common database in the years 2004-2006; information on isolates from blood cultures was extracted and linked to HDR to identify the causative microorganisms and their influence on outcomes in such small subset of the cohort.\nThe study was carried out on data routinely collected by health services and linkage was performed on anonymized records without any possibility of identification of individuals. The study was approved by the Institutional Review Board of the Vicenza Hospital.\nContinuous variables are presented as medians with interquantile ranges; categorical variables are presented as frequencies/percentages. The presence of time trends across the time periods was assessed by means of the Chi-square test for linear trend or a non parametric trend test derived from the Wilcoxon rank-sum test, as appropriate. The influence of predictive variables on short term mortality (dependent variable = vital status at 90 days) was evaluated by computing the Odds Ratio (OR) with 95% Confidence Intervals (CI). Variables resulting statistically significant at univariate analysis were selected by two models of stepwise logistic regression, including or excluding the Charlson index (which already takes into account diseases reported in the index admission). Statistical analyses were performed using commercially available software (Stata version 9.1; SAS version 9.1).", "After excluding prevalent cases, 1,863 residents in the Veneto Region were hospitalized for IE in the period 2000-2008. The number of incident IE increased from 562 in 2000-2002 to 700 in 2006-2008 (+25%), with a corresponding crude rate rising from 4.1 to 4.9 per 100,000 person-years (+17%; p = 0.003). Table 1 shows the distribution of demographic characteristics, risk factors and comorbidities in the selected population, and significant changes through 2000-2008. The male to female ratio was 1.7:1, and 60% of subjects were aged 65 and older with a significant increase across the study period. The median age (interquantile range) was 68 years (57-77) in the whole cohort, increasing from 66 years (54-74) in 2000-2002 to 70 years (58-78) in 2006-2008 (p < 0.001). Globally, the rate of IE in people aged 65 and older was 20.3 per 100,000 person-years in male and 10.9 per 100,000 in female subjects. About two out five subjects were hospitalized in the three months preceding the index admission; in 12% a diagnostic code for congestive heart failure was reported in the year before the onset of IE. Information retrieved both from the index and from prior hospitalizations showed that a relevant proportion of the cohort was affected by diabetes mellitus (16%), cancer (10%), and chronic renal failure (8%).\nDemographic and clinical characteristics of hospitalized subjects with infective endocarditis in 2000-2002, 2003-2005, 2006-2008, and p value of the Chi square test for linear trend across the study periods\nTable 2 shows the sparse microbiological data retrieved from diagnostic codes of HDR in the whole cohort (available in 502 subjects), and from blood isolates in the few hospitals participating to the microbiological archive (available in 106 subjects). Although raw (ICD9-CM codes can be hardly translated into a microbiological classification), limited, and related to different subsets of the cohort, the two sources of information are consistent in indicating that about 40% of IE were due to Staphylococci (mainly S. aureus), followed by Streptococci, Enterococci and Gram negatives.\nMicrobiological data retrieved from secondary diagnostic codes in discharge records (n = 502), and from blood culture samples (six hospitals, 2004-2006: n = 106)\nList of ICD-9-CM codes adopted:\n38.1 Staphylococcal septicemia; 38.10 Staphylococcal septicemia, unspecified; 38.11 Staphylococcus aureus septicaemia; 38.19 Other staphylococcal septicemia\nBacterial infection in conditions classified elsewhere and of unspecified site: 41.1 Staphylococcus; 41.10 Staphylococcus, unspecified; 41.11 Staphylococcus aureus; 41.19 Other Staphylococcus\n38.0 Streptococcal septicemia\nBacterial infection in conditions classified elsewhere and of unspecified site: 41.0 Streptococcus; 41.04 Group D [Enterococcus]\n38.4 Septicemia due to other gram-negative organisms\nBacterial infection in conditions classified elsewhere and of unspecified site: 41.3 K. pneumoniae; 41.4 E. coli; 41.5 H. influenzae; 41.6 P. mirabilis; 41.7 Pseudomonas\n38.2 Pneumococcal septicaemia; 38.3 Septicemia due to anaerobes\nBacterial infection in conditions classified elsewhere and of unspecified site: 41.2 Pneumococcus; 41.8 Other specified bacterial infections\n98.84 gonococcal endocarditis; 112.81 candidal endocarditis\nTable 3 shows surgical treatment and outcomes of IE. The median length of stay increased from 23 to 33 days when including hospital transfers, with a rising trend through the study period. If the analysis was restricted to survivors, the median stay was 24 and 35 days excluding or including hospital transfers, respectively. Furthermore, a substantial proportion of subjects surviving the first hospitalization were re-admitted (transfers excluded) with a diagnostic code of IE within 1 year. 23% of subjects underwent a cardiac valve procedure in the index admission or in the following year; such percentage heavily decreased with age (< 55 yrs = 36%; 55-64 yrs = 32%, 65-74 yrs = 23%; ≥75 yrs = 10%; p for trend <0.001). Among subjects submitted to cardiac valve procedures, 37% were treated in the index admission, 24% following an hospital transfer, and 38% in a subsequent readmission within 1 year. The inhospital mortality was 14.3% when excluding hospital transfers, with only a slight and non significant increase through study years; it rose to 18.5% when transfers were accounted for. The linkage with mortality data showed that a substantial proportion of deaths happened between 30 and 90 days from the admission for IE (Figure 1); an increasing time trend in overall mortality was observed at 90 and 365 days of follow-up (Table 3).\nHospitalization data and follow-up of subjects with infective endocarditis (IE) in 2000-2002, 2003-2005, 2006-2008, and p value of the test for linear trend across the study periods\n*data limited to 2006-2007\nSurvival curves of subjects with infective endocarditis according to age class.\nFigure 1 shows survival of subjects according to age: the survival curves of younger patients flatten after 60-90 days, while curves of older patients (especially those aged over 74) continue to drop until the end of follow-up probably due to several comorbidities.\nThe raise of incident cases observed among residents in the Veneto Region and the increase in 90-day fatality produced a growing mortality rate associated to IE, from 0.7 to 1.1 per 100,000 person-years (+65%; p < 0.001).\nTable 4 shows the association of demographic and clinical variables already reported in Table 1 with 90 days mortality; the risk of death increased with age and the Charlson comorbidity index, in subjects hospitalized in the previous three months (only at univariate analysis), in those with a previous diagnosis of heart failure, and in patients with a mention of cancer and chronic renal failure. In the subsets of the cohort with information on microbiology, infections due to Staphylococci (n = 210/502 according to HDR) or to S. aureus (n = 31/106 according to blood isolates) were significantly associated to mortality, with an estimated OR (95% CI) equal to 3.0 (1.8-5.0) and to 4.2 (1.5-11.4), respectively.\nVariables associated to 90-days mortality: odds ratio (OR) with 95% Confidence Interval (CI), at univariate analysis, and at stepwise logistic regression including (Model 1) or excluding (Model2) the Charlson index.", "Our study demonstrates an increasing trend in incidence and mortality for IE in the Veneto Region over the last decade. The main limit of the study is the lack of validation of IE tracked by ICD9-CM discharge codes: such disease misclassification could have led to the inclusion of false positive cases and the exclusion of false negatives. Discharge diagnoses were chosen based on a pilot investigation carried out in recent years in a single Veneto hospital by chart review of discharges with a broader extraction of diagnostic codes, including as well as 421.x, 98.84, 112.81, also 93.2 (syphilitic endocarditis), 391.1 (acute rheumatic endocarditis), 424.9x (endocarditis, valve unspecified), 966.61 (infection and inflammatory reaction due to cardiac device implant and graft). IE diagnoses were validated according to the modified Duke criteria [13], demonstrating a positive predictive value and a sensitivity for IE of the selection applied in the present study equal to 91/123 = 81% and 91/98 = 93%, respectively (Pellizzer, personnel communication). To what extent such findings are applicable to the entire region and study period remains uncertain, but disease misclassification does not probably account for such sharp time trends found in our study; moreover, the incidence rate (4.4 per 100000 person-years) is in the range reported in literature and in particular it's similar to incidences estimated by multicenter prospective surveys conducted in other regions of Northern Italy [14,15]. Furthermore, since the period analyzed to exclude prevalent cases increases over time (minimum = 365 days in 2000), possibly not all prevalent cases have been deleted from early study years, leading to a small underestimation of the real increase of IE incidence; this effect was tested using a constant wash-out time of 365 days through the study period.\nOur investigation is a large population-based survey spanning over several years that through record-linkage allows for a complete follow-up of IE cases as regards multiple outcomes: surgical interventions, hospital re-admissions, short-term and middle-term mortality. Descriptive data on incidence, surgical treatment, case-fatality obtained by the record-linkage system are within the range reported in the international literature. It must be remarked that in our analyses day 0 is the date of admission; as a consequence, health care-associated cases would have altered figures on length of stay and survival. However survival curves (except for the oldest age group) flatten after the first 60-90 days from admission; therefore 90-day mortality seems to be a good estimate of short term mortality associated to IE. Although recent studies report similar mortality rates in patients diagnosed and treated in tertiary care hospitals and those referred to tertiary centres from other hospitals [16,17], our data show that some usual outcome measures, such as simple in hospital mortality, could be inadequate since they miss a relevant burden of deaths occurring after discharge at home or in subsequent hospitalizations.\nOn the other hand, constraints in clinical data (risk factors, comorbidities, diagnostic work-out and medical therapy) and the almost complete lack of microbiological data derivable from HDR limit our interpretation. Although more accurate diagnostic techniques (such as a higher use of transesophageal echocardiography) can play a role in rising incidence rates [18,19], we cannot exclude that population at risk itself is increasing, considering the increasing share of elderly subjects among patients with IE.\nAccording to literature, we confirm a high mortality associated to IE, particularly in older ages and in subjects with comorbidities [2,10,20-22]; a novel finding is the poor prognosis in subjects with prior hospitalizations for heart failure. However, the increasing trend in mortality can only partially be explained by the growing number of affected elderly patients. It must be enlighten that some well identified predictors of poor prognosis such as some echocardiographic findings (size of vegetations, presence of periannular complications, severity of valve dysfunction at diagnosis) and the presence of foreign material [2,18,23-25] couldn't be deduced from HDR. Moreover, we couldn't get differences in treatment protocols over time and among different hospitals.\nIn particular, our survey lacks information on indications for surgery; we found that 23% of patients underwent surgical treatment, with a probability sharply decreasing with age. Recent multicentric studies reported surgery in about half of cases [1,26]. However, some studies included only centres with cardiac surgery units, and the percentage of surgical therapy decreased when restricting analysis to patients admitted directly to study sites [1]; moreover they dealt with a younger population (median age under 60). Among population-based studies, the share of patients submitted to surgery ranged from 12.8% to 40% [9,14,15], with the exception of a survey in France where 49% of subjects were surgically treated [27]. Although there are still some areas of debate and individualized factors must be considered, it's been advocated that early surgery could be worth even in patients with high operative risk [27]; in general, in all IE cases, early surgical consultation is recommended [2,26].\nThe high proportion of patients with previous hospitalization may explain the consistent - although limited - microbiological data, being S. aureus the leading microorganism isolated. We also found that infections caused by Staphylococci are significantly associated to mortality, a finding consistent with literature [1,3,4,11,14,28]. Whether an increasing number and severity of comorbidities in subjects developing IE or, as previously hypothized [1,4,11], an increasing share of IE due to Staphylococci, could have led to the observed raise in mortality remains uncertain and deserves further investigations on a clinical basis.", "The study demonstrates an increasing incidence and mortality for IE over the last decade. Some usual outcome measures (in hospital mortality) miss a relevant burden of deaths occurring after discharge. Analyses through electronic archives allow to draw a region-wide picture of IE, overcoming those referral biases that unavoidably affect single clinic or multicentric studies, and therefore represent a first fundamental step to detect critical issues related to infective endocarditis.", "The authors declare that they have no competing interests.", "GP and SP designed the study and revised the manuscript, ES and UF collected data and performed statistical analyses, UF and DB wrote the first draft of the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2334/11/48/prepub\n" ]

[ null, "methods", null, null, null, null, null, null ]

[]

[ "Age Factors", "Agglutination Tests", "Anti-Infective Agents", "Body Weight", "Child, Preschool", "Cross-Sectional Studies", "Diarrhea", "Disk Diffusion Antimicrobial Tests", "Enzyme-Linked Immunosorbent Assay", "Feces", "Female", "Hospitalization", "Humans", "Infant", "Infant, Newborn", "Male", "Nutritional Status", "Practice Guidelines as Topic", "Risk Factors", "Surveys and Questionnaires", "Tanzania" ]

[ "Background", "Study design and settings", "Interviews", "Weight measurements", "Determination of nutritional status", "Collection and transportation of stool", "Isolation and identification of bacteria pathogens", "Antimicrobial susceptibility testing of bacterial isolates", "Detection of viruses", "Detection of intestinal protozoa Cryptosporidium parvum and Giardia lamblia", "Ethical considerations", "Data analysis", "Results", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Infective diarrhoea is one of the leading cause of morbidity and mortality among children under five years in the developing world [1] and can be caused by a wide range of viruses, bacteria, or parasites [2,3]. The prevalence of the different enteric pathogens varies with the geographical area [2].\nThere are relatively few studies on this aspect that have been conducted in Tanzania. Most of these studies have focused on few aetiological agents of diarrhoea and were conducted in late eighties and early nineties [4-6]. There is one recent study conducted in Ifakara, Tanzania in 2004, which shows the prevalence of different aetiological agents of diarrhoea, i.e. viruses, bacteria and parasites among children with diarrhoea. In that study diarrhoeagenic Escherichia coli (DEC) was the predominant enteropathogen. Moreover, Shigella spp. and rotavirus were more prevalent in the dry season than in the rainy season and Giardia lamblia was more prevalent in the rainy season [7].\nFor quite some time children with diarrhoea in Tanzania have been and are being treated empirically with erythromycin and co-trimoxazole according to the Integrated Management of Childhood Illness (IMCI) guidelines [8]. However current information regarding antimicrobial susceptibility pattern of bacteria causing diarrhoea in children is limited and thus it is uncertain whether the recommended antibiotics are still effective.\nThe present study aimed at detection of enteric pathogens in children with diarrhoea to provide an update on the spectrum and age specific causes of diarrhoea. The study also determined the antimicrobial resistance pattern of bacterial pathogens.", "This cross-sectional study was conducted in Dar es Salaam, Tanzania between December 2005 and February 2006. Participants were children ≤5 years, who during the study period, were admitted due to diarrhoea at Muhimbili National Referral Hospital (MNH), Amana, Mwananyamala and Temeke Municipal Hospitals in Dar es Salaam, Tanzania. Enrolment was subjected to obtaining an informed verbal consent from parent or guardian who accompanied the child.", "A structured questionnaire was used to obtain information of the children from the parents/guardians regarding age, sex, duration and description of the stool (watery, mucoid, or bloody) and the use of antibiotics prior to hospitalization. The definition and forms of diarrhoea were according to WHO guidelines [8].", "On admission infants less than two years of age were weighed using a 25 kg Salter hanging scales (CMS Weighing equipment, High Holborn, London, United Kingdom). Children over two years were weighed on scales calibrated before each session. Weight of children was recorded to the nearest kilogram.", "Weight-for age Z-scores were calculated using EPI Info (USD, Inc., Stone Mountain, GA). According to WHO criteria children were considered to be undernourished if the Z-scores were less than -2SD[9].", "Stool specimens were collected using wide mouthed sterile plastic containers and transported to the Microbiology laboratory at Muhimbili National Hospital within two hours of collection. A portion of specimen was stored at -20°C until further analysis at Haukeland University Hospital in Bergen, Norway.", "Bacterial pathogens E. coli, V. cholerae, Salmonella spp and Shigella spp. were isolated and identified by conventional methods [10]. Identification of DEC was done as previously described [11].", "Susceptibility testing was performed by disk diffusion method according to Clinical and Laboratory Standards Institute guidelines (CLSI) [12]. Escherichia coli ATCC 25922 was used as a reference control strain. Antibiotics tested were ampicillin 10 μg, amoxycillin and clavulanic acid (Augmentin) 20/10 μg, erythromycin 15 μg, ciprofloxacin 100 μg, gentamicin 10 μg, cefotaxime 30 μg, cephalothin 30 μg, co-trimoxazole 25 μg, chloramphenicol 30 μg and tetracycline 30 μg (Remel, Lenexa, USA).\nAccording to the size of the zones of inhibition, the organisms were classified as sensitive, intermediate or resistant to a specific antibiotic according to CLSI guidelines [12]. For the purpose of this study intermediate sensitivity was considered as sensitive.", "The presence of four enteric viruses: rotavirus, norovirus, adenovirus and astrovirus were detected by commercially available enzyme linked immunosorbent assay (ELISA) according to manufacturer's instructions (Dako Ltd., Ely, United Kingdom). The tests detected specific antigens for group A rotaviruses, norovirus, adenoviruses type 40 and 41 and astroviruses, respectively.", "The presence of C. parvum and G. lamblia antigens was tested from frozen stool samples using ImmunoCard STAT! Rapid Assay (Meridian Bioscience, Belgium). This test simultaneously detects and distinguishes between C. parvum and G. lamblia antigens in aqueous extracts of patient stool specimens.", "Ethical clearance was obtained from the Institutional Review Board of MUHAS, Dar es Salaam. Informed verbal consent was obtained from parents/guardians of the children before enrolment. Children were treated according to IMCI guidelines [8].", "The Statistical Package for the Social Sciences (SPSS for windows, version 10.0) was used for statistical analysis. Assuming the data follows a normal distribution, comparison of proportions and statistical significance were tested by using the Chi-square test. A p value less than 0.05 was considered statistically significant.", "A total of 280 children (172 boys and 108 girls) aged 0-60 months with diarrhoea were enrolled. Of these 148(52.9%) were from Amana Hospital followed by MNH 47(16.8%), Temeke Hospital 47(16.8%) and Mwananyamala Hospital 38 (13.6%) (Table 1).\nSocial demographic data, breastfeeding and nutritional status and type of diarrhoea of the study group\nTwo hundred and thirty five (83.9%) children presented mainly with acute watery followed by persistent diarrhoea in 27 (9.6%), while 28 (6.4%) had dysentery. Of the 51 children aged below six months, only six (11.8%) were exclusively breast-fed. About 97% of the children aged 7 to 12 months were still being breast-fed. Seventy nine children (28.2%) were malnourished.\nAt least one enteric pathogen was detected in 188 (67.1%) patients. Of these 93 (33.2%), 87 (32.2%) and 52 (19.2%) were bacteria, viruses and parasites respectively. Mixed infections with up to four different pathogens were detected in 56 (20.7%) of all cases. Most (75.9%) of the co-infections were with two pathogens, followed by three pathogens (20.4%) and a few (3.7%) were with four pathogens.\nThe prevalence of different enteric pathogens is shown in Table 2. DEC was the most common pathogen, isolated in 64 children (22.9%), followed by C. parvum detected in 51 (18.9%), rotavirus in 49 (18.1%) and norovirus in 37 (13.7%). V. cholerae and Shigella spp accounted for 5.7% and 5.4% of all cases of diarrhoea respectively. Of the DEC strains detected, 41(64.1%), 13(20.3%) and 10(15.6%) were EAEC, EPEC and ETEC respectively. All ETEC strains harboured only the stable toxin. There were no EIEC or EHEC detected in this group.\nIsolation frequency of enteric pathogens from under fives with diarrhoea in Dar es Salaam\nAll V. cholerae isolates were sero group OI strains of Ogawa. Of the seven Salmonella isolated, four were Salmonella Typhimurium, two were Salmonella Enteritidis and one was Salmonella Typhi. Salmonella Paratyphi was not detected. Of the fifteen Shigella isolates, 10 were Shigella flexneri and five were Shigella dysenteriae. Shigella sonnei and Shigella boydii were not detected.\nTable 3 shows that bacteria, viruses, parasites and mixed infection were highly prevalent in young infants aged 0-6 months (25.0%-56.8%) and 7-12 months (26.1%-52.3%) (p > 0.05), the prevalence of all pathogens decreasing with increasing age. When stratifying bacterial pathogens with age groups, DEC, was significantly higher (39.2%) in the age group of 0-6 months than the rest of the age groups (p = 0.01) while V. cholerae was significantly higher in older children >25 months (p = 0.01).\nAssociation between risk factors and enteric pathogens isolated among under fives with diarrhoea in Dar es Salaam\nOverall, 86.2% - 97.4% of all the pathogens caused acute watery diarrhoea, with the exception of Shigella species which accounted for 40% of cases of dysentery (p < 0.05). Thirty-five (85.4%) of the 41 children with EAEC presented with acute watery diarrhoea, while five (12.2%) and one (2.4%) presented with persistent diarrhoea and dysentery respectively. Among children harbouring EPEC, ten (76.9%) had acute watery and three (23.1%) had persistent diarrhoea. Six (60.0%), two (20.0%) and two (20.0%) of the children with ETEC had acute watery diarrhoea, persistent diarrhoea and dysentery, respectively. Among the 92 children with diarrhoea whose aetiogical agents could not be detected; 87(94.6%) and 5(5.4%) presented with acute and persistent diarrhoea respectively.\nTable 4 summarizes the antimicrobial resistance pattern of different bacterial pathogens. Overall, bacterial pathogens isolated showed no resistance to ciprofloxacin and cefotaxime except for DEC which showed 4.7% resistance to cefotaxime. Shigella spp, V. cholerae and DEC showed moderate to high rates of resistance to erythromycin, ampicillin, chloramphenicol and tetracycline (56.2-100%). V. cholerae showed full susceptibility to co-trimoxazole (100%), while DEC and Shigella spp showed high rate of resistance to co-trimoxazole; 90.6% and 93.3% respectively.\nAntimicrobial resistance pattern of bacterial enteric pathogens isolated from under fives with diarrhoea in Dar es Salaam", "In this study the prevalence of diarrhoea with an identified aetiology was 67.1% and consistent with previous findings from Ifakara, Tanzania [7]. Pathogens were not detected in 92(32.9%) of the children. This could be partly due to potential enteric pathogens such as helminths, Campylobacter spp, Yersinia enterocolitica and Aeromonas spp which were not searched for, or because of the fact that some of the children had a history of antibiotic use prior to admission.\nWe found bacterial pathogens in 33.3%, enteric viruses in 32.2% and the protozoas in 19.2% of patients. However when age stratification was done, the main cause of diarrohea in children aged 0-6 months were bacteria, predominantly DEC, while diarrhoea in children aged 7-12 months was more often due to viruses, mainly rotavirus and norovirus. Vibrio cholerae was isolated mainly in children aged above two. This age related pattern of pathogens is consistent with reports from studies conducted in other developing countries [2] and should be taken into account when considering appropriate management of childhood diarrhoea in Dar es Salaam.\nFifty six children (20.7%) had multiple infections with up to four associated pathogens, a figure comparable with findings in other developing countries [13,14]. Dual infections raise the question of whether a single pathogen is responsible for illness, or whether several pathogens act in synergy. Further studies must be performed in order to obtain a better understanding of these infections. The predominance of DEC (64.1%) among all bacterial isolates is consistent with previous reports from Tanzania and other developing countries [7,15], underlining their importance as the main bacterial causes of diarrhoea in children aged less than five years and the need to type them in order to know the types that cause diarrhoea in our setting.\nSurprisingly, viruses were detected in four children with dysentery. One them had mixed infection with ETEC strain of DEC while the other three had no bacteria detected. Viruses do not normally cause bloody diarrhoea, the possible reason for this finding could be that enteric bacteria such as Shigella spp were not isolated because all the four children had history of taking antibiotics prior to the study or dysentery could have been caused by other enteric pathogens such Entamoeba histolytica which was not searched for in this study.\nOverall, Shigella spp were detected in 5.7% of cases of diarrhoea but the frequency was significantly higher (40%) in children with bloody diarrhoea. The predominance of S. flexneri (66.7% of Shigella isolates) found in this study is consistent with the finding in Ifakara Tanzania [7].\nThe prevalence of Salmonella spp found in this study (5.7%) falls within the reported range of 1-5% of gastroenteritis cases in most developing countries [8] and the serotypes were mainly S. Enteritidis (28.6%) and S. Typhimurium (57.1%).\nThe high prevalence V. cholerae in the present study (5.7%) was due to an outbreak of cholera in the city during the time of study.\nThe prevalence of cryptosporidiosis reported in this study (18.9%) is higher than the 8.5% reported previously in Dar es Salaam [6]. The difference could be due to difference in HIV prevalence during the two study periods given the association between cryptosporidiosis and HIV infection [6]. However we did not screen for HIV infection in the current study. The difference could also be due to different methods used for detection of cryptosporidiosis. We found G. lamblia in only a minority of diarrhoea cases (1.9%), probably due to their limited role in childhood diarrhoea among children of Dar es Salaam.\nWe found no relationship between the enteric pathogens detected and the lack of exclusive breast feeding. This could be due to the fact that only a few children (11.2%) were exclusively breast-fed.\nOur findings show that most bacterial pathogens isolated were sensitive to ciprofloxacin, cefotaxime, gentamicin, amoxicillin-clavulanic acid and cephalothin, probably due to their seldom use [16-19]. On the contrary, high rate of resistance was seen in commonly prescribed antibiotics such as ampicillin, tetracycline, including co-trimoxazole which are currently recommended for empirical treatment [8]. The high rate of resistance towards these antibiotics may be due to their overuse because they are readily available over the counter. The 4.7% prevalence of resistance of DEC to cefotaxime is noteworthy because third generation cephalosporin resistance is usually caused by expression of extended-spectrum beta-lactamase (ESBL) enzymes which may be the case in these strains although specific detection of ESBL production was not performed. ESBL producing-strains have been reported previously in septicaemic as well as Intensive Care Unit patients from Muhimbili National Hospital [20,21].\nThe antimicrobial resistance of V. cholerae observed in this study when compared with those of Urrasa et al [17] does show an increase in rate the of resistance to ampicillin (17% and 53.4% in 1997 and 1999 respectively versus 75% in the present study), erythromycin (18.1% and 54% in 1997 and 1999 respectively versus 75%) and tetracycline (6.45 and 41.1%% in 1997 and 1999 respectively versus 93.7%). The study also found a decrease in the rate of resistance to co-trimoxazole (96.8% and 96.6% in 1997 and 1999 respectively versus 0% in the present study). The low rate of resistance of V. cholerae to co-trimoxazole supports the current national policy of treating cholera with co-trimoxazole. However there is a need for continuous surveillance to monitor and track potential development of resistance.", "During the dry season, acute watery diarrhoea is the most common type of diarrhoea in children under five years in Dar es Salaam and is predominantly due to DEC, C. parvum, rotaviruses and noroviruses. High rate of resistance of bacteria to co-trimoxazole and erythromycin which are currently recommended for empiric treatment of diarrhoea and other antimicrobial agents, calls for continued antibiotic surveillance.", "The authors declare that they have no competing interests.", "SJM was the principal investigator, who conceived and designed study and was responsible for collection of specimens and clinical information as well as data analysis. Laboratory investigations were performed by SJM under the guidance of NG and HM. MIM, JK, HM, HM SYM and NL assisted in the development of the research proposal, data analysis and preparation of the manuscript. All authors have read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2431/11/19/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study design and settings", "Interviews", "Weight measurements", "Determination of nutritional status", "Collection and transportation of stool", "Isolation and identification of bacteria pathogens", "Antimicrobial susceptibility testing of bacterial isolates", "Detection of viruses", "Detection of intestinal protozoa Cryptosporidium parvum and Giardia lamblia", "Ethical considerations", "Data analysis", "Results", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Infective diarrhoea is one of the leading cause of morbidity and mortality among children under five years in the developing world [1] and can be caused by a wide range of viruses, bacteria, or parasites [2,3]. The prevalence of the different enteric pathogens varies with the geographical area [2].\nThere are relatively few studies on this aspect that have been conducted in Tanzania. Most of these studies have focused on few aetiological agents of diarrhoea and were conducted in late eighties and early nineties [4-6]. There is one recent study conducted in Ifakara, Tanzania in 2004, which shows the prevalence of different aetiological agents of diarrhoea, i.e. viruses, bacteria and parasites among children with diarrhoea. In that study diarrhoeagenic Escherichia coli (DEC) was the predominant enteropathogen. Moreover, Shigella spp. and rotavirus were more prevalent in the dry season than in the rainy season and Giardia lamblia was more prevalent in the rainy season [7].\nFor quite some time children with diarrhoea in Tanzania have been and are being treated empirically with erythromycin and co-trimoxazole according to the Integrated Management of Childhood Illness (IMCI) guidelines [8]. However current information regarding antimicrobial susceptibility pattern of bacteria causing diarrhoea in children is limited and thus it is uncertain whether the recommended antibiotics are still effective.\nThe present study aimed at detection of enteric pathogens in children with diarrhoea to provide an update on the spectrum and age specific causes of diarrhoea. The study also determined the antimicrobial resistance pattern of bacterial pathogens.", "[SUBTITLE] Study design and settings [SUBSECTION] This cross-sectional study was conducted in Dar es Salaam, Tanzania between December 2005 and February 2006. Participants were children ≤5 years, who during the study period, were admitted due to diarrhoea at Muhimbili National Referral Hospital (MNH), Amana, Mwananyamala and Temeke Municipal Hospitals in Dar es Salaam, Tanzania. Enrolment was subjected to obtaining an informed verbal consent from parent or guardian who accompanied the child.\nThis cross-sectional study was conducted in Dar es Salaam, Tanzania between December 2005 and February 2006. Participants were children ≤5 years, who during the study period, were admitted due to diarrhoea at Muhimbili National Referral Hospital (MNH), Amana, Mwananyamala and Temeke Municipal Hospitals in Dar es Salaam, Tanzania. Enrolment was subjected to obtaining an informed verbal consent from parent or guardian who accompanied the child.\n[SUBTITLE] Interviews [SUBSECTION] A structured questionnaire was used to obtain information of the children from the parents/guardians regarding age, sex, duration and description of the stool (watery, mucoid, or bloody) and the use of antibiotics prior to hospitalization. The definition and forms of diarrhoea were according to WHO guidelines [8].\nA structured questionnaire was used to obtain information of the children from the parents/guardians regarding age, sex, duration and description of the stool (watery, mucoid, or bloody) and the use of antibiotics prior to hospitalization. The definition and forms of diarrhoea were according to WHO guidelines [8].\n[SUBTITLE] Weight measurements [SUBSECTION] On admission infants less than two years of age were weighed using a 25 kg Salter hanging scales (CMS Weighing equipment, High Holborn, London, United Kingdom). Children over two years were weighed on scales calibrated before each session. Weight of children was recorded to the nearest kilogram.\nOn admission infants less than two years of age were weighed using a 25 kg Salter hanging scales (CMS Weighing equipment, High Holborn, London, United Kingdom). Children over two years were weighed on scales calibrated before each session. Weight of children was recorded to the nearest kilogram.\n[SUBTITLE] Determination of nutritional status [SUBSECTION] Weight-for age Z-scores were calculated using EPI Info (USD, Inc., Stone Mountain, GA). According to WHO criteria children were considered to be undernourished if the Z-scores were less than -2SD[9].\nWeight-for age Z-scores were calculated using EPI Info (USD, Inc., Stone Mountain, GA). According to WHO criteria children were considered to be undernourished if the Z-scores were less than -2SD[9].\n[SUBTITLE] Collection and transportation of stool [SUBSECTION] Stool specimens were collected using wide mouthed sterile plastic containers and transported to the Microbiology laboratory at Muhimbili National Hospital within two hours of collection. A portion of specimen was stored at -20°C until further analysis at Haukeland University Hospital in Bergen, Norway.\nStool specimens were collected using wide mouthed sterile plastic containers and transported to the Microbiology laboratory at Muhimbili National Hospital within two hours of collection. A portion of specimen was stored at -20°C until further analysis at Haukeland University Hospital in Bergen, Norway.\n[SUBTITLE] Isolation and identification of bacteria pathogens [SUBSECTION] Bacterial pathogens E. coli, V. cholerae, Salmonella spp and Shigella spp. were isolated and identified by conventional methods [10]. Identification of DEC was done as previously described [11].\nBacterial pathogens E. coli, V. cholerae, Salmonella spp and Shigella spp. were isolated and identified by conventional methods [10]. Identification of DEC was done as previously described [11].\n[SUBTITLE] Antimicrobial susceptibility testing of bacterial isolates [SUBSECTION] Susceptibility testing was performed by disk diffusion method according to Clinical and Laboratory Standards Institute guidelines (CLSI) [12]. Escherichia coli ATCC 25922 was used as a reference control strain. Antibiotics tested were ampicillin 10 μg, amoxycillin and clavulanic acid (Augmentin) 20/10 μg, erythromycin 15 μg, ciprofloxacin 100 μg, gentamicin 10 μg, cefotaxime 30 μg, cephalothin 30 μg, co-trimoxazole 25 μg, chloramphenicol 30 μg and tetracycline 30 μg (Remel, Lenexa, USA).\nAccording to the size of the zones of inhibition, the organisms were classified as sensitive, intermediate or resistant to a specific antibiotic according to CLSI guidelines [12]. For the purpose of this study intermediate sensitivity was considered as sensitive.\nSusceptibility testing was performed by disk diffusion method according to Clinical and Laboratory Standards Institute guidelines (CLSI) [12]. Escherichia coli ATCC 25922 was used as a reference control strain. Antibiotics tested were ampicillin 10 μg, amoxycillin and clavulanic acid (Augmentin) 20/10 μg, erythromycin 15 μg, ciprofloxacin 100 μg, gentamicin 10 μg, cefotaxime 30 μg, cephalothin 30 μg, co-trimoxazole 25 μg, chloramphenicol 30 μg and tetracycline 30 μg (Remel, Lenexa, USA).\nAccording to the size of the zones of inhibition, the organisms were classified as sensitive, intermediate or resistant to a specific antibiotic according to CLSI guidelines [12]. For the purpose of this study intermediate sensitivity was considered as sensitive.\n[SUBTITLE] Detection of viruses [SUBSECTION] The presence of four enteric viruses: rotavirus, norovirus, adenovirus and astrovirus were detected by commercially available enzyme linked immunosorbent assay (ELISA) according to manufacturer's instructions (Dako Ltd., Ely, United Kingdom). The tests detected specific antigens for group A rotaviruses, norovirus, adenoviruses type 40 and 41 and astroviruses, respectively.\nThe presence of four enteric viruses: rotavirus, norovirus, adenovirus and astrovirus were detected by commercially available enzyme linked immunosorbent assay (ELISA) according to manufacturer's instructions (Dako Ltd., Ely, United Kingdom). The tests detected specific antigens for group A rotaviruses, norovirus, adenoviruses type 40 and 41 and astroviruses, respectively.\n[SUBTITLE] Detection of intestinal protozoa Cryptosporidium parvum and Giardia lamblia [SUBSECTION] The presence of C. parvum and G. lamblia antigens was tested from frozen stool samples using ImmunoCard STAT! Rapid Assay (Meridian Bioscience, Belgium). This test simultaneously detects and distinguishes between C. parvum and G. lamblia antigens in aqueous extracts of patient stool specimens.\nThe presence of C. parvum and G. lamblia antigens was tested from frozen stool samples using ImmunoCard STAT! Rapid Assay (Meridian Bioscience, Belgium). This test simultaneously detects and distinguishes between C. parvum and G. lamblia antigens in aqueous extracts of patient stool specimens.\n[SUBTITLE] Ethical considerations [SUBSECTION] Ethical clearance was obtained from the Institutional Review Board of MUHAS, Dar es Salaam. Informed verbal consent was obtained from parents/guardians of the children before enrolment. Children were treated according to IMCI guidelines [8].\nEthical clearance was obtained from the Institutional Review Board of MUHAS, Dar es Salaam. Informed verbal consent was obtained from parents/guardians of the children before enrolment. Children were treated according to IMCI guidelines [8].\n[SUBTITLE] Data analysis [SUBSECTION] The Statistical Package for the Social Sciences (SPSS for windows, version 10.0) was used for statistical analysis. Assuming the data follows a normal distribution, comparison of proportions and statistical significance were tested by using the Chi-square test. A p value less than 0.05 was considered statistically significant.\nThe Statistical Package for the Social Sciences (SPSS for windows, version 10.0) was used for statistical analysis. Assuming the data follows a normal distribution, comparison of proportions and statistical significance were tested by using the Chi-square test. A p value less than 0.05 was considered statistically significant.", "This cross-sectional study was conducted in Dar es Salaam, Tanzania between December 2005 and February 2006. Participants were children ≤5 years, who during the study period, were admitted due to diarrhoea at Muhimbili National Referral Hospital (MNH), Amana, Mwananyamala and Temeke Municipal Hospitals in Dar es Salaam, Tanzania. Enrolment was subjected to obtaining an informed verbal consent from parent or guardian who accompanied the child.", "A structured questionnaire was used to obtain information of the children from the parents/guardians regarding age, sex, duration and description of the stool (watery, mucoid, or bloody) and the use of antibiotics prior to hospitalization. The definition and forms of diarrhoea were according to WHO guidelines [8].", "On admission infants less than two years of age were weighed using a 25 kg Salter hanging scales (CMS Weighing equipment, High Holborn, London, United Kingdom). Children over two years were weighed on scales calibrated before each session. Weight of children was recorded to the nearest kilogram.", "Weight-for age Z-scores were calculated using EPI Info (USD, Inc., Stone Mountain, GA). According to WHO criteria children were considered to be undernourished if the Z-scores were less than -2SD[9].", "Stool specimens were collected using wide mouthed sterile plastic containers and transported to the Microbiology laboratory at Muhimbili National Hospital within two hours of collection. A portion of specimen was stored at -20°C until further analysis at Haukeland University Hospital in Bergen, Norway.", "Bacterial pathogens E. coli, V. cholerae, Salmonella spp and Shigella spp. were isolated and identified by conventional methods [10]. Identification of DEC was done as previously described [11].", "Susceptibility testing was performed by disk diffusion method according to Clinical and Laboratory Standards Institute guidelines (CLSI) [12]. Escherichia coli ATCC 25922 was used as a reference control strain. Antibiotics tested were ampicillin 10 μg, amoxycillin and clavulanic acid (Augmentin) 20/10 μg, erythromycin 15 μg, ciprofloxacin 100 μg, gentamicin 10 μg, cefotaxime 30 μg, cephalothin 30 μg, co-trimoxazole 25 μg, chloramphenicol 30 μg and tetracycline 30 μg (Remel, Lenexa, USA).\nAccording to the size of the zones of inhibition, the organisms were classified as sensitive, intermediate or resistant to a specific antibiotic according to CLSI guidelines [12]. For the purpose of this study intermediate sensitivity was considered as sensitive.", "The presence of four enteric viruses: rotavirus, norovirus, adenovirus and astrovirus were detected by commercially available enzyme linked immunosorbent assay (ELISA) according to manufacturer's instructions (Dako Ltd., Ely, United Kingdom). The tests detected specific antigens for group A rotaviruses, norovirus, adenoviruses type 40 and 41 and astroviruses, respectively.", "The presence of C. parvum and G. lamblia antigens was tested from frozen stool samples using ImmunoCard STAT! Rapid Assay (Meridian Bioscience, Belgium). This test simultaneously detects and distinguishes between C. parvum and G. lamblia antigens in aqueous extracts of patient stool specimens.", "Ethical clearance was obtained from the Institutional Review Board of MUHAS, Dar es Salaam. Informed verbal consent was obtained from parents/guardians of the children before enrolment. Children were treated according to IMCI guidelines [8].", "The Statistical Package for the Social Sciences (SPSS for windows, version 10.0) was used for statistical analysis. Assuming the data follows a normal distribution, comparison of proportions and statistical significance were tested by using the Chi-square test. A p value less than 0.05 was considered statistically significant.", "A total of 280 children (172 boys and 108 girls) aged 0-60 months with diarrhoea were enrolled. Of these 148(52.9%) were from Amana Hospital followed by MNH 47(16.8%), Temeke Hospital 47(16.8%) and Mwananyamala Hospital 38 (13.6%) (Table 1).\nSocial demographic data, breastfeeding and nutritional status and type of diarrhoea of the study group\nTwo hundred and thirty five (83.9%) children presented mainly with acute watery followed by persistent diarrhoea in 27 (9.6%), while 28 (6.4%) had dysentery. Of the 51 children aged below six months, only six (11.8%) were exclusively breast-fed. About 97% of the children aged 7 to 12 months were still being breast-fed. Seventy nine children (28.2%) were malnourished.\nAt least one enteric pathogen was detected in 188 (67.1%) patients. Of these 93 (33.2%), 87 (32.2%) and 52 (19.2%) were bacteria, viruses and parasites respectively. Mixed infections with up to four different pathogens were detected in 56 (20.7%) of all cases. Most (75.9%) of the co-infections were with two pathogens, followed by three pathogens (20.4%) and a few (3.7%) were with four pathogens.\nThe prevalence of different enteric pathogens is shown in Table 2. DEC was the most common pathogen, isolated in 64 children (22.9%), followed by C. parvum detected in 51 (18.9%), rotavirus in 49 (18.1%) and norovirus in 37 (13.7%). V. cholerae and Shigella spp accounted for 5.7% and 5.4% of all cases of diarrhoea respectively. Of the DEC strains detected, 41(64.1%), 13(20.3%) and 10(15.6%) were EAEC, EPEC and ETEC respectively. All ETEC strains harboured only the stable toxin. There were no EIEC or EHEC detected in this group.\nIsolation frequency of enteric pathogens from under fives with diarrhoea in Dar es Salaam\nAll V. cholerae isolates were sero group OI strains of Ogawa. Of the seven Salmonella isolated, four were Salmonella Typhimurium, two were Salmonella Enteritidis and one was Salmonella Typhi. Salmonella Paratyphi was not detected. Of the fifteen Shigella isolates, 10 were Shigella flexneri and five were Shigella dysenteriae. Shigella sonnei and Shigella boydii were not detected.\nTable 3 shows that bacteria, viruses, parasites and mixed infection were highly prevalent in young infants aged 0-6 months (25.0%-56.8%) and 7-12 months (26.1%-52.3%) (p > 0.05), the prevalence of all pathogens decreasing with increasing age. When stratifying bacterial pathogens with age groups, DEC, was significantly higher (39.2%) in the age group of 0-6 months than the rest of the age groups (p = 0.01) while V. cholerae was significantly higher in older children >25 months (p = 0.01).\nAssociation between risk factors and enteric pathogens isolated among under fives with diarrhoea in Dar es Salaam\nOverall, 86.2% - 97.4% of all the pathogens caused acute watery diarrhoea, with the exception of Shigella species which accounted for 40% of cases of dysentery (p < 0.05). Thirty-five (85.4%) of the 41 children with EAEC presented with acute watery diarrhoea, while five (12.2%) and one (2.4%) presented with persistent diarrhoea and dysentery respectively. Among children harbouring EPEC, ten (76.9%) had acute watery and three (23.1%) had persistent diarrhoea. Six (60.0%), two (20.0%) and two (20.0%) of the children with ETEC had acute watery diarrhoea, persistent diarrhoea and dysentery, respectively. Among the 92 children with diarrhoea whose aetiogical agents could not be detected; 87(94.6%) and 5(5.4%) presented with acute and persistent diarrhoea respectively.\nTable 4 summarizes the antimicrobial resistance pattern of different bacterial pathogens. Overall, bacterial pathogens isolated showed no resistance to ciprofloxacin and cefotaxime except for DEC which showed 4.7% resistance to cefotaxime. Shigella spp, V. cholerae and DEC showed moderate to high rates of resistance to erythromycin, ampicillin, chloramphenicol and tetracycline (56.2-100%). V. cholerae showed full susceptibility to co-trimoxazole (100%), while DEC and Shigella spp showed high rate of resistance to co-trimoxazole; 90.6% and 93.3% respectively.\nAntimicrobial resistance pattern of bacterial enteric pathogens isolated from under fives with diarrhoea in Dar es Salaam", "In this study the prevalence of diarrhoea with an identified aetiology was 67.1% and consistent with previous findings from Ifakara, Tanzania [7]. Pathogens were not detected in 92(32.9%) of the children. This could be partly due to potential enteric pathogens such as helminths, Campylobacter spp, Yersinia enterocolitica and Aeromonas spp which were not searched for, or because of the fact that some of the children had a history of antibiotic use prior to admission.\nWe found bacterial pathogens in 33.3%, enteric viruses in 32.2% and the protozoas in 19.2% of patients. However when age stratification was done, the main cause of diarrohea in children aged 0-6 months were bacteria, predominantly DEC, while diarrhoea in children aged 7-12 months was more often due to viruses, mainly rotavirus and norovirus. Vibrio cholerae was isolated mainly in children aged above two. This age related pattern of pathogens is consistent with reports from studies conducted in other developing countries [2] and should be taken into account when considering appropriate management of childhood diarrhoea in Dar es Salaam.\nFifty six children (20.7%) had multiple infections with up to four associated pathogens, a figure comparable with findings in other developing countries [13,14]. Dual infections raise the question of whether a single pathogen is responsible for illness, or whether several pathogens act in synergy. Further studies must be performed in order to obtain a better understanding of these infections. The predominance of DEC (64.1%) among all bacterial isolates is consistent with previous reports from Tanzania and other developing countries [7,15], underlining their importance as the main bacterial causes of diarrhoea in children aged less than five years and the need to type them in order to know the types that cause diarrhoea in our setting.\nSurprisingly, viruses were detected in four children with dysentery. One them had mixed infection with ETEC strain of DEC while the other three had no bacteria detected. Viruses do not normally cause bloody diarrhoea, the possible reason for this finding could be that enteric bacteria such as Shigella spp were not isolated because all the four children had history of taking antibiotics prior to the study or dysentery could have been caused by other enteric pathogens such Entamoeba histolytica which was not searched for in this study.\nOverall, Shigella spp were detected in 5.7% of cases of diarrhoea but the frequency was significantly higher (40%) in children with bloody diarrhoea. The predominance of S. flexneri (66.7% of Shigella isolates) found in this study is consistent with the finding in Ifakara Tanzania [7].\nThe prevalence of Salmonella spp found in this study (5.7%) falls within the reported range of 1-5% of gastroenteritis cases in most developing countries [8] and the serotypes were mainly S. Enteritidis (28.6%) and S. Typhimurium (57.1%).\nThe high prevalence V. cholerae in the present study (5.7%) was due to an outbreak of cholera in the city during the time of study.\nThe prevalence of cryptosporidiosis reported in this study (18.9%) is higher than the 8.5% reported previously in Dar es Salaam [6]. The difference could be due to difference in HIV prevalence during the two study periods given the association between cryptosporidiosis and HIV infection [6]. However we did not screen for HIV infection in the current study. The difference could also be due to different methods used for detection of cryptosporidiosis. We found G. lamblia in only a minority of diarrhoea cases (1.9%), probably due to their limited role in childhood diarrhoea among children of Dar es Salaam.\nWe found no relationship between the enteric pathogens detected and the lack of exclusive breast feeding. This could be due to the fact that only a few children (11.2%) were exclusively breast-fed.\nOur findings show that most bacterial pathogens isolated were sensitive to ciprofloxacin, cefotaxime, gentamicin, amoxicillin-clavulanic acid and cephalothin, probably due to their seldom use [16-19]. On the contrary, high rate of resistance was seen in commonly prescribed antibiotics such as ampicillin, tetracycline, including co-trimoxazole which are currently recommended for empirical treatment [8]. The high rate of resistance towards these antibiotics may be due to their overuse because they are readily available over the counter. The 4.7% prevalence of resistance of DEC to cefotaxime is noteworthy because third generation cephalosporin resistance is usually caused by expression of extended-spectrum beta-lactamase (ESBL) enzymes which may be the case in these strains although specific detection of ESBL production was not performed. ESBL producing-strains have been reported previously in septicaemic as well as Intensive Care Unit patients from Muhimbili National Hospital [20,21].\nThe antimicrobial resistance of V. cholerae observed in this study when compared with those of Urrasa et al [17] does show an increase in rate the of resistance to ampicillin (17% and 53.4% in 1997 and 1999 respectively versus 75% in the present study), erythromycin (18.1% and 54% in 1997 and 1999 respectively versus 75%) and tetracycline (6.45 and 41.1%% in 1997 and 1999 respectively versus 93.7%). The study also found a decrease in the rate of resistance to co-trimoxazole (96.8% and 96.6% in 1997 and 1999 respectively versus 0% in the present study). The low rate of resistance of V. cholerae to co-trimoxazole supports the current national policy of treating cholera with co-trimoxazole. However there is a need for continuous surveillance to monitor and track potential development of resistance.", "During the dry season, acute watery diarrhoea is the most common type of diarrhoea in children under five years in Dar es Salaam and is predominantly due to DEC, C. parvum, rotaviruses and noroviruses. High rate of resistance of bacteria to co-trimoxazole and erythromycin which are currently recommended for empiric treatment of diarrhoea and other antimicrobial agents, calls for continued antibiotic surveillance.", "The authors declare that they have no competing interests.", "SJM was the principal investigator, who conceived and designed study and was responsible for collection of specimens and clinical information as well as data analysis. Laboratory investigations were performed by SJM under the guidance of NG and HM. MIM, JK, HM, HM SYM and NL assisted in the development of the research proposal, data analysis and preparation of the manuscript. All authors have read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2431/11/19/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adolescent", "Adult", "Burnout, Professional", "Female", "Health Surveys", "Hearing Loss", "Humans", "Male", "Middle Aged", "Occupational Exposure", "Quality of Life", "Self Concept", "Sleep", "Stress, Psychological", "Sweden", "Tinnitus", "Young Adult" ]

[ "Background", "Population", "Questionnaire", "Statistical analyses", "Ethical approval", "Results", "Work-related stressors/threats", "Self-rated health", "Long-term illness, inconvenience after an accident, any handicap or other weakness", "Sleep quality", "Burnout", "Long-lasting stress", "Performance-based self-esteem", "Discussion", "Gender issues", "Occupational consequences", "Self-rated health", "Sleep quality", "Burnout and symptoms of long-lasting stress", "Performance-based self-esteem", "Strengths and limitations", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Hearing problems are the most common sensory deficit in human populations, with hearing loss alone affecting more than 250 million people worldwide [1]. In 2002, the WHO estimated hearing loss to be the 13th most frequent burden of disease in medium- and high-income countries, and it is projected to become among the top ten by the year 2030 [2]. Most epidemiological studies report prevalence figures of 10-15% for both hearing loss [1,3] and tinnitus [4,5] respectively. These figures may however not be inferable to hearing problems in a broader sense, since the prevalence is most often only calculated for either hearing loss or tinnitus. Consequently, a large Swedish study has recently addressed both issues concurrently, finding that approximately 32% of the Swedish working population suffer from either hearing complaints, tinnitus or both [6]. The study was also the first to provide evidence for a negative association between self-rated socioeconomic status (SES) and prevalence of hearing problems. It is well established that lower SES is associated with higher stress levels [7-9], so this finding may indirectly imply a relationship between stress and hearing problems.\nWhile the deleterious effects of mechanical stress (i.e. noise) on hearing have been studied extensively in both animal models [10] and human populations [11,12], the notion of emotional stress as a modulator of the auditory system is rather novel. A complex set of pathways of the stress response have been identified, involving both sympathetic stimulation of adrenergic α-receptors within the cochlea [13,14], as well as neuro-endocrine responses primarily aimed at engaging the hypothalamic-pituitary-adrenal (HPA) axis [15]. Current research suggests that acute stress may protect the cochlea [16-19], whereas chronic stress exposure seems to be harmful to hearing [20]. The importance of a normal functioning of the HPA-axis for healthy hearing is supported by clinical studies showing that patients with tinnitus display signs of an impaired HPA-axis along with a higher degree of perceived stress, compared to non-tinnitus patients [21-23]. Additionally, Hasson et al. [24] recently found that symphony orchestra musicians with hearing problems exhibited lower heart rate variability (high frequency power), indicating an impaired ability to \"unwind\" and activate the parasympathetic system.\nHuman and animal studies are providing more evidence of an association between stress and hearing problems. It also needs be considered that long-term hearing problems can be stressful. Regarding the high prevalence of hearing problems in human populations and the negative future projections of the WHO, further investigations of the relationship between stress and hearing problems are warranted. Therefore, the aim of this study is to assess the prevalence of two common hearing problems, i.e. hearing complaints and tinnitus, in relation to different work- and health-related stressors. More specifically, we will study if prevalence differs with poor self-rated health, poorer sleep, higher burnout scores (work-related), and more symptoms of long-term stress and higher levels of performance-based self-esteem.", "The Swedish Work Environment Survey (SWES) is conducted biennially by Statistics Sweden (SCB) and consists of subsamples of gainfully employed people, aged 16-64 years, from the Labor Force Survey (LSF). These individuals were first sampled into the LFS through stratification by county of birth, sex, citizenship, and inferred employment status. The respondents to SWES 2003 and 2005 were invited to enroll in the Swedish Longitudinal Occupational Survey of Health (SLOSH) [25], which was initiated by the Stress Research Institute in 2006. The second data collection was conducted in April 2008 by Statistics Sweden, on behalf of the Stress Research Institute at Stockholm University. A total of 18,734 individuals were mailed self-completion questionnaires in 2008, out of which 9,756 (52%) working individuals responded. The total response rate of the study was however 11,441 (61%), including non-working participants (not analyzed in the present study). More detailed information about the cohort, response rate and characteristics of responders vs. non-responders has been published elsewhere [6]. There was no difference between responders and non-responders with regard to county of birth and citizenship.", "Apart from socioeconomic status and demographic factors, it also included approximately 120 questions about psychosocial and physical work-environment, lifestyle, as well as physical and mental health.\nHearing problems were assessed with three questions. Tinnitus. Have you during the most recent time experienced sound in any of the ears, without there being an external source (so-called tinnitus) lasting more than five minutes? (No, Yes sometimes, Yes often, Yes all the time). Tinnitus severity. How much do you feel that the tinnitus sounds worry, bother or upset you? (Not at all, A little, Moderately, Severely). The questions about tinnitus were adapted from Davis [26] and Palmer et al. [27]. Hearing complaints. How difficult is it for you to (without hearing aid) hear what is said in a conversation between several persons? (Not difficult at all, Not very difficult, Quite difficult, Very difficult). In this study, hearing complaints reflects difficulties in communicating. The questions about hearing complaints were derived and adapted from Statistics Sweden and have been used in population studies for several years.\nA new variable, \"hearing problems\", was computed based on the existence or non-existence of either tinnitus or hearing complaints or both. This consequently yielded three groups; those without hearing problems, those with either tinnitus or hearing complaints or those suffering from both. The cut-off for tinnitus was \"yes, sometimes\" or more often, and for hearing complaints \"quite difficult\" or \"very difficult\".\nWork-related stressors/threats. Risks of being moved to another work/job against ones will, threats of getting fired were derived from the Swedish Labor Force Survey (LFS). Threats of bankruptcy were constructed for SLOSH 2008 to assess a threat particularly important for self-employed, a group who contacted the research group and expressed feelings of neglect in the SLOSH 2006 survey. The question was formulated: \"Are you subjected to any of the following risks or threats in your work?\" Response alternatives were yes/no.\nSelf-rated health was assessed with the single item \"How would you rate your general state of health?\" This question has been widely used in research [28-30] and the respondents answered on a Likert scale from ranging from 1 (very poor) to 5 (very good). Since only few participants had very poor SRH the categories quite poor and very poor were merged in the analyses.\nLong-term illness, inconvenience after an accident, any handicap or other weakness. One question was asked about long-term illness, inconvenience after an accident or other weakness: \"Do you have any prolonged sickness, accident-related complaints, a disability or other weakness?\" This question was derived from the WOLFF (WOrk, Lipids and Fibrinogen-follow-up) [31,32] questionnaire and response alternatives were yes/no.\nSleep quality. Sleep quality was assessed by the single item: \"How is your sleep quality in general?\" This item was derived from the Karolinska Sleep Questionnaire [33]. Response ranged from 1 (very poor) to 5 (very good) on a 5-graded Likert scale.\nBurnout was assessed with the Maslach Burnout Inventory general survey (MBI-GS) using the emotional exhaustion subscale [34]. The scale consists of five items, derived from the Maslach Burnout Inventory human services survey (MBI-HSS) in unmodified form. Scorings reach from 1 (every day) to 6 (a few times a year or less/never). Cronbach's alpha and stability for the subscale have been reported to be satisfactory. Strong support for the construct validity of the Swedish translation of the MBI-HSS has been found [35]. The index was calculated on the basis that 4 out of 5 items had to be answered in order be included in the index.\nLong lasting stress (LLS) was assessed with 11 items reflecting stress arousal symptoms but not stress reactions. The participants were asked how they felt during the last three months with regard to both physiological (e.g. \"I sweat easily even though I do not exert myself physically\") and cognitive-behavioral symptoms (e.g. \"I have worrying thoughts\"; \"I often feel tense\"). The four response alternatives reached from \"Not at all\" to \"Nearly all the time\". The scale was introduced in the 2008 SLOSH questionnaire and is currently being validated (data not yet published). A factor analysis yielded one factor of interest. This factor included 7 of the 11 items and only the cognitive-behavioral symptoms. Factor loadings ranged from .675 ± 798 and a Chronbach's α of .863. The 7 included items were (including factor loading, FL):\nA) I have days when I feel geared up all the time (FL = .675). B) I have days when I feel very pressured, on the verge of what I can handle (FL = .737). C) I find it hard to relax during my leisure time (FL = .787). D) I often feel tense (FL = .798). E) I often have disturbing thoughts (FL = .768). F) I often feel restless (FL = .720). G) I do not feel rested after taking it easy for a few days (FL = .699). The correlation between LLS and MBI-GS is r = .64, p < .0001, two-tailed. The index was calculated on the basis that 6 out of 7 items had to be answered in order be included in the index.\nPerformance based self-esteem [36,37] was assessed with four items (e.g. \"At times, I have to be better than others to be good enough myself\") with five response alternatives with the end-point labels \"Fully disagree\" to \"Fully agree\". Cronbach's alpha is between .85 and .89 and the one-year stability has been found to be satisfying. The index was calculated on the basis that 3 out of 4 items had to be answered in order be included in the index.", "The programs SPSS 18.0 and SAS 9.2 were used for statistical analyses. Prevalence was calculated via frequency plots and crosstabs were used for calculation of χ, Kendall's tau-b, and specific prevalence within different groups. Kendall's tau-b is a correlation analysis that illustrates the direction and magnitude of association between two variables. Multivariate analyses, proportional odds model (also called ordered logistic regression), were used to calculate possible interacting or confounding effects of age, gender and SES. Comparisons were made between those having no hearing problems compared to those with either tinnitus or hearing loss or both tinnitus and hearing loss. Statistical significance was set at p < 0.05 level.", "The regional ethics committee in Stockholm approved the research project and all participants gave their informed consent to participate.", "In a previous analysis of the present data it was shown that 31% of the working population report either hearing complaints or tinnitus (i.e. 25%) or both (i.e. 6%) and the prevalence of hearing problems increased with age, was higher among men and in persons with lower self-rated SES, and co-varied with exposure to noise at work [6]. In light of this background we now present the association between work- and health-related stressors and hearing problems.\nOverall, the results describe an association between hearing problems and work- and health-related stressors. All the results were controlled for possible confounding effect of age, gender or SES with multivariate analyses. These analyses showed no confounding effect of these variables. The demographics of the population in the present study are as follows: 4,462 (46%) men and 5,294 (54%) women. The mean age was 48.6 (± 10.8) for men and 48.2 (± 10.5) for women. Age distribution was: under 40 years 1,148 (26%) men and 1,314 (25%) women; 41 ± 51 years 1,202 (27%) men and 1,520 (29%) women; 51-60 years 1,415 (32%) men and 1,754 (33%) women; 60 years or older 697 (26%) men and 706 (13%) women. Marital status: married 2,491 (56%) men and 2,946 (56%) women; unmarried 1,494 (34%) men and 1,516 (29%) women; divorced 444 (10%) men and 726 (14%) women; widow 33 (1%) and 106 (2%) women. With regard to highest completed educational level, 1,016 (10%) had no gymnasium, 4,510 (46%) had gymnasium, 619 (6%) had undergraduate studies of two years or less, 3,472 (36%) had undergraduate studies of three years or more and 134 (1%) hade post graduate studies.\n[SUBTITLE] Work-related stressors/threats [SUBSECTION] Table 1 illustrates the prevalence of hearing problems in relation to different work-related stressors. There was a statistically significant difference in the prevalence of hearing problems between those who were and were not exposed to work-related stressors or threats such as risk of being moved to another work/job against ones will (χ2 = 54.704df = 2, p < 0.0001) and threats of getting fired (χ2 = 27.095df = 2, p < 0.0001). There was however no statistically significant difference between those who were exposed to threats of bankruptcy compared to those who were not.\nPrevalence of hearing problems in relation to different work-related stressors.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems when exposed to a threat of being moved to another job against ones will (yes vs. no) were 1.39 for men (p < 0.001) compared to the adjusted odds ratios which were 1.43 (p < 0.001). For women, the corresponding values were 1.7 (p < 0.001) and when adjusted the value was 1.74 (p < 0.001).\nTable 1 illustrates the prevalence of hearing problems in relation to different work-related stressors. There was a statistically significant difference in the prevalence of hearing problems between those who were and were not exposed to work-related stressors or threats such as risk of being moved to another work/job against ones will (χ2 = 54.704df = 2, p < 0.0001) and threats of getting fired (χ2 = 27.095df = 2, p < 0.0001). There was however no statistically significant difference between those who were exposed to threats of bankruptcy compared to those who were not.\nPrevalence of hearing problems in relation to different work-related stressors.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems when exposed to a threat of being moved to another job against ones will (yes vs. no) were 1.39 for men (p < 0.001) compared to the adjusted odds ratios which were 1.43 (p < 0.001). For women, the corresponding values were 1.7 (p < 0.001) and when adjusted the value was 1.74 (p < 0.001).\n[SUBTITLE] Self-rated health [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of self-rated health (for all: χ2 = 262.522df = 6, p < 0.0001, for women χ2 = 141.535df = 6, p < 0.0001, for men: χ2 = 123.306df = 6, p < 0.0001). Figure 1 clearly demonstrates that poorer SRH is associated with a higher prevalence of hearing problems. The association is negative (for all: Kendall's τ-b = -0.139 p < 0.0001, for women: Kendall's τ-b = -0.142 p < 0.0001, for men: Kendall's τ-b = -0.135 p < 0.0001) and indicates increasing prevalence of hearing problems among those with poorer health.\nPrevalence of hearing problems in percent in relation to different ratings of SRH. The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. The unadjusted odds ratio of having hearing problems when reporting poor vs. good SRH were 3.89 for men (p < 0.001) compared to the adjusted odds ratios which were 3.29 (p < 0.001). For women, the corresponding values were 3.81 (p < 0.001) and when adjusted the value was 3.49 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of self-rated health (for all: χ2 = 262.522df = 6, p < 0.0001, for women χ2 = 141.535df = 6, p < 0.0001, for men: χ2 = 123.306df = 6, p < 0.0001). Figure 1 clearly demonstrates that poorer SRH is associated with a higher prevalence of hearing problems. The association is negative (for all: Kendall's τ-b = -0.139 p < 0.0001, for women: Kendall's τ-b = -0.142 p < 0.0001, for men: Kendall's τ-b = -0.135 p < 0.0001) and indicates increasing prevalence of hearing problems among those with poorer health.\nPrevalence of hearing problems in percent in relation to different ratings of SRH. The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. The unadjusted odds ratio of having hearing problems when reporting poor vs. good SRH were 3.89 for men (p < 0.001) compared to the adjusted odds ratios which were 3.29 (p < 0.001). For women, the corresponding values were 3.81 (p < 0.001) and when adjusted the value was 3.49 (p < 0.001).\n[SUBTITLE] Long-term illness, inconvenience after an accident, any handicap or other weakness [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between individuals with different handicaps and those without (χ2 = 181.650df = 2, p < 0.0001). Those with long-term handicaps and illnesses reported more hearing problems (Table 2).\nPrevalence of hearing problems in relation to long-term illness, inconvenience after an accident, any handicap or other weakness.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems after long-term illness (yes vs. no) were 1.92 for men (p < 0.001) compared to the adjusted odds ratios which were 1.76 (p < 0.001). For women, the corresponding values were 1.72 (p < 0.001) and when adjusted the value was 1.64 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between individuals with different handicaps and those without (χ2 = 181.650df = 2, p < 0.0001). Those with long-term handicaps and illnesses reported more hearing problems (Table 2).\nPrevalence of hearing problems in relation to long-term illness, inconvenience after an accident, any handicap or other weakness.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems after long-term illness (yes vs. no) were 1.92 for men (p < 0.001) compared to the adjusted odds ratios which were 1.76 (p < 0.001). For women, the corresponding values were 1.72 (p < 0.001) and when adjusted the value was 1.64 (p < 0.001).\n[SUBTITLE] Sleep quality [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of sleep quality (for all: χ2 = 169.875df = 8, p < 0.0001, for women χ2 = 112.509df = 8, p < 0.0001, for men: χ2 = 75.068df = 8, p < 0.0001). The association was negative (for all: Kendall's τ-b = -0.116 p < 0.0001, for women: Kendall's τ-b = -0.126 p < 0.0001, for men: Kendall's τ-b = -0.113 p < 0.0001) and Figure 2 demonstrates that poorer sleep quality is associated with a higher prevalence of hearing problems. As tinnitus and hearing complaints may differ with regard to sleep, both of these variables were also analyzed separately. The prevalence of sleeping problems was significantly higher among those reporting tinnitus (χ2 = 126.884df = 4, p < 0.0001, Kendall's τ-b = -0.106 p < 0.0001) compared to those reporting hearing complaints (χ2 = 76.145df = 4, p < 0.0001, Kendall's τ-b = -0.081 p < 0.0001, see Table 3) and the associations were negative for both.\nPrevalence of hearing problems in percent in relation to different ratings of sleep quality. The Kendall's τ-b value is indicated by a \"τ\".\nPrevalence of sleeping problems among those with tinnitus and hearing loss respectively.\nNumber of participants and row percent in parentheses showing increasing prevalence of tinnitus with poorer sleep.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. When the unadjusted odds ratio of having hearing problems when reporting poor vs. good sleep quality were 2.74 for men (p < 0.001) compared to the adjusted odds ratios which were 2.67 (p < 0.001). For women, the corresponding values were 3.51 (p < 0.001) and when adjusted the value was 3.24 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of sleep quality (for all: χ2 = 169.875df = 8, p < 0.0001, for women χ2 = 112.509df = 8, p < 0.0001, for men: χ2 = 75.068df = 8, p < 0.0001). The association was negative (for all: Kendall's τ-b = -0.116 p < 0.0001, for women: Kendall's τ-b = -0.126 p < 0.0001, for men: Kendall's τ-b = -0.113 p < 0.0001) and Figure 2 demonstrates that poorer sleep quality is associated with a higher prevalence of hearing problems. As tinnitus and hearing complaints may differ with regard to sleep, both of these variables were also analyzed separately. The prevalence of sleeping problems was significantly higher among those reporting tinnitus (χ2 = 126.884df = 4, p < 0.0001, Kendall's τ-b = -0.106 p < 0.0001) compared to those reporting hearing complaints (χ2 = 76.145df = 4, p < 0.0001, Kendall's τ-b = -0.081 p < 0.0001, see Table 3) and the associations were negative for both.\nPrevalence of hearing problems in percent in relation to different ratings of sleep quality. The Kendall's τ-b value is indicated by a \"τ\".\nPrevalence of sleeping problems among those with tinnitus and hearing loss respectively.\nNumber of participants and row percent in parentheses showing increasing prevalence of tinnitus with poorer sleep.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. When the unadjusted odds ratio of having hearing problems when reporting poor vs. good sleep quality were 2.74 for men (p < 0.001) compared to the adjusted odds ratios which were 2.67 (p < 0.001). For women, the corresponding values were 3.51 (p < 0.001) and when adjusted the value was 3.24 (p < 0.001).\n[SUBTITLE] Burnout [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between those with higher burnout scores compared to those with lower scores (χ2 = 214.473df = 6, p < 0.0001, for women χ2 = 159.205df = 6, p < 0.0001, for men: χ2 = 98.935df = 6, p < 0.0001). Hearing problems were significantly more prevalent among those with higher burnout scores. Multivariate analyses showed no age, gender or SES related differences in prevalence increases with increasing burnout scores. The association was positive (for all: Kendall's τ-b = 0.129 p < 0.0001, for women: Kendall's τ-b = 0.150 p < 0.0001, for men: Kendall's τ-b = 0.129 p < 0.0001) and Figure 3 demonstrates that higher burnout scores are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to different burnout scores (higher quartiles indicate more severe burnout symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For example, the unadjusted odds ratio of having hearing problems when being in the highest vs. lowest burnout quartile were 2.36 for men (p < 0.001) compared to the adjusted odds ratios which were 2.63 (p < 0.001). For women, the corresponding values were 2.80 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between those with higher burnout scores compared to those with lower scores (χ2 = 214.473df = 6, p < 0.0001, for women χ2 = 159.205df = 6, p < 0.0001, for men: χ2 = 98.935df = 6, p < 0.0001). Hearing problems were significantly more prevalent among those with higher burnout scores. Multivariate analyses showed no age, gender or SES related differences in prevalence increases with increasing burnout scores. The association was positive (for all: Kendall's τ-b = 0.129 p < 0.0001, for women: Kendall's τ-b = 0.150 p < 0.0001, for men: Kendall's τ-b = 0.129 p < 0.0001) and Figure 3 demonstrates that higher burnout scores are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to different burnout scores (higher quartiles indicate more severe burnout symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For example, the unadjusted odds ratio of having hearing problems when being in the highest vs. lowest burnout quartile were 2.36 for men (p < 0.001) compared to the adjusted odds ratios which were 2.63 (p < 0.001). For women, the corresponding values were 2.80 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).\n[SUBTITLE] Long-lasting stress [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between those with more symptoms of long-lasting stress scores compared to those with less (χ2 = 196.855df = 6, p < 0.0001, for women χ2 = 145.608df = 6, p < 0.0001, for men: χ2 = 90.613df = 6, p < 0.0001). Hearing problems were more prevalent among those with more symptoms of long-lasting stress. Similarly to the pattern for burnout, the prevalence increase was higher for women than for men, even if it was less pronounced for this variable. The association was positive (for all: Kendall's τ-b = 0.127 p < 0.0001, for women: Kendall's τ-b = 0.148 p < 0.0001, for men: Kendall's τ-b = 0.126 p < 0.0001) and Figure 4 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to symptoms of long-lasting stress (higher quartiles indicate more stress symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest long-lasting stress quartile were 2.06 for men (p < 0.001) compared to the adjusted odds ratios which were 2.42 (p < 0.001). For women, the corresponding values were 2.61 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between those with more symptoms of long-lasting stress scores compared to those with less (χ2 = 196.855df = 6, p < 0.0001, for women χ2 = 145.608df = 6, p < 0.0001, for men: χ2 = 90.613df = 6, p < 0.0001). Hearing problems were more prevalent among those with more symptoms of long-lasting stress. Similarly to the pattern for burnout, the prevalence increase was higher for women than for men, even if it was less pronounced for this variable. The association was positive (for all: Kendall's τ-b = 0.127 p < 0.0001, for women: Kendall's τ-b = 0.148 p < 0.0001, for men: Kendall's τ-b = 0.126 p < 0.0001) and Figure 4 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to symptoms of long-lasting stress (higher quartiles indicate more stress symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest long-lasting stress quartile were 2.06 for men (p < 0.001) compared to the adjusted odds ratios which were 2.42 (p < 0.001). For women, the corresponding values were 2.61 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).\n[SUBTITLE] Performance-based self-esteem [SUBSECTION] A statistically significant difference was found in the prevalence of hearing problems between those with higher and lower levels of PBS (χ2 = 39.946df = 6, p < 0.0001, for women χ2 = 36.410df = 6, p < 0.0001, for men: χ2 = 15.181df = 6, p < 0.05). Hearing problems were more prevalent among those with higher levels of PBS. The association was positive and linear and more pronounced for women than for men (for all: Kendall's τ-b = 0.056 p < 0.0001, for women: Kendall's τ-b = 0.073 p < 0.0001, for men: Kendall's τ-b = 0.043 p < 0.01) and Figure 5 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to performance-based self-esteem (higher quartiles indicate higher PBS). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest PBS quartile were 1.26 for men (p < 0.001) compared to the adjusted odds ratios which were 1.41 (p < 0.001). For women, the corresponding values were 1.64 (p < 0.001) and when adjusted the value was 1.80 (p < 0.001).\nA statistically significant difference was found in the prevalence of hearing problems between those with higher and lower levels of PBS (χ2 = 39.946df = 6, p < 0.0001, for women χ2 = 36.410df = 6, p < 0.0001, for men: χ2 = 15.181df = 6, p < 0.05). Hearing problems were more prevalent among those with higher levels of PBS. The association was positive and linear and more pronounced for women than for men (for all: Kendall's τ-b = 0.056 p < 0.0001, for women: Kendall's τ-b = 0.073 p < 0.0001, for men: Kendall's τ-b = 0.043 p < 0.01) and Figure 5 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to performance-based self-esteem (higher quartiles indicate higher PBS). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest PBS quartile were 1.26 for men (p < 0.001) compared to the adjusted odds ratios which were 1.41 (p < 0.001). For women, the corresponding values were 1.64 (p < 0.001) and when adjusted the value was 1.80 (p < 0.001).", "Table 1 illustrates the prevalence of hearing problems in relation to different work-related stressors. There was a statistically significant difference in the prevalence of hearing problems between those who were and were not exposed to work-related stressors or threats such as risk of being moved to another work/job against ones will (χ2 = 54.704df = 2, p < 0.0001) and threats of getting fired (χ2 = 27.095df = 2, p < 0.0001). There was however no statistically significant difference between those who were exposed to threats of bankruptcy compared to those who were not.\nPrevalence of hearing problems in relation to different work-related stressors.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems when exposed to a threat of being moved to another job against ones will (yes vs. no) were 1.39 for men (p < 0.001) compared to the adjusted odds ratios which were 1.43 (p < 0.001). For women, the corresponding values were 1.7 (p < 0.001) and when adjusted the value was 1.74 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of self-rated health (for all: χ2 = 262.522df = 6, p < 0.0001, for women χ2 = 141.535df = 6, p < 0.0001, for men: χ2 = 123.306df = 6, p < 0.0001). Figure 1 clearly demonstrates that poorer SRH is associated with a higher prevalence of hearing problems. The association is negative (for all: Kendall's τ-b = -0.139 p < 0.0001, for women: Kendall's τ-b = -0.142 p < 0.0001, for men: Kendall's τ-b = -0.135 p < 0.0001) and indicates increasing prevalence of hearing problems among those with poorer health.\nPrevalence of hearing problems in percent in relation to different ratings of SRH. The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. The unadjusted odds ratio of having hearing problems when reporting poor vs. good SRH were 3.89 for men (p < 0.001) compared to the adjusted odds ratios which were 3.29 (p < 0.001). For women, the corresponding values were 3.81 (p < 0.001) and when adjusted the value was 3.49 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between individuals with different handicaps and those without (χ2 = 181.650df = 2, p < 0.0001). Those with long-term handicaps and illnesses reported more hearing problems (Table 2).\nPrevalence of hearing problems in relation to long-term illness, inconvenience after an accident, any handicap or other weakness.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems after long-term illness (yes vs. no) were 1.92 for men (p < 0.001) compared to the adjusted odds ratios which were 1.76 (p < 0.001). For women, the corresponding values were 1.72 (p < 0.001) and when adjusted the value was 1.64 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of sleep quality (for all: χ2 = 169.875df = 8, p < 0.0001, for women χ2 = 112.509df = 8, p < 0.0001, for men: χ2 = 75.068df = 8, p < 0.0001). The association was negative (for all: Kendall's τ-b = -0.116 p < 0.0001, for women: Kendall's τ-b = -0.126 p < 0.0001, for men: Kendall's τ-b = -0.113 p < 0.0001) and Figure 2 demonstrates that poorer sleep quality is associated with a higher prevalence of hearing problems. As tinnitus and hearing complaints may differ with regard to sleep, both of these variables were also analyzed separately. The prevalence of sleeping problems was significantly higher among those reporting tinnitus (χ2 = 126.884df = 4, p < 0.0001, Kendall's τ-b = -0.106 p < 0.0001) compared to those reporting hearing complaints (χ2 = 76.145df = 4, p < 0.0001, Kendall's τ-b = -0.081 p < 0.0001, see Table 3) and the associations were negative for both.\nPrevalence of hearing problems in percent in relation to different ratings of sleep quality. The Kendall's τ-b value is indicated by a \"τ\".\nPrevalence of sleeping problems among those with tinnitus and hearing loss respectively.\nNumber of participants and row percent in parentheses showing increasing prevalence of tinnitus with poorer sleep.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. When the unadjusted odds ratio of having hearing problems when reporting poor vs. good sleep quality were 2.74 for men (p < 0.001) compared to the adjusted odds ratios which were 2.67 (p < 0.001). For women, the corresponding values were 3.51 (p < 0.001) and when adjusted the value was 3.24 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between those with higher burnout scores compared to those with lower scores (χ2 = 214.473df = 6, p < 0.0001, for women χ2 = 159.205df = 6, p < 0.0001, for men: χ2 = 98.935df = 6, p < 0.0001). Hearing problems were significantly more prevalent among those with higher burnout scores. Multivariate analyses showed no age, gender or SES related differences in prevalence increases with increasing burnout scores. The association was positive (for all: Kendall's τ-b = 0.129 p < 0.0001, for women: Kendall's τ-b = 0.150 p < 0.0001, for men: Kendall's τ-b = 0.129 p < 0.0001) and Figure 3 demonstrates that higher burnout scores are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to different burnout scores (higher quartiles indicate more severe burnout symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For example, the unadjusted odds ratio of having hearing problems when being in the highest vs. lowest burnout quartile were 2.36 for men (p < 0.001) compared to the adjusted odds ratios which were 2.63 (p < 0.001). For women, the corresponding values were 2.80 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between those with more symptoms of long-lasting stress scores compared to those with less (χ2 = 196.855df = 6, p < 0.0001, for women χ2 = 145.608df = 6, p < 0.0001, for men: χ2 = 90.613df = 6, p < 0.0001). Hearing problems were more prevalent among those with more symptoms of long-lasting stress. Similarly to the pattern for burnout, the prevalence increase was higher for women than for men, even if it was less pronounced for this variable. The association was positive (for all: Kendall's τ-b = 0.127 p < 0.0001, for women: Kendall's τ-b = 0.148 p < 0.0001, for men: Kendall's τ-b = 0.126 p < 0.0001) and Figure 4 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to symptoms of long-lasting stress (higher quartiles indicate more stress symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest long-lasting stress quartile were 2.06 for men (p < 0.001) compared to the adjusted odds ratios which were 2.42 (p < 0.001). For women, the corresponding values were 2.61 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).", "A statistically significant difference was found in the prevalence of hearing problems between those with higher and lower levels of PBS (χ2 = 39.946df = 6, p < 0.0001, for women χ2 = 36.410df = 6, p < 0.0001, for men: χ2 = 15.181df = 6, p < 0.05). Hearing problems were more prevalent among those with higher levels of PBS. The association was positive and linear and more pronounced for women than for men (for all: Kendall's τ-b = 0.056 p < 0.0001, for women: Kendall's τ-b = 0.073 p < 0.0001, for men: Kendall's τ-b = 0.043 p < 0.01) and Figure 5 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to performance-based self-esteem (higher quartiles indicate higher PBS). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest PBS quartile were 1.26 for men (p < 0.001) compared to the adjusted odds ratios which were 1.41 (p < 0.001). For women, the corresponding values were 1.64 (p < 0.001) and when adjusted the value was 1.80 (p < 0.001).", "The aim of the present study was to assess the prevalence of two common hearing problems, i.e. hearing complaints and tinnitus, in relation to different work-, life- and health-related stressors. The results demonstrate a clear and mostly linear relationship between higher prevalence of hearing problems (tinnitus or hearing complaints or both) and different stressors, symptoms of ill health and stress as well as poor sleep. Thus, the salient features of the present study illustrate that occupational stressors, poorer self-rated health, long-term illness, poorer sleep quality, higher burnout scores, more symptoms of long-lasting stress, and higher performance-based self-esteem are statistically significantly associated with a higher prevalence of hearing problems. To our knowledge, this is the first time that such vivid and systematic associations have been found for hearing problems (hearing complaints and tinnitus) in an extensive human population.\n[SUBTITLE] Gender issues [SUBSECTION] The findings were consistent for both males and females, and men showed a slightly higher prevalence of hearing problems in absolute levels for all items analyzed except for performance-based self-esteem. Previously it has been found that the prevalence of hearing problems is higher among men and that prevalence increases with age [38]. The present findings elaborate on the previous findings and emphasize the correlation between poorer health, stress-related symptoms and hearing problems that affect men to a greater extent than women. Since hearing problems are becoming a major public health issue it is imperative to understand the factors that are directly and indirectly related to the underlying cause. The findings from the present study shed light on new attributes, biologically- and stress-related, that may be associated with the prevalence of hearing problems. The gender differences need to be interpreted in light of the fact that men and women may be exposed to different types of work environments and exposures that may affect the prevalence of hearing problems.\nIt is also noteworthy to mention that in the present study we have assessed two different factors for hearing problems. First we analyzed the prevalence of hearing problems either for tinnitus or hearing complaints and secondly for both tinnitus and hearing complaints. This assessment is more realistic since it is common for individuals who have hearing complaints to have tinnitus and vice versa. In a previous study, the prevalence of tinnitus was higher for men than women (5% vs. 3%, resp.)[6]. When analyzed for how often the tinnitus occurred, among those who answered that they suffer all the time, the prevalence for men was 10% vs. 5% for women. Interestingly, the effects of higher burnout scores, more symptoms of long-lasting stress and poor sleep quality also demonstrated a male dominance.\nThe findings were consistent for both males and females, and men showed a slightly higher prevalence of hearing problems in absolute levels for all items analyzed except for performance-based self-esteem. Previously it has been found that the prevalence of hearing problems is higher among men and that prevalence increases with age [38]. The present findings elaborate on the previous findings and emphasize the correlation between poorer health, stress-related symptoms and hearing problems that affect men to a greater extent than women. Since hearing problems are becoming a major public health issue it is imperative to understand the factors that are directly and indirectly related to the underlying cause. The findings from the present study shed light on new attributes, biologically- and stress-related, that may be associated with the prevalence of hearing problems. The gender differences need to be interpreted in light of the fact that men and women may be exposed to different types of work environments and exposures that may affect the prevalence of hearing problems.\nIt is also noteworthy to mention that in the present study we have assessed two different factors for hearing problems. First we analyzed the prevalence of hearing problems either for tinnitus or hearing complaints and secondly for both tinnitus and hearing complaints. This assessment is more realistic since it is common for individuals who have hearing complaints to have tinnitus and vice versa. In a previous study, the prevalence of tinnitus was higher for men than women (5% vs. 3%, resp.)[6]. When analyzed for how often the tinnitus occurred, among those who answered that they suffer all the time, the prevalence for men was 10% vs. 5% for women. Interestingly, the effects of higher burnout scores, more symptoms of long-lasting stress and poor sleep quality also demonstrated a male dominance.\n[SUBTITLE] Occupational consequences [SUBSECTION] Hearing problems were also more common among individuals exposed to occupational stressors, i.e. employment-related stress such as risk of being moved to another work/job against ones will and threats of getting fired. Threats of bankruptcy were not significantly associated with a higher prevalence of hearing problems. These problems include difficulties in communicating, lower awareness for important surrounding sounds such as telephone calls or warning signals or oversensitivity to sounds (hyperacusis) and even reduce productivity due to emotional exhaustion. These burdens become exaggerated for individuals with long-lasting stress symptoms and high burnout scores, as demonstrated in the present study. This is important for many reasons including the public awareness and the clarification that long-lasting stress and burnout symptoms are risk factors correlated with hearing problems.\nHearing problems were also more common among individuals exposed to occupational stressors, i.e. employment-related stress such as risk of being moved to another work/job against ones will and threats of getting fired. Threats of bankruptcy were not significantly associated with a higher prevalence of hearing problems. These problems include difficulties in communicating, lower awareness for important surrounding sounds such as telephone calls or warning signals or oversensitivity to sounds (hyperacusis) and even reduce productivity due to emotional exhaustion. These burdens become exaggerated for individuals with long-lasting stress symptoms and high burnout scores, as demonstrated in the present study. This is important for many reasons including the public awareness and the clarification that long-lasting stress and burnout symptoms are risk factors correlated with hearing problems.\n[SUBTITLE] Self-rated health [SUBSECTION] Self-rated health (SRH) is one of the most widely used single measures of perceived current health status [28]. There is extensive evidence suggesting that SRH is a potent predictor of future mortality and morbidity [39,40], functional decline and disability, as well as the utilization of health care [39,41]. Most previous studies have shown that SRH is an independent predictor of future health outcomes, even after adjusting for self-ratings of other health-related measures, physician-reported health status, behavioral and psychosocial risk factors, socioeconomic status and environmental factors. Nevertheless, debate still continues about what SRH really represents [39,42,43].\nIt has been proposed that SRH represents an individual's general perception of health, including biological, psychological and social dimensions. Therefore SRH might be more sensitive in health monitoring than external measures of health [44]. Furthermore, it has been indicated that risk associated with poor SRH status is higher than that associated with poor objective health measures [45]. On the other hand, Kaplan & Camacho [46] found that objective health status has a stronger relationship with mortality than SRH.\nA poorer self-rated health was correlated to hearing problems. Poor self-rated health can be influenced by several factors including social and marital status (Lindström, 2009), type of employment [47], sleep quality, sense of coherence, self-esteem, social support [28], higher cytokine levels [48], allostatic load [49] and unhealthy habits [50]. Unhealthy habits could include careless use of hearing protectors when in noisy environments or being exposed to excessive sound stimulation over long durations. Moreover, self-rated health is a reflection of the individual's quality of life and social abilities [51]. Depending on the severity of a particular hearing problem, consequent problems in communication and social interactions would result. This can cause problems in daily life as well as in the work environment. Since self-rated health is a compilation of biological, psychological and social assessments it may be a more accurate account of the consequences of hearing disabilities of the individual compared to, for example, pure tone audiometry.\nSelf-rated health (SRH) is one of the most widely used single measures of perceived current health status [28]. There is extensive evidence suggesting that SRH is a potent predictor of future mortality and morbidity [39,40], functional decline and disability, as well as the utilization of health care [39,41]. Most previous studies have shown that SRH is an independent predictor of future health outcomes, even after adjusting for self-ratings of other health-related measures, physician-reported health status, behavioral and psychosocial risk factors, socioeconomic status and environmental factors. Nevertheless, debate still continues about what SRH really represents [39,42,43].\nIt has been proposed that SRH represents an individual's general perception of health, including biological, psychological and social dimensions. Therefore SRH might be more sensitive in health monitoring than external measures of health [44]. Furthermore, it has been indicated that risk associated with poor SRH status is higher than that associated with poor objective health measures [45]. On the other hand, Kaplan & Camacho [46] found that objective health status has a stronger relationship with mortality than SRH.\nA poorer self-rated health was correlated to hearing problems. Poor self-rated health can be influenced by several factors including social and marital status (Lindström, 2009), type of employment [47], sleep quality, sense of coherence, self-esteem, social support [28], higher cytokine levels [48], allostatic load [49] and unhealthy habits [50]. Unhealthy habits could include careless use of hearing protectors when in noisy environments or being exposed to excessive sound stimulation over long durations. Moreover, self-rated health is a reflection of the individual's quality of life and social abilities [51]. Depending on the severity of a particular hearing problem, consequent problems in communication and social interactions would result. This can cause problems in daily life as well as in the work environment. Since self-rated health is a compilation of biological, psychological and social assessments it may be a more accurate account of the consequences of hearing disabilities of the individual compared to, for example, pure tone audiometry.\n[SUBTITLE] Sleep quality [SUBSECTION] Poor sleep quality was found to be associated with a higher prevalence of hearing problems in both men and women. In a study aimed at evaluating age specific prevalence of general symptoms it was found that five symptoms increased with increasing age [52]. These symptoms included insomnia, leg pain, joint pain, eye problems and impaired hearing. In another population study, patients with temporomandibular disorders were characterized for the auditory health [53]. The subjects who displayed auditory problems such as tinnitus, and perceived hearing complaints were found to be significantly more likely to consider themselves in poor health and have sleep disturbances than those without auditory problems. Two more studies have found correlations between sleep quality and hearing loss as well as tinnitus [54,55]. Thus, a correlation between poorer sleep quality and hearing problems is apparent in several different populations. In the present study, tinnitus was the more common disturbance when trying to sleep, probably since the perceived sound would increase annoyance and stress reactions. However, hearing complaints and tinnitus are often related co-morbid symptoms [38,56,57] and in the present study worse sleep quality was associated with higher prevalence of hearing complaints, albeit to a lesser extent than tinnitus.\nIn a previous study we found that individuals with hearing problems have a poorer ability to unwind or activate their parasympathetic system [24]. The present study confirms and strengthens this association since problems unwinding are strongly related to sleeping problems.\nPoor sleep quality was found to be associated with a higher prevalence of hearing problems in both men and women. In a study aimed at evaluating age specific prevalence of general symptoms it was found that five symptoms increased with increasing age [52]. These symptoms included insomnia, leg pain, joint pain, eye problems and impaired hearing. In another population study, patients with temporomandibular disorders were characterized for the auditory health [53]. The subjects who displayed auditory problems such as tinnitus, and perceived hearing complaints were found to be significantly more likely to consider themselves in poor health and have sleep disturbances than those without auditory problems. Two more studies have found correlations between sleep quality and hearing loss as well as tinnitus [54,55]. Thus, a correlation between poorer sleep quality and hearing problems is apparent in several different populations. In the present study, tinnitus was the more common disturbance when trying to sleep, probably since the perceived sound would increase annoyance and stress reactions. However, hearing complaints and tinnitus are often related co-morbid symptoms [38,56,57] and in the present study worse sleep quality was associated with higher prevalence of hearing complaints, albeit to a lesser extent than tinnitus.\nIn a previous study we found that individuals with hearing problems have a poorer ability to unwind or activate their parasympathetic system [24]. The present study confirms and strengthens this association since problems unwinding are strongly related to sleeping problems.\n[SUBTITLE] Burnout and symptoms of long-lasting stress [SUBSECTION] It is by now well established that stress-related disorders, such as burnout, are associated with several forms of co-morbidity [58-62], e.g. long-term pain and psychiatric disorders. Now hearing problems may be added to the list. Our clinical experiences from a stress clinic shows that patients are often referred from audiologists. The present study showed clear associations between burnout as well as symptoms of long-lasting stress and hearing problems. Thus, hearing problems should be taken into account clinically in the diagnosis and treatment of stress-related disorders and vice versa. Furthermore, hearing problems were also more common among those with long-term illness, pains, inconveniences or handicaps. This strengthens the hypothesis about high levels of co-morbidity among individuals with hearing problems.\nThe relation between long-term stress and tinnitus has been previously explored and tinnitus sufferers often report enhanced problems by stress and fatigue [63]. It remains to be determined if tinnitus is a direct or indirect response of the auditory system to stress. Non-auditory-related brain regions, such as the limbic system have been shown to be activated during tinnitus [64]. The involvement of limbic areas during tinnitus offers a neuroanatomical correlate for stress-related tinnitus since the limbic region is a key region for regulating stress responses. Moreover, a major therapeutic strategy for tinnitus patients includes relaxation programs which have proven successful for many sufferers. Thus, taking into consideration the results of individuals with hearing problems (hearing complaints and tinnitus) of the present study and findings from prior tinnitus studies, it is becoming more apparent that stress can increase the prevalence of hearing problems. Furthermore, it has been shown that individuals with hearing problems have a worsened ability to unwind and activate the parasympathetic system [24].\nIt is by now well established that stress-related disorders, such as burnout, are associated with several forms of co-morbidity [58-62], e.g. long-term pain and psychiatric disorders. Now hearing problems may be added to the list. Our clinical experiences from a stress clinic shows that patients are often referred from audiologists. The present study showed clear associations between burnout as well as symptoms of long-lasting stress and hearing problems. Thus, hearing problems should be taken into account clinically in the diagnosis and treatment of stress-related disorders and vice versa. Furthermore, hearing problems were also more common among those with long-term illness, pains, inconveniences or handicaps. This strengthens the hypothesis about high levels of co-morbidity among individuals with hearing problems.\nThe relation between long-term stress and tinnitus has been previously explored and tinnitus sufferers often report enhanced problems by stress and fatigue [63]. It remains to be determined if tinnitus is a direct or indirect response of the auditory system to stress. Non-auditory-related brain regions, such as the limbic system have been shown to be activated during tinnitus [64]. The involvement of limbic areas during tinnitus offers a neuroanatomical correlate for stress-related tinnitus since the limbic region is a key region for regulating stress responses. Moreover, a major therapeutic strategy for tinnitus patients includes relaxation programs which have proven successful for many sufferers. Thus, taking into consideration the results of individuals with hearing problems (hearing complaints and tinnitus) of the present study and findings from prior tinnitus studies, it is becoming more apparent that stress can increase the prevalence of hearing problems. Furthermore, it has been shown that individuals with hearing problems have a worsened ability to unwind and activate the parasympathetic system [24].\n[SUBTITLE] Performance-based self-esteem [SUBSECTION] In the present study there was a higher prevalence of hearing problems among those with higher PBS. High PBS is associated with high but vulnerable engagement, as well as higher prevalence of sickness presenteeism. When combined with high scores on a burnout scale it is a strong predictor of burnout and increased risk (OR = 2.84, CI 95% 1.61-5.01) for long-term sick leave [65]. This finding point to the direction that personality factors, especially PBS, may be important when assessing the risks for having hearing problems. It also indicates that hearing problems are multidimensional as they, apart from different stressors, also may be associated with other personality factors. These factors are by and of themselves complicated and multidimensional phenomena that have not yet been completely described. Therefore, more research should be directed at studying these interactions.\nIn the present study there was a higher prevalence of hearing problems among those with higher PBS. High PBS is associated with high but vulnerable engagement, as well as higher prevalence of sickness presenteeism. When combined with high scores on a burnout scale it is a strong predictor of burnout and increased risk (OR = 2.84, CI 95% 1.61-5.01) for long-term sick leave [65]. This finding point to the direction that personality factors, especially PBS, may be important when assessing the risks for having hearing problems. It also indicates that hearing problems are multidimensional as they, apart from different stressors, also may be associated with other personality factors. These factors are by and of themselves complicated and multidimensional phenomena that have not yet been completely described. Therefore, more research should be directed at studying these interactions.\n[SUBTITLE] Strengths and limitations [SUBSECTION] One of the major strengths with this study is the large sample size and that the sample is representative for the general Swedish working population. One weakness is the cross-sectional design, which does not allow conclusions about causality. Accordingly, prospective studies, which are forthcoming from our group, will be needed to illuminate the extent to which the observed associations are causal. Also, the study population is typical of a post-industrial country, with a high proportion of participants having a high educational level. It is possible that associations would be different in a population with a higher proportion of blue collar workers.\nThe assessment of hearing complaints and tinnitus via questionnaires has its advantages and disadvantages and there is no consensus as to which method is most valid. For example, the subjective evaluation of hearing problems may be difficult to interpret when there is a mild hearing loss. However, individuals having constant difficulties in understanding speech (especially in background noise) are usually well aware of their problem. It must be pointed out that an individual who has difficulties in understanding speech in background noise can have a completely normal audiogram. Thus, there are also limitations to audiological testing. This is not to say that an audiological test would not complement our subjective ratings, but we argue that addressing the question if there are \"hearing problems\" is acceptable. In fact, rating scales are commonly utilized to assess hearing problems [5,66] and it has been shown that the single question: 'Do you feel that you have a hearing loss?' was the most sensitive to assess hearing loss compared with pure-tone air conduction audiometry.\nOne of the major strengths with this study is the large sample size and that the sample is representative for the general Swedish working population. One weakness is the cross-sectional design, which does not allow conclusions about causality. Accordingly, prospective studies, which are forthcoming from our group, will be needed to illuminate the extent to which the observed associations are causal. Also, the study population is typical of a post-industrial country, with a high proportion of participants having a high educational level. It is possible that associations would be different in a population with a higher proportion of blue collar workers.\nThe assessment of hearing complaints and tinnitus via questionnaires has its advantages and disadvantages and there is no consensus as to which method is most valid. For example, the subjective evaluation of hearing problems may be difficult to interpret when there is a mild hearing loss. However, individuals having constant difficulties in understanding speech (especially in background noise) are usually well aware of their problem. It must be pointed out that an individual who has difficulties in understanding speech in background noise can have a completely normal audiogram. Thus, there are also limitations to audiological testing. This is not to say that an audiological test would not complement our subjective ratings, but we argue that addressing the question if there are \"hearing problems\" is acceptable. In fact, rating scales are commonly utilized to assess hearing problems [5,66] and it has been shown that the single question: 'Do you feel that you have a hearing loss?' was the most sensitive to assess hearing loss compared with pure-tone air conduction audiometry.", "The findings were consistent for both males and females, and men showed a slightly higher prevalence of hearing problems in absolute levels for all items analyzed except for performance-based self-esteem. Previously it has been found that the prevalence of hearing problems is higher among men and that prevalence increases with age [38]. The present findings elaborate on the previous findings and emphasize the correlation between poorer health, stress-related symptoms and hearing problems that affect men to a greater extent than women. Since hearing problems are becoming a major public health issue it is imperative to understand the factors that are directly and indirectly related to the underlying cause. The findings from the present study shed light on new attributes, biologically- and stress-related, that may be associated with the prevalence of hearing problems. The gender differences need to be interpreted in light of the fact that men and women may be exposed to different types of work environments and exposures that may affect the prevalence of hearing problems.\nIt is also noteworthy to mention that in the present study we have assessed two different factors for hearing problems. First we analyzed the prevalence of hearing problems either for tinnitus or hearing complaints and secondly for both tinnitus and hearing complaints. This assessment is more realistic since it is common for individuals who have hearing complaints to have tinnitus and vice versa. In a previous study, the prevalence of tinnitus was higher for men than women (5% vs. 3%, resp.)[6]. When analyzed for how often the tinnitus occurred, among those who answered that they suffer all the time, the prevalence for men was 10% vs. 5% for women. Interestingly, the effects of higher burnout scores, more symptoms of long-lasting stress and poor sleep quality also demonstrated a male dominance.", "Hearing problems were also more common among individuals exposed to occupational stressors, i.e. employment-related stress such as risk of being moved to another work/job against ones will and threats of getting fired. Threats of bankruptcy were not significantly associated with a higher prevalence of hearing problems. These problems include difficulties in communicating, lower awareness for important surrounding sounds such as telephone calls or warning signals or oversensitivity to sounds (hyperacusis) and even reduce productivity due to emotional exhaustion. These burdens become exaggerated for individuals with long-lasting stress symptoms and high burnout scores, as demonstrated in the present study. This is important for many reasons including the public awareness and the clarification that long-lasting stress and burnout symptoms are risk factors correlated with hearing problems.", "Self-rated health (SRH) is one of the most widely used single measures of perceived current health status [28]. There is extensive evidence suggesting that SRH is a potent predictor of future mortality and morbidity [39,40], functional decline and disability, as well as the utilization of health care [39,41]. Most previous studies have shown that SRH is an independent predictor of future health outcomes, even after adjusting for self-ratings of other health-related measures, physician-reported health status, behavioral and psychosocial risk factors, socioeconomic status and environmental factors. Nevertheless, debate still continues about what SRH really represents [39,42,43].\nIt has been proposed that SRH represents an individual's general perception of health, including biological, psychological and social dimensions. Therefore SRH might be more sensitive in health monitoring than external measures of health [44]. Furthermore, it has been indicated that risk associated with poor SRH status is higher than that associated with poor objective health measures [45]. On the other hand, Kaplan & Camacho [46] found that objective health status has a stronger relationship with mortality than SRH.\nA poorer self-rated health was correlated to hearing problems. Poor self-rated health can be influenced by several factors including social and marital status (Lindström, 2009), type of employment [47], sleep quality, sense of coherence, self-esteem, social support [28], higher cytokine levels [48], allostatic load [49] and unhealthy habits [50]. Unhealthy habits could include careless use of hearing protectors when in noisy environments or being exposed to excessive sound stimulation over long durations. Moreover, self-rated health is a reflection of the individual's quality of life and social abilities [51]. Depending on the severity of a particular hearing problem, consequent problems in communication and social interactions would result. This can cause problems in daily life as well as in the work environment. Since self-rated health is a compilation of biological, psychological and social assessments it may be a more accurate account of the consequences of hearing disabilities of the individual compared to, for example, pure tone audiometry.", "Poor sleep quality was found to be associated with a higher prevalence of hearing problems in both men and women. In a study aimed at evaluating age specific prevalence of general symptoms it was found that five symptoms increased with increasing age [52]. These symptoms included insomnia, leg pain, joint pain, eye problems and impaired hearing. In another population study, patients with temporomandibular disorders were characterized for the auditory health [53]. The subjects who displayed auditory problems such as tinnitus, and perceived hearing complaints were found to be significantly more likely to consider themselves in poor health and have sleep disturbances than those without auditory problems. Two more studies have found correlations between sleep quality and hearing loss as well as tinnitus [54,55]. Thus, a correlation between poorer sleep quality and hearing problems is apparent in several different populations. In the present study, tinnitus was the more common disturbance when trying to sleep, probably since the perceived sound would increase annoyance and stress reactions. However, hearing complaints and tinnitus are often related co-morbid symptoms [38,56,57] and in the present study worse sleep quality was associated with higher prevalence of hearing complaints, albeit to a lesser extent than tinnitus.\nIn a previous study we found that individuals with hearing problems have a poorer ability to unwind or activate their parasympathetic system [24]. The present study confirms and strengthens this association since problems unwinding are strongly related to sleeping problems.", "It is by now well established that stress-related disorders, such as burnout, are associated with several forms of co-morbidity [58-62], e.g. long-term pain and psychiatric disorders. Now hearing problems may be added to the list. Our clinical experiences from a stress clinic shows that patients are often referred from audiologists. The present study showed clear associations between burnout as well as symptoms of long-lasting stress and hearing problems. Thus, hearing problems should be taken into account clinically in the diagnosis and treatment of stress-related disorders and vice versa. Furthermore, hearing problems were also more common among those with long-term illness, pains, inconveniences or handicaps. This strengthens the hypothesis about high levels of co-morbidity among individuals with hearing problems.\nThe relation between long-term stress and tinnitus has been previously explored and tinnitus sufferers often report enhanced problems by stress and fatigue [63]. It remains to be determined if tinnitus is a direct or indirect response of the auditory system to stress. Non-auditory-related brain regions, such as the limbic system have been shown to be activated during tinnitus [64]. The involvement of limbic areas during tinnitus offers a neuroanatomical correlate for stress-related tinnitus since the limbic region is a key region for regulating stress responses. Moreover, a major therapeutic strategy for tinnitus patients includes relaxation programs which have proven successful for many sufferers. Thus, taking into consideration the results of individuals with hearing problems (hearing complaints and tinnitus) of the present study and findings from prior tinnitus studies, it is becoming more apparent that stress can increase the prevalence of hearing problems. Furthermore, it has been shown that individuals with hearing problems have a worsened ability to unwind and activate the parasympathetic system [24].", "In the present study there was a higher prevalence of hearing problems among those with higher PBS. High PBS is associated with high but vulnerable engagement, as well as higher prevalence of sickness presenteeism. When combined with high scores on a burnout scale it is a strong predictor of burnout and increased risk (OR = 2.84, CI 95% 1.61-5.01) for long-term sick leave [65]. This finding point to the direction that personality factors, especially PBS, may be important when assessing the risks for having hearing problems. It also indicates that hearing problems are multidimensional as they, apart from different stressors, also may be associated with other personality factors. These factors are by and of themselves complicated and multidimensional phenomena that have not yet been completely described. Therefore, more research should be directed at studying these interactions.", "One of the major strengths with this study is the large sample size and that the sample is representative for the general Swedish working population. One weakness is the cross-sectional design, which does not allow conclusions about causality. Accordingly, prospective studies, which are forthcoming from our group, will be needed to illuminate the extent to which the observed associations are causal. Also, the study population is typical of a post-industrial country, with a high proportion of participants having a high educational level. It is possible that associations would be different in a population with a higher proportion of blue collar workers.\nThe assessment of hearing complaints and tinnitus via questionnaires has its advantages and disadvantages and there is no consensus as to which method is most valid. For example, the subjective evaluation of hearing problems may be difficult to interpret when there is a mild hearing loss. However, individuals having constant difficulties in understanding speech (especially in background noise) are usually well aware of their problem. It must be pointed out that an individual who has difficulties in understanding speech in background noise can have a completely normal audiogram. Thus, there are also limitations to audiological testing. This is not to say that an audiological test would not complement our subjective ratings, but we argue that addressing the question if there are \"hearing problems\" is acceptable. In fact, rating scales are commonly utilized to assess hearing problems [5,66] and it has been shown that the single question: 'Do you feel that you have a hearing loss?' was the most sensitive to assess hearing loss compared with pure-tone air conduction audiometry.", "In conclusion, the present study unambiguously demonstrates associations between hearing problems and occupational stressors, poorer self-rated health, higher burnout scores, more symptoms of long-lasting stress, long-term illness and poorer sleep quality. The interaction between the prevalence of hearing problems and the above mentioned features have not been previously described for the auditory system. These findings also indicate that hearing problems are multidimensional, which warrants further investigations of possible predictors.", "The authors declare that they have no competing interests.", "DH conducted the statistical analyses and drafted the manuscript together with BC, MBW, with substantial and essential input from CL and TT. TT was PI of the project and CL data manager. All authors have read and approved the final version of the manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2458/11/130/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Population", "Questionnaire", "Statistical analyses", "Ethical approval", "Results", "Work-related stressors/threats", "Self-rated health", "Long-term illness, inconvenience after an accident, any handicap or other weakness", "Sleep quality", "Burnout", "Long-lasting stress", "Performance-based self-esteem", "Discussion", "Gender issues", "Occupational consequences", "Self-rated health", "Sleep quality", "Burnout and symptoms of long-lasting stress", "Performance-based self-esteem", "Strengths and limitations", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Hearing problems are the most common sensory deficit in human populations, with hearing loss alone affecting more than 250 million people worldwide [1]. In 2002, the WHO estimated hearing loss to be the 13th most frequent burden of disease in medium- and high-income countries, and it is projected to become among the top ten by the year 2030 [2]. Most epidemiological studies report prevalence figures of 10-15% for both hearing loss [1,3] and tinnitus [4,5] respectively. These figures may however not be inferable to hearing problems in a broader sense, since the prevalence is most often only calculated for either hearing loss or tinnitus. Consequently, a large Swedish study has recently addressed both issues concurrently, finding that approximately 32% of the Swedish working population suffer from either hearing complaints, tinnitus or both [6]. The study was also the first to provide evidence for a negative association between self-rated socioeconomic status (SES) and prevalence of hearing problems. It is well established that lower SES is associated with higher stress levels [7-9], so this finding may indirectly imply a relationship between stress and hearing problems.\nWhile the deleterious effects of mechanical stress (i.e. noise) on hearing have been studied extensively in both animal models [10] and human populations [11,12], the notion of emotional stress as a modulator of the auditory system is rather novel. A complex set of pathways of the stress response have been identified, involving both sympathetic stimulation of adrenergic α-receptors within the cochlea [13,14], as well as neuro-endocrine responses primarily aimed at engaging the hypothalamic-pituitary-adrenal (HPA) axis [15]. Current research suggests that acute stress may protect the cochlea [16-19], whereas chronic stress exposure seems to be harmful to hearing [20]. The importance of a normal functioning of the HPA-axis for healthy hearing is supported by clinical studies showing that patients with tinnitus display signs of an impaired HPA-axis along with a higher degree of perceived stress, compared to non-tinnitus patients [21-23]. Additionally, Hasson et al. [24] recently found that symphony orchestra musicians with hearing problems exhibited lower heart rate variability (high frequency power), indicating an impaired ability to \"unwind\" and activate the parasympathetic system.\nHuman and animal studies are providing more evidence of an association between stress and hearing problems. It also needs be considered that long-term hearing problems can be stressful. Regarding the high prevalence of hearing problems in human populations and the negative future projections of the WHO, further investigations of the relationship between stress and hearing problems are warranted. Therefore, the aim of this study is to assess the prevalence of two common hearing problems, i.e. hearing complaints and tinnitus, in relation to different work- and health-related stressors. More specifically, we will study if prevalence differs with poor self-rated health, poorer sleep, higher burnout scores (work-related), and more symptoms of long-term stress and higher levels of performance-based self-esteem.", "[SUBTITLE] Population [SUBSECTION] The Swedish Work Environment Survey (SWES) is conducted biennially by Statistics Sweden (SCB) and consists of subsamples of gainfully employed people, aged 16-64 years, from the Labor Force Survey (LSF). These individuals were first sampled into the LFS through stratification by county of birth, sex, citizenship, and inferred employment status. The respondents to SWES 2003 and 2005 were invited to enroll in the Swedish Longitudinal Occupational Survey of Health (SLOSH) [25], which was initiated by the Stress Research Institute in 2006. The second data collection was conducted in April 2008 by Statistics Sweden, on behalf of the Stress Research Institute at Stockholm University. A total of 18,734 individuals were mailed self-completion questionnaires in 2008, out of which 9,756 (52%) working individuals responded. The total response rate of the study was however 11,441 (61%), including non-working participants (not analyzed in the present study). More detailed information about the cohort, response rate and characteristics of responders vs. non-responders has been published elsewhere [6]. There was no difference between responders and non-responders with regard to county of birth and citizenship.\nThe Swedish Work Environment Survey (SWES) is conducted biennially by Statistics Sweden (SCB) and consists of subsamples of gainfully employed people, aged 16-64 years, from the Labor Force Survey (LSF). These individuals were first sampled into the LFS through stratification by county of birth, sex, citizenship, and inferred employment status. The respondents to SWES 2003 and 2005 were invited to enroll in the Swedish Longitudinal Occupational Survey of Health (SLOSH) [25], which was initiated by the Stress Research Institute in 2006. The second data collection was conducted in April 2008 by Statistics Sweden, on behalf of the Stress Research Institute at Stockholm University. A total of 18,734 individuals were mailed self-completion questionnaires in 2008, out of which 9,756 (52%) working individuals responded. The total response rate of the study was however 11,441 (61%), including non-working participants (not analyzed in the present study). More detailed information about the cohort, response rate and characteristics of responders vs. non-responders has been published elsewhere [6]. There was no difference between responders and non-responders with regard to county of birth and citizenship.\n[SUBTITLE] Questionnaire [SUBSECTION] Apart from socioeconomic status and demographic factors, it also included approximately 120 questions about psychosocial and physical work-environment, lifestyle, as well as physical and mental health.\nHearing problems were assessed with three questions. Tinnitus. Have you during the most recent time experienced sound in any of the ears, without there being an external source (so-called tinnitus) lasting more than five minutes? (No, Yes sometimes, Yes often, Yes all the time). Tinnitus severity. How much do you feel that the tinnitus sounds worry, bother or upset you? (Not at all, A little, Moderately, Severely). The questions about tinnitus were adapted from Davis [26] and Palmer et al. [27]. Hearing complaints. How difficult is it for you to (without hearing aid) hear what is said in a conversation between several persons? (Not difficult at all, Not very difficult, Quite difficult, Very difficult). In this study, hearing complaints reflects difficulties in communicating. The questions about hearing complaints were derived and adapted from Statistics Sweden and have been used in population studies for several years.\nA new variable, \"hearing problems\", was computed based on the existence or non-existence of either tinnitus or hearing complaints or both. This consequently yielded three groups; those without hearing problems, those with either tinnitus or hearing complaints or those suffering from both. The cut-off for tinnitus was \"yes, sometimes\" or more often, and for hearing complaints \"quite difficult\" or \"very difficult\".\nWork-related stressors/threats. Risks of being moved to another work/job against ones will, threats of getting fired were derived from the Swedish Labor Force Survey (LFS). Threats of bankruptcy were constructed for SLOSH 2008 to assess a threat particularly important for self-employed, a group who contacted the research group and expressed feelings of neglect in the SLOSH 2006 survey. The question was formulated: \"Are you subjected to any of the following risks or threats in your work?\" Response alternatives were yes/no.\nSelf-rated health was assessed with the single item \"How would you rate your general state of health?\" This question has been widely used in research [28-30] and the respondents answered on a Likert scale from ranging from 1 (very poor) to 5 (very good). Since only few participants had very poor SRH the categories quite poor and very poor were merged in the analyses.\nLong-term illness, inconvenience after an accident, any handicap or other weakness. One question was asked about long-term illness, inconvenience after an accident or other weakness: \"Do you have any prolonged sickness, accident-related complaints, a disability or other weakness?\" This question was derived from the WOLFF (WOrk, Lipids and Fibrinogen-follow-up) [31,32] questionnaire and response alternatives were yes/no.\nSleep quality. Sleep quality was assessed by the single item: \"How is your sleep quality in general?\" This item was derived from the Karolinska Sleep Questionnaire [33]. Response ranged from 1 (very poor) to 5 (very good) on a 5-graded Likert scale.\nBurnout was assessed with the Maslach Burnout Inventory general survey (MBI-GS) using the emotional exhaustion subscale [34]. The scale consists of five items, derived from the Maslach Burnout Inventory human services survey (MBI-HSS) in unmodified form. Scorings reach from 1 (every day) to 6 (a few times a year or less/never). Cronbach's alpha and stability for the subscale have been reported to be satisfactory. Strong support for the construct validity of the Swedish translation of the MBI-HSS has been found [35]. The index was calculated on the basis that 4 out of 5 items had to be answered in order be included in the index.\nLong lasting stress (LLS) was assessed with 11 items reflecting stress arousal symptoms but not stress reactions. The participants were asked how they felt during the last three months with regard to both physiological (e.g. \"I sweat easily even though I do not exert myself physically\") and cognitive-behavioral symptoms (e.g. \"I have worrying thoughts\"; \"I often feel tense\"). The four response alternatives reached from \"Not at all\" to \"Nearly all the time\". The scale was introduced in the 2008 SLOSH questionnaire and is currently being validated (data not yet published). A factor analysis yielded one factor of interest. This factor included 7 of the 11 items and only the cognitive-behavioral symptoms. Factor loadings ranged from .675 ± 798 and a Chronbach's α of .863. The 7 included items were (including factor loading, FL):\nA) I have days when I feel geared up all the time (FL = .675). B) I have days when I feel very pressured, on the verge of what I can handle (FL = .737). C) I find it hard to relax during my leisure time (FL = .787). D) I often feel tense (FL = .798). E) I often have disturbing thoughts (FL = .768). F) I often feel restless (FL = .720). G) I do not feel rested after taking it easy for a few days (FL = .699). The correlation between LLS and MBI-GS is r = .64, p < .0001, two-tailed. The index was calculated on the basis that 6 out of 7 items had to be answered in order be included in the index.\nPerformance based self-esteem [36,37] was assessed with four items (e.g. \"At times, I have to be better than others to be good enough myself\") with five response alternatives with the end-point labels \"Fully disagree\" to \"Fully agree\". Cronbach's alpha is between .85 and .89 and the one-year stability has been found to be satisfying. The index was calculated on the basis that 3 out of 4 items had to be answered in order be included in the index.\nApart from socioeconomic status and demographic factors, it also included approximately 120 questions about psychosocial and physical work-environment, lifestyle, as well as physical and mental health.\nHearing problems were assessed with three questions. Tinnitus. Have you during the most recent time experienced sound in any of the ears, without there being an external source (so-called tinnitus) lasting more than five minutes? (No, Yes sometimes, Yes often, Yes all the time). Tinnitus severity. How much do you feel that the tinnitus sounds worry, bother or upset you? (Not at all, A little, Moderately, Severely). The questions about tinnitus were adapted from Davis [26] and Palmer et al. [27]. Hearing complaints. How difficult is it for you to (without hearing aid) hear what is said in a conversation between several persons? (Not difficult at all, Not very difficult, Quite difficult, Very difficult). In this study, hearing complaints reflects difficulties in communicating. The questions about hearing complaints were derived and adapted from Statistics Sweden and have been used in population studies for several years.\nA new variable, \"hearing problems\", was computed based on the existence or non-existence of either tinnitus or hearing complaints or both. This consequently yielded three groups; those without hearing problems, those with either tinnitus or hearing complaints or those suffering from both. The cut-off for tinnitus was \"yes, sometimes\" or more often, and for hearing complaints \"quite difficult\" or \"very difficult\".\nWork-related stressors/threats. Risks of being moved to another work/job against ones will, threats of getting fired were derived from the Swedish Labor Force Survey (LFS). Threats of bankruptcy were constructed for SLOSH 2008 to assess a threat particularly important for self-employed, a group who contacted the research group and expressed feelings of neglect in the SLOSH 2006 survey. The question was formulated: \"Are you subjected to any of the following risks or threats in your work?\" Response alternatives were yes/no.\nSelf-rated health was assessed with the single item \"How would you rate your general state of health?\" This question has been widely used in research [28-30] and the respondents answered on a Likert scale from ranging from 1 (very poor) to 5 (very good). Since only few participants had very poor SRH the categories quite poor and very poor were merged in the analyses.\nLong-term illness, inconvenience after an accident, any handicap or other weakness. One question was asked about long-term illness, inconvenience after an accident or other weakness: \"Do you have any prolonged sickness, accident-related complaints, a disability or other weakness?\" This question was derived from the WOLFF (WOrk, Lipids and Fibrinogen-follow-up) [31,32] questionnaire and response alternatives were yes/no.\nSleep quality. Sleep quality was assessed by the single item: \"How is your sleep quality in general?\" This item was derived from the Karolinska Sleep Questionnaire [33]. Response ranged from 1 (very poor) to 5 (very good) on a 5-graded Likert scale.\nBurnout was assessed with the Maslach Burnout Inventory general survey (MBI-GS) using the emotional exhaustion subscale [34]. The scale consists of five items, derived from the Maslach Burnout Inventory human services survey (MBI-HSS) in unmodified form. Scorings reach from 1 (every day) to 6 (a few times a year or less/never). Cronbach's alpha and stability for the subscale have been reported to be satisfactory. Strong support for the construct validity of the Swedish translation of the MBI-HSS has been found [35]. The index was calculated on the basis that 4 out of 5 items had to be answered in order be included in the index.\nLong lasting stress (LLS) was assessed with 11 items reflecting stress arousal symptoms but not stress reactions. The participants were asked how they felt during the last three months with regard to both physiological (e.g. \"I sweat easily even though I do not exert myself physically\") and cognitive-behavioral symptoms (e.g. \"I have worrying thoughts\"; \"I often feel tense\"). The four response alternatives reached from \"Not at all\" to \"Nearly all the time\". The scale was introduced in the 2008 SLOSH questionnaire and is currently being validated (data not yet published). A factor analysis yielded one factor of interest. This factor included 7 of the 11 items and only the cognitive-behavioral symptoms. Factor loadings ranged from .675 ± 798 and a Chronbach's α of .863. The 7 included items were (including factor loading, FL):\nA) I have days when I feel geared up all the time (FL = .675). B) I have days when I feel very pressured, on the verge of what I can handle (FL = .737). C) I find it hard to relax during my leisure time (FL = .787). D) I often feel tense (FL = .798). E) I often have disturbing thoughts (FL = .768). F) I often feel restless (FL = .720). G) I do not feel rested after taking it easy for a few days (FL = .699). The correlation between LLS and MBI-GS is r = .64, p < .0001, two-tailed. The index was calculated on the basis that 6 out of 7 items had to be answered in order be included in the index.\nPerformance based self-esteem [36,37] was assessed with four items (e.g. \"At times, I have to be better than others to be good enough myself\") with five response alternatives with the end-point labels \"Fully disagree\" to \"Fully agree\". Cronbach's alpha is between .85 and .89 and the one-year stability has been found to be satisfying. The index was calculated on the basis that 3 out of 4 items had to be answered in order be included in the index.\n[SUBTITLE] Statistical analyses [SUBSECTION] The programs SPSS 18.0 and SAS 9.2 were used for statistical analyses. Prevalence was calculated via frequency plots and crosstabs were used for calculation of χ, Kendall's tau-b, and specific prevalence within different groups. Kendall's tau-b is a correlation analysis that illustrates the direction and magnitude of association between two variables. Multivariate analyses, proportional odds model (also called ordered logistic regression), were used to calculate possible interacting or confounding effects of age, gender and SES. Comparisons were made between those having no hearing problems compared to those with either tinnitus or hearing loss or both tinnitus and hearing loss. Statistical significance was set at p < 0.05 level.\nThe programs SPSS 18.0 and SAS 9.2 were used for statistical analyses. Prevalence was calculated via frequency plots and crosstabs were used for calculation of χ, Kendall's tau-b, and specific prevalence within different groups. Kendall's tau-b is a correlation analysis that illustrates the direction and magnitude of association between two variables. Multivariate analyses, proportional odds model (also called ordered logistic regression), were used to calculate possible interacting or confounding effects of age, gender and SES. Comparisons were made between those having no hearing problems compared to those with either tinnitus or hearing loss or both tinnitus and hearing loss. Statistical significance was set at p < 0.05 level.\n[SUBTITLE] Ethical approval [SUBSECTION] The regional ethics committee in Stockholm approved the research project and all participants gave their informed consent to participate.\nThe regional ethics committee in Stockholm approved the research project and all participants gave their informed consent to participate.", "The Swedish Work Environment Survey (SWES) is conducted biennially by Statistics Sweden (SCB) and consists of subsamples of gainfully employed people, aged 16-64 years, from the Labor Force Survey (LSF). These individuals were first sampled into the LFS through stratification by county of birth, sex, citizenship, and inferred employment status. The respondents to SWES 2003 and 2005 were invited to enroll in the Swedish Longitudinal Occupational Survey of Health (SLOSH) [25], which was initiated by the Stress Research Institute in 2006. The second data collection was conducted in April 2008 by Statistics Sweden, on behalf of the Stress Research Institute at Stockholm University. A total of 18,734 individuals were mailed self-completion questionnaires in 2008, out of which 9,756 (52%) working individuals responded. The total response rate of the study was however 11,441 (61%), including non-working participants (not analyzed in the present study). More detailed information about the cohort, response rate and characteristics of responders vs. non-responders has been published elsewhere [6]. There was no difference between responders and non-responders with regard to county of birth and citizenship.", "Apart from socioeconomic status and demographic factors, it also included approximately 120 questions about psychosocial and physical work-environment, lifestyle, as well as physical and mental health.\nHearing problems were assessed with three questions. Tinnitus. Have you during the most recent time experienced sound in any of the ears, without there being an external source (so-called tinnitus) lasting more than five minutes? (No, Yes sometimes, Yes often, Yes all the time). Tinnitus severity. How much do you feel that the tinnitus sounds worry, bother or upset you? (Not at all, A little, Moderately, Severely). The questions about tinnitus were adapted from Davis [26] and Palmer et al. [27]. Hearing complaints. How difficult is it for you to (without hearing aid) hear what is said in a conversation between several persons? (Not difficult at all, Not very difficult, Quite difficult, Very difficult). In this study, hearing complaints reflects difficulties in communicating. The questions about hearing complaints were derived and adapted from Statistics Sweden and have been used in population studies for several years.\nA new variable, \"hearing problems\", was computed based on the existence or non-existence of either tinnitus or hearing complaints or both. This consequently yielded three groups; those without hearing problems, those with either tinnitus or hearing complaints or those suffering from both. The cut-off for tinnitus was \"yes, sometimes\" or more often, and for hearing complaints \"quite difficult\" or \"very difficult\".\nWork-related stressors/threats. Risks of being moved to another work/job against ones will, threats of getting fired were derived from the Swedish Labor Force Survey (LFS). Threats of bankruptcy were constructed for SLOSH 2008 to assess a threat particularly important for self-employed, a group who contacted the research group and expressed feelings of neglect in the SLOSH 2006 survey. The question was formulated: \"Are you subjected to any of the following risks or threats in your work?\" Response alternatives were yes/no.\nSelf-rated health was assessed with the single item \"How would you rate your general state of health?\" This question has been widely used in research [28-30] and the respondents answered on a Likert scale from ranging from 1 (very poor) to 5 (very good). Since only few participants had very poor SRH the categories quite poor and very poor were merged in the analyses.\nLong-term illness, inconvenience after an accident, any handicap or other weakness. One question was asked about long-term illness, inconvenience after an accident or other weakness: \"Do you have any prolonged sickness, accident-related complaints, a disability or other weakness?\" This question was derived from the WOLFF (WOrk, Lipids and Fibrinogen-follow-up) [31,32] questionnaire and response alternatives were yes/no.\nSleep quality. Sleep quality was assessed by the single item: \"How is your sleep quality in general?\" This item was derived from the Karolinska Sleep Questionnaire [33]. Response ranged from 1 (very poor) to 5 (very good) on a 5-graded Likert scale.\nBurnout was assessed with the Maslach Burnout Inventory general survey (MBI-GS) using the emotional exhaustion subscale [34]. The scale consists of five items, derived from the Maslach Burnout Inventory human services survey (MBI-HSS) in unmodified form. Scorings reach from 1 (every day) to 6 (a few times a year or less/never). Cronbach's alpha and stability for the subscale have been reported to be satisfactory. Strong support for the construct validity of the Swedish translation of the MBI-HSS has been found [35]. The index was calculated on the basis that 4 out of 5 items had to be answered in order be included in the index.\nLong lasting stress (LLS) was assessed with 11 items reflecting stress arousal symptoms but not stress reactions. The participants were asked how they felt during the last three months with regard to both physiological (e.g. \"I sweat easily even though I do not exert myself physically\") and cognitive-behavioral symptoms (e.g. \"I have worrying thoughts\"; \"I often feel tense\"). The four response alternatives reached from \"Not at all\" to \"Nearly all the time\". The scale was introduced in the 2008 SLOSH questionnaire and is currently being validated (data not yet published). A factor analysis yielded one factor of interest. This factor included 7 of the 11 items and only the cognitive-behavioral symptoms. Factor loadings ranged from .675 ± 798 and a Chronbach's α of .863. The 7 included items were (including factor loading, FL):\nA) I have days when I feel geared up all the time (FL = .675). B) I have days when I feel very pressured, on the verge of what I can handle (FL = .737). C) I find it hard to relax during my leisure time (FL = .787). D) I often feel tense (FL = .798). E) I often have disturbing thoughts (FL = .768). F) I often feel restless (FL = .720). G) I do not feel rested after taking it easy for a few days (FL = .699). The correlation between LLS and MBI-GS is r = .64, p < .0001, two-tailed. The index was calculated on the basis that 6 out of 7 items had to be answered in order be included in the index.\nPerformance based self-esteem [36,37] was assessed with four items (e.g. \"At times, I have to be better than others to be good enough myself\") with five response alternatives with the end-point labels \"Fully disagree\" to \"Fully agree\". Cronbach's alpha is between .85 and .89 and the one-year stability has been found to be satisfying. The index was calculated on the basis that 3 out of 4 items had to be answered in order be included in the index.", "The programs SPSS 18.0 and SAS 9.2 were used for statistical analyses. Prevalence was calculated via frequency plots and crosstabs were used for calculation of χ, Kendall's tau-b, and specific prevalence within different groups. Kendall's tau-b is a correlation analysis that illustrates the direction and magnitude of association between two variables. Multivariate analyses, proportional odds model (also called ordered logistic regression), were used to calculate possible interacting or confounding effects of age, gender and SES. Comparisons were made between those having no hearing problems compared to those with either tinnitus or hearing loss or both tinnitus and hearing loss. Statistical significance was set at p < 0.05 level.", "The regional ethics committee in Stockholm approved the research project and all participants gave their informed consent to participate.", "In a previous analysis of the present data it was shown that 31% of the working population report either hearing complaints or tinnitus (i.e. 25%) or both (i.e. 6%) and the prevalence of hearing problems increased with age, was higher among men and in persons with lower self-rated SES, and co-varied with exposure to noise at work [6]. In light of this background we now present the association between work- and health-related stressors and hearing problems.\nOverall, the results describe an association between hearing problems and work- and health-related stressors. All the results were controlled for possible confounding effect of age, gender or SES with multivariate analyses. These analyses showed no confounding effect of these variables. The demographics of the population in the present study are as follows: 4,462 (46%) men and 5,294 (54%) women. The mean age was 48.6 (± 10.8) for men and 48.2 (± 10.5) for women. Age distribution was: under 40 years 1,148 (26%) men and 1,314 (25%) women; 41 ± 51 years 1,202 (27%) men and 1,520 (29%) women; 51-60 years 1,415 (32%) men and 1,754 (33%) women; 60 years or older 697 (26%) men and 706 (13%) women. Marital status: married 2,491 (56%) men and 2,946 (56%) women; unmarried 1,494 (34%) men and 1,516 (29%) women; divorced 444 (10%) men and 726 (14%) women; widow 33 (1%) and 106 (2%) women. With regard to highest completed educational level, 1,016 (10%) had no gymnasium, 4,510 (46%) had gymnasium, 619 (6%) had undergraduate studies of two years or less, 3,472 (36%) had undergraduate studies of three years or more and 134 (1%) hade post graduate studies.\n[SUBTITLE] Work-related stressors/threats [SUBSECTION] Table 1 illustrates the prevalence of hearing problems in relation to different work-related stressors. There was a statistically significant difference in the prevalence of hearing problems between those who were and were not exposed to work-related stressors or threats such as risk of being moved to another work/job against ones will (χ2 = 54.704df = 2, p < 0.0001) and threats of getting fired (χ2 = 27.095df = 2, p < 0.0001). There was however no statistically significant difference between those who were exposed to threats of bankruptcy compared to those who were not.\nPrevalence of hearing problems in relation to different work-related stressors.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems when exposed to a threat of being moved to another job against ones will (yes vs. no) were 1.39 for men (p < 0.001) compared to the adjusted odds ratios which were 1.43 (p < 0.001). For women, the corresponding values were 1.7 (p < 0.001) and when adjusted the value was 1.74 (p < 0.001).\nTable 1 illustrates the prevalence of hearing problems in relation to different work-related stressors. There was a statistically significant difference in the prevalence of hearing problems between those who were and were not exposed to work-related stressors or threats such as risk of being moved to another work/job against ones will (χ2 = 54.704df = 2, p < 0.0001) and threats of getting fired (χ2 = 27.095df = 2, p < 0.0001). There was however no statistically significant difference between those who were exposed to threats of bankruptcy compared to those who were not.\nPrevalence of hearing problems in relation to different work-related stressors.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems when exposed to a threat of being moved to another job against ones will (yes vs. no) were 1.39 for men (p < 0.001) compared to the adjusted odds ratios which were 1.43 (p < 0.001). For women, the corresponding values were 1.7 (p < 0.001) and when adjusted the value was 1.74 (p < 0.001).\n[SUBTITLE] Self-rated health [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of self-rated health (for all: χ2 = 262.522df = 6, p < 0.0001, for women χ2 = 141.535df = 6, p < 0.0001, for men: χ2 = 123.306df = 6, p < 0.0001). Figure 1 clearly demonstrates that poorer SRH is associated with a higher prevalence of hearing problems. The association is negative (for all: Kendall's τ-b = -0.139 p < 0.0001, for women: Kendall's τ-b = -0.142 p < 0.0001, for men: Kendall's τ-b = -0.135 p < 0.0001) and indicates increasing prevalence of hearing problems among those with poorer health.\nPrevalence of hearing problems in percent in relation to different ratings of SRH. The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. The unadjusted odds ratio of having hearing problems when reporting poor vs. good SRH were 3.89 for men (p < 0.001) compared to the adjusted odds ratios which were 3.29 (p < 0.001). For women, the corresponding values were 3.81 (p < 0.001) and when adjusted the value was 3.49 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of self-rated health (for all: χ2 = 262.522df = 6, p < 0.0001, for women χ2 = 141.535df = 6, p < 0.0001, for men: χ2 = 123.306df = 6, p < 0.0001). Figure 1 clearly demonstrates that poorer SRH is associated with a higher prevalence of hearing problems. The association is negative (for all: Kendall's τ-b = -0.139 p < 0.0001, for women: Kendall's τ-b = -0.142 p < 0.0001, for men: Kendall's τ-b = -0.135 p < 0.0001) and indicates increasing prevalence of hearing problems among those with poorer health.\nPrevalence of hearing problems in percent in relation to different ratings of SRH. The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. The unadjusted odds ratio of having hearing problems when reporting poor vs. good SRH were 3.89 for men (p < 0.001) compared to the adjusted odds ratios which were 3.29 (p < 0.001). For women, the corresponding values were 3.81 (p < 0.001) and when adjusted the value was 3.49 (p < 0.001).\n[SUBTITLE] Long-term illness, inconvenience after an accident, any handicap or other weakness [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between individuals with different handicaps and those without (χ2 = 181.650df = 2, p < 0.0001). Those with long-term handicaps and illnesses reported more hearing problems (Table 2).\nPrevalence of hearing problems in relation to long-term illness, inconvenience after an accident, any handicap or other weakness.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems after long-term illness (yes vs. no) were 1.92 for men (p < 0.001) compared to the adjusted odds ratios which were 1.76 (p < 0.001). For women, the corresponding values were 1.72 (p < 0.001) and when adjusted the value was 1.64 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between individuals with different handicaps and those without (χ2 = 181.650df = 2, p < 0.0001). Those with long-term handicaps and illnesses reported more hearing problems (Table 2).\nPrevalence of hearing problems in relation to long-term illness, inconvenience after an accident, any handicap or other weakness.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems after long-term illness (yes vs. no) were 1.92 for men (p < 0.001) compared to the adjusted odds ratios which were 1.76 (p < 0.001). For women, the corresponding values were 1.72 (p < 0.001) and when adjusted the value was 1.64 (p < 0.001).\n[SUBTITLE] Sleep quality [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of sleep quality (for all: χ2 = 169.875df = 8, p < 0.0001, for women χ2 = 112.509df = 8, p < 0.0001, for men: χ2 = 75.068df = 8, p < 0.0001). The association was negative (for all: Kendall's τ-b = -0.116 p < 0.0001, for women: Kendall's τ-b = -0.126 p < 0.0001, for men: Kendall's τ-b = -0.113 p < 0.0001) and Figure 2 demonstrates that poorer sleep quality is associated with a higher prevalence of hearing problems. As tinnitus and hearing complaints may differ with regard to sleep, both of these variables were also analyzed separately. The prevalence of sleeping problems was significantly higher among those reporting tinnitus (χ2 = 126.884df = 4, p < 0.0001, Kendall's τ-b = -0.106 p < 0.0001) compared to those reporting hearing complaints (χ2 = 76.145df = 4, p < 0.0001, Kendall's τ-b = -0.081 p < 0.0001, see Table 3) and the associations were negative for both.\nPrevalence of hearing problems in percent in relation to different ratings of sleep quality. The Kendall's τ-b value is indicated by a \"τ\".\nPrevalence of sleeping problems among those with tinnitus and hearing loss respectively.\nNumber of participants and row percent in parentheses showing increasing prevalence of tinnitus with poorer sleep.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. When the unadjusted odds ratio of having hearing problems when reporting poor vs. good sleep quality were 2.74 for men (p < 0.001) compared to the adjusted odds ratios which were 2.67 (p < 0.001). For women, the corresponding values were 3.51 (p < 0.001) and when adjusted the value was 3.24 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of sleep quality (for all: χ2 = 169.875df = 8, p < 0.0001, for women χ2 = 112.509df = 8, p < 0.0001, for men: χ2 = 75.068df = 8, p < 0.0001). The association was negative (for all: Kendall's τ-b = -0.116 p < 0.0001, for women: Kendall's τ-b = -0.126 p < 0.0001, for men: Kendall's τ-b = -0.113 p < 0.0001) and Figure 2 demonstrates that poorer sleep quality is associated with a higher prevalence of hearing problems. As tinnitus and hearing complaints may differ with regard to sleep, both of these variables were also analyzed separately. The prevalence of sleeping problems was significantly higher among those reporting tinnitus (χ2 = 126.884df = 4, p < 0.0001, Kendall's τ-b = -0.106 p < 0.0001) compared to those reporting hearing complaints (χ2 = 76.145df = 4, p < 0.0001, Kendall's τ-b = -0.081 p < 0.0001, see Table 3) and the associations were negative for both.\nPrevalence of hearing problems in percent in relation to different ratings of sleep quality. The Kendall's τ-b value is indicated by a \"τ\".\nPrevalence of sleeping problems among those with tinnitus and hearing loss respectively.\nNumber of participants and row percent in parentheses showing increasing prevalence of tinnitus with poorer sleep.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. When the unadjusted odds ratio of having hearing problems when reporting poor vs. good sleep quality were 2.74 for men (p < 0.001) compared to the adjusted odds ratios which were 2.67 (p < 0.001). For women, the corresponding values were 3.51 (p < 0.001) and when adjusted the value was 3.24 (p < 0.001).\n[SUBTITLE] Burnout [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between those with higher burnout scores compared to those with lower scores (χ2 = 214.473df = 6, p < 0.0001, for women χ2 = 159.205df = 6, p < 0.0001, for men: χ2 = 98.935df = 6, p < 0.0001). Hearing problems were significantly more prevalent among those with higher burnout scores. Multivariate analyses showed no age, gender or SES related differences in prevalence increases with increasing burnout scores. The association was positive (for all: Kendall's τ-b = 0.129 p < 0.0001, for women: Kendall's τ-b = 0.150 p < 0.0001, for men: Kendall's τ-b = 0.129 p < 0.0001) and Figure 3 demonstrates that higher burnout scores are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to different burnout scores (higher quartiles indicate more severe burnout symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For example, the unadjusted odds ratio of having hearing problems when being in the highest vs. lowest burnout quartile were 2.36 for men (p < 0.001) compared to the adjusted odds ratios which were 2.63 (p < 0.001). For women, the corresponding values were 2.80 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between those with higher burnout scores compared to those with lower scores (χ2 = 214.473df = 6, p < 0.0001, for women χ2 = 159.205df = 6, p < 0.0001, for men: χ2 = 98.935df = 6, p < 0.0001). Hearing problems were significantly more prevalent among those with higher burnout scores. Multivariate analyses showed no age, gender or SES related differences in prevalence increases with increasing burnout scores. The association was positive (for all: Kendall's τ-b = 0.129 p < 0.0001, for women: Kendall's τ-b = 0.150 p < 0.0001, for men: Kendall's τ-b = 0.129 p < 0.0001) and Figure 3 demonstrates that higher burnout scores are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to different burnout scores (higher quartiles indicate more severe burnout symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For example, the unadjusted odds ratio of having hearing problems when being in the highest vs. lowest burnout quartile were 2.36 for men (p < 0.001) compared to the adjusted odds ratios which were 2.63 (p < 0.001). For women, the corresponding values were 2.80 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).\n[SUBTITLE] Long-lasting stress [SUBSECTION] There was a statistically significant difference in the prevalence of hearing problems between those with more symptoms of long-lasting stress scores compared to those with less (χ2 = 196.855df = 6, p < 0.0001, for women χ2 = 145.608df = 6, p < 0.0001, for men: χ2 = 90.613df = 6, p < 0.0001). Hearing problems were more prevalent among those with more symptoms of long-lasting stress. Similarly to the pattern for burnout, the prevalence increase was higher for women than for men, even if it was less pronounced for this variable. The association was positive (for all: Kendall's τ-b = 0.127 p < 0.0001, for women: Kendall's τ-b = 0.148 p < 0.0001, for men: Kendall's τ-b = 0.126 p < 0.0001) and Figure 4 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to symptoms of long-lasting stress (higher quartiles indicate more stress symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest long-lasting stress quartile were 2.06 for men (p < 0.001) compared to the adjusted odds ratios which were 2.42 (p < 0.001). For women, the corresponding values were 2.61 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).\nThere was a statistically significant difference in the prevalence of hearing problems between those with more symptoms of long-lasting stress scores compared to those with less (χ2 = 196.855df = 6, p < 0.0001, for women χ2 = 145.608df = 6, p < 0.0001, for men: χ2 = 90.613df = 6, p < 0.0001). Hearing problems were more prevalent among those with more symptoms of long-lasting stress. Similarly to the pattern for burnout, the prevalence increase was higher for women than for men, even if it was less pronounced for this variable. The association was positive (for all: Kendall's τ-b = 0.127 p < 0.0001, for women: Kendall's τ-b = 0.148 p < 0.0001, for men: Kendall's τ-b = 0.126 p < 0.0001) and Figure 4 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to symptoms of long-lasting stress (higher quartiles indicate more stress symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest long-lasting stress quartile were 2.06 for men (p < 0.001) compared to the adjusted odds ratios which were 2.42 (p < 0.001). For women, the corresponding values were 2.61 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).\n[SUBTITLE] Performance-based self-esteem [SUBSECTION] A statistically significant difference was found in the prevalence of hearing problems between those with higher and lower levels of PBS (χ2 = 39.946df = 6, p < 0.0001, for women χ2 = 36.410df = 6, p < 0.0001, for men: χ2 = 15.181df = 6, p < 0.05). Hearing problems were more prevalent among those with higher levels of PBS. The association was positive and linear and more pronounced for women than for men (for all: Kendall's τ-b = 0.056 p < 0.0001, for women: Kendall's τ-b = 0.073 p < 0.0001, for men: Kendall's τ-b = 0.043 p < 0.01) and Figure 5 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to performance-based self-esteem (higher quartiles indicate higher PBS). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest PBS quartile were 1.26 for men (p < 0.001) compared to the adjusted odds ratios which were 1.41 (p < 0.001). For women, the corresponding values were 1.64 (p < 0.001) and when adjusted the value was 1.80 (p < 0.001).\nA statistically significant difference was found in the prevalence of hearing problems between those with higher and lower levels of PBS (χ2 = 39.946df = 6, p < 0.0001, for women χ2 = 36.410df = 6, p < 0.0001, for men: χ2 = 15.181df = 6, p < 0.05). Hearing problems were more prevalent among those with higher levels of PBS. The association was positive and linear and more pronounced for women than for men (for all: Kendall's τ-b = 0.056 p < 0.0001, for women: Kendall's τ-b = 0.073 p < 0.0001, for men: Kendall's τ-b = 0.043 p < 0.01) and Figure 5 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to performance-based self-esteem (higher quartiles indicate higher PBS). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest PBS quartile were 1.26 for men (p < 0.001) compared to the adjusted odds ratios which were 1.41 (p < 0.001). For women, the corresponding values were 1.64 (p < 0.001) and when adjusted the value was 1.80 (p < 0.001).", "Table 1 illustrates the prevalence of hearing problems in relation to different work-related stressors. There was a statistically significant difference in the prevalence of hearing problems between those who were and were not exposed to work-related stressors or threats such as risk of being moved to another work/job against ones will (χ2 = 54.704df = 2, p < 0.0001) and threats of getting fired (χ2 = 27.095df = 2, p < 0.0001). There was however no statistically significant difference between those who were exposed to threats of bankruptcy compared to those who were not.\nPrevalence of hearing problems in relation to different work-related stressors.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems when exposed to a threat of being moved to another job against ones will (yes vs. no) were 1.39 for men (p < 0.001) compared to the adjusted odds ratios which were 1.43 (p < 0.001). For women, the corresponding values were 1.7 (p < 0.001) and when adjusted the value was 1.74 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of self-rated health (for all: χ2 = 262.522df = 6, p < 0.0001, for women χ2 = 141.535df = 6, p < 0.0001, for men: χ2 = 123.306df = 6, p < 0.0001). Figure 1 clearly demonstrates that poorer SRH is associated with a higher prevalence of hearing problems. The association is negative (for all: Kendall's τ-b = -0.139 p < 0.0001, for women: Kendall's τ-b = -0.142 p < 0.0001, for men: Kendall's τ-b = -0.135 p < 0.0001) and indicates increasing prevalence of hearing problems among those with poorer health.\nPrevalence of hearing problems in percent in relation to different ratings of SRH. The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. The unadjusted odds ratio of having hearing problems when reporting poor vs. good SRH were 3.89 for men (p < 0.001) compared to the adjusted odds ratios which were 3.29 (p < 0.001). For women, the corresponding values were 3.81 (p < 0.001) and when adjusted the value was 3.49 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between individuals with different handicaps and those without (χ2 = 181.650df = 2, p < 0.0001). Those with long-term handicaps and illnesses reported more hearing problems (Table 2).\nPrevalence of hearing problems in relation to long-term illness, inconvenience after an accident, any handicap or other weakness.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For instance, the unadjusted odds ratio of having hearing problems after long-term illness (yes vs. no) were 1.92 for men (p < 0.001) compared to the adjusted odds ratios which were 1.76 (p < 0.001). For women, the corresponding values were 1.72 (p < 0.001) and when adjusted the value was 1.64 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between individuals with different levels of sleep quality (for all: χ2 = 169.875df = 8, p < 0.0001, for women χ2 = 112.509df = 8, p < 0.0001, for men: χ2 = 75.068df = 8, p < 0.0001). The association was negative (for all: Kendall's τ-b = -0.116 p < 0.0001, for women: Kendall's τ-b = -0.126 p < 0.0001, for men: Kendall's τ-b = -0.113 p < 0.0001) and Figure 2 demonstrates that poorer sleep quality is associated with a higher prevalence of hearing problems. As tinnitus and hearing complaints may differ with regard to sleep, both of these variables were also analyzed separately. The prevalence of sleeping problems was significantly higher among those reporting tinnitus (χ2 = 126.884df = 4, p < 0.0001, Kendall's τ-b = -0.106 p < 0.0001) compared to those reporting hearing complaints (χ2 = 76.145df = 4, p < 0.0001, Kendall's τ-b = -0.081 p < 0.0001, see Table 3) and the associations were negative for both.\nPrevalence of hearing problems in percent in relation to different ratings of sleep quality. The Kendall's τ-b value is indicated by a \"τ\".\nPrevalence of sleeping problems among those with tinnitus and hearing loss respectively.\nNumber of participants and row percent in parentheses showing increasing prevalence of tinnitus with poorer sleep.\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. When the unadjusted odds ratio of having hearing problems when reporting poor vs. good sleep quality were 2.74 for men (p < 0.001) compared to the adjusted odds ratios which were 2.67 (p < 0.001). For women, the corresponding values were 3.51 (p < 0.001) and when adjusted the value was 3.24 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between those with higher burnout scores compared to those with lower scores (χ2 = 214.473df = 6, p < 0.0001, for women χ2 = 159.205df = 6, p < 0.0001, for men: χ2 = 98.935df = 6, p < 0.0001). Hearing problems were significantly more prevalent among those with higher burnout scores. Multivariate analyses showed no age, gender or SES related differences in prevalence increases with increasing burnout scores. The association was positive (for all: Kendall's τ-b = 0.129 p < 0.0001, for women: Kendall's τ-b = 0.150 p < 0.0001, for men: Kendall's τ-b = 0.129 p < 0.0001) and Figure 3 demonstrates that higher burnout scores are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to different burnout scores (higher quartiles indicate more severe burnout symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nThe proportional odds model did not exhibit any differences in odds ratios when age and socioeconomic status were included in models for neither women nor men. For example, the unadjusted odds ratio of having hearing problems when being in the highest vs. lowest burnout quartile were 2.36 for men (p < 0.001) compared to the adjusted odds ratios which were 2.63 (p < 0.001). For women, the corresponding values were 2.80 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).", "There was a statistically significant difference in the prevalence of hearing problems between those with more symptoms of long-lasting stress scores compared to those with less (χ2 = 196.855df = 6, p < 0.0001, for women χ2 = 145.608df = 6, p < 0.0001, for men: χ2 = 90.613df = 6, p < 0.0001). Hearing problems were more prevalent among those with more symptoms of long-lasting stress. Similarly to the pattern for burnout, the prevalence increase was higher for women than for men, even if it was less pronounced for this variable. The association was positive (for all: Kendall's τ-b = 0.127 p < 0.0001, for women: Kendall's τ-b = 0.148 p < 0.0001, for men: Kendall's τ-b = 0.126 p < 0.0001) and Figure 4 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to symptoms of long-lasting stress (higher quartiles indicate more stress symptoms). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest long-lasting stress quartile were 2.06 for men (p < 0.001) compared to the adjusted odds ratios which were 2.42 (p < 0.001). For women, the corresponding values were 2.61 (p < 0.001) and when adjusted the value was 2.79 (p < 0.001).", "A statistically significant difference was found in the prevalence of hearing problems between those with higher and lower levels of PBS (χ2 = 39.946df = 6, p < 0.0001, for women χ2 = 36.410df = 6, p < 0.0001, for men: χ2 = 15.181df = 6, p < 0.05). Hearing problems were more prevalent among those with higher levels of PBS. The association was positive and linear and more pronounced for women than for men (for all: Kendall's τ-b = 0.056 p < 0.0001, for women: Kendall's τ-b = 0.073 p < 0.0001, for men: Kendall's τ-b = 0.043 p < 0.01) and Figure 5 demonstrates that more symptoms of long-lasting stress are associated with a higher prevalence of hearing problems.\nPrevalence of hearing problems in percent in relation to performance-based self-esteem (higher quartiles indicate higher PBS). The Kendall's τ-b value is indicated by a \"τ\".\nNo major differences in odds ratios were found when age and socioeconomic status were included in models for women or men. The unadjusted odds ratio of having hearing problems when being in the highest vs. lowest PBS quartile were 1.26 for men (p < 0.001) compared to the adjusted odds ratios which were 1.41 (p < 0.001). For women, the corresponding values were 1.64 (p < 0.001) and when adjusted the value was 1.80 (p < 0.001).", "The aim of the present study was to assess the prevalence of two common hearing problems, i.e. hearing complaints and tinnitus, in relation to different work-, life- and health-related stressors. The results demonstrate a clear and mostly linear relationship between higher prevalence of hearing problems (tinnitus or hearing complaints or both) and different stressors, symptoms of ill health and stress as well as poor sleep. Thus, the salient features of the present study illustrate that occupational stressors, poorer self-rated health, long-term illness, poorer sleep quality, higher burnout scores, more symptoms of long-lasting stress, and higher performance-based self-esteem are statistically significantly associated with a higher prevalence of hearing problems. To our knowledge, this is the first time that such vivid and systematic associations have been found for hearing problems (hearing complaints and tinnitus) in an extensive human population.\n[SUBTITLE] Gender issues [SUBSECTION] The findings were consistent for both males and females, and men showed a slightly higher prevalence of hearing problems in absolute levels for all items analyzed except for performance-based self-esteem. Previously it has been found that the prevalence of hearing problems is higher among men and that prevalence increases with age [38]. The present findings elaborate on the previous findings and emphasize the correlation between poorer health, stress-related symptoms and hearing problems that affect men to a greater extent than women. Since hearing problems are becoming a major public health issue it is imperative to understand the factors that are directly and indirectly related to the underlying cause. The findings from the present study shed light on new attributes, biologically- and stress-related, that may be associated with the prevalence of hearing problems. The gender differences need to be interpreted in light of the fact that men and women may be exposed to different types of work environments and exposures that may affect the prevalence of hearing problems.\nIt is also noteworthy to mention that in the present study we have assessed two different factors for hearing problems. First we analyzed the prevalence of hearing problems either for tinnitus or hearing complaints and secondly for both tinnitus and hearing complaints. This assessment is more realistic since it is common for individuals who have hearing complaints to have tinnitus and vice versa. In a previous study, the prevalence of tinnitus was higher for men than women (5% vs. 3%, resp.)[6]. When analyzed for how often the tinnitus occurred, among those who answered that they suffer all the time, the prevalence for men was 10% vs. 5% for women. Interestingly, the effects of higher burnout scores, more symptoms of long-lasting stress and poor sleep quality also demonstrated a male dominance.\nThe findings were consistent for both males and females, and men showed a slightly higher prevalence of hearing problems in absolute levels for all items analyzed except for performance-based self-esteem. Previously it has been found that the prevalence of hearing problems is higher among men and that prevalence increases with age [38]. The present findings elaborate on the previous findings and emphasize the correlation between poorer health, stress-related symptoms and hearing problems that affect men to a greater extent than women. Since hearing problems are becoming a major public health issue it is imperative to understand the factors that are directly and indirectly related to the underlying cause. The findings from the present study shed light on new attributes, biologically- and stress-related, that may be associated with the prevalence of hearing problems. The gender differences need to be interpreted in light of the fact that men and women may be exposed to different types of work environments and exposures that may affect the prevalence of hearing problems.\nIt is also noteworthy to mention that in the present study we have assessed two different factors for hearing problems. First we analyzed the prevalence of hearing problems either for tinnitus or hearing complaints and secondly for both tinnitus and hearing complaints. This assessment is more realistic since it is common for individuals who have hearing complaints to have tinnitus and vice versa. In a previous study, the prevalence of tinnitus was higher for men than women (5% vs. 3%, resp.)[6]. When analyzed for how often the tinnitus occurred, among those who answered that they suffer all the time, the prevalence for men was 10% vs. 5% for women. Interestingly, the effects of higher burnout scores, more symptoms of long-lasting stress and poor sleep quality also demonstrated a male dominance.\n[SUBTITLE] Occupational consequences [SUBSECTION] Hearing problems were also more common among individuals exposed to occupational stressors, i.e. employment-related stress such as risk of being moved to another work/job against ones will and threats of getting fired. Threats of bankruptcy were not significantly associated with a higher prevalence of hearing problems. These problems include difficulties in communicating, lower awareness for important surrounding sounds such as telephone calls or warning signals or oversensitivity to sounds (hyperacusis) and even reduce productivity due to emotional exhaustion. These burdens become exaggerated for individuals with long-lasting stress symptoms and high burnout scores, as demonstrated in the present study. This is important for many reasons including the public awareness and the clarification that long-lasting stress and burnout symptoms are risk factors correlated with hearing problems.\nHearing problems were also more common among individuals exposed to occupational stressors, i.e. employment-related stress such as risk of being moved to another work/job against ones will and threats of getting fired. Threats of bankruptcy were not significantly associated with a higher prevalence of hearing problems. These problems include difficulties in communicating, lower awareness for important surrounding sounds such as telephone calls or warning signals or oversensitivity to sounds (hyperacusis) and even reduce productivity due to emotional exhaustion. These burdens become exaggerated for individuals with long-lasting stress symptoms and high burnout scores, as demonstrated in the present study. This is important for many reasons including the public awareness and the clarification that long-lasting stress and burnout symptoms are risk factors correlated with hearing problems.\n[SUBTITLE] Self-rated health [SUBSECTION] Self-rated health (SRH) is one of the most widely used single measures of perceived current health status [28]. There is extensive evidence suggesting that SRH is a potent predictor of future mortality and morbidity [39,40], functional decline and disability, as well as the utilization of health care [39,41]. Most previous studies have shown that SRH is an independent predictor of future health outcomes, even after adjusting for self-ratings of other health-related measures, physician-reported health status, behavioral and psychosocial risk factors, socioeconomic status and environmental factors. Nevertheless, debate still continues about what SRH really represents [39,42,43].\nIt has been proposed that SRH represents an individual's general perception of health, including biological, psychological and social dimensions. Therefore SRH might be more sensitive in health monitoring than external measures of health [44]. Furthermore, it has been indicated that risk associated with poor SRH status is higher than that associated with poor objective health measures [45]. On the other hand, Kaplan & Camacho [46] found that objective health status has a stronger relationship with mortality than SRH.\nA poorer self-rated health was correlated to hearing problems. Poor self-rated health can be influenced by several factors including social and marital status (Lindström, 2009), type of employment [47], sleep quality, sense of coherence, self-esteem, social support [28], higher cytokine levels [48], allostatic load [49] and unhealthy habits [50]. Unhealthy habits could include careless use of hearing protectors when in noisy environments or being exposed to excessive sound stimulation over long durations. Moreover, self-rated health is a reflection of the individual's quality of life and social abilities [51]. Depending on the severity of a particular hearing problem, consequent problems in communication and social interactions would result. This can cause problems in daily life as well as in the work environment. Since self-rated health is a compilation of biological, psychological and social assessments it may be a more accurate account of the consequences of hearing disabilities of the individual compared to, for example, pure tone audiometry.\nSelf-rated health (SRH) is one of the most widely used single measures of perceived current health status [28]. There is extensive evidence suggesting that SRH is a potent predictor of future mortality and morbidity [39,40], functional decline and disability, as well as the utilization of health care [39,41]. Most previous studies have shown that SRH is an independent predictor of future health outcomes, even after adjusting for self-ratings of other health-related measures, physician-reported health status, behavioral and psychosocial risk factors, socioeconomic status and environmental factors. Nevertheless, debate still continues about what SRH really represents [39,42,43].\nIt has been proposed that SRH represents an individual's general perception of health, including biological, psychological and social dimensions. Therefore SRH might be more sensitive in health monitoring than external measures of health [44]. Furthermore, it has been indicated that risk associated with poor SRH status is higher than that associated with poor objective health measures [45]. On the other hand, Kaplan & Camacho [46] found that objective health status has a stronger relationship with mortality than SRH.\nA poorer self-rated health was correlated to hearing problems. Poor self-rated health can be influenced by several factors including social and marital status (Lindström, 2009), type of employment [47], sleep quality, sense of coherence, self-esteem, social support [28], higher cytokine levels [48], allostatic load [49] and unhealthy habits [50]. Unhealthy habits could include careless use of hearing protectors when in noisy environments or being exposed to excessive sound stimulation over long durations. Moreover, self-rated health is a reflection of the individual's quality of life and social abilities [51]. Depending on the severity of a particular hearing problem, consequent problems in communication and social interactions would result. This can cause problems in daily life as well as in the work environment. Since self-rated health is a compilation of biological, psychological and social assessments it may be a more accurate account of the consequences of hearing disabilities of the individual compared to, for example, pure tone audiometry.\n[SUBTITLE] Sleep quality [SUBSECTION] Poor sleep quality was found to be associated with a higher prevalence of hearing problems in both men and women. In a study aimed at evaluating age specific prevalence of general symptoms it was found that five symptoms increased with increasing age [52]. These symptoms included insomnia, leg pain, joint pain, eye problems and impaired hearing. In another population study, patients with temporomandibular disorders were characterized for the auditory health [53]. The subjects who displayed auditory problems such as tinnitus, and perceived hearing complaints were found to be significantly more likely to consider themselves in poor health and have sleep disturbances than those without auditory problems. Two more studies have found correlations between sleep quality and hearing loss as well as tinnitus [54,55]. Thus, a correlation between poorer sleep quality and hearing problems is apparent in several different populations. In the present study, tinnitus was the more common disturbance when trying to sleep, probably since the perceived sound would increase annoyance and stress reactions. However, hearing complaints and tinnitus are often related co-morbid symptoms [38,56,57] and in the present study worse sleep quality was associated with higher prevalence of hearing complaints, albeit to a lesser extent than tinnitus.\nIn a previous study we found that individuals with hearing problems have a poorer ability to unwind or activate their parasympathetic system [24]. The present study confirms and strengthens this association since problems unwinding are strongly related to sleeping problems.\nPoor sleep quality was found to be associated with a higher prevalence of hearing problems in both men and women. In a study aimed at evaluating age specific prevalence of general symptoms it was found that five symptoms increased with increasing age [52]. These symptoms included insomnia, leg pain, joint pain, eye problems and impaired hearing. In another population study, patients with temporomandibular disorders were characterized for the auditory health [53]. The subjects who displayed auditory problems such as tinnitus, and perceived hearing complaints were found to be significantly more likely to consider themselves in poor health and have sleep disturbances than those without auditory problems. Two more studies have found correlations between sleep quality and hearing loss as well as tinnitus [54,55]. Thus, a correlation between poorer sleep quality and hearing problems is apparent in several different populations. In the present study, tinnitus was the more common disturbance when trying to sleep, probably since the perceived sound would increase annoyance and stress reactions. However, hearing complaints and tinnitus are often related co-morbid symptoms [38,56,57] and in the present study worse sleep quality was associated with higher prevalence of hearing complaints, albeit to a lesser extent than tinnitus.\nIn a previous study we found that individuals with hearing problems have a poorer ability to unwind or activate their parasympathetic system [24]. The present study confirms and strengthens this association since problems unwinding are strongly related to sleeping problems.\n[SUBTITLE] Burnout and symptoms of long-lasting stress [SUBSECTION] It is by now well established that stress-related disorders, such as burnout, are associated with several forms of co-morbidity [58-62], e.g. long-term pain and psychiatric disorders. Now hearing problems may be added to the list. Our clinical experiences from a stress clinic shows that patients are often referred from audiologists. The present study showed clear associations between burnout as well as symptoms of long-lasting stress and hearing problems. Thus, hearing problems should be taken into account clinically in the diagnosis and treatment of stress-related disorders and vice versa. Furthermore, hearing problems were also more common among those with long-term illness, pains, inconveniences or handicaps. This strengthens the hypothesis about high levels of co-morbidity among individuals with hearing problems.\nThe relation between long-term stress and tinnitus has been previously explored and tinnitus sufferers often report enhanced problems by stress and fatigue [63]. It remains to be determined if tinnitus is a direct or indirect response of the auditory system to stress. Non-auditory-related brain regions, such as the limbic system have been shown to be activated during tinnitus [64]. The involvement of limbic areas during tinnitus offers a neuroanatomical correlate for stress-related tinnitus since the limbic region is a key region for regulating stress responses. Moreover, a major therapeutic strategy for tinnitus patients includes relaxation programs which have proven successful for many sufferers. Thus, taking into consideration the results of individuals with hearing problems (hearing complaints and tinnitus) of the present study and findings from prior tinnitus studies, it is becoming more apparent that stress can increase the prevalence of hearing problems. Furthermore, it has been shown that individuals with hearing problems have a worsened ability to unwind and activate the parasympathetic system [24].\nIt is by now well established that stress-related disorders, such as burnout, are associated with several forms of co-morbidity [58-62], e.g. long-term pain and psychiatric disorders. Now hearing problems may be added to the list. Our clinical experiences from a stress clinic shows that patients are often referred from audiologists. The present study showed clear associations between burnout as well as symptoms of long-lasting stress and hearing problems. Thus, hearing problems should be taken into account clinically in the diagnosis and treatment of stress-related disorders and vice versa. Furthermore, hearing problems were also more common among those with long-term illness, pains, inconveniences or handicaps. This strengthens the hypothesis about high levels of co-morbidity among individuals with hearing problems.\nThe relation between long-term stress and tinnitus has been previously explored and tinnitus sufferers often report enhanced problems by stress and fatigue [63]. It remains to be determined if tinnitus is a direct or indirect response of the auditory system to stress. Non-auditory-related brain regions, such as the limbic system have been shown to be activated during tinnitus [64]. The involvement of limbic areas during tinnitus offers a neuroanatomical correlate for stress-related tinnitus since the limbic region is a key region for regulating stress responses. Moreover, a major therapeutic strategy for tinnitus patients includes relaxation programs which have proven successful for many sufferers. Thus, taking into consideration the results of individuals with hearing problems (hearing complaints and tinnitus) of the present study and findings from prior tinnitus studies, it is becoming more apparent that stress can increase the prevalence of hearing problems. Furthermore, it has been shown that individuals with hearing problems have a worsened ability to unwind and activate the parasympathetic system [24].\n[SUBTITLE] Performance-based self-esteem [SUBSECTION] In the present study there was a higher prevalence of hearing problems among those with higher PBS. High PBS is associated with high but vulnerable engagement, as well as higher prevalence of sickness presenteeism. When combined with high scores on a burnout scale it is a strong predictor of burnout and increased risk (OR = 2.84, CI 95% 1.61-5.01) for long-term sick leave [65]. This finding point to the direction that personality factors, especially PBS, may be important when assessing the risks for having hearing problems. It also indicates that hearing problems are multidimensional as they, apart from different stressors, also may be associated with other personality factors. These factors are by and of themselves complicated and multidimensional phenomena that have not yet been completely described. Therefore, more research should be directed at studying these interactions.\nIn the present study there was a higher prevalence of hearing problems among those with higher PBS. High PBS is associated with high but vulnerable engagement, as well as higher prevalence of sickness presenteeism. When combined with high scores on a burnout scale it is a strong predictor of burnout and increased risk (OR = 2.84, CI 95% 1.61-5.01) for long-term sick leave [65]. This finding point to the direction that personality factors, especially PBS, may be important when assessing the risks for having hearing problems. It also indicates that hearing problems are multidimensional as they, apart from different stressors, also may be associated with other personality factors. These factors are by and of themselves complicated and multidimensional phenomena that have not yet been completely described. Therefore, more research should be directed at studying these interactions.\n[SUBTITLE] Strengths and limitations [SUBSECTION] One of the major strengths with this study is the large sample size and that the sample is representative for the general Swedish working population. One weakness is the cross-sectional design, which does not allow conclusions about causality. Accordingly, prospective studies, which are forthcoming from our group, will be needed to illuminate the extent to which the observed associations are causal. Also, the study population is typical of a post-industrial country, with a high proportion of participants having a high educational level. It is possible that associations would be different in a population with a higher proportion of blue collar workers.\nThe assessment of hearing complaints and tinnitus via questionnaires has its advantages and disadvantages and there is no consensus as to which method is most valid. For example, the subjective evaluation of hearing problems may be difficult to interpret when there is a mild hearing loss. However, individuals having constant difficulties in understanding speech (especially in background noise) are usually well aware of their problem. It must be pointed out that an individual who has difficulties in understanding speech in background noise can have a completely normal audiogram. Thus, there are also limitations to audiological testing. This is not to say that an audiological test would not complement our subjective ratings, but we argue that addressing the question if there are \"hearing problems\" is acceptable. In fact, rating scales are commonly utilized to assess hearing problems [5,66] and it has been shown that the single question: 'Do you feel that you have a hearing loss?' was the most sensitive to assess hearing loss compared with pure-tone air conduction audiometry.\nOne of the major strengths with this study is the large sample size and that the sample is representative for the general Swedish working population. One weakness is the cross-sectional design, which does not allow conclusions about causality. Accordingly, prospective studies, which are forthcoming from our group, will be needed to illuminate the extent to which the observed associations are causal. Also, the study population is typical of a post-industrial country, with a high proportion of participants having a high educational level. It is possible that associations would be different in a population with a higher proportion of blue collar workers.\nThe assessment of hearing complaints and tinnitus via questionnaires has its advantages and disadvantages and there is no consensus as to which method is most valid. For example, the subjective evaluation of hearing problems may be difficult to interpret when there is a mild hearing loss. However, individuals having constant difficulties in understanding speech (especially in background noise) are usually well aware of their problem. It must be pointed out that an individual who has difficulties in understanding speech in background noise can have a completely normal audiogram. Thus, there are also limitations to audiological testing. This is not to say that an audiological test would not complement our subjective ratings, but we argue that addressing the question if there are \"hearing problems\" is acceptable. In fact, rating scales are commonly utilized to assess hearing problems [5,66] and it has been shown that the single question: 'Do you feel that you have a hearing loss?' was the most sensitive to assess hearing loss compared with pure-tone air conduction audiometry.", "The findings were consistent for both males and females, and men showed a slightly higher prevalence of hearing problems in absolute levels for all items analyzed except for performance-based self-esteem. Previously it has been found that the prevalence of hearing problems is higher among men and that prevalence increases with age [38]. The present findings elaborate on the previous findings and emphasize the correlation between poorer health, stress-related symptoms and hearing problems that affect men to a greater extent than women. Since hearing problems are becoming a major public health issue it is imperative to understand the factors that are directly and indirectly related to the underlying cause. The findings from the present study shed light on new attributes, biologically- and stress-related, that may be associated with the prevalence of hearing problems. The gender differences need to be interpreted in light of the fact that men and women may be exposed to different types of work environments and exposures that may affect the prevalence of hearing problems.\nIt is also noteworthy to mention that in the present study we have assessed two different factors for hearing problems. First we analyzed the prevalence of hearing problems either for tinnitus or hearing complaints and secondly for both tinnitus and hearing complaints. This assessment is more realistic since it is common for individuals who have hearing complaints to have tinnitus and vice versa. In a previous study, the prevalence of tinnitus was higher for men than women (5% vs. 3%, resp.)[6]. When analyzed for how often the tinnitus occurred, among those who answered that they suffer all the time, the prevalence for men was 10% vs. 5% for women. Interestingly, the effects of higher burnout scores, more symptoms of long-lasting stress and poor sleep quality also demonstrated a male dominance.", "Hearing problems were also more common among individuals exposed to occupational stressors, i.e. employment-related stress such as risk of being moved to another work/job against ones will and threats of getting fired. Threats of bankruptcy were not significantly associated with a higher prevalence of hearing problems. These problems include difficulties in communicating, lower awareness for important surrounding sounds such as telephone calls or warning signals or oversensitivity to sounds (hyperacusis) and even reduce productivity due to emotional exhaustion. These burdens become exaggerated for individuals with long-lasting stress symptoms and high burnout scores, as demonstrated in the present study. This is important for many reasons including the public awareness and the clarification that long-lasting stress and burnout symptoms are risk factors correlated with hearing problems.", "Self-rated health (SRH) is one of the most widely used single measures of perceived current health status [28]. There is extensive evidence suggesting that SRH is a potent predictor of future mortality and morbidity [39,40], functional decline and disability, as well as the utilization of health care [39,41]. Most previous studies have shown that SRH is an independent predictor of future health outcomes, even after adjusting for self-ratings of other health-related measures, physician-reported health status, behavioral and psychosocial risk factors, socioeconomic status and environmental factors. Nevertheless, debate still continues about what SRH really represents [39,42,43].\nIt has been proposed that SRH represents an individual's general perception of health, including biological, psychological and social dimensions. Therefore SRH might be more sensitive in health monitoring than external measures of health [44]. Furthermore, it has been indicated that risk associated with poor SRH status is higher than that associated with poor objective health measures [45]. On the other hand, Kaplan & Camacho [46] found that objective health status has a stronger relationship with mortality than SRH.\nA poorer self-rated health was correlated to hearing problems. Poor self-rated health can be influenced by several factors including social and marital status (Lindström, 2009), type of employment [47], sleep quality, sense of coherence, self-esteem, social support [28], higher cytokine levels [48], allostatic load [49] and unhealthy habits [50]. Unhealthy habits could include careless use of hearing protectors when in noisy environments or being exposed to excessive sound stimulation over long durations. Moreover, self-rated health is a reflection of the individual's quality of life and social abilities [51]. Depending on the severity of a particular hearing problem, consequent problems in communication and social interactions would result. This can cause problems in daily life as well as in the work environment. Since self-rated health is a compilation of biological, psychological and social assessments it may be a more accurate account of the consequences of hearing disabilities of the individual compared to, for example, pure tone audiometry.", "Poor sleep quality was found to be associated with a higher prevalence of hearing problems in both men and women. In a study aimed at evaluating age specific prevalence of general symptoms it was found that five symptoms increased with increasing age [52]. These symptoms included insomnia, leg pain, joint pain, eye problems and impaired hearing. In another population study, patients with temporomandibular disorders were characterized for the auditory health [53]. The subjects who displayed auditory problems such as tinnitus, and perceived hearing complaints were found to be significantly more likely to consider themselves in poor health and have sleep disturbances than those without auditory problems. Two more studies have found correlations between sleep quality and hearing loss as well as tinnitus [54,55]. Thus, a correlation between poorer sleep quality and hearing problems is apparent in several different populations. In the present study, tinnitus was the more common disturbance when trying to sleep, probably since the perceived sound would increase annoyance and stress reactions. However, hearing complaints and tinnitus are often related co-morbid symptoms [38,56,57] and in the present study worse sleep quality was associated with higher prevalence of hearing complaints, albeit to a lesser extent than tinnitus.\nIn a previous study we found that individuals with hearing problems have a poorer ability to unwind or activate their parasympathetic system [24]. The present study confirms and strengthens this association since problems unwinding are strongly related to sleeping problems.", "It is by now well established that stress-related disorders, such as burnout, are associated with several forms of co-morbidity [58-62], e.g. long-term pain and psychiatric disorders. Now hearing problems may be added to the list. Our clinical experiences from a stress clinic shows that patients are often referred from audiologists. The present study showed clear associations between burnout as well as symptoms of long-lasting stress and hearing problems. Thus, hearing problems should be taken into account clinically in the diagnosis and treatment of stress-related disorders and vice versa. Furthermore, hearing problems were also more common among those with long-term illness, pains, inconveniences or handicaps. This strengthens the hypothesis about high levels of co-morbidity among individuals with hearing problems.\nThe relation between long-term stress and tinnitus has been previously explored and tinnitus sufferers often report enhanced problems by stress and fatigue [63]. It remains to be determined if tinnitus is a direct or indirect response of the auditory system to stress. Non-auditory-related brain regions, such as the limbic system have been shown to be activated during tinnitus [64]. The involvement of limbic areas during tinnitus offers a neuroanatomical correlate for stress-related tinnitus since the limbic region is a key region for regulating stress responses. Moreover, a major therapeutic strategy for tinnitus patients includes relaxation programs which have proven successful for many sufferers. Thus, taking into consideration the results of individuals with hearing problems (hearing complaints and tinnitus) of the present study and findings from prior tinnitus studies, it is becoming more apparent that stress can increase the prevalence of hearing problems. Furthermore, it has been shown that individuals with hearing problems have a worsened ability to unwind and activate the parasympathetic system [24].", "In the present study there was a higher prevalence of hearing problems among those with higher PBS. High PBS is associated with high but vulnerable engagement, as well as higher prevalence of sickness presenteeism. When combined with high scores on a burnout scale it is a strong predictor of burnout and increased risk (OR = 2.84, CI 95% 1.61-5.01) for long-term sick leave [65]. This finding point to the direction that personality factors, especially PBS, may be important when assessing the risks for having hearing problems. It also indicates that hearing problems are multidimensional as they, apart from different stressors, also may be associated with other personality factors. These factors are by and of themselves complicated and multidimensional phenomena that have not yet been completely described. Therefore, more research should be directed at studying these interactions.", "One of the major strengths with this study is the large sample size and that the sample is representative for the general Swedish working population. One weakness is the cross-sectional design, which does not allow conclusions about causality. Accordingly, prospective studies, which are forthcoming from our group, will be needed to illuminate the extent to which the observed associations are causal. Also, the study population is typical of a post-industrial country, with a high proportion of participants having a high educational level. It is possible that associations would be different in a population with a higher proportion of blue collar workers.\nThe assessment of hearing complaints and tinnitus via questionnaires has its advantages and disadvantages and there is no consensus as to which method is most valid. For example, the subjective evaluation of hearing problems may be difficult to interpret when there is a mild hearing loss. However, individuals having constant difficulties in understanding speech (especially in background noise) are usually well aware of their problem. It must be pointed out that an individual who has difficulties in understanding speech in background noise can have a completely normal audiogram. Thus, there are also limitations to audiological testing. This is not to say that an audiological test would not complement our subjective ratings, but we argue that addressing the question if there are \"hearing problems\" is acceptable. In fact, rating scales are commonly utilized to assess hearing problems [5,66] and it has been shown that the single question: 'Do you feel that you have a hearing loss?' was the most sensitive to assess hearing loss compared with pure-tone air conduction audiometry.", "In conclusion, the present study unambiguously demonstrates associations between hearing problems and occupational stressors, poorer self-rated health, higher burnout scores, more symptoms of long-lasting stress, long-term illness and poorer sleep quality. The interaction between the prevalence of hearing problems and the above mentioned features have not been previously described for the auditory system. These findings also indicate that hearing problems are multidimensional, which warrants further investigations of possible predictors.", "The authors declare that they have no competing interests.", "DH conducted the statistical analyses and drafted the manuscript together with BC, MBW, with substantial and essential input from CL and TT. TT was PI of the project and CL data manager. All authors have read and approved the final version of the manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2458/11/130/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adult", "Aged", "Aged, 80 and over", "Cross-Sectional Studies", "Female", "Humans", "Insurance Claim Review", "Male", "Managed Care Programs", "Middle Aged", "Pulmonary Disease, Chronic Obstructive", "Retrospective Studies", "United States" ]

[ "Background", "Study design and data source", "Identification of patients with COPD", "Population demographics and comorbid conditions", "Complexity", "Exacerbations", "Utilization of healthcare services", "Data analysis", "Results", "Population characteristics", "Comparison of the commercial and Medicare COPD populations", "COPD exacerbations and health service utilization by disease complexity", "Discussion", "Comorbid conditions", "Exacerbations", "Utilization of healthcare services", "Conclusions", "Competing interests", "Authors' contributions", "Disclosure", "Pre-publication history" ]

[ "Chronic obstructive pulmonary disease (COPD) is the fourth leading cause of death among US adults and is projected to be the third by 2020 [1-3], although the disease is both preventable and treatable [4-6]. With this projected increased burden on healthcare systems, data describing how COPD patients are currently managed, together with information on COPD patients' healthcare utilization, are needed to inform healthcare organizations and providers. A small study of 1522 COPD patients in a health maintenance organization demonstrated that COPD patients had healthcare utilization and associated costs of more than twice those of age- and sex-matched controls [7]. Similarly, a larger study of over 100,000 patients aged ≥65 years with COPD or asthma showed the utilization of healthcare resources by these older COPD patients to be extremely high, both during hospitalization and after discharge [8]. The high prevalence of comorbidities in patients with COPD, especially respiratory conditions and cardiovascular disease, increases the morbidity associated with COPD [9-11].\nBecause of the burden imposed on healthcare systems and payers by patients with COPD, a means of identifying COPD patients who have higher healthcare utilization and associated higher costs is needed. The identification and stratification of COPD patients at risk of complications might facilitate management of these patients, improve care, and reduce costs. Furthermore, since COPD exacerbations, especially those requiring hospitalization, account for significant healthcare utilization, a method to accurately document COPD exacerbations using claims data would be very useful. Finally, comparing an older Medicare population with a working age commercial population will provide important information on how age influences the healthcare utilization in patients with COPD.\nThe purpose of this project is to describe a unique methodology that identifies COPD patients in a large managed care database and documents their demographics, comorbid conditions, and COPD exacerbations. We stratified these COPD patients by means of a novel algorithm of disease complexity (high, moderate, low complexity), used as a proxy of disease severity, and then examined the relationship between complexity of illness and key indicators of healthcare utilization and exacerbations. Furthermore, data were analyzed based on age group to determine if there are differences in COPD patients of an older Medicare (generally ≥65 years of age) population and a younger employer-based (< 65 years) population.", "This was a retrospective, cross-sectional analysis of US managed care administrative claims data from multiple health plans during the one-year study period, July 1, 2004, to June 30, 2005. The study population was extracted from a dataset of 12.4 million covered lives maintained by PharMetrics Inc (Watertown, MA, USA) from 19 health plans across the US: 3.2 million from the Northeast, 6.4 million from the Midwest, 1.8 million from the South, and 0.7 million from the West. The plans varied in size: 6 were <200,000 covered lives, 9 were between 200,001 and 1 million covered lives, and 4 were over 1 million covered lives. We evaluated the 7.79 million members who were continuously eligible during the study period for potential inclusion in the COPD cohort.\nAs differences likely exist between older and younger individuals, data from Medicare plans were evaluated separately from commercial health plans. The commercial population represented the employer-based managed care product offerings of Health Maintenance Organization, Preferred Provider Organization, and Point of Service plans. Due to the fragmentation of healthcare claims for the population aged older than 64 years, these patients were only included in Medicare analyses if they were continuously enrolled in a Medicare Risk product for the entire time span once becoming 65 years old.", "Patients were identified as having COPD if they were aged ≥40 years and had any one of the following:\n1. One inpatient hospitalization or one emergency room encounter with a COPD diagnosis (491.x [chronic bronchitis], 492.x [emphysema], or 496 [chronic airway obstruction]) listed in any position as a discharge diagnosis; or\n2. Two professional claims, with different dates of services, with a COPD diagnosis listed in any position; or\n3. A COPD-related surgical procedure (e.g. lung volume reduction) listed on either a professional or facility claim.", "The age and gender were determined for patients identified with COPD from claims data. Comorbid conditions were determined if claims data included diagnostic, procedures, and services codes (2004 ICD-9 CM, CPT-4, and HCPCS codes, respectively) for predetermined conditions (see Additional Files) during the reporting period. Condition frequencies were determined for respiratory and for non-respiratory comorbid conditions.", "A claims-based classification of COPD complexity was created to serve as a surrogate for COPD disease severity. Comorbid respiratory conditions and medical procedures at any time during the study period were used to assign patients to one of three disease complexity levels (high, moderate, or low) based on selected diagnostic, procedures and services codes (2004 ICD-9, CPT-4, and HCPCS), as detailed in Additional File 1. Examples of code descriptions that resulted in a patient being classified as high complexity include a claim for cor pulmonale, tuberculosis, or malignant neoplasm. Examples for moderate complexity include pneumonia, cyanosis, bronchoscopy or dependence on supplemental oxygen. If a COPD patient did not have any comorbid condition for high or moderate complexity (Additional Files), they were classified as low complexity.", "An algorithm was developed to identify and stratify COPD exacerbations using claims data (Table 1). Although there is controversy about the definition of a COPD exacerbation, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) defines an exacerbation as: \"an event in the natural course of the disease (COPD) characterized by a change in the patient's baseline dyspnea, cough and/or sputum that is beyond normal day-to-day variation, is acute in onset, and may warrant a change in regular medication\" [1]. An exacerbation in this study was defined, as outlined in Table 1, by either a primary diagnosis recorded by a medical provider and coded in claims or a medication for an oral antibiotic commonly used for respiratory infection, or systemic steroid received by a patient during any 14-day time period. Exacerbation severity was classified by site of care reflecting resource utilization.\nAlgorithm for identification and classification of exacerbations in COPD patients\nTotal exacerbations included exacerbations of any severity (inpatient, emergency room, ambulatory by qualifying diagnosis, or ambulatory supported by drug therapy). Exacerbation frequency is reported for both total and hospitalized exacerbations as percent of patients experiencing ≥1 exacerbation(s), percent of patients experiencing multiple (≥2) exacerbations, and number of exacerbations/COPD patient.", "The healthcare services assessed in this study included tests, procedures (including surgeries), and visits considered to be related with COPD (see Additional File 1). The most common or relevant of these are included in this report. These services were identified based on 2004 CPT-4 and HCPCS codes collected in claims. Hospital admissions and emergency room visits for any reason were identified based on the presence of a facility claim for an inpatient hospital stay or an outpatient emergency room visit.", "DTEC™software (Pfizer, New York, NY, USA, Version 3.3) was used to integrate administrative data and claims files, identify and stratify patients with COPD, as well as characterize demographics, comorbidities, and utilization of healthcare services. All of these analyses were specified prior to study initiation, and programmed in the software. These analyses, including the algorithms for COPD complexity stratification and exacerbation identification, were developed by a panel of experts that included pulmonologists, outcomes researchers and claims-based research consultants. They are based upon information from accepted guidelines [1,12] but also incorporate previous experiences of the panel in claims-based research. While DTEC™, a proprietary software program, was used for this analysis, the algorithms for COPD disease identification and stratification included in the software are specifically outlined and included in the body of this article or Additional File 1 so that they may be used in other claims querying systems.\nClaims data during the 12-month study period were analyzed, and are presented as means with standard deviations. Categorical data are presented as numbers and percentages. Mean data were compared using the Student's t-test for normally distributed values and the Wilcoxon rank-sum test for non-normally distributed values. Categorical data were compared between commercial and Medicare populations before stratifying for complexity, using Chi-square test. Data were analyzed using GraphPad Prism® statistical software (version 5.0 for Windows; GraphPad Software Inc., CA, USA). Because of large sample sizes a p-value of 0.01 was designated as being statistically significant for all comparisons. The database was compiled in accordance with all aspects of the Health Information Portability and Accountability Act (HIPAA) of 1996.", "[SUBTITLE] Population characteristics [SUBSECTION] Eligible patients from the commercial or Medicare cohorts were identified as having COPD, and then stratified by complexity as shown in Figure 1. From the 7,671,018 health plan members in the commercial dataset, 42,565 (0.55%) met the criteria for COPD (Table 2). The median age was 56 years and 51.4% of patients were female. From the 115,652 health plan members in the Medicare dataset, 8,507 (7.4%) were identified as having COPD. In this dataset, the median age was 75 years and 53.1% were female. Although over half of both cohorts were classified with moderate- or high complexity disease, a larger proportion of the older Medicare cohort were classified as either high or moderate complexity (63.4%) compared with the younger commercial cohort (55.0%). This was mostly due to a higher percentage of high complexity patients in the Medicare cohort (Table 2).\nStratification of patients in the study.\nDemographic characteristics for the commercial and Medicare data sets, stratified by complexity\nEligible patients from the commercial or Medicare cohorts were identified as having COPD, and then stratified by complexity as shown in Figure 1. From the 7,671,018 health plan members in the commercial dataset, 42,565 (0.55%) met the criteria for COPD (Table 2). The median age was 56 years and 51.4% of patients were female. From the 115,652 health plan members in the Medicare dataset, 8,507 (7.4%) were identified as having COPD. In this dataset, the median age was 75 years and 53.1% were female. Although over half of both cohorts were classified with moderate- or high complexity disease, a larger proportion of the older Medicare cohort were classified as either high or moderate complexity (63.4%) compared with the younger commercial cohort (55.0%). This was mostly due to a higher percentage of high complexity patients in the Medicare cohort (Table 2).\nStratification of patients in the study.\nDemographic characteristics for the commercial and Medicare data sets, stratified by complexity\n[SUBTITLE] Comparison of the commercial and Medicare COPD populations [SUBSECTION] Specific differences were identified in the prevalence of comorbid conditions between the commercial and Medicare data sets (Table 3). For example, most cardiovascular complications were significantly less common in the younger commercial cohort than the older Medicare cohort, including hypertension, ischemic heart disease, and heart failure. Those in the younger commercial cohort were significantly more likely to have episodes of upper respiratory tract complaints including allergic rhinitis and sinusitis compared with the older Medicare cohort, who were more likely to be diagnosed with pneumonia. Furthermore, a significantly higher proportion of patients in the younger commercial database compared with the older Medicare dataset were documented to have current tobacco use during the study period.\nPrevalence (%) of comorbid conditions in COPD patients from commercial and Medicare data sets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nMost COPD patients in both the younger commercial cohort and the older Medicare cohort experienced at least 1 exacerbation during the study period (Table 4). Patients in the commercial cohort were more likely than Medicare members to experience an exacerbation of any type (commercial 70.6% and Medicare 61.0%), and the number of exacerbations per-patient in the commercial population was higher than in the Medicare population (2.13 vs. 1.58 per patient per year). However, patients in the older Medicare group were more likely than commercial patients to experience an exacerbation that led to hospitalization (19.4% vs. 13.9%). Furthermore, Medicare patients had more hospitalized exacerbations (0.25 per patient per year) than commercial patients (0.17 per patient per year). Multiple exacerbations (≥2) occurred more commonly in commercial patients (47.3% of commercial patients and 35.9% of Medicare patients), but multiple hospitalized exacerbations occurred in 2.3% and 3.6% of patients in each group, respectively.\nPrevalence (%) of patients experiencing ≥1 exacerbation and health services utilization among COPD patients in the commercial and Medicare datasets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nImportant differences were identified in health service utilization between cohorts (Table 4). While nearly all of both populations experienced at least 1 office visit consultation, Medicare patients were significantly more likely to have been hospitalized for any reason than commercial patients. Although standard chest X-ray, ECG's, chest/thorax CT/MRIs, and heart echo exams were more likely in Medicare patients, more of the younger commercial patients had pulmonary function testing, cardiac catheterization/coronary angiography, and outpatient respiratory therapy. Only 39.0% of the total cohort had pulmonary function testing during the study period (40.0% commercial vs. 33.2% Medicare).\nSpecific differences were identified in the prevalence of comorbid conditions between the commercial and Medicare data sets (Table 3). For example, most cardiovascular complications were significantly less common in the younger commercial cohort than the older Medicare cohort, including hypertension, ischemic heart disease, and heart failure. Those in the younger commercial cohort were significantly more likely to have episodes of upper respiratory tract complaints including allergic rhinitis and sinusitis compared with the older Medicare cohort, who were more likely to be diagnosed with pneumonia. Furthermore, a significantly higher proportion of patients in the younger commercial database compared with the older Medicare dataset were documented to have current tobacco use during the study period.\nPrevalence (%) of comorbid conditions in COPD patients from commercial and Medicare data sets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nMost COPD patients in both the younger commercial cohort and the older Medicare cohort experienced at least 1 exacerbation during the study period (Table 4). Patients in the commercial cohort were more likely than Medicare members to experience an exacerbation of any type (commercial 70.6% and Medicare 61.0%), and the number of exacerbations per-patient in the commercial population was higher than in the Medicare population (2.13 vs. 1.58 per patient per year). However, patients in the older Medicare group were more likely than commercial patients to experience an exacerbation that led to hospitalization (19.4% vs. 13.9%). Furthermore, Medicare patients had more hospitalized exacerbations (0.25 per patient per year) than commercial patients (0.17 per patient per year). Multiple exacerbations (≥2) occurred more commonly in commercial patients (47.3% of commercial patients and 35.9% of Medicare patients), but multiple hospitalized exacerbations occurred in 2.3% and 3.6% of patients in each group, respectively.\nPrevalence (%) of patients experiencing ≥1 exacerbation and health services utilization among COPD patients in the commercial and Medicare datasets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nImportant differences were identified in health service utilization between cohorts (Table 4). While nearly all of both populations experienced at least 1 office visit consultation, Medicare patients were significantly more likely to have been hospitalized for any reason than commercial patients. Although standard chest X-ray, ECG's, chest/thorax CT/MRIs, and heart echo exams were more likely in Medicare patients, more of the younger commercial patients had pulmonary function testing, cardiac catheterization/coronary angiography, and outpatient respiratory therapy. Only 39.0% of the total cohort had pulmonary function testing during the study period (40.0% commercial vs. 33.2% Medicare).\n[SUBTITLE] COPD exacerbations and health service utilization by disease complexity [SUBSECTION] Evaluation of exacerbations by complexity group demonstrates a general increase in the percentage of patients experiencing total exacerbations and hospitalized exacerbations with increasing disease complexity in both age cohorts (Figure 2). In addition, a higher proportion of patients with high complexity disease in both data sets experienced multiple (≥2) exacerbations (61.7% high-complexity commercial patients; 49.0% high-complexity Medicare) than patients with moderate- (commercial 56.9%, Medicare 41.6%) or low-complexity disease (commercial 33.4%, Medicare 20.5%), highlighting the relationship between complexity and exacerbations. In addition, for both the commercial and Medicare data sets, high-complexity patients had higher numbers of exacerbations per patient (mean 3.08 commercial patients; 2.22 Medicare patients) compared with moderate (commercial 2.47; Medicare 1.76) or low (commercial 1.40; Medicare 0.91) complexity. The majority of patients experiencing multiple (≥2) exacerbations requiring hospitalization were classified as high complexity (7.9% high complexity commercial patients; 10.6% Medicare), with ≤2% of moderate-complexity patients (1.8% and 2.1%) and <1% of low-complexity patients (0% and 0.01%), respectively.\nCOPD Patients Registering an inpatient exacerbation or experiencing any exacerbation, stratified by complexity, for (A) commercial, and (B) Medicare cohorts. †Any exacerbation includes inpatient hospitalized exacerbation, an ambulatory exacerbation with qualifying diagnosis or ambulatory exacerbation with qualifying drug therapy.\nUtilization of healthcare services was also observed to increase with an increase in complexity (Table 5). High-complexity patients had more than twice the pulmonary function tests per-patient compared with low-complexity patients, in both data sets. Patients with high complexity illness had more office/consultation visits than those with low complexity illness (5.7 more office/consultations in the commercial group and 3.6 more in the Medicare data set).\nProportion (%) of COPD patients in the Medicare and Commercial Cohorts Using ≥1 healthcare services, and mean number of each form of health service utilized per COPD patient, stratified by complexity of illness\nEvaluation of exacerbations by complexity group demonstrates a general increase in the percentage of patients experiencing total exacerbations and hospitalized exacerbations with increasing disease complexity in both age cohorts (Figure 2). In addition, a higher proportion of patients with high complexity disease in both data sets experienced multiple (≥2) exacerbations (61.7% high-complexity commercial patients; 49.0% high-complexity Medicare) than patients with moderate- (commercial 56.9%, Medicare 41.6%) or low-complexity disease (commercial 33.4%, Medicare 20.5%), highlighting the relationship between complexity and exacerbations. In addition, for both the commercial and Medicare data sets, high-complexity patients had higher numbers of exacerbations per patient (mean 3.08 commercial patients; 2.22 Medicare patients) compared with moderate (commercial 2.47; Medicare 1.76) or low (commercial 1.40; Medicare 0.91) complexity. The majority of patients experiencing multiple (≥2) exacerbations requiring hospitalization were classified as high complexity (7.9% high complexity commercial patients; 10.6% Medicare), with ≤2% of moderate-complexity patients (1.8% and 2.1%) and <1% of low-complexity patients (0% and 0.01%), respectively.\nCOPD Patients Registering an inpatient exacerbation or experiencing any exacerbation, stratified by complexity, for (A) commercial, and (B) Medicare cohorts. †Any exacerbation includes inpatient hospitalized exacerbation, an ambulatory exacerbation with qualifying diagnosis or ambulatory exacerbation with qualifying drug therapy.\nUtilization of healthcare services was also observed to increase with an increase in complexity (Table 5). High-complexity patients had more than twice the pulmonary function tests per-patient compared with low-complexity patients, in both data sets. Patients with high complexity illness had more office/consultation visits than those with low complexity illness (5.7 more office/consultations in the commercial group and 3.6 more in the Medicare data set).\nProportion (%) of COPD patients in the Medicare and Commercial Cohorts Using ≥1 healthcare services, and mean number of each form of health service utilized per COPD patient, stratified by complexity of illness", "Eligible patients from the commercial or Medicare cohorts were identified as having COPD, and then stratified by complexity as shown in Figure 1. From the 7,671,018 health plan members in the commercial dataset, 42,565 (0.55%) met the criteria for COPD (Table 2). The median age was 56 years and 51.4% of patients were female. From the 115,652 health plan members in the Medicare dataset, 8,507 (7.4%) were identified as having COPD. In this dataset, the median age was 75 years and 53.1% were female. Although over half of both cohorts were classified with moderate- or high complexity disease, a larger proportion of the older Medicare cohort were classified as either high or moderate complexity (63.4%) compared with the younger commercial cohort (55.0%). This was mostly due to a higher percentage of high complexity patients in the Medicare cohort (Table 2).\nStratification of patients in the study.\nDemographic characteristics for the commercial and Medicare data sets, stratified by complexity", "Specific differences were identified in the prevalence of comorbid conditions between the commercial and Medicare data sets (Table 3). For example, most cardiovascular complications were significantly less common in the younger commercial cohort than the older Medicare cohort, including hypertension, ischemic heart disease, and heart failure. Those in the younger commercial cohort were significantly more likely to have episodes of upper respiratory tract complaints including allergic rhinitis and sinusitis compared with the older Medicare cohort, who were more likely to be diagnosed with pneumonia. Furthermore, a significantly higher proportion of patients in the younger commercial database compared with the older Medicare dataset were documented to have current tobacco use during the study period.\nPrevalence (%) of comorbid conditions in COPD patients from commercial and Medicare data sets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nMost COPD patients in both the younger commercial cohort and the older Medicare cohort experienced at least 1 exacerbation during the study period (Table 4). Patients in the commercial cohort were more likely than Medicare members to experience an exacerbation of any type (commercial 70.6% and Medicare 61.0%), and the number of exacerbations per-patient in the commercial population was higher than in the Medicare population (2.13 vs. 1.58 per patient per year). However, patients in the older Medicare group were more likely than commercial patients to experience an exacerbation that led to hospitalization (19.4% vs. 13.9%). Furthermore, Medicare patients had more hospitalized exacerbations (0.25 per patient per year) than commercial patients (0.17 per patient per year). Multiple exacerbations (≥2) occurred more commonly in commercial patients (47.3% of commercial patients and 35.9% of Medicare patients), but multiple hospitalized exacerbations occurred in 2.3% and 3.6% of patients in each group, respectively.\nPrevalence (%) of patients experiencing ≥1 exacerbation and health services utilization among COPD patients in the commercial and Medicare datasets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nImportant differences were identified in health service utilization between cohorts (Table 4). While nearly all of both populations experienced at least 1 office visit consultation, Medicare patients were significantly more likely to have been hospitalized for any reason than commercial patients. Although standard chest X-ray, ECG's, chest/thorax CT/MRIs, and heart echo exams were more likely in Medicare patients, more of the younger commercial patients had pulmonary function testing, cardiac catheterization/coronary angiography, and outpatient respiratory therapy. Only 39.0% of the total cohort had pulmonary function testing during the study period (40.0% commercial vs. 33.2% Medicare).", "Evaluation of exacerbations by complexity group demonstrates a general increase in the percentage of patients experiencing total exacerbations and hospitalized exacerbations with increasing disease complexity in both age cohorts (Figure 2). In addition, a higher proportion of patients with high complexity disease in both data sets experienced multiple (≥2) exacerbations (61.7% high-complexity commercial patients; 49.0% high-complexity Medicare) than patients with moderate- (commercial 56.9%, Medicare 41.6%) or low-complexity disease (commercial 33.4%, Medicare 20.5%), highlighting the relationship between complexity and exacerbations. In addition, for both the commercial and Medicare data sets, high-complexity patients had higher numbers of exacerbations per patient (mean 3.08 commercial patients; 2.22 Medicare patients) compared with moderate (commercial 2.47; Medicare 1.76) or low (commercial 1.40; Medicare 0.91) complexity. The majority of patients experiencing multiple (≥2) exacerbations requiring hospitalization were classified as high complexity (7.9% high complexity commercial patients; 10.6% Medicare), with ≤2% of moderate-complexity patients (1.8% and 2.1%) and <1% of low-complexity patients (0% and 0.01%), respectively.\nCOPD Patients Registering an inpatient exacerbation or experiencing any exacerbation, stratified by complexity, for (A) commercial, and (B) Medicare cohorts. †Any exacerbation includes inpatient hospitalized exacerbation, an ambulatory exacerbation with qualifying diagnosis or ambulatory exacerbation with qualifying drug therapy.\nUtilization of healthcare services was also observed to increase with an increase in complexity (Table 5). High-complexity patients had more than twice the pulmonary function tests per-patient compared with low-complexity patients, in both data sets. Patients with high complexity illness had more office/consultation visits than those with low complexity illness (5.7 more office/consultations in the commercial group and 3.6 more in the Medicare data set).\nProportion (%) of COPD patients in the Medicare and Commercial Cohorts Using ≥1 healthcare services, and mean number of each form of health service utilized per COPD patient, stratified by complexity of illness", "This large, retrospective, cross-sectional analysis of US managed care administrative claims data describes a practical method to identify COPD patients using claims data. Furthermore, this methodology defines the complexity of their illness as a proxy of disease severity, and documents their exacerbations. The finding of a progressive increase in the prevalence of COPD exacerbations and utilization of health services for COPD patients from low to moderate to high complexity in both the commercial and Medicare data sets validates the utility of the complexity algorithm. Patients with COPD classified as high complexity had the highest health services utilization and were most likely to experience an exacerbation.\n[SUBTITLE] Comorbid conditions [SUBSECTION] Our complexity classification was derived based on claims identifying selected comorbid conditions or medical procedures. Patients with COPD are typically thought to have comorbid diseases and conditions that contribute to the high burden of their disease [7,10,13,14]. Our data highlight that 4-27% of COPD patients had a respiratory comorbid condition, but up to 72% had a non-respiratory comorbid condition. Consequently over half of both age cohorts were stratified as moderate or high complexity. Our findings concur with other data [7,10], highlighting the high incidence of comorbidities in COPD patients. For example a study of 200 patients showed that patients with COPD had an average of 3.7 chronic medical conditions (including lung disease), compared with 1.8 chronic medical conditions for the controls [7]. Furthermore, a review by Sin et al., highlighted that a large proportion of patients with COPD have comorbid cardiovascular disease, depression, muscle wasting, reduced fat-free mass, osteopenia, and chronic infections [10].\nComorbidities in COPD patients contribute towards the high morbidity and mortality [14]. In particular, cardiovascular disease is a common cause of death in COPD patients [15-17], with co-prevalence estimates in COPD patients of 43% (mean age 68.8 years [14]) and 45% (mean age 67.5 years [7]). In the present study, similar prevalence estimates for key cardiovascular disease risk factors were found in commercial patients (median age 56 years), although higher prevalence in older patients, with 72% of the Medicare cohort (median age 75 years) having comorbid hypertension and 47% having dyslipidemia. In addition, heart failure (HF) is a risk factor for mortality in COPD patients, and particularly in those experiencing an exacerbation [14,18]. In the present study nearly 40% of the older Medicare cohort had HF, a proportion in agreement with a study of a similarly aged population among whom 40% had HF [18].\nOur complexity classification was derived based on claims identifying selected comorbid conditions or medical procedures. Patients with COPD are typically thought to have comorbid diseases and conditions that contribute to the high burden of their disease [7,10,13,14]. Our data highlight that 4-27% of COPD patients had a respiratory comorbid condition, but up to 72% had a non-respiratory comorbid condition. Consequently over half of both age cohorts were stratified as moderate or high complexity. Our findings concur with other data [7,10], highlighting the high incidence of comorbidities in COPD patients. For example a study of 200 patients showed that patients with COPD had an average of 3.7 chronic medical conditions (including lung disease), compared with 1.8 chronic medical conditions for the controls [7]. Furthermore, a review by Sin et al., highlighted that a large proportion of patients with COPD have comorbid cardiovascular disease, depression, muscle wasting, reduced fat-free mass, osteopenia, and chronic infections [10].\nComorbidities in COPD patients contribute towards the high morbidity and mortality [14]. In particular, cardiovascular disease is a common cause of death in COPD patients [15-17], with co-prevalence estimates in COPD patients of 43% (mean age 68.8 years [14]) and 45% (mean age 67.5 years [7]). In the present study, similar prevalence estimates for key cardiovascular disease risk factors were found in commercial patients (median age 56 years), although higher prevalence in older patients, with 72% of the Medicare cohort (median age 75 years) having comorbid hypertension and 47% having dyslipidemia. In addition, heart failure (HF) is a risk factor for mortality in COPD patients, and particularly in those experiencing an exacerbation [14,18]. In the present study nearly 40% of the older Medicare cohort had HF, a proportion in agreement with a study of a similarly aged population among whom 40% had HF [18].\n[SUBTITLE] Exacerbations [SUBSECTION] Our study describes a practical way to identify COPD exacerbations using claims data, and to classify them by site of care. In the current study the majority of patients, regardless of age, experienced COPD exacerbations, which often involved hospitalization or an emergency room visit. Others have used healthcare utilization to define the staging of COPD exacerbations, in order to incorporate those parameters considered most appropriate to base sub-classification [19]. Indeed, classification of an exacerbation as mild, moderate, or severe was deemed to be strongly related to the patient's underlying condition [19]. Accordingly, for a patient with severe COPD just a small change in lung function may present as a moderate-to-severe exacerbation because it necessitates physician intervention and increased healthcare utilization [19]. We stratified exacerbations by the site of care (inpatient hospitalization, emergency room visit, and ambulatory exacerbations), which enabled more detailed information regarding the patients' site of healthcare utilization to be accounted for. This is also somewhat similar to symptom-based classification of exacerbations often used in clinical trials which capture patients whose condition has changed enough to require a change in treatment, an emergency room visit, or hospitalization.\nPatients in the older Medicare cohort were more likely to have moderate- or high complexity illness compared with the younger commercial dataset. This might be expected, as lung function and general health declines with age [20,21]. Our data also highlight that these patients qualified as high complexity were more likely to have multiple exacerbations and exacerbations requiring hospitalization compared with those of moderate or low complexity. As data were de-identified, it was not possible to confirm exacerbation information with corresponding medical records or clinical observations in this analysis. However, by distinguishing these high-risk patients from the overall COPD population, it might be ultimately possible to specifically target high-risk patients to limit the severity of their exacerbations with appropriate adjustment of therapy and/or monitoring of their comorbid conditions.\nAlthough this study did not evaluate healthcare costs, the data generated could be used for future economic analyses and modeling. Others have reported the economic burden that exacerbations place on the healthcare system. Indeed, the estimated costs of exacerbations have been found to vary widely across studies from approx. $88 to $7,757 per exacerbation (2007 US dollars) [22]. Furthermore, exacerbations accounted for 35%-45% of the total per capita healthcare costs for COPD in one study, with costs increasing with severity of exacerbations [23]. Accordingly, investigating how actual or projected costs of COPD may change through identification of patients by exacerbations and subsequent stratification by complexity, as described in the present study, would be an interesting area to investigate further.\nOur study describes a practical way to identify COPD exacerbations using claims data, and to classify them by site of care. In the current study the majority of patients, regardless of age, experienced COPD exacerbations, which often involved hospitalization or an emergency room visit. Others have used healthcare utilization to define the staging of COPD exacerbations, in order to incorporate those parameters considered most appropriate to base sub-classification [19]. Indeed, classification of an exacerbation as mild, moderate, or severe was deemed to be strongly related to the patient's underlying condition [19]. Accordingly, for a patient with severe COPD just a small change in lung function may present as a moderate-to-severe exacerbation because it necessitates physician intervention and increased healthcare utilization [19]. We stratified exacerbations by the site of care (inpatient hospitalization, emergency room visit, and ambulatory exacerbations), which enabled more detailed information regarding the patients' site of healthcare utilization to be accounted for. This is also somewhat similar to symptom-based classification of exacerbations often used in clinical trials which capture patients whose condition has changed enough to require a change in treatment, an emergency room visit, or hospitalization.\nPatients in the older Medicare cohort were more likely to have moderate- or high complexity illness compared with the younger commercial dataset. This might be expected, as lung function and general health declines with age [20,21]. Our data also highlight that these patients qualified as high complexity were more likely to have multiple exacerbations and exacerbations requiring hospitalization compared with those of moderate or low complexity. As data were de-identified, it was not possible to confirm exacerbation information with corresponding medical records or clinical observations in this analysis. However, by distinguishing these high-risk patients from the overall COPD population, it might be ultimately possible to specifically target high-risk patients to limit the severity of their exacerbations with appropriate adjustment of therapy and/or monitoring of their comorbid conditions.\nAlthough this study did not evaluate healthcare costs, the data generated could be used for future economic analyses and modeling. Others have reported the economic burden that exacerbations place on the healthcare system. Indeed, the estimated costs of exacerbations have been found to vary widely across studies from approx. $88 to $7,757 per exacerbation (2007 US dollars) [22]. Furthermore, exacerbations accounted for 35%-45% of the total per capita healthcare costs for COPD in one study, with costs increasing with severity of exacerbations [23]. Accordingly, investigating how actual or projected costs of COPD may change through identification of patients by exacerbations and subsequent stratification by complexity, as described in the present study, would be an interesting area to investigate further.\n[SUBTITLE] Utilization of healthcare services [SUBSECTION] We identified groups of COPD patients that are high users of healthcare services. For example, over half of the Medicare COPD patients and almost 40% of commercial COPD patients were hospitalized at least once for any reason in a 1-year period. Evidence of high outpatient utilization is characterized by the finding that virtually all patients in both populations (≥96%) had at least 1 office visit/consultation with a mean of over 10 per patient during the year. Furthermore, our method of stratifying patients found differential healthcare utilization; our high complexity patients had the highest health services utilization across almost all of the services monitored. As our method of stratifying patients to high and medium complexity was dependent on their comorbid conditions and the procedures they received, it is logical to expect that these patients had higher utilization compared with those of low complexity.\nWe do not know if our high complexity group had higher healthcare utilization in subsequent years, and this would be an important subject for future investigations. Interestingly, although GOLD guidelines state that lung function testing with spirometry is essential for the diagnosis and management of COPD, and indeed can provide a useful description of the severity of pathological changes in COPD [1], less than half of the population had a pulmonary function test carried out during the study year.\nOur data should be interpreted in light of some important limitations. These analyses are retrospective and descriptive in nature; there were no control groups. The large sample sizes resulted in statistically significant differences between cohorts that may not be clinically significant. However, we adjusted for the large sample size by defining significance as p < 0.01, rather than the conventional level of p < 0.05. Administrative claims data are primarily generated for reimbursement purposes rather than research purposes. Consequently, the accuracy of claims data is dependent on the precision and timing of the coding associated with their use. As such, some comorbid conditions, such as obesity and tobacco use, are often under-reported in claims data and likely are under-reported here as well. Furthermore, confusion over the differential diagnosis of asthma and COPD could have led to some patients in our COPD cohorts who actually had asthma rather than COPD. In fact, at least 1 claim with a diagnosis of asthma was recorded in 27% and 21% of the commercial and Medicare COPD cohorts, respectively. Findings of cross-sectional studies have shown a similar overlap of up to 30% between people who have a clinical diagnosis of COPD and asthma [24]. While some of these claims could represent coexisting asthma and COPD, the potential for diagnosis coding errors needs to be taken into consideration. The possibility that other respiratory conditions, such as bronchiectases, hypoventilation-obesity, or overlap syndrome, were miscoded as COPD should also be considered. Furthermore, there is a possibility that asthma attacks were miscoded as COPD exacerbations. Because the study period was limited to 1 year, it is possible for the data to provide false positives or negatives in the population selection criteria and overstate or understate the clinical severity of the disease being considered. In this case, individuals with COPD who did not have COPD-related claims within the observation period are not included in the analysis. Since these patients are likely to have less severe disease, this would result in the overall prevalence of COPD being understated, while the burden per patient with COPD overstated. On the other hand, because patients had to be continuously enrolled during the entire study year to be eligible for inclusion in this cross-sectional analysis, those who either left their health plan or died were excluded from the analysis. Excluding those who died could have removed those who were most ill and utilized the most healthcare resources. This could be especially important in the elderly cohort.\nIt is also possible that differences in exacerbations and healthcare utilization between different groups could be due to unrecognized confounders, such as duration of COPD, current smoking status, and type and duration of medication use. In addition, given that the study was a cross-sectional analysis, no cause and effect analyses could be conducted. We broadly compared an older Medicare population with a younger employer-based commercial population, although 7% of the Medicare population was <65 years and we were unable to extract these patients from the data set. However, as over 90% of the patients were ≥65 years we feel our observations provide useful information on how age influences the burden of COPD on managed care resources\nDespite these limitations, the methodology presented here provides a practical way for healthcare providers to identify and stratify COPD patients and identify those experiencing exacerbations within a large managed care database. In turn, this might help healthcare providers prioritize patients at risk for future exacerbations and resource utilization, to help ensure that those COPD patients with the greatest need for close monitoring receive optimal care.\nWe identified groups of COPD patients that are high users of healthcare services. For example, over half of the Medicare COPD patients and almost 40% of commercial COPD patients were hospitalized at least once for any reason in a 1-year period. Evidence of high outpatient utilization is characterized by the finding that virtually all patients in both populations (≥96%) had at least 1 office visit/consultation with a mean of over 10 per patient during the year. Furthermore, our method of stratifying patients found differential healthcare utilization; our high complexity patients had the highest health services utilization across almost all of the services monitored. As our method of stratifying patients to high and medium complexity was dependent on their comorbid conditions and the procedures they received, it is logical to expect that these patients had higher utilization compared with those of low complexity.\nWe do not know if our high complexity group had higher healthcare utilization in subsequent years, and this would be an important subject for future investigations. Interestingly, although GOLD guidelines state that lung function testing with spirometry is essential for the diagnosis and management of COPD, and indeed can provide a useful description of the severity of pathological changes in COPD [1], less than half of the population had a pulmonary function test carried out during the study year.\nOur data should be interpreted in light of some important limitations. These analyses are retrospective and descriptive in nature; there were no control groups. The large sample sizes resulted in statistically significant differences between cohorts that may not be clinically significant. However, we adjusted for the large sample size by defining significance as p < 0.01, rather than the conventional level of p < 0.05. Administrative claims data are primarily generated for reimbursement purposes rather than research purposes. Consequently, the accuracy of claims data is dependent on the precision and timing of the coding associated with their use. As such, some comorbid conditions, such as obesity and tobacco use, are often under-reported in claims data and likely are under-reported here as well. Furthermore, confusion over the differential diagnosis of asthma and COPD could have led to some patients in our COPD cohorts who actually had asthma rather than COPD. In fact, at least 1 claim with a diagnosis of asthma was recorded in 27% and 21% of the commercial and Medicare COPD cohorts, respectively. Findings of cross-sectional studies have shown a similar overlap of up to 30% between people who have a clinical diagnosis of COPD and asthma [24]. While some of these claims could represent coexisting asthma and COPD, the potential for diagnosis coding errors needs to be taken into consideration. The possibility that other respiratory conditions, such as bronchiectases, hypoventilation-obesity, or overlap syndrome, were miscoded as COPD should also be considered. Furthermore, there is a possibility that asthma attacks were miscoded as COPD exacerbations. Because the study period was limited to 1 year, it is possible for the data to provide false positives or negatives in the population selection criteria and overstate or understate the clinical severity of the disease being considered. In this case, individuals with COPD who did not have COPD-related claims within the observation period are not included in the analysis. Since these patients are likely to have less severe disease, this would result in the overall prevalence of COPD being understated, while the burden per patient with COPD overstated. On the other hand, because patients had to be continuously enrolled during the entire study year to be eligible for inclusion in this cross-sectional analysis, those who either left their health plan or died were excluded from the analysis. Excluding those who died could have removed those who were most ill and utilized the most healthcare resources. This could be especially important in the elderly cohort.\nIt is also possible that differences in exacerbations and healthcare utilization between different groups could be due to unrecognized confounders, such as duration of COPD, current smoking status, and type and duration of medication use. In addition, given that the study was a cross-sectional analysis, no cause and effect analyses could be conducted. We broadly compared an older Medicare population with a younger employer-based commercial population, although 7% of the Medicare population was <65 years and we were unable to extract these patients from the data set. However, as over 90% of the patients were ≥65 years we feel our observations provide useful information on how age influences the burden of COPD on managed care resources\nDespite these limitations, the methodology presented here provides a practical way for healthcare providers to identify and stratify COPD patients and identify those experiencing exacerbations within a large managed care database. In turn, this might help healthcare providers prioritize patients at risk for future exacerbations and resource utilization, to help ensure that those COPD patients with the greatest need for close monitoring receive optimal care.", "Our complexity classification was derived based on claims identifying selected comorbid conditions or medical procedures. Patients with COPD are typically thought to have comorbid diseases and conditions that contribute to the high burden of their disease [7,10,13,14]. Our data highlight that 4-27% of COPD patients had a respiratory comorbid condition, but up to 72% had a non-respiratory comorbid condition. Consequently over half of both age cohorts were stratified as moderate or high complexity. Our findings concur with other data [7,10], highlighting the high incidence of comorbidities in COPD patients. For example a study of 200 patients showed that patients with COPD had an average of 3.7 chronic medical conditions (including lung disease), compared with 1.8 chronic medical conditions for the controls [7]. Furthermore, a review by Sin et al., highlighted that a large proportion of patients with COPD have comorbid cardiovascular disease, depression, muscle wasting, reduced fat-free mass, osteopenia, and chronic infections [10].\nComorbidities in COPD patients contribute towards the high morbidity and mortality [14]. In particular, cardiovascular disease is a common cause of death in COPD patients [15-17], with co-prevalence estimates in COPD patients of 43% (mean age 68.8 years [14]) and 45% (mean age 67.5 years [7]). In the present study, similar prevalence estimates for key cardiovascular disease risk factors were found in commercial patients (median age 56 years), although higher prevalence in older patients, with 72% of the Medicare cohort (median age 75 years) having comorbid hypertension and 47% having dyslipidemia. In addition, heart failure (HF) is a risk factor for mortality in COPD patients, and particularly in those experiencing an exacerbation [14,18]. In the present study nearly 40% of the older Medicare cohort had HF, a proportion in agreement with a study of a similarly aged population among whom 40% had HF [18].", "Our study describes a practical way to identify COPD exacerbations using claims data, and to classify them by site of care. In the current study the majority of patients, regardless of age, experienced COPD exacerbations, which often involved hospitalization or an emergency room visit. Others have used healthcare utilization to define the staging of COPD exacerbations, in order to incorporate those parameters considered most appropriate to base sub-classification [19]. Indeed, classification of an exacerbation as mild, moderate, or severe was deemed to be strongly related to the patient's underlying condition [19]. Accordingly, for a patient with severe COPD just a small change in lung function may present as a moderate-to-severe exacerbation because it necessitates physician intervention and increased healthcare utilization [19]. We stratified exacerbations by the site of care (inpatient hospitalization, emergency room visit, and ambulatory exacerbations), which enabled more detailed information regarding the patients' site of healthcare utilization to be accounted for. This is also somewhat similar to symptom-based classification of exacerbations often used in clinical trials which capture patients whose condition has changed enough to require a change in treatment, an emergency room visit, or hospitalization.\nPatients in the older Medicare cohort were more likely to have moderate- or high complexity illness compared with the younger commercial dataset. This might be expected, as lung function and general health declines with age [20,21]. Our data also highlight that these patients qualified as high complexity were more likely to have multiple exacerbations and exacerbations requiring hospitalization compared with those of moderate or low complexity. As data were de-identified, it was not possible to confirm exacerbation information with corresponding medical records or clinical observations in this analysis. However, by distinguishing these high-risk patients from the overall COPD population, it might be ultimately possible to specifically target high-risk patients to limit the severity of their exacerbations with appropriate adjustment of therapy and/or monitoring of their comorbid conditions.\nAlthough this study did not evaluate healthcare costs, the data generated could be used for future economic analyses and modeling. Others have reported the economic burden that exacerbations place on the healthcare system. Indeed, the estimated costs of exacerbations have been found to vary widely across studies from approx. $88 to $7,757 per exacerbation (2007 US dollars) [22]. Furthermore, exacerbations accounted for 35%-45% of the total per capita healthcare costs for COPD in one study, with costs increasing with severity of exacerbations [23]. Accordingly, investigating how actual or projected costs of COPD may change through identification of patients by exacerbations and subsequent stratification by complexity, as described in the present study, would be an interesting area to investigate further.", "We identified groups of COPD patients that are high users of healthcare services. For example, over half of the Medicare COPD patients and almost 40% of commercial COPD patients were hospitalized at least once for any reason in a 1-year period. Evidence of high outpatient utilization is characterized by the finding that virtually all patients in both populations (≥96%) had at least 1 office visit/consultation with a mean of over 10 per patient during the year. Furthermore, our method of stratifying patients found differential healthcare utilization; our high complexity patients had the highest health services utilization across almost all of the services monitored. As our method of stratifying patients to high and medium complexity was dependent on their comorbid conditions and the procedures they received, it is logical to expect that these patients had higher utilization compared with those of low complexity.\nWe do not know if our high complexity group had higher healthcare utilization in subsequent years, and this would be an important subject for future investigations. Interestingly, although GOLD guidelines state that lung function testing with spirometry is essential for the diagnosis and management of COPD, and indeed can provide a useful description of the severity of pathological changes in COPD [1], less than half of the population had a pulmonary function test carried out during the study year.\nOur data should be interpreted in light of some important limitations. These analyses are retrospective and descriptive in nature; there were no control groups. The large sample sizes resulted in statistically significant differences between cohorts that may not be clinically significant. However, we adjusted for the large sample size by defining significance as p < 0.01, rather than the conventional level of p < 0.05. Administrative claims data are primarily generated for reimbursement purposes rather than research purposes. Consequently, the accuracy of claims data is dependent on the precision and timing of the coding associated with their use. As such, some comorbid conditions, such as obesity and tobacco use, are often under-reported in claims data and likely are under-reported here as well. Furthermore, confusion over the differential diagnosis of asthma and COPD could have led to some patients in our COPD cohorts who actually had asthma rather than COPD. In fact, at least 1 claim with a diagnosis of asthma was recorded in 27% and 21% of the commercial and Medicare COPD cohorts, respectively. Findings of cross-sectional studies have shown a similar overlap of up to 30% between people who have a clinical diagnosis of COPD and asthma [24]. While some of these claims could represent coexisting asthma and COPD, the potential for diagnosis coding errors needs to be taken into consideration. The possibility that other respiratory conditions, such as bronchiectases, hypoventilation-obesity, or overlap syndrome, were miscoded as COPD should also be considered. Furthermore, there is a possibility that asthma attacks were miscoded as COPD exacerbations. Because the study period was limited to 1 year, it is possible for the data to provide false positives or negatives in the population selection criteria and overstate or understate the clinical severity of the disease being considered. In this case, individuals with COPD who did not have COPD-related claims within the observation period are not included in the analysis. Since these patients are likely to have less severe disease, this would result in the overall prevalence of COPD being understated, while the burden per patient with COPD overstated. On the other hand, because patients had to be continuously enrolled during the entire study year to be eligible for inclusion in this cross-sectional analysis, those who either left their health plan or died were excluded from the analysis. Excluding those who died could have removed those who were most ill and utilized the most healthcare resources. This could be especially important in the elderly cohort.\nIt is also possible that differences in exacerbations and healthcare utilization between different groups could be due to unrecognized confounders, such as duration of COPD, current smoking status, and type and duration of medication use. In addition, given that the study was a cross-sectional analysis, no cause and effect analyses could be conducted. We broadly compared an older Medicare population with a younger employer-based commercial population, although 7% of the Medicare population was <65 years and we were unable to extract these patients from the data set. However, as over 90% of the patients were ≥65 years we feel our observations provide useful information on how age influences the burden of COPD on managed care resources\nDespite these limitations, the methodology presented here provides a practical way for healthcare providers to identify and stratify COPD patients and identify those experiencing exacerbations within a large managed care database. In turn, this might help healthcare providers prioritize patients at risk for future exacerbations and resource utilization, to help ensure that those COPD patients with the greatest need for close monitoring receive optimal care.", "We present a unique and practical method for identifying patients with COPD, determining disease severity - as \"complexity of illness\" - and documenting exacerbations using claims data. Our data highlight important differences in comorbidities, exacerbations, and healthcare utilizations in older (aged ≥65 years) compared with younger (aged <65 years) COPD patients. Furthermore, by stratifying COPD patients based on diagnostic, procedures, and services codes, we have demonstrated that patients stratified as having high- or moderate complexity disease experienced a higher number of exacerbations than those with low complexity disease. Additionally, by stratifying by complexity of illness we show linearity between complexity of illness and utilization of healthcare services and hospitalizations.\nIdentification of COPD patients at higher risk of complications using complexity stratification and/or identifying those experiencing exacerbations may serve to improve patient management, and ultimately reduce the burden of disease to the patient, and to the healthcare systems supporting them.", "Dr. Mapel was a paid consultant to Pfizer Inc in connection with conduct of the analysis and development of the manuscript. Dr. Mapel has served as a consultant to and received research funding from Pfizer Pharmaceuticals, GlaxoSmithKline, and AstraZeneca. Drs. Woodruff, Marton, and Dutro, are employees of, and own stock in Pfizer Inc. Dr. Make has participated in advisory boards and received honoraria for speaking from Pfizer Inc within the past five years.", "All authors participated in the design of the study, and contributed to drafting the manuscript. All authors read and approved the final manuscript.", "Parts of these data were presented as a poster at the American Thoracic Society 103rd International Conference, May 22, 2007, San Francisco, CA, USA.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1472-6963/11/43/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study design and data source", "Identification of patients with COPD", "Population demographics and comorbid conditions", "Complexity", "Exacerbations", "Utilization of healthcare services", "Data analysis", "Results", "Population characteristics", "Comparison of the commercial and Medicare COPD populations", "COPD exacerbations and health service utilization by disease complexity", "Discussion", "Comorbid conditions", "Exacerbations", "Utilization of healthcare services", "Conclusions", "Competing interests", "Authors' contributions", "Disclosure", "Pre-publication history", "Supplementary Material" ]

[ "Chronic obstructive pulmonary disease (COPD) is the fourth leading cause of death among US adults and is projected to be the third by 2020 [1-3], although the disease is both preventable and treatable [4-6]. With this projected increased burden on healthcare systems, data describing how COPD patients are currently managed, together with information on COPD patients' healthcare utilization, are needed to inform healthcare organizations and providers. A small study of 1522 COPD patients in a health maintenance organization demonstrated that COPD patients had healthcare utilization and associated costs of more than twice those of age- and sex-matched controls [7]. Similarly, a larger study of over 100,000 patients aged ≥65 years with COPD or asthma showed the utilization of healthcare resources by these older COPD patients to be extremely high, both during hospitalization and after discharge [8]. The high prevalence of comorbidities in patients with COPD, especially respiratory conditions and cardiovascular disease, increases the morbidity associated with COPD [9-11].\nBecause of the burden imposed on healthcare systems and payers by patients with COPD, a means of identifying COPD patients who have higher healthcare utilization and associated higher costs is needed. The identification and stratification of COPD patients at risk of complications might facilitate management of these patients, improve care, and reduce costs. Furthermore, since COPD exacerbations, especially those requiring hospitalization, account for significant healthcare utilization, a method to accurately document COPD exacerbations using claims data would be very useful. Finally, comparing an older Medicare population with a working age commercial population will provide important information on how age influences the healthcare utilization in patients with COPD.\nThe purpose of this project is to describe a unique methodology that identifies COPD patients in a large managed care database and documents their demographics, comorbid conditions, and COPD exacerbations. We stratified these COPD patients by means of a novel algorithm of disease complexity (high, moderate, low complexity), used as a proxy of disease severity, and then examined the relationship between complexity of illness and key indicators of healthcare utilization and exacerbations. Furthermore, data were analyzed based on age group to determine if there are differences in COPD patients of an older Medicare (generally ≥65 years of age) population and a younger employer-based (< 65 years) population.", "[SUBTITLE] Study design and data source [SUBSECTION] This was a retrospective, cross-sectional analysis of US managed care administrative claims data from multiple health plans during the one-year study period, July 1, 2004, to June 30, 2005. The study population was extracted from a dataset of 12.4 million covered lives maintained by PharMetrics Inc (Watertown, MA, USA) from 19 health plans across the US: 3.2 million from the Northeast, 6.4 million from the Midwest, 1.8 million from the South, and 0.7 million from the West. The plans varied in size: 6 were <200,000 covered lives, 9 were between 200,001 and 1 million covered lives, and 4 were over 1 million covered lives. We evaluated the 7.79 million members who were continuously eligible during the study period for potential inclusion in the COPD cohort.\nAs differences likely exist between older and younger individuals, data from Medicare plans were evaluated separately from commercial health plans. The commercial population represented the employer-based managed care product offerings of Health Maintenance Organization, Preferred Provider Organization, and Point of Service plans. Due to the fragmentation of healthcare claims for the population aged older than 64 years, these patients were only included in Medicare analyses if they were continuously enrolled in a Medicare Risk product for the entire time span once becoming 65 years old.\nThis was a retrospective, cross-sectional analysis of US managed care administrative claims data from multiple health plans during the one-year study period, July 1, 2004, to June 30, 2005. The study population was extracted from a dataset of 12.4 million covered lives maintained by PharMetrics Inc (Watertown, MA, USA) from 19 health plans across the US: 3.2 million from the Northeast, 6.4 million from the Midwest, 1.8 million from the South, and 0.7 million from the West. The plans varied in size: 6 were <200,000 covered lives, 9 were between 200,001 and 1 million covered lives, and 4 were over 1 million covered lives. We evaluated the 7.79 million members who were continuously eligible during the study period for potential inclusion in the COPD cohort.\nAs differences likely exist between older and younger individuals, data from Medicare plans were evaluated separately from commercial health plans. The commercial population represented the employer-based managed care product offerings of Health Maintenance Organization, Preferred Provider Organization, and Point of Service plans. Due to the fragmentation of healthcare claims for the population aged older than 64 years, these patients were only included in Medicare analyses if they were continuously enrolled in a Medicare Risk product for the entire time span once becoming 65 years old.\n[SUBTITLE] Identification of patients with COPD [SUBSECTION] Patients were identified as having COPD if they were aged ≥40 years and had any one of the following:\n1. One inpatient hospitalization or one emergency room encounter with a COPD diagnosis (491.x [chronic bronchitis], 492.x [emphysema], or 496 [chronic airway obstruction]) listed in any position as a discharge diagnosis; or\n2. Two professional claims, with different dates of services, with a COPD diagnosis listed in any position; or\n3. A COPD-related surgical procedure (e.g. lung volume reduction) listed on either a professional or facility claim.\nPatients were identified as having COPD if they were aged ≥40 years and had any one of the following:\n1. One inpatient hospitalization or one emergency room encounter with a COPD diagnosis (491.x [chronic bronchitis], 492.x [emphysema], or 496 [chronic airway obstruction]) listed in any position as a discharge diagnosis; or\n2. Two professional claims, with different dates of services, with a COPD diagnosis listed in any position; or\n3. A COPD-related surgical procedure (e.g. lung volume reduction) listed on either a professional or facility claim.\n[SUBTITLE] Population demographics and comorbid conditions [SUBSECTION] The age and gender were determined for patients identified with COPD from claims data. Comorbid conditions were determined if claims data included diagnostic, procedures, and services codes (2004 ICD-9 CM, CPT-4, and HCPCS codes, respectively) for predetermined conditions (see Additional Files) during the reporting period. Condition frequencies were determined for respiratory and for non-respiratory comorbid conditions.\nThe age and gender were determined for patients identified with COPD from claims data. Comorbid conditions were determined if claims data included diagnostic, procedures, and services codes (2004 ICD-9 CM, CPT-4, and HCPCS codes, respectively) for predetermined conditions (see Additional Files) during the reporting period. Condition frequencies were determined for respiratory and for non-respiratory comorbid conditions.\n[SUBTITLE] Complexity [SUBSECTION] A claims-based classification of COPD complexity was created to serve as a surrogate for COPD disease severity. Comorbid respiratory conditions and medical procedures at any time during the study period were used to assign patients to one of three disease complexity levels (high, moderate, or low) based on selected diagnostic, procedures and services codes (2004 ICD-9, CPT-4, and HCPCS), as detailed in Additional File 1. Examples of code descriptions that resulted in a patient being classified as high complexity include a claim for cor pulmonale, tuberculosis, or malignant neoplasm. Examples for moderate complexity include pneumonia, cyanosis, bronchoscopy or dependence on supplemental oxygen. If a COPD patient did not have any comorbid condition for high or moderate complexity (Additional Files), they were classified as low complexity.\nA claims-based classification of COPD complexity was created to serve as a surrogate for COPD disease severity. Comorbid respiratory conditions and medical procedures at any time during the study period were used to assign patients to one of three disease complexity levels (high, moderate, or low) based on selected diagnostic, procedures and services codes (2004 ICD-9, CPT-4, and HCPCS), as detailed in Additional File 1. Examples of code descriptions that resulted in a patient being classified as high complexity include a claim for cor pulmonale, tuberculosis, or malignant neoplasm. Examples for moderate complexity include pneumonia, cyanosis, bronchoscopy or dependence on supplemental oxygen. If a COPD patient did not have any comorbid condition for high or moderate complexity (Additional Files), they were classified as low complexity.\n[SUBTITLE] Exacerbations [SUBSECTION] An algorithm was developed to identify and stratify COPD exacerbations using claims data (Table 1). Although there is controversy about the definition of a COPD exacerbation, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) defines an exacerbation as: \"an event in the natural course of the disease (COPD) characterized by a change in the patient's baseline dyspnea, cough and/or sputum that is beyond normal day-to-day variation, is acute in onset, and may warrant a change in regular medication\" [1]. An exacerbation in this study was defined, as outlined in Table 1, by either a primary diagnosis recorded by a medical provider and coded in claims or a medication for an oral antibiotic commonly used for respiratory infection, or systemic steroid received by a patient during any 14-day time period. Exacerbation severity was classified by site of care reflecting resource utilization.\nAlgorithm for identification and classification of exacerbations in COPD patients\nTotal exacerbations included exacerbations of any severity (inpatient, emergency room, ambulatory by qualifying diagnosis, or ambulatory supported by drug therapy). Exacerbation frequency is reported for both total and hospitalized exacerbations as percent of patients experiencing ≥1 exacerbation(s), percent of patients experiencing multiple (≥2) exacerbations, and number of exacerbations/COPD patient.\nAn algorithm was developed to identify and stratify COPD exacerbations using claims data (Table 1). Although there is controversy about the definition of a COPD exacerbation, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) defines an exacerbation as: \"an event in the natural course of the disease (COPD) characterized by a change in the patient's baseline dyspnea, cough and/or sputum that is beyond normal day-to-day variation, is acute in onset, and may warrant a change in regular medication\" [1]. An exacerbation in this study was defined, as outlined in Table 1, by either a primary diagnosis recorded by a medical provider and coded in claims or a medication for an oral antibiotic commonly used for respiratory infection, or systemic steroid received by a patient during any 14-day time period. Exacerbation severity was classified by site of care reflecting resource utilization.\nAlgorithm for identification and classification of exacerbations in COPD patients\nTotal exacerbations included exacerbations of any severity (inpatient, emergency room, ambulatory by qualifying diagnosis, or ambulatory supported by drug therapy). Exacerbation frequency is reported for both total and hospitalized exacerbations as percent of patients experiencing ≥1 exacerbation(s), percent of patients experiencing multiple (≥2) exacerbations, and number of exacerbations/COPD patient.\n[SUBTITLE] Utilization of healthcare services [SUBSECTION] The healthcare services assessed in this study included tests, procedures (including surgeries), and visits considered to be related with COPD (see Additional File 1). The most common or relevant of these are included in this report. These services were identified based on 2004 CPT-4 and HCPCS codes collected in claims. Hospital admissions and emergency room visits for any reason were identified based on the presence of a facility claim for an inpatient hospital stay or an outpatient emergency room visit.\nThe healthcare services assessed in this study included tests, procedures (including surgeries), and visits considered to be related with COPD (see Additional File 1). The most common or relevant of these are included in this report. These services were identified based on 2004 CPT-4 and HCPCS codes collected in claims. Hospital admissions and emergency room visits for any reason were identified based on the presence of a facility claim for an inpatient hospital stay or an outpatient emergency room visit.\n[SUBTITLE] Data analysis [SUBSECTION] DTEC™software (Pfizer, New York, NY, USA, Version 3.3) was used to integrate administrative data and claims files, identify and stratify patients with COPD, as well as characterize demographics, comorbidities, and utilization of healthcare services. All of these analyses were specified prior to study initiation, and programmed in the software. These analyses, including the algorithms for COPD complexity stratification and exacerbation identification, were developed by a panel of experts that included pulmonologists, outcomes researchers and claims-based research consultants. They are based upon information from accepted guidelines [1,12] but also incorporate previous experiences of the panel in claims-based research. While DTEC™, a proprietary software program, was used for this analysis, the algorithms for COPD disease identification and stratification included in the software are specifically outlined and included in the body of this article or Additional File 1 so that they may be used in other claims querying systems.\nClaims data during the 12-month study period were analyzed, and are presented as means with standard deviations. Categorical data are presented as numbers and percentages. Mean data were compared using the Student's t-test for normally distributed values and the Wilcoxon rank-sum test for non-normally distributed values. Categorical data were compared between commercial and Medicare populations before stratifying for complexity, using Chi-square test. Data were analyzed using GraphPad Prism® statistical software (version 5.0 for Windows; GraphPad Software Inc., CA, USA). Because of large sample sizes a p-value of 0.01 was designated as being statistically significant for all comparisons. The database was compiled in accordance with all aspects of the Health Information Portability and Accountability Act (HIPAA) of 1996.\nDTEC™software (Pfizer, New York, NY, USA, Version 3.3) was used to integrate administrative data and claims files, identify and stratify patients with COPD, as well as characterize demographics, comorbidities, and utilization of healthcare services. All of these analyses were specified prior to study initiation, and programmed in the software. These analyses, including the algorithms for COPD complexity stratification and exacerbation identification, were developed by a panel of experts that included pulmonologists, outcomes researchers and claims-based research consultants. They are based upon information from accepted guidelines [1,12] but also incorporate previous experiences of the panel in claims-based research. While DTEC™, a proprietary software program, was used for this analysis, the algorithms for COPD disease identification and stratification included in the software are specifically outlined and included in the body of this article or Additional File 1 so that they may be used in other claims querying systems.\nClaims data during the 12-month study period were analyzed, and are presented as means with standard deviations. Categorical data are presented as numbers and percentages. Mean data were compared using the Student's t-test for normally distributed values and the Wilcoxon rank-sum test for non-normally distributed values. Categorical data were compared between commercial and Medicare populations before stratifying for complexity, using Chi-square test. Data were analyzed using GraphPad Prism® statistical software (version 5.0 for Windows; GraphPad Software Inc., CA, USA). Because of large sample sizes a p-value of 0.01 was designated as being statistically significant for all comparisons. The database was compiled in accordance with all aspects of the Health Information Portability and Accountability Act (HIPAA) of 1996.", "This was a retrospective, cross-sectional analysis of US managed care administrative claims data from multiple health plans during the one-year study period, July 1, 2004, to June 30, 2005. The study population was extracted from a dataset of 12.4 million covered lives maintained by PharMetrics Inc (Watertown, MA, USA) from 19 health plans across the US: 3.2 million from the Northeast, 6.4 million from the Midwest, 1.8 million from the South, and 0.7 million from the West. The plans varied in size: 6 were <200,000 covered lives, 9 were between 200,001 and 1 million covered lives, and 4 were over 1 million covered lives. We evaluated the 7.79 million members who were continuously eligible during the study period for potential inclusion in the COPD cohort.\nAs differences likely exist between older and younger individuals, data from Medicare plans were evaluated separately from commercial health plans. The commercial population represented the employer-based managed care product offerings of Health Maintenance Organization, Preferred Provider Organization, and Point of Service plans. Due to the fragmentation of healthcare claims for the population aged older than 64 years, these patients were only included in Medicare analyses if they were continuously enrolled in a Medicare Risk product for the entire time span once becoming 65 years old.", "Patients were identified as having COPD if they were aged ≥40 years and had any one of the following:\n1. One inpatient hospitalization or one emergency room encounter with a COPD diagnosis (491.x [chronic bronchitis], 492.x [emphysema], or 496 [chronic airway obstruction]) listed in any position as a discharge diagnosis; or\n2. Two professional claims, with different dates of services, with a COPD diagnosis listed in any position; or\n3. A COPD-related surgical procedure (e.g. lung volume reduction) listed on either a professional or facility claim.", "The age and gender were determined for patients identified with COPD from claims data. Comorbid conditions were determined if claims data included diagnostic, procedures, and services codes (2004 ICD-9 CM, CPT-4, and HCPCS codes, respectively) for predetermined conditions (see Additional Files) during the reporting period. Condition frequencies were determined for respiratory and for non-respiratory comorbid conditions.", "A claims-based classification of COPD complexity was created to serve as a surrogate for COPD disease severity. Comorbid respiratory conditions and medical procedures at any time during the study period were used to assign patients to one of three disease complexity levels (high, moderate, or low) based on selected diagnostic, procedures and services codes (2004 ICD-9, CPT-4, and HCPCS), as detailed in Additional File 1. Examples of code descriptions that resulted in a patient being classified as high complexity include a claim for cor pulmonale, tuberculosis, or malignant neoplasm. Examples for moderate complexity include pneumonia, cyanosis, bronchoscopy or dependence on supplemental oxygen. If a COPD patient did not have any comorbid condition for high or moderate complexity (Additional Files), they were classified as low complexity.", "An algorithm was developed to identify and stratify COPD exacerbations using claims data (Table 1). Although there is controversy about the definition of a COPD exacerbation, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) defines an exacerbation as: \"an event in the natural course of the disease (COPD) characterized by a change in the patient's baseline dyspnea, cough and/or sputum that is beyond normal day-to-day variation, is acute in onset, and may warrant a change in regular medication\" [1]. An exacerbation in this study was defined, as outlined in Table 1, by either a primary diagnosis recorded by a medical provider and coded in claims or a medication for an oral antibiotic commonly used for respiratory infection, or systemic steroid received by a patient during any 14-day time period. Exacerbation severity was classified by site of care reflecting resource utilization.\nAlgorithm for identification and classification of exacerbations in COPD patients\nTotal exacerbations included exacerbations of any severity (inpatient, emergency room, ambulatory by qualifying diagnosis, or ambulatory supported by drug therapy). Exacerbation frequency is reported for both total and hospitalized exacerbations as percent of patients experiencing ≥1 exacerbation(s), percent of patients experiencing multiple (≥2) exacerbations, and number of exacerbations/COPD patient.", "The healthcare services assessed in this study included tests, procedures (including surgeries), and visits considered to be related with COPD (see Additional File 1). The most common or relevant of these are included in this report. These services were identified based on 2004 CPT-4 and HCPCS codes collected in claims. Hospital admissions and emergency room visits for any reason were identified based on the presence of a facility claim for an inpatient hospital stay or an outpatient emergency room visit.", "DTEC™software (Pfizer, New York, NY, USA, Version 3.3) was used to integrate administrative data and claims files, identify and stratify patients with COPD, as well as characterize demographics, comorbidities, and utilization of healthcare services. All of these analyses were specified prior to study initiation, and programmed in the software. These analyses, including the algorithms for COPD complexity stratification and exacerbation identification, were developed by a panel of experts that included pulmonologists, outcomes researchers and claims-based research consultants. They are based upon information from accepted guidelines [1,12] but also incorporate previous experiences of the panel in claims-based research. While DTEC™, a proprietary software program, was used for this analysis, the algorithms for COPD disease identification and stratification included in the software are specifically outlined and included in the body of this article or Additional File 1 so that they may be used in other claims querying systems.\nClaims data during the 12-month study period were analyzed, and are presented as means with standard deviations. Categorical data are presented as numbers and percentages. Mean data were compared using the Student's t-test for normally distributed values and the Wilcoxon rank-sum test for non-normally distributed values. Categorical data were compared between commercial and Medicare populations before stratifying for complexity, using Chi-square test. Data were analyzed using GraphPad Prism® statistical software (version 5.0 for Windows; GraphPad Software Inc., CA, USA). Because of large sample sizes a p-value of 0.01 was designated as being statistically significant for all comparisons. The database was compiled in accordance with all aspects of the Health Information Portability and Accountability Act (HIPAA) of 1996.", "[SUBTITLE] Population characteristics [SUBSECTION] Eligible patients from the commercial or Medicare cohorts were identified as having COPD, and then stratified by complexity as shown in Figure 1. From the 7,671,018 health plan members in the commercial dataset, 42,565 (0.55%) met the criteria for COPD (Table 2). The median age was 56 years and 51.4% of patients were female. From the 115,652 health plan members in the Medicare dataset, 8,507 (7.4%) were identified as having COPD. In this dataset, the median age was 75 years and 53.1% were female. Although over half of both cohorts were classified with moderate- or high complexity disease, a larger proportion of the older Medicare cohort were classified as either high or moderate complexity (63.4%) compared with the younger commercial cohort (55.0%). This was mostly due to a higher percentage of high complexity patients in the Medicare cohort (Table 2).\nStratification of patients in the study.\nDemographic characteristics for the commercial and Medicare data sets, stratified by complexity\nEligible patients from the commercial or Medicare cohorts were identified as having COPD, and then stratified by complexity as shown in Figure 1. From the 7,671,018 health plan members in the commercial dataset, 42,565 (0.55%) met the criteria for COPD (Table 2). The median age was 56 years and 51.4% of patients were female. From the 115,652 health plan members in the Medicare dataset, 8,507 (7.4%) were identified as having COPD. In this dataset, the median age was 75 years and 53.1% were female. Although over half of both cohorts were classified with moderate- or high complexity disease, a larger proportion of the older Medicare cohort were classified as either high or moderate complexity (63.4%) compared with the younger commercial cohort (55.0%). This was mostly due to a higher percentage of high complexity patients in the Medicare cohort (Table 2).\nStratification of patients in the study.\nDemographic characteristics for the commercial and Medicare data sets, stratified by complexity\n[SUBTITLE] Comparison of the commercial and Medicare COPD populations [SUBSECTION] Specific differences were identified in the prevalence of comorbid conditions between the commercial and Medicare data sets (Table 3). For example, most cardiovascular complications were significantly less common in the younger commercial cohort than the older Medicare cohort, including hypertension, ischemic heart disease, and heart failure. Those in the younger commercial cohort were significantly more likely to have episodes of upper respiratory tract complaints including allergic rhinitis and sinusitis compared with the older Medicare cohort, who were more likely to be diagnosed with pneumonia. Furthermore, a significantly higher proportion of patients in the younger commercial database compared with the older Medicare dataset were documented to have current tobacco use during the study period.\nPrevalence (%) of comorbid conditions in COPD patients from commercial and Medicare data sets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nMost COPD patients in both the younger commercial cohort and the older Medicare cohort experienced at least 1 exacerbation during the study period (Table 4). Patients in the commercial cohort were more likely than Medicare members to experience an exacerbation of any type (commercial 70.6% and Medicare 61.0%), and the number of exacerbations per-patient in the commercial population was higher than in the Medicare population (2.13 vs. 1.58 per patient per year). However, patients in the older Medicare group were more likely than commercial patients to experience an exacerbation that led to hospitalization (19.4% vs. 13.9%). Furthermore, Medicare patients had more hospitalized exacerbations (0.25 per patient per year) than commercial patients (0.17 per patient per year). Multiple exacerbations (≥2) occurred more commonly in commercial patients (47.3% of commercial patients and 35.9% of Medicare patients), but multiple hospitalized exacerbations occurred in 2.3% and 3.6% of patients in each group, respectively.\nPrevalence (%) of patients experiencing ≥1 exacerbation and health services utilization among COPD patients in the commercial and Medicare datasets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nImportant differences were identified in health service utilization between cohorts (Table 4). While nearly all of both populations experienced at least 1 office visit consultation, Medicare patients were significantly more likely to have been hospitalized for any reason than commercial patients. Although standard chest X-ray, ECG's, chest/thorax CT/MRIs, and heart echo exams were more likely in Medicare patients, more of the younger commercial patients had pulmonary function testing, cardiac catheterization/coronary angiography, and outpatient respiratory therapy. Only 39.0% of the total cohort had pulmonary function testing during the study period (40.0% commercial vs. 33.2% Medicare).\nSpecific differences were identified in the prevalence of comorbid conditions between the commercial and Medicare data sets (Table 3). For example, most cardiovascular complications were significantly less common in the younger commercial cohort than the older Medicare cohort, including hypertension, ischemic heart disease, and heart failure. Those in the younger commercial cohort were significantly more likely to have episodes of upper respiratory tract complaints including allergic rhinitis and sinusitis compared with the older Medicare cohort, who were more likely to be diagnosed with pneumonia. Furthermore, a significantly higher proportion of patients in the younger commercial database compared with the older Medicare dataset were documented to have current tobacco use during the study period.\nPrevalence (%) of comorbid conditions in COPD patients from commercial and Medicare data sets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nMost COPD patients in both the younger commercial cohort and the older Medicare cohort experienced at least 1 exacerbation during the study period (Table 4). Patients in the commercial cohort were more likely than Medicare members to experience an exacerbation of any type (commercial 70.6% and Medicare 61.0%), and the number of exacerbations per-patient in the commercial population was higher than in the Medicare population (2.13 vs. 1.58 per patient per year). However, patients in the older Medicare group were more likely than commercial patients to experience an exacerbation that led to hospitalization (19.4% vs. 13.9%). Furthermore, Medicare patients had more hospitalized exacerbations (0.25 per patient per year) than commercial patients (0.17 per patient per year). Multiple exacerbations (≥2) occurred more commonly in commercial patients (47.3% of commercial patients and 35.9% of Medicare patients), but multiple hospitalized exacerbations occurred in 2.3% and 3.6% of patients in each group, respectively.\nPrevalence (%) of patients experiencing ≥1 exacerbation and health services utilization among COPD patients in the commercial and Medicare datasets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nImportant differences were identified in health service utilization between cohorts (Table 4). While nearly all of both populations experienced at least 1 office visit consultation, Medicare patients were significantly more likely to have been hospitalized for any reason than commercial patients. Although standard chest X-ray, ECG's, chest/thorax CT/MRIs, and heart echo exams were more likely in Medicare patients, more of the younger commercial patients had pulmonary function testing, cardiac catheterization/coronary angiography, and outpatient respiratory therapy. Only 39.0% of the total cohort had pulmonary function testing during the study period (40.0% commercial vs. 33.2% Medicare).\n[SUBTITLE] COPD exacerbations and health service utilization by disease complexity [SUBSECTION] Evaluation of exacerbations by complexity group demonstrates a general increase in the percentage of patients experiencing total exacerbations and hospitalized exacerbations with increasing disease complexity in both age cohorts (Figure 2). In addition, a higher proportion of patients with high complexity disease in both data sets experienced multiple (≥2) exacerbations (61.7% high-complexity commercial patients; 49.0% high-complexity Medicare) than patients with moderate- (commercial 56.9%, Medicare 41.6%) or low-complexity disease (commercial 33.4%, Medicare 20.5%), highlighting the relationship between complexity and exacerbations. In addition, for both the commercial and Medicare data sets, high-complexity patients had higher numbers of exacerbations per patient (mean 3.08 commercial patients; 2.22 Medicare patients) compared with moderate (commercial 2.47; Medicare 1.76) or low (commercial 1.40; Medicare 0.91) complexity. The majority of patients experiencing multiple (≥2) exacerbations requiring hospitalization were classified as high complexity (7.9% high complexity commercial patients; 10.6% Medicare), with ≤2% of moderate-complexity patients (1.8% and 2.1%) and <1% of low-complexity patients (0% and 0.01%), respectively.\nCOPD Patients Registering an inpatient exacerbation or experiencing any exacerbation, stratified by complexity, for (A) commercial, and (B) Medicare cohorts. †Any exacerbation includes inpatient hospitalized exacerbation, an ambulatory exacerbation with qualifying diagnosis or ambulatory exacerbation with qualifying drug therapy.\nUtilization of healthcare services was also observed to increase with an increase in complexity (Table 5). High-complexity patients had more than twice the pulmonary function tests per-patient compared with low-complexity patients, in both data sets. Patients with high complexity illness had more office/consultation visits than those with low complexity illness (5.7 more office/consultations in the commercial group and 3.6 more in the Medicare data set).\nProportion (%) of COPD patients in the Medicare and Commercial Cohorts Using ≥1 healthcare services, and mean number of each form of health service utilized per COPD patient, stratified by complexity of illness\nEvaluation of exacerbations by complexity group demonstrates a general increase in the percentage of patients experiencing total exacerbations and hospitalized exacerbations with increasing disease complexity in both age cohorts (Figure 2). In addition, a higher proportion of patients with high complexity disease in both data sets experienced multiple (≥2) exacerbations (61.7% high-complexity commercial patients; 49.0% high-complexity Medicare) than patients with moderate- (commercial 56.9%, Medicare 41.6%) or low-complexity disease (commercial 33.4%, Medicare 20.5%), highlighting the relationship between complexity and exacerbations. In addition, for both the commercial and Medicare data sets, high-complexity patients had higher numbers of exacerbations per patient (mean 3.08 commercial patients; 2.22 Medicare patients) compared with moderate (commercial 2.47; Medicare 1.76) or low (commercial 1.40; Medicare 0.91) complexity. The majority of patients experiencing multiple (≥2) exacerbations requiring hospitalization were classified as high complexity (7.9% high complexity commercial patients; 10.6% Medicare), with ≤2% of moderate-complexity patients (1.8% and 2.1%) and <1% of low-complexity patients (0% and 0.01%), respectively.\nCOPD Patients Registering an inpatient exacerbation or experiencing any exacerbation, stratified by complexity, for (A) commercial, and (B) Medicare cohorts. †Any exacerbation includes inpatient hospitalized exacerbation, an ambulatory exacerbation with qualifying diagnosis or ambulatory exacerbation with qualifying drug therapy.\nUtilization of healthcare services was also observed to increase with an increase in complexity (Table 5). High-complexity patients had more than twice the pulmonary function tests per-patient compared with low-complexity patients, in both data sets. Patients with high complexity illness had more office/consultation visits than those with low complexity illness (5.7 more office/consultations in the commercial group and 3.6 more in the Medicare data set).\nProportion (%) of COPD patients in the Medicare and Commercial Cohorts Using ≥1 healthcare services, and mean number of each form of health service utilized per COPD patient, stratified by complexity of illness", "Eligible patients from the commercial or Medicare cohorts were identified as having COPD, and then stratified by complexity as shown in Figure 1. From the 7,671,018 health plan members in the commercial dataset, 42,565 (0.55%) met the criteria for COPD (Table 2). The median age was 56 years and 51.4% of patients were female. From the 115,652 health plan members in the Medicare dataset, 8,507 (7.4%) were identified as having COPD. In this dataset, the median age was 75 years and 53.1% were female. Although over half of both cohorts were classified with moderate- or high complexity disease, a larger proportion of the older Medicare cohort were classified as either high or moderate complexity (63.4%) compared with the younger commercial cohort (55.0%). This was mostly due to a higher percentage of high complexity patients in the Medicare cohort (Table 2).\nStratification of patients in the study.\nDemographic characteristics for the commercial and Medicare data sets, stratified by complexity", "Specific differences were identified in the prevalence of comorbid conditions between the commercial and Medicare data sets (Table 3). For example, most cardiovascular complications were significantly less common in the younger commercial cohort than the older Medicare cohort, including hypertension, ischemic heart disease, and heart failure. Those in the younger commercial cohort were significantly more likely to have episodes of upper respiratory tract complaints including allergic rhinitis and sinusitis compared with the older Medicare cohort, who were more likely to be diagnosed with pneumonia. Furthermore, a significantly higher proportion of patients in the younger commercial database compared with the older Medicare dataset were documented to have current tobacco use during the study period.\nPrevalence (%) of comorbid conditions in COPD patients from commercial and Medicare data sets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nMost COPD patients in both the younger commercial cohort and the older Medicare cohort experienced at least 1 exacerbation during the study period (Table 4). Patients in the commercial cohort were more likely than Medicare members to experience an exacerbation of any type (commercial 70.6% and Medicare 61.0%), and the number of exacerbations per-patient in the commercial population was higher than in the Medicare population (2.13 vs. 1.58 per patient per year). However, patients in the older Medicare group were more likely than commercial patients to experience an exacerbation that led to hospitalization (19.4% vs. 13.9%). Furthermore, Medicare patients had more hospitalized exacerbations (0.25 per patient per year) than commercial patients (0.17 per patient per year). Multiple exacerbations (≥2) occurred more commonly in commercial patients (47.3% of commercial patients and 35.9% of Medicare patients), but multiple hospitalized exacerbations occurred in 2.3% and 3.6% of patients in each group, respectively.\nPrevalence (%) of patients experiencing ≥1 exacerbation and health services utilization among COPD patients in the commercial and Medicare datasets*\n*All of the differences comparing prevalence of comorbid conditions between commercial and Medicare populations were significant (p < 0.01), using two-tailed, chi-squared analysis, except for those indicated by italic text.\nImportant differences were identified in health service utilization between cohorts (Table 4). While nearly all of both populations experienced at least 1 office visit consultation, Medicare patients were significantly more likely to have been hospitalized for any reason than commercial patients. Although standard chest X-ray, ECG's, chest/thorax CT/MRIs, and heart echo exams were more likely in Medicare patients, more of the younger commercial patients had pulmonary function testing, cardiac catheterization/coronary angiography, and outpatient respiratory therapy. Only 39.0% of the total cohort had pulmonary function testing during the study period (40.0% commercial vs. 33.2% Medicare).", "Evaluation of exacerbations by complexity group demonstrates a general increase in the percentage of patients experiencing total exacerbations and hospitalized exacerbations with increasing disease complexity in both age cohorts (Figure 2). In addition, a higher proportion of patients with high complexity disease in both data sets experienced multiple (≥2) exacerbations (61.7% high-complexity commercial patients; 49.0% high-complexity Medicare) than patients with moderate- (commercial 56.9%, Medicare 41.6%) or low-complexity disease (commercial 33.4%, Medicare 20.5%), highlighting the relationship between complexity and exacerbations. In addition, for both the commercial and Medicare data sets, high-complexity patients had higher numbers of exacerbations per patient (mean 3.08 commercial patients; 2.22 Medicare patients) compared with moderate (commercial 2.47; Medicare 1.76) or low (commercial 1.40; Medicare 0.91) complexity. The majority of patients experiencing multiple (≥2) exacerbations requiring hospitalization were classified as high complexity (7.9% high complexity commercial patients; 10.6% Medicare), with ≤2% of moderate-complexity patients (1.8% and 2.1%) and <1% of low-complexity patients (0% and 0.01%), respectively.\nCOPD Patients Registering an inpatient exacerbation or experiencing any exacerbation, stratified by complexity, for (A) commercial, and (B) Medicare cohorts. †Any exacerbation includes inpatient hospitalized exacerbation, an ambulatory exacerbation with qualifying diagnosis or ambulatory exacerbation with qualifying drug therapy.\nUtilization of healthcare services was also observed to increase with an increase in complexity (Table 5). High-complexity patients had more than twice the pulmonary function tests per-patient compared with low-complexity patients, in both data sets. Patients with high complexity illness had more office/consultation visits than those with low complexity illness (5.7 more office/consultations in the commercial group and 3.6 more in the Medicare data set).\nProportion (%) of COPD patients in the Medicare and Commercial Cohorts Using ≥1 healthcare services, and mean number of each form of health service utilized per COPD patient, stratified by complexity of illness", "This large, retrospective, cross-sectional analysis of US managed care administrative claims data describes a practical method to identify COPD patients using claims data. Furthermore, this methodology defines the complexity of their illness as a proxy of disease severity, and documents their exacerbations. The finding of a progressive increase in the prevalence of COPD exacerbations and utilization of health services for COPD patients from low to moderate to high complexity in both the commercial and Medicare data sets validates the utility of the complexity algorithm. Patients with COPD classified as high complexity had the highest health services utilization and were most likely to experience an exacerbation.\n[SUBTITLE] Comorbid conditions [SUBSECTION] Our complexity classification was derived based on claims identifying selected comorbid conditions or medical procedures. Patients with COPD are typically thought to have comorbid diseases and conditions that contribute to the high burden of their disease [7,10,13,14]. Our data highlight that 4-27% of COPD patients had a respiratory comorbid condition, but up to 72% had a non-respiratory comorbid condition. Consequently over half of both age cohorts were stratified as moderate or high complexity. Our findings concur with other data [7,10], highlighting the high incidence of comorbidities in COPD patients. For example a study of 200 patients showed that patients with COPD had an average of 3.7 chronic medical conditions (including lung disease), compared with 1.8 chronic medical conditions for the controls [7]. Furthermore, a review by Sin et al., highlighted that a large proportion of patients with COPD have comorbid cardiovascular disease, depression, muscle wasting, reduced fat-free mass, osteopenia, and chronic infections [10].\nComorbidities in COPD patients contribute towards the high morbidity and mortality [14]. In particular, cardiovascular disease is a common cause of death in COPD patients [15-17], with co-prevalence estimates in COPD patients of 43% (mean age 68.8 years [14]) and 45% (mean age 67.5 years [7]). In the present study, similar prevalence estimates for key cardiovascular disease risk factors were found in commercial patients (median age 56 years), although higher prevalence in older patients, with 72% of the Medicare cohort (median age 75 years) having comorbid hypertension and 47% having dyslipidemia. In addition, heart failure (HF) is a risk factor for mortality in COPD patients, and particularly in those experiencing an exacerbation [14,18]. In the present study nearly 40% of the older Medicare cohort had HF, a proportion in agreement with a study of a similarly aged population among whom 40% had HF [18].\nOur complexity classification was derived based on claims identifying selected comorbid conditions or medical procedures. Patients with COPD are typically thought to have comorbid diseases and conditions that contribute to the high burden of their disease [7,10,13,14]. Our data highlight that 4-27% of COPD patients had a respiratory comorbid condition, but up to 72% had a non-respiratory comorbid condition. Consequently over half of both age cohorts were stratified as moderate or high complexity. Our findings concur with other data [7,10], highlighting the high incidence of comorbidities in COPD patients. For example a study of 200 patients showed that patients with COPD had an average of 3.7 chronic medical conditions (including lung disease), compared with 1.8 chronic medical conditions for the controls [7]. Furthermore, a review by Sin et al., highlighted that a large proportion of patients with COPD have comorbid cardiovascular disease, depression, muscle wasting, reduced fat-free mass, osteopenia, and chronic infections [10].\nComorbidities in COPD patients contribute towards the high morbidity and mortality [14]. In particular, cardiovascular disease is a common cause of death in COPD patients [15-17], with co-prevalence estimates in COPD patients of 43% (mean age 68.8 years [14]) and 45% (mean age 67.5 years [7]). In the present study, similar prevalence estimates for key cardiovascular disease risk factors were found in commercial patients (median age 56 years), although higher prevalence in older patients, with 72% of the Medicare cohort (median age 75 years) having comorbid hypertension and 47% having dyslipidemia. In addition, heart failure (HF) is a risk factor for mortality in COPD patients, and particularly in those experiencing an exacerbation [14,18]. In the present study nearly 40% of the older Medicare cohort had HF, a proportion in agreement with a study of a similarly aged population among whom 40% had HF [18].\n[SUBTITLE] Exacerbations [SUBSECTION] Our study describes a practical way to identify COPD exacerbations using claims data, and to classify them by site of care. In the current study the majority of patients, regardless of age, experienced COPD exacerbations, which often involved hospitalization or an emergency room visit. Others have used healthcare utilization to define the staging of COPD exacerbations, in order to incorporate those parameters considered most appropriate to base sub-classification [19]. Indeed, classification of an exacerbation as mild, moderate, or severe was deemed to be strongly related to the patient's underlying condition [19]. Accordingly, for a patient with severe COPD just a small change in lung function may present as a moderate-to-severe exacerbation because it necessitates physician intervention and increased healthcare utilization [19]. We stratified exacerbations by the site of care (inpatient hospitalization, emergency room visit, and ambulatory exacerbations), which enabled more detailed information regarding the patients' site of healthcare utilization to be accounted for. This is also somewhat similar to symptom-based classification of exacerbations often used in clinical trials which capture patients whose condition has changed enough to require a change in treatment, an emergency room visit, or hospitalization.\nPatients in the older Medicare cohort were more likely to have moderate- or high complexity illness compared with the younger commercial dataset. This might be expected, as lung function and general health declines with age [20,21]. Our data also highlight that these patients qualified as high complexity were more likely to have multiple exacerbations and exacerbations requiring hospitalization compared with those of moderate or low complexity. As data were de-identified, it was not possible to confirm exacerbation information with corresponding medical records or clinical observations in this analysis. However, by distinguishing these high-risk patients from the overall COPD population, it might be ultimately possible to specifically target high-risk patients to limit the severity of their exacerbations with appropriate adjustment of therapy and/or monitoring of their comorbid conditions.\nAlthough this study did not evaluate healthcare costs, the data generated could be used for future economic analyses and modeling. Others have reported the economic burden that exacerbations place on the healthcare system. Indeed, the estimated costs of exacerbations have been found to vary widely across studies from approx. $88 to $7,757 per exacerbation (2007 US dollars) [22]. Furthermore, exacerbations accounted for 35%-45% of the total per capita healthcare costs for COPD in one study, with costs increasing with severity of exacerbations [23]. Accordingly, investigating how actual or projected costs of COPD may change through identification of patients by exacerbations and subsequent stratification by complexity, as described in the present study, would be an interesting area to investigate further.\nOur study describes a practical way to identify COPD exacerbations using claims data, and to classify them by site of care. In the current study the majority of patients, regardless of age, experienced COPD exacerbations, which often involved hospitalization or an emergency room visit. Others have used healthcare utilization to define the staging of COPD exacerbations, in order to incorporate those parameters considered most appropriate to base sub-classification [19]. Indeed, classification of an exacerbation as mild, moderate, or severe was deemed to be strongly related to the patient's underlying condition [19]. Accordingly, for a patient with severe COPD just a small change in lung function may present as a moderate-to-severe exacerbation because it necessitates physician intervention and increased healthcare utilization [19]. We stratified exacerbations by the site of care (inpatient hospitalization, emergency room visit, and ambulatory exacerbations), which enabled more detailed information regarding the patients' site of healthcare utilization to be accounted for. This is also somewhat similar to symptom-based classification of exacerbations often used in clinical trials which capture patients whose condition has changed enough to require a change in treatment, an emergency room visit, or hospitalization.\nPatients in the older Medicare cohort were more likely to have moderate- or high complexity illness compared with the younger commercial dataset. This might be expected, as lung function and general health declines with age [20,21]. Our data also highlight that these patients qualified as high complexity were more likely to have multiple exacerbations and exacerbations requiring hospitalization compared with those of moderate or low complexity. As data were de-identified, it was not possible to confirm exacerbation information with corresponding medical records or clinical observations in this analysis. However, by distinguishing these high-risk patients from the overall COPD population, it might be ultimately possible to specifically target high-risk patients to limit the severity of their exacerbations with appropriate adjustment of therapy and/or monitoring of their comorbid conditions.\nAlthough this study did not evaluate healthcare costs, the data generated could be used for future economic analyses and modeling. Others have reported the economic burden that exacerbations place on the healthcare system. Indeed, the estimated costs of exacerbations have been found to vary widely across studies from approx. $88 to $7,757 per exacerbation (2007 US dollars) [22]. Furthermore, exacerbations accounted for 35%-45% of the total per capita healthcare costs for COPD in one study, with costs increasing with severity of exacerbations [23]. Accordingly, investigating how actual or projected costs of COPD may change through identification of patients by exacerbations and subsequent stratification by complexity, as described in the present study, would be an interesting area to investigate further.\n[SUBTITLE] Utilization of healthcare services [SUBSECTION] We identified groups of COPD patients that are high users of healthcare services. For example, over half of the Medicare COPD patients and almost 40% of commercial COPD patients were hospitalized at least once for any reason in a 1-year period. Evidence of high outpatient utilization is characterized by the finding that virtually all patients in both populations (≥96%) had at least 1 office visit/consultation with a mean of over 10 per patient during the year. Furthermore, our method of stratifying patients found differential healthcare utilization; our high complexity patients had the highest health services utilization across almost all of the services monitored. As our method of stratifying patients to high and medium complexity was dependent on their comorbid conditions and the procedures they received, it is logical to expect that these patients had higher utilization compared with those of low complexity.\nWe do not know if our high complexity group had higher healthcare utilization in subsequent years, and this would be an important subject for future investigations. Interestingly, although GOLD guidelines state that lung function testing with spirometry is essential for the diagnosis and management of COPD, and indeed can provide a useful description of the severity of pathological changes in COPD [1], less than half of the population had a pulmonary function test carried out during the study year.\nOur data should be interpreted in light of some important limitations. These analyses are retrospective and descriptive in nature; there were no control groups. The large sample sizes resulted in statistically significant differences between cohorts that may not be clinically significant. However, we adjusted for the large sample size by defining significance as p < 0.01, rather than the conventional level of p < 0.05. Administrative claims data are primarily generated for reimbursement purposes rather than research purposes. Consequently, the accuracy of claims data is dependent on the precision and timing of the coding associated with their use. As such, some comorbid conditions, such as obesity and tobacco use, are often under-reported in claims data and likely are under-reported here as well. Furthermore, confusion over the differential diagnosis of asthma and COPD could have led to some patients in our COPD cohorts who actually had asthma rather than COPD. In fact, at least 1 claim with a diagnosis of asthma was recorded in 27% and 21% of the commercial and Medicare COPD cohorts, respectively. Findings of cross-sectional studies have shown a similar overlap of up to 30% between people who have a clinical diagnosis of COPD and asthma [24]. While some of these claims could represent coexisting asthma and COPD, the potential for diagnosis coding errors needs to be taken into consideration. The possibility that other respiratory conditions, such as bronchiectases, hypoventilation-obesity, or overlap syndrome, were miscoded as COPD should also be considered. Furthermore, there is a possibility that asthma attacks were miscoded as COPD exacerbations. Because the study period was limited to 1 year, it is possible for the data to provide false positives or negatives in the population selection criteria and overstate or understate the clinical severity of the disease being considered. In this case, individuals with COPD who did not have COPD-related claims within the observation period are not included in the analysis. Since these patients are likely to have less severe disease, this would result in the overall prevalence of COPD being understated, while the burden per patient with COPD overstated. On the other hand, because patients had to be continuously enrolled during the entire study year to be eligible for inclusion in this cross-sectional analysis, those who either left their health plan or died were excluded from the analysis. Excluding those who died could have removed those who were most ill and utilized the most healthcare resources. This could be especially important in the elderly cohort.\nIt is also possible that differences in exacerbations and healthcare utilization between different groups could be due to unrecognized confounders, such as duration of COPD, current smoking status, and type and duration of medication use. In addition, given that the study was a cross-sectional analysis, no cause and effect analyses could be conducted. We broadly compared an older Medicare population with a younger employer-based commercial population, although 7% of the Medicare population was <65 years and we were unable to extract these patients from the data set. However, as over 90% of the patients were ≥65 years we feel our observations provide useful information on how age influences the burden of COPD on managed care resources\nDespite these limitations, the methodology presented here provides a practical way for healthcare providers to identify and stratify COPD patients and identify those experiencing exacerbations within a large managed care database. In turn, this might help healthcare providers prioritize patients at risk for future exacerbations and resource utilization, to help ensure that those COPD patients with the greatest need for close monitoring receive optimal care.\nWe identified groups of COPD patients that are high users of healthcare services. For example, over half of the Medicare COPD patients and almost 40% of commercial COPD patients were hospitalized at least once for any reason in a 1-year period. Evidence of high outpatient utilization is characterized by the finding that virtually all patients in both populations (≥96%) had at least 1 office visit/consultation with a mean of over 10 per patient during the year. Furthermore, our method of stratifying patients found differential healthcare utilization; our high complexity patients had the highest health services utilization across almost all of the services monitored. As our method of stratifying patients to high and medium complexity was dependent on their comorbid conditions and the procedures they received, it is logical to expect that these patients had higher utilization compared with those of low complexity.\nWe do not know if our high complexity group had higher healthcare utilization in subsequent years, and this would be an important subject for future investigations. Interestingly, although GOLD guidelines state that lung function testing with spirometry is essential for the diagnosis and management of COPD, and indeed can provide a useful description of the severity of pathological changes in COPD [1], less than half of the population had a pulmonary function test carried out during the study year.\nOur data should be interpreted in light of some important limitations. These analyses are retrospective and descriptive in nature; there were no control groups. The large sample sizes resulted in statistically significant differences between cohorts that may not be clinically significant. However, we adjusted for the large sample size by defining significance as p < 0.01, rather than the conventional level of p < 0.05. Administrative claims data are primarily generated for reimbursement purposes rather than research purposes. Consequently, the accuracy of claims data is dependent on the precision and timing of the coding associated with their use. As such, some comorbid conditions, such as obesity and tobacco use, are often under-reported in claims data and likely are under-reported here as well. Furthermore, confusion over the differential diagnosis of asthma and COPD could have led to some patients in our COPD cohorts who actually had asthma rather than COPD. In fact, at least 1 claim with a diagnosis of asthma was recorded in 27% and 21% of the commercial and Medicare COPD cohorts, respectively. Findings of cross-sectional studies have shown a similar overlap of up to 30% between people who have a clinical diagnosis of COPD and asthma [24]. While some of these claims could represent coexisting asthma and COPD, the potential for diagnosis coding errors needs to be taken into consideration. The possibility that other respiratory conditions, such as bronchiectases, hypoventilation-obesity, or overlap syndrome, were miscoded as COPD should also be considered. Furthermore, there is a possibility that asthma attacks were miscoded as COPD exacerbations. Because the study period was limited to 1 year, it is possible for the data to provide false positives or negatives in the population selection criteria and overstate or understate the clinical severity of the disease being considered. In this case, individuals with COPD who did not have COPD-related claims within the observation period are not included in the analysis. Since these patients are likely to have less severe disease, this would result in the overall prevalence of COPD being understated, while the burden per patient with COPD overstated. On the other hand, because patients had to be continuously enrolled during the entire study year to be eligible for inclusion in this cross-sectional analysis, those who either left their health plan or died were excluded from the analysis. Excluding those who died could have removed those who were most ill and utilized the most healthcare resources. This could be especially important in the elderly cohort.\nIt is also possible that differences in exacerbations and healthcare utilization between different groups could be due to unrecognized confounders, such as duration of COPD, current smoking status, and type and duration of medication use. In addition, given that the study was a cross-sectional analysis, no cause and effect analyses could be conducted. We broadly compared an older Medicare population with a younger employer-based commercial population, although 7% of the Medicare population was <65 years and we were unable to extract these patients from the data set. However, as over 90% of the patients were ≥65 years we feel our observations provide useful information on how age influences the burden of COPD on managed care resources\nDespite these limitations, the methodology presented here provides a practical way for healthcare providers to identify and stratify COPD patients and identify those experiencing exacerbations within a large managed care database. In turn, this might help healthcare providers prioritize patients at risk for future exacerbations and resource utilization, to help ensure that those COPD patients with the greatest need for close monitoring receive optimal care.", "Our complexity classification was derived based on claims identifying selected comorbid conditions or medical procedures. Patients with COPD are typically thought to have comorbid diseases and conditions that contribute to the high burden of their disease [7,10,13,14]. Our data highlight that 4-27% of COPD patients had a respiratory comorbid condition, but up to 72% had a non-respiratory comorbid condition. Consequently over half of both age cohorts were stratified as moderate or high complexity. Our findings concur with other data [7,10], highlighting the high incidence of comorbidities in COPD patients. For example a study of 200 patients showed that patients with COPD had an average of 3.7 chronic medical conditions (including lung disease), compared with 1.8 chronic medical conditions for the controls [7]. Furthermore, a review by Sin et al., highlighted that a large proportion of patients with COPD have comorbid cardiovascular disease, depression, muscle wasting, reduced fat-free mass, osteopenia, and chronic infections [10].\nComorbidities in COPD patients contribute towards the high morbidity and mortality [14]. In particular, cardiovascular disease is a common cause of death in COPD patients [15-17], with co-prevalence estimates in COPD patients of 43% (mean age 68.8 years [14]) and 45% (mean age 67.5 years [7]). In the present study, similar prevalence estimates for key cardiovascular disease risk factors were found in commercial patients (median age 56 years), although higher prevalence in older patients, with 72% of the Medicare cohort (median age 75 years) having comorbid hypertension and 47% having dyslipidemia. In addition, heart failure (HF) is a risk factor for mortality in COPD patients, and particularly in those experiencing an exacerbation [14,18]. In the present study nearly 40% of the older Medicare cohort had HF, a proportion in agreement with a study of a similarly aged population among whom 40% had HF [18].", "Our study describes a practical way to identify COPD exacerbations using claims data, and to classify them by site of care. In the current study the majority of patients, regardless of age, experienced COPD exacerbations, which often involved hospitalization or an emergency room visit. Others have used healthcare utilization to define the staging of COPD exacerbations, in order to incorporate those parameters considered most appropriate to base sub-classification [19]. Indeed, classification of an exacerbation as mild, moderate, or severe was deemed to be strongly related to the patient's underlying condition [19]. Accordingly, for a patient with severe COPD just a small change in lung function may present as a moderate-to-severe exacerbation because it necessitates physician intervention and increased healthcare utilization [19]. We stratified exacerbations by the site of care (inpatient hospitalization, emergency room visit, and ambulatory exacerbations), which enabled more detailed information regarding the patients' site of healthcare utilization to be accounted for. This is also somewhat similar to symptom-based classification of exacerbations often used in clinical trials which capture patients whose condition has changed enough to require a change in treatment, an emergency room visit, or hospitalization.\nPatients in the older Medicare cohort were more likely to have moderate- or high complexity illness compared with the younger commercial dataset. This might be expected, as lung function and general health declines with age [20,21]. Our data also highlight that these patients qualified as high complexity were more likely to have multiple exacerbations and exacerbations requiring hospitalization compared with those of moderate or low complexity. As data were de-identified, it was not possible to confirm exacerbation information with corresponding medical records or clinical observations in this analysis. However, by distinguishing these high-risk patients from the overall COPD population, it might be ultimately possible to specifically target high-risk patients to limit the severity of their exacerbations with appropriate adjustment of therapy and/or monitoring of their comorbid conditions.\nAlthough this study did not evaluate healthcare costs, the data generated could be used for future economic analyses and modeling. Others have reported the economic burden that exacerbations place on the healthcare system. Indeed, the estimated costs of exacerbations have been found to vary widely across studies from approx. $88 to $7,757 per exacerbation (2007 US dollars) [22]. Furthermore, exacerbations accounted for 35%-45% of the total per capita healthcare costs for COPD in one study, with costs increasing with severity of exacerbations [23]. Accordingly, investigating how actual or projected costs of COPD may change through identification of patients by exacerbations and subsequent stratification by complexity, as described in the present study, would be an interesting area to investigate further.", "We identified groups of COPD patients that are high users of healthcare services. For example, over half of the Medicare COPD patients and almost 40% of commercial COPD patients were hospitalized at least once for any reason in a 1-year period. Evidence of high outpatient utilization is characterized by the finding that virtually all patients in both populations (≥96%) had at least 1 office visit/consultation with a mean of over 10 per patient during the year. Furthermore, our method of stratifying patients found differential healthcare utilization; our high complexity patients had the highest health services utilization across almost all of the services monitored. As our method of stratifying patients to high and medium complexity was dependent on their comorbid conditions and the procedures they received, it is logical to expect that these patients had higher utilization compared with those of low complexity.\nWe do not know if our high complexity group had higher healthcare utilization in subsequent years, and this would be an important subject for future investigations. Interestingly, although GOLD guidelines state that lung function testing with spirometry is essential for the diagnosis and management of COPD, and indeed can provide a useful description of the severity of pathological changes in COPD [1], less than half of the population had a pulmonary function test carried out during the study year.\nOur data should be interpreted in light of some important limitations. These analyses are retrospective and descriptive in nature; there were no control groups. The large sample sizes resulted in statistically significant differences between cohorts that may not be clinically significant. However, we adjusted for the large sample size by defining significance as p < 0.01, rather than the conventional level of p < 0.05. Administrative claims data are primarily generated for reimbursement purposes rather than research purposes. Consequently, the accuracy of claims data is dependent on the precision and timing of the coding associated with their use. As such, some comorbid conditions, such as obesity and tobacco use, are often under-reported in claims data and likely are under-reported here as well. Furthermore, confusion over the differential diagnosis of asthma and COPD could have led to some patients in our COPD cohorts who actually had asthma rather than COPD. In fact, at least 1 claim with a diagnosis of asthma was recorded in 27% and 21% of the commercial and Medicare COPD cohorts, respectively. Findings of cross-sectional studies have shown a similar overlap of up to 30% between people who have a clinical diagnosis of COPD and asthma [24]. While some of these claims could represent coexisting asthma and COPD, the potential for diagnosis coding errors needs to be taken into consideration. The possibility that other respiratory conditions, such as bronchiectases, hypoventilation-obesity, or overlap syndrome, were miscoded as COPD should also be considered. Furthermore, there is a possibility that asthma attacks were miscoded as COPD exacerbations. Because the study period was limited to 1 year, it is possible for the data to provide false positives or negatives in the population selection criteria and overstate or understate the clinical severity of the disease being considered. In this case, individuals with COPD who did not have COPD-related claims within the observation period are not included in the analysis. Since these patients are likely to have less severe disease, this would result in the overall prevalence of COPD being understated, while the burden per patient with COPD overstated. On the other hand, because patients had to be continuously enrolled during the entire study year to be eligible for inclusion in this cross-sectional analysis, those who either left their health plan or died were excluded from the analysis. Excluding those who died could have removed those who were most ill and utilized the most healthcare resources. This could be especially important in the elderly cohort.\nIt is also possible that differences in exacerbations and healthcare utilization between different groups could be due to unrecognized confounders, such as duration of COPD, current smoking status, and type and duration of medication use. In addition, given that the study was a cross-sectional analysis, no cause and effect analyses could be conducted. We broadly compared an older Medicare population with a younger employer-based commercial population, although 7% of the Medicare population was <65 years and we were unable to extract these patients from the data set. However, as over 90% of the patients were ≥65 years we feel our observations provide useful information on how age influences the burden of COPD on managed care resources\nDespite these limitations, the methodology presented here provides a practical way for healthcare providers to identify and stratify COPD patients and identify those experiencing exacerbations within a large managed care database. In turn, this might help healthcare providers prioritize patients at risk for future exacerbations and resource utilization, to help ensure that those COPD patients with the greatest need for close monitoring receive optimal care.", "We present a unique and practical method for identifying patients with COPD, determining disease severity - as \"complexity of illness\" - and documenting exacerbations using claims data. Our data highlight important differences in comorbidities, exacerbations, and healthcare utilizations in older (aged ≥65 years) compared with younger (aged <65 years) COPD patients. Furthermore, by stratifying COPD patients based on diagnostic, procedures, and services codes, we have demonstrated that patients stratified as having high- or moderate complexity disease experienced a higher number of exacerbations than those with low complexity disease. Additionally, by stratifying by complexity of illness we show linearity between complexity of illness and utilization of healthcare services and hospitalizations.\nIdentification of COPD patients at higher risk of complications using complexity stratification and/or identifying those experiencing exacerbations may serve to improve patient management, and ultimately reduce the burden of disease to the patient, and to the healthcare systems supporting them.", "Dr. Mapel was a paid consultant to Pfizer Inc in connection with conduct of the analysis and development of the manuscript. Dr. Mapel has served as a consultant to and received research funding from Pfizer Pharmaceuticals, GlaxoSmithKline, and AstraZeneca. Drs. Woodruff, Marton, and Dutro, are employees of, and own stock in Pfizer Inc. Dr. Make has participated in advisory boards and received honoraria for speaking from Pfizer Inc within the past five years.", "All authors participated in the design of the study, and contributed to drafting the manuscript. All authors read and approved the final manuscript.", "Parts of these data were presented as a poster at the American Thoracic Society 103rd International Conference, May 22, 2007, San Francisco, CA, USA.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1472-6963/11/43/prepub\n", "Classifying COPD patients by complexity of illness. 1A: The analysis included the following comorbid conditions and health care services that were predetermined and evaluated by the presence of diagnostic, procedures, and services codes (2004 ICD-9 CM, CPT-4, and HCPCS codes). Diagnosis codes used to define COPD complexity level. 1B: Comorbid respiratory conditions and medical procedures at any time during the 1-year study period (July 1, 2004, to June 30, 2005) (see Additional File 1A) were used to assign patients to 1 of 3 disease complexity levels (high, moderate, or low) based on selected diagnostic, procedures and services codes (2004 ICD-9, CPT-4, and HCPCS), as detailed. If a COPD patient did not have any comorbid condition for high or moderate complexity, they were classified as low complexity.\nClick here for file" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Allergens", "Animals", "Asthma", "Bronchoalveolar Lavage Fluid", "Chemokines, CXC", "Cockroaches", "Cytokines", "Disease Models, Animal", "Female", "Immunoglobulin E", "Immunoglobulin G", "Lipopolysaccharides", "Lung", "Mice", "Mice, Inbred BALB C", "Pneumonia", "Severity of Illness Index", "Tumor Necrosis Factor-alpha" ]

[ "Background", "Animals", "Reduction of LPS content in House Dust Extract", "LPS Assay", "Asthma induction protocol", "Timepoints for data collection", "Airways Hyperresponsiveness and Airway Resistance", "Bronchoalveolar lavage and lung homogenate preparation", "Myeloperoxidase and Eosinophil Peroxidase Assays", "ELISA", "Cysteinyl-Leukotriene Immunoassay", "Statistical Analysis", "Results", "Depletion of LPS from the house dust extract", "TNFα production in BAL fluid post final challenge", "Role of LPS in antibody production", "Inflammatory cell recruitment in the BAL fluid post allergen challenge", "Induction of airways hyperresponsiveness after allergen challenge", "Th1 and Th2 cytokine production in the BAL post allergen challenge", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Asthma is a chronic allergic disorder characterized by IgE production, airway eosinophilia and bronchial hyperresponsiveness[2]. Numerous studies have shown that allergens are not solely responsible for the severity of the asthmatic response (reviewed in[3]). Microbial components, such as lipopolysaccharide (LPS) from Gram negative bacteria, are ubiquitously present in the environment, including the ambient air. In addition to being potent activators of innate immunity, these compounds have been shown to modulate asthma severity[4]. The presence of microbial components often is associated with environmental and household cleanliness, however the hygiene hypothesis postulates that the lack of exposure to these pathogens at a young age increases susceptibility to allergen sensitization, accounting for the dramatic increases in allergic diseases in the developed world [5,6].\nSeveral studies have been carried out investigating the role LPS plays in allergen sensitization. These studies suggest that in ovalbumin (OVA) models, increasing doses of LPS protect against eosinophilia and AHR[7,8]. Recently, more evidence has emerged implicating reduced tumor necrosis factor-α (TNFα) production as the mechanism responsible for this protection. Eisenbarth. et. al. showed that OVA sensitization in toll-like receptor-4 (TLR4) deficient mice did not result in IgE production or eosinophil recruitment, however these responses could be restored by exogenous administration of TNFα shortly after sensitization[8]. Additionally, mice deficient in TNF-receptor associated factor-1 do not develop eosinophilia or AHR in response to OVA sensitization and challenge[9].\nWhile LPS plays a key role in allergen sensitization, the synergistic effect of LPS and allergen during the effector phase of asthma remains unclear. We sought to determine whether the removal of LPS from house dust obtained from the home environment can protect against the severity of murine airways-inflammation. We employed a clinically relevant model of asthma-like pulmonary inflammation based on immunization with a CRA[1] extract containing LPS, followed by challenge with a house dust extract (HDE) collected from the home of an asthmatic child. This HDE contains high levels of both CRA and LPS, and can induce the cardinal features of asthma-like inflammation, such as airway eosinophilia, airways hyperresponsiveness and IgE production[10,11]. In order to investigate the role of LPS at the time of allergen challenge, LPS was removed from the HDE using a commercially available column. The current study shows that reduction of the LPS content in the HDE exacerbates cytokine production, reduces IgE, but does not alter AHR, airway eosinophilia or neutrophil recruitment.", "Female BALB/c mice 9-12 weeks old were purchased from Jackson Laboratories (Bar Harbor, ME) and maintained under standard laboratory conditions. The mice were housed in a temperature and humidity controlled room with 12 hour light/dark cycles. Food and water were allowed ad libitum. All experiments were performed according to the National Institutes of Health guidelines and were approved by the Boston University and University of Michigan Institutional Animal Care and Use Committees.", "LPS was removed from the house dust extract (HDE) using the EndoTrap Blue column (Profos AG, Regensburg, Germany), by modifying the manufacturer's protocol. HDE was diluted 1:1 in sterile PBS. The depletion column was equilibrated with the provided equilibration buffer modified to contain 400 mM NaCl. The HDE mixture was applied to the column and eluted using the provided elution buffer. The eluted mixture was immediately aliquoted and stored at -70°C until use. Concentrations of Bla g (Blatella germanica) 1 and Bla g2 and LPS were measured after application to the column, and appropriate dilutions prepared for in vivo administration.", "LPS in the house dust extract and cockroach allergen preparations was assayed in pyrogen free water using the endpoint Limulus amoebocyte lysate (LAL) assay (Lonza, Basel Switzerland, 50-647 U). 96-well microplates and substrate solutions were warmed to 37°C. 50 μL of sample and standard were added to the plate in duplicate followed by 50 μL LAL regent. The plate was incubated at 37°C for 10 min. 100 μL of substrate solution was then added to each well and the plate was incubated at 37°C for 6 minutes. The reaction was stopped with 50 μL 25% glacial acetic acid and the absorbance was read at 405 nm.", "House dust collection and processing was performed as previously described[11]. We used a commercial CRA preparation for immunizations since there are limited amounts of HDE available. Additionally, our lab has demonstrated that asthma-like inflammation induced by the HDE is CRA specific[12] and CRA administration will produce asthma-like pulmonary inflammation[13,14]. For these reasons, the commercial German cockroach-allergen[1] (Greer Laboratories, LeNoir, NC, Item # B46) was used for immunization[12]. The CRA (61.9 ng) was diluted to a final volume of 50 μL in sterile phosphate buffered saline (PBS) immediately prior to use. This mixture was emulsified in 50 μL TiterMax Gold adjuvant (CytRx, Norcross, GA). Each mouse was injected i.p. with 100 μL of the adjuvant/allergen mixture on day 0. On day 14 and day 21, mice were challenged by direct intratracheal installation of 50 μL of the intact HDE or LPS-reduced HDE, each containing 61.9 ng CRA [15]. Briefly, mice were lightly anesthetized and suspended by their front incisors on a vertical board. Their tails were taped down to support the body weight. The tongue was gently extended and the liquid was placed at the base of the oropharynx so that it was inhaled.", "Animals were euthanized by cervical dislocation following ketamine/xylazine anesthesia at 0, 2 or 24 hours post final allergen challenge. At the time of sacrifice, plasma, bronchoalveolar lavage, and lung homogenates were prepared. The 2 and 24 hour timepoints represent the early phase and late phase of the asthmatic response based on prior publications[11]. The 0 hour timepoint refers to animals sacrificed on day 21 without receiving the second pulmonary challenge. This timepoint was included to determine if sensitization and the first challenge caused any lasting inflammatory reaction in either group. Airway hyperresponsiveness [10] was measured 4 hours post final challenge, based on our observations of robust AHR induction at this timepoint (data not shown). These mice were then sacrificed at 24 hours.", "Airways changes were measured either with direct invasive techniques (Flexivent, Scireq Scientific Respiratory Equipment, Montreal, Canada) or using unrestrained whole body plethysmography (Buxco Systems, Troy, NY). For whole body plethysmography, mice were placed in the instrument chamber and allowed to acclimate for at least 5 minutes. Baseline measurements were recorded for 5 minutes. Mice were then challenged for 2 minutes with aerosolized PBS and increasing doses of methacholine (Sigma, St. Louis, MO). Each aerosol challenge was followed by 5 minutes of monitoring and data collection. The partial pressure difference between the experimental and reference chambers represented the PenH parameter, and the data presented as the percent increase above baseline PenH measurements.\nThese data were further verified using invasive pulmonary function tests. For measurement of mouse airway resistance, mice were anesthetized with an i.p. injection of 1:5 diluted pentobarbital (Nembutal®, 0.016 ml/g body weight, Ovation Pharmaceutical, Deerfield, IL). The paralytic was pancuronium (Sigma-Aldrich, St. Louis, MO) at 0.5 micrograms per gram body weight. Once adequate surgical sedation was established, determined by a firm squeeze of the foot pad, a tracheotomy was performed by insertion of an 18 g polyethylene cannula into the distal trachea. The mouse was then placed on the FlexiVent mechanical ventilator (Scireq Scientific Respiratory Equipment, Montreal, Canada) and ventilated at 190 breaths per minute with positive-end expiratory pressure set at 3 cmH2O. Measurement of airway resistance in response to increasing concentrations of aerosolized methacholine was obtained through periodic computer-generated \"snapshot 150\" forced-maneuver interruptions in ventilation. Data are then presented as resistance change from baseline (cmH2O per milliliter per second).", "Mice were exanguinated and BAL performed by cannulating the trachea. The lung was lavaged with 2, 1 mL aliquots of warm Hank's Buffered Salt Solution (HBSS, Gibco, Grand Island, NY). Both aliquots were centrifuged and the supernatant of the first wash removed and frozen at -20°C for cytokine analysis. The supernatant from the second aliquot was discarded and the cell pellets were resuspended and combined. Total cell counts were obtained using a Beckman-Coulter particle counter model ZF (Coulter Electronics Inc., Hialeah, FL). Cytospin preparations were stained with Diff-Quick and 300 cell differential counts were performed to determine the absolute numbers of inflammatory cells. The right lung was removed, placed in ice cold protease inhibitor cocktail (Roche, Indianapolis, IN) containing 0.00005% Triton X-100 in PBS, and homogenized with 3, 10 second passes in a Brinkmann Polytron PT3000 homogenizer. An aliquot was removed and sonicated in hexadecyltrimethylammonium bromide (HTAB) buffer for myeloperoxidase assay. A separate aliquot was removed and sonicated in 0.5% cetyltrimethylammoniumchloride (CTAC) (Sigma, St. Louis, MO) for eosinophil-specific peroxidase assay. The homogenized and sonicated mixtures were centrifuged at 15,000 g for 15 min. The homogenate supernatant was removed and stored at -20°C for cytokine analysis and the supernatant from the sonicated fractions was used immediately for peroxidase assays.", "Myeloperoxidase (MPO) and eosinophil peroxidase (EPO) assays were performed as described previously[16], with some modifications. EPO was performed by diluting the supernatant of the mixture sonicated in CTAC 1:3 in 10 mM HEPES, pH 8, in quadruplicate in a 96-well plate. 150 μL ice cold stop solution (4N H2SO4 + 2 mM resorcinol) was added to 2 of the sample wells. 75 μL of substrate solution containing 6 mM KBr, 1.5 mM o-phenylenediamine (Sigma, St. Louis, MO, P9029), and 0.3% H2O2 in 50 mM HEPES, pH8, was added to the two remaining sample wells and incubated in the dark for 30 seconds. 150 μL ice cold stop solution was added and the absorbance read at 490 nm.\nMPO was measured by diluting the supernatant of the mixture sonicated in HTAB 1:5 in 10 mM citrate buffer, pH 5 in quadruplicate. 150 μL ice cold stop solution (4N H2SO4) was added to 2 of the sample wells. 75 μL substrate solution containing 0.3 mM 3,3',5,5'-Tetramethylbenzidine, 120 mM resorcinol (an eosinophil peroxidase inhibitor), and 0.007% H2O2 in ddH2O was added to the two remaining sample wells and incubated in the dark for 2 minutes. 150 μL ice cold stop solution was added and the absorbance read at 450 nm. Data is expressed as ΔOD reflecting the difference in absorbance between the average of the sample and background wells.", "Cytokines and chemokines were measured by ELISA as previously described[17]. Matched antibody pairs and recombinant standards were purchased from R&D Systems (Minneapolis, MN). Lung homogenate samples were assayed with the addition of 20% normal lung homogenate to the standards to adjust for the increased background caused by non-specific matrix effects. Antibody ELISA reagents (IgG and IgE) were purchased from Bethyl Laboratories (Montgomery, TX) and assays were performed by the same standard protocols as for cytokines and chemokines. Cockroach allergen ELISAs were performed as previously described[12] with antibody pairs and recombinant standards from Indoor Biotechnologies (Charlottesville, VA).\nCRA-specific IgE was measured by coating ELISA plates with CRA (Greer Labs) over-night at 4°C. Plates were washed and non-specific binding blocked for 2.5 hours at room temperature on an orbital shaker. Plates were washed an incubated with 1:10 dilutions of plasma overnight at 4°C. Plates were washed and incubated with HRP-goat anti mouse IgE for 1 hour at room temperature on an orbital shaker. Plates were developed using 3,3',5,5'-Tetramethylbenzidine as substrate and read as previously described[12]. Results are represented as ΔOD (OD465 - OD590). Since CRA-specific IgE standards are not currently available, plasma from a naïve mouse was run in parallel in each assay for baseline.", "Cysteinyl leukotrienes in the BAL fluid were measured by an Enzyme-linked Immunoassay Kit (Cayman Chemicals, Ann Arbor, MI) according to the manufacturer instructions. All samples were run at two dilutions. %B/B0 values in the linear range of the standard curve were accepted. Sample values that did not fall in this range were appropriately diluted and rerun.", "All data are represented as mean ± SEM. Statistical significance was determined by unpaired Student's t-test or One-way ANOVA with Turkey's post test using GraphPad Prism version 4.0.3. (GraphPad Software, San Diego, CA). Statistical significance was achieved when p < 0.05.", "[SUBTITLE] Depletion of LPS from the house dust extract [SUBSECTION] Dust was collected from the homes of asthmatic children as described previously[11]. This dust was shown to contain significantly higher concentrations of cockroach allergens Bla g1 and Bla g2 in comparison to other indoor and outdoor allergens[12]. This extract, when used in previous studies at a 1:10 dilution (containing a total of 61.9 ng CRA) successfully induced the hallmark features of asthma-like inflammation, including airways hyperresponsiveness, eosinophilia and plasma IgE[18-20]. In the present study, we specifically sought to determine whether removal of the potent innate immune activating agent LPS from the HDE at the time of allergen challenge would attenuate the severity of pulmonary inflammation in sensitized mice. To address this question, LPS in the HDE was reduced as described in materials and methods. Unfortunately, our previous attempts at LPS removal using polymyxin B columns (Pierce, Rockford, IL) also removed 99.9% of Bla g1 from the HDE, and these preparations were not appropriate for the present studies (Table 1). Although use of the EndoTrap column in this study slightly reduced CRA concentrations, through appropriate dilution, we were able to maintain the same total amount of 61.9 ng CRA for all treatments (Table 2).\nTotal amounts of CRA and LPS in the intact HDE and in the HDE treated with Polymyxin B or EndoTrap to deplete LPS.\nAmounts of CRA and LPS present in both the immunization and challenge mixtures.\nDust was collected from the homes of asthmatic children as described previously[11]. This dust was shown to contain significantly higher concentrations of cockroach allergens Bla g1 and Bla g2 in comparison to other indoor and outdoor allergens[12]. This extract, when used in previous studies at a 1:10 dilution (containing a total of 61.9 ng CRA) successfully induced the hallmark features of asthma-like inflammation, including airways hyperresponsiveness, eosinophilia and plasma IgE[18-20]. In the present study, we specifically sought to determine whether removal of the potent innate immune activating agent LPS from the HDE at the time of allergen challenge would attenuate the severity of pulmonary inflammation in sensitized mice. To address this question, LPS in the HDE was reduced as described in materials and methods. Unfortunately, our previous attempts at LPS removal using polymyxin B columns (Pierce, Rockford, IL) also removed 99.9% of Bla g1 from the HDE, and these preparations were not appropriate for the present studies (Table 1). Although use of the EndoTrap column in this study slightly reduced CRA concentrations, through appropriate dilution, we were able to maintain the same total amount of 61.9 ng CRA for all treatments (Table 2).\nTotal amounts of CRA and LPS in the intact HDE and in the HDE treated with Polymyxin B or EndoTrap to deplete LPS.\nAmounts of CRA and LPS present in both the immunization and challenge mixtures.\n[SUBTITLE] TNFα production in BAL fluid post final challenge [SUBSECTION] Sensitization to cockroach allergens plays a key role in the development of childhood asthma[21], and the level of LPS present in the home has also been shown to have a major effect on the severity of respiratory diseases[5,10,22]. Asthma-like inflammation was induced by i.p. immunization with 61.9 ng CRA, containing 11.6 ng of LPS. Mice were challenged on days 14 and 21 with either the intact HDE (182.1 ng of LPS) or LPS-reduced HDE (3.0 ng of LPS). Although different allergen mixtures are used for sensitization and challenge, we have previously shown that HDE immunization specifically induces an allergic response to CRA[12]. Consistent with the low levels of LPS present in the LPS-reduced HDE, TNFα production was significantly decreased in this group at 2 hours post final challenge (Figure 1). We have previously shown that this early timepoint reflects the point at which TNFα production peaks in the BAL fluid[19].\nTNFα production in bronchoalveolar lavage fluid. TNFα was measured by ELISA at the timepoints indicated. The 0 h sample was collected prior to the second pulmonary challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ***p ≤ 0.0001 comparing the LPS-reduced HDE group to the HDE group by Student's t-test.\nSensitization to cockroach allergens plays a key role in the development of childhood asthma[21], and the level of LPS present in the home has also been shown to have a major effect on the severity of respiratory diseases[5,10,22]. Asthma-like inflammation was induced by i.p. immunization with 61.9 ng CRA, containing 11.6 ng of LPS. Mice were challenged on days 14 and 21 with either the intact HDE (182.1 ng of LPS) or LPS-reduced HDE (3.0 ng of LPS). Although different allergen mixtures are used for sensitization and challenge, we have previously shown that HDE immunization specifically induces an allergic response to CRA[12]. Consistent with the low levels of LPS present in the LPS-reduced HDE, TNFα production was significantly decreased in this group at 2 hours post final challenge (Figure 1). We have previously shown that this early timepoint reflects the point at which TNFα production peaks in the BAL fluid[19].\nTNFα production in bronchoalveolar lavage fluid. TNFα was measured by ELISA at the timepoints indicated. The 0 h sample was collected prior to the second pulmonary challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ***p ≤ 0.0001 comparing the LPS-reduced HDE group to the HDE group by Student's t-test.\n[SUBTITLE] Role of LPS in antibody production [SUBSECTION] The presence or absence of TNFα has been shown to affect the antibody response to allergen[8]. We sought to determine whether the reduced levels of TNFα produced in response to challenge with the LPS-reduced HDE had an effect on antibody production in this model. Intact HDE challenge resulted in robust IgE production while challenge with LPS-reduced HDE did not induce IgE production beyond that seen in mice receiving only the CRA immunization (Figure 2A). Despite the global increase in IgE, measurement of CRA-specific IgE showed no difference between LPS-reduced HDE challenge and HDE challenge (Figure 2B). IgG1 production was similar to total IgE production with reduced levels in the mice challenged with the LPS-reduced HDE compared to immunized-only mice (Figure 2C). These changes in immunoglobulins were not found with all classes, since LPS-reduced HDE challenge induced significantly higher levels of plasma IgG2a compared to the immunized-only group (Figure 2D).\nPlasma antibody levels in immunized and challenged mice. Plasma levels of A) IgE, B) CRA-specific IgE, C) IgG1, and D) IgG2a were measured by standard ELISA from EDTA-plasma collected 24 hours post final challenge (day 22). The immunized-only group was sensitized on Day 0 and plasma was collected on day 22. Data are represented as the mean ± SEM. Statistical significance was assessed by one-way ANOVA followed by Turkey's post test as indicated in each panel. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ** p < 0.001.\nThe presence or absence of TNFα has been shown to affect the antibody response to allergen[8]. We sought to determine whether the reduced levels of TNFα produced in response to challenge with the LPS-reduced HDE had an effect on antibody production in this model. Intact HDE challenge resulted in robust IgE production while challenge with LPS-reduced HDE did not induce IgE production beyond that seen in mice receiving only the CRA immunization (Figure 2A). Despite the global increase in IgE, measurement of CRA-specific IgE showed no difference between LPS-reduced HDE challenge and HDE challenge (Figure 2B). IgG1 production was similar to total IgE production with reduced levels in the mice challenged with the LPS-reduced HDE compared to immunized-only mice (Figure 2C). These changes in immunoglobulins were not found with all classes, since LPS-reduced HDE challenge induced significantly higher levels of plasma IgG2a compared to the immunized-only group (Figure 2D).\nPlasma antibody levels in immunized and challenged mice. Plasma levels of A) IgE, B) CRA-specific IgE, C) IgG1, and D) IgG2a were measured by standard ELISA from EDTA-plasma collected 24 hours post final challenge (day 22). The immunized-only group was sensitized on Day 0 and plasma was collected on day 22. Data are represented as the mean ± SEM. Statistical significance was assessed by one-way ANOVA followed by Turkey's post test as indicated in each panel. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ** p < 0.001.\n[SUBTITLE] Inflammatory cell recruitment in the BAL fluid post allergen challenge [SUBSECTION] The composition of inflammatory cells in the BAL fluid was assessed at various timepoints after the final intratracheal challenge. Cytospin preparations demonstrate distinct inflammatory cell morphology. Eosinophils were characterized by a ring-shaped nucleus with pink cytoplasmic amyloid staining, while neutrophils show a lobular nucleus with no cytoplasmic staining. Similar numbers of eosinophils and neutrophils in the mice challenged with the intact HDE or the LPS-reduced HDE, in contrast to that of naïve mice containing predominantly macrophages (Figure 3). We observed a slight reduction in eosinophil numbers in the BAL fluid of the LPS-reduced HDE group, but this change did not reach statistical significance (Figure 4A). In the BAL fluid, the eosinophil-specific chemoattractant eotaxin-1 (CCL11) was significantly increased at 24 hours post final challenge in the LPS-reduced HDE mice, however no difference in eotaxin-2 (CCL24) was observed (Figure 4). We also assayed eosinophil-specific peroxidase activity (EPO) in the lungs of mice (after BAL) as a measure of eosinophils within the lung interstitium. No significant difference in EPO was measured between the two groups (Figure 4D). A statistically significant decrease in eotaxin-1 (CCL11) was measured in the lung homogenate of the LPS-reduced HDE group (Figure 4E), but no difference was measured in eotaxin-2 (CCL24) between groups (Figure 4F). While statistically significant differences in eotaxin-1 are observed in both the BAL and lung homogenate, it should be noted that eotaxin-2 is produced at much higher concentrations in both the alveolar compartment and the lung tissue, rendering an equivalent biological outcome (eosinophil recruitment) in both groups.\nCytospin preparations of cells recovered from BAL fluid. A) Naïve mice, B) HDE challenged mice and C) LPS-reduced HDE challenged mice at 24 hours post final challenge. Each is represented at 1000×. Representative eosinophils (Eo) and neutrophils (Ne) are indicated in the figure.\nEosinophil recruitment and production of eosinophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at the timepoints indicated were made and 300 cell differential counts performed to determine the absolute numbers of eosinophils (A). Concentrations of the chemokines B) CCL11 and C) CCL24 in the BAL fluid were measured by ELISA. D) Eosinophil peroxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CCL11 and F) CCL24 in the lung homogenate were assayed by ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment, in some data points the symbol is larger than the SEM. * p < 0.05 and *** p ≤ 0.0001 HDE vs. LPS-reduced HDE by Student's t-test.\nWe also assayed the number of neutrophils recruited in response to allergen challenge. Interestingly, comparable numbers of neutrophils were present in both groups despite reduced levels of LPS in the LPS-reduced HDE (Figure 5A). This suggests that other components of the HDE are able to activate the innate immune response in order to recruit neutrophils. The concentrations of the neutrophil chemoattractants CXCL1 and CXCL2 were assayed as the mechanism of neutrophil recruitment into the alveolar space and lung tissue. BAL levels of CXCL1 are significantly decreased at 2 hours in mice challenged with the LPS-reduced HDE, however no difference was seen in CXCL2 (Figure 5). Despite this slight decrease in chemokine production, we postulate that the concentrations at which they are being produced is adequate to recruit equivalent numbers of neutrophils shown in Figure 5A.\nNeutrophil recruitment and production of neutrophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at 24 hours post final challenge were made and 300 cell differential counts performed to determine the absolute numbers of A) neutrophils. Concentrations of chemokines B) CXCL1 and C) CXCL2 in the BAL fluid were measured by ELISA. D) Myeloperoxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CXCL1 and F) CXCL2 in the lung homogenate were assayed by standard ELISA. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05, ** p ≤ 0.001 and *** p ≤ 0.0001 comparing the LPS-reduced HDE group to HDE the group by Student's t-test.\nMyeloperoxidase activity (MPO) in the lung tissue was not changed between groups. (Figure 5D). Significant increases were measured in both CXCL1 and CXCL2 in the lung homogenate in the LPS-depleted HDE at all timepoints assayed (Figure 5E). This again suggests that although the chemokines were lower in the intact HDE, the concentrations were sufficient to sequester neutrophils within the lung interstitium. No differences in macrophage or lymphocyte numbers were seen in the BAL and no difference was seen in neutrophil numbers at 2 h post final challenge (data not shown).\nThe composition of inflammatory cells in the BAL fluid was assessed at various timepoints after the final intratracheal challenge. Cytospin preparations demonstrate distinct inflammatory cell morphology. Eosinophils were characterized by a ring-shaped nucleus with pink cytoplasmic amyloid staining, while neutrophils show a lobular nucleus with no cytoplasmic staining. Similar numbers of eosinophils and neutrophils in the mice challenged with the intact HDE or the LPS-reduced HDE, in contrast to that of naïve mice containing predominantly macrophages (Figure 3). We observed a slight reduction in eosinophil numbers in the BAL fluid of the LPS-reduced HDE group, but this change did not reach statistical significance (Figure 4A). In the BAL fluid, the eosinophil-specific chemoattractant eotaxin-1 (CCL11) was significantly increased at 24 hours post final challenge in the LPS-reduced HDE mice, however no difference in eotaxin-2 (CCL24) was observed (Figure 4). We also assayed eosinophil-specific peroxidase activity (EPO) in the lungs of mice (after BAL) as a measure of eosinophils within the lung interstitium. No significant difference in EPO was measured between the two groups (Figure 4D). A statistically significant decrease in eotaxin-1 (CCL11) was measured in the lung homogenate of the LPS-reduced HDE group (Figure 4E), but no difference was measured in eotaxin-2 (CCL24) between groups (Figure 4F). While statistically significant differences in eotaxin-1 are observed in both the BAL and lung homogenate, it should be noted that eotaxin-2 is produced at much higher concentrations in both the alveolar compartment and the lung tissue, rendering an equivalent biological outcome (eosinophil recruitment) in both groups.\nCytospin preparations of cells recovered from BAL fluid. A) Naïve mice, B) HDE challenged mice and C) LPS-reduced HDE challenged mice at 24 hours post final challenge. Each is represented at 1000×. Representative eosinophils (Eo) and neutrophils (Ne) are indicated in the figure.\nEosinophil recruitment and production of eosinophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at the timepoints indicated were made and 300 cell differential counts performed to determine the absolute numbers of eosinophils (A). Concentrations of the chemokines B) CCL11 and C) CCL24 in the BAL fluid were measured by ELISA. D) Eosinophil peroxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CCL11 and F) CCL24 in the lung homogenate were assayed by ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment, in some data points the symbol is larger than the SEM. * p < 0.05 and *** p ≤ 0.0001 HDE vs. LPS-reduced HDE by Student's t-test.\nWe also assayed the number of neutrophils recruited in response to allergen challenge. Interestingly, comparable numbers of neutrophils were present in both groups despite reduced levels of LPS in the LPS-reduced HDE (Figure 5A). This suggests that other components of the HDE are able to activate the innate immune response in order to recruit neutrophils. The concentrations of the neutrophil chemoattractants CXCL1 and CXCL2 were assayed as the mechanism of neutrophil recruitment into the alveolar space and lung tissue. BAL levels of CXCL1 are significantly decreased at 2 hours in mice challenged with the LPS-reduced HDE, however no difference was seen in CXCL2 (Figure 5). Despite this slight decrease in chemokine production, we postulate that the concentrations at which they are being produced is adequate to recruit equivalent numbers of neutrophils shown in Figure 5A.\nNeutrophil recruitment and production of neutrophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at 24 hours post final challenge were made and 300 cell differential counts performed to determine the absolute numbers of A) neutrophils. Concentrations of chemokines B) CXCL1 and C) CXCL2 in the BAL fluid were measured by ELISA. D) Myeloperoxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CXCL1 and F) CXCL2 in the lung homogenate were assayed by standard ELISA. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05, ** p ≤ 0.001 and *** p ≤ 0.0001 comparing the LPS-reduced HDE group to HDE the group by Student's t-test.\nMyeloperoxidase activity (MPO) in the lung tissue was not changed between groups. (Figure 5D). Significant increases were measured in both CXCL1 and CXCL2 in the lung homogenate in the LPS-depleted HDE at all timepoints assayed (Figure 5E). This again suggests that although the chemokines were lower in the intact HDE, the concentrations were sufficient to sequester neutrophils within the lung interstitium. No differences in macrophage or lymphocyte numbers were seen in the BAL and no difference was seen in neutrophil numbers at 2 h post final challenge (data not shown).\n[SUBTITLE] Induction of airways hyperresponsiveness after allergen challenge [SUBSECTION] Several studies have shown that both LPS inhalation and pulmonary TNFα administration induce AHR, however little data exist regarding the synergistic role of allergen and endotoxin on AHR[23-25]. Therefore, we sought to determine whether the levels of LPS in the HDE would affect AHR severity. Challenge with increasing concentrations of aerosolized methacholine 4 hours post final challenge did not result in any difference in AHR between groups (Figure 6A). The 4 hour timepoint was assessed based on observations in our laboratory showing robust AHR induction at this timepoint. No increase in AHR was measured comparing normal mice and sensitized-only mice (data not shown). Given the controversy associated with non-invasive whole body plethysmography for assessing AHR, we confirmed our findings using invasive lung function tests. As shown in Figure 6B, invasive tests demonstrated no significant difference in airway resistance between groups at all methacholine concentrations tested. We also assessed BAL cysteinyl leukotriene (Cys-LT) concentrations at 2 hours as a mechanism for AHR, and found no significant change in BAL Cys-LT in mice challenged with LPS-reduced HDE compared to HDE challenge (Figure 6C).\nAirways hyperresponsiveness, airway resistance and cysteinyl leukotrienes post allergen challenge. A) For non-invasive measurement of AHR, mice were challenged with increasing concentrations of aerosolized methacholine at 4 hours after the final challenge and PenH values are represented as percent above baseline. B) For invasive measurements, mice were challenged with methacholine and pulmonary resistance is expressed as R cmH2O.s/mL. C) Cysteinyl leukotriene concentrations were measured in the BAL fluid 2 hours post final challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment.\nSeveral studies have shown that both LPS inhalation and pulmonary TNFα administration induce AHR, however little data exist regarding the synergistic role of allergen and endotoxin on AHR[23-25]. Therefore, we sought to determine whether the levels of LPS in the HDE would affect AHR severity. Challenge with increasing concentrations of aerosolized methacholine 4 hours post final challenge did not result in any difference in AHR between groups (Figure 6A). The 4 hour timepoint was assessed based on observations in our laboratory showing robust AHR induction at this timepoint. No increase in AHR was measured comparing normal mice and sensitized-only mice (data not shown). Given the controversy associated with non-invasive whole body plethysmography for assessing AHR, we confirmed our findings using invasive lung function tests. As shown in Figure 6B, invasive tests demonstrated no significant difference in airway resistance between groups at all methacholine concentrations tested. We also assessed BAL cysteinyl leukotriene (Cys-LT) concentrations at 2 hours as a mechanism for AHR, and found no significant change in BAL Cys-LT in mice challenged with LPS-reduced HDE compared to HDE challenge (Figure 6C).\nAirways hyperresponsiveness, airway resistance and cysteinyl leukotrienes post allergen challenge. A) For non-invasive measurement of AHR, mice were challenged with increasing concentrations of aerosolized methacholine at 4 hours after the final challenge and PenH values are represented as percent above baseline. B) For invasive measurements, mice were challenged with methacholine and pulmonary resistance is expressed as R cmH2O.s/mL. C) Cysteinyl leukotriene concentrations were measured in the BAL fluid 2 hours post final challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment.\n[SUBTITLE] Th1 and Th2 cytokine production in the BAL post allergen challenge [SUBSECTION] It has been suggested previously that the presence of TNFα can induce increased levels of Th2 cytokines in the BAL fluid of allergen challenged mice[19]. Further, asthma is largely characterized as a Th2 mediated disease, accompanied by high IL-4 and IL-5 production and low IFNγ[26]. To determine whether the reduction in TNFα seen in LPS-reduced HDE challenged mice affected cytokine production, we assayed both Th1 and Th2 cytokines 24 hours post final challenge. In conflict with the Th2 paradigm, IFNγ gamma was significantly increased in the LPS-reduced HDE mice (Figure 7A). Surprisingly, IL-5 and IL-13 were significantly increased in the BAL from LPS-reduced HDE challenged mice although there was no difference in IL-4 levels (Figure 7B, C). IL-17 has also been implicated as a key cytokine in the asthmatic response, and is also modulated by LPS levels [27,28]. However, we measured no significant differences in IL-17 levels between groups in the BAL fluid (data not shown).\nTh1 and Th2 cytokine production in the BAL fluid at 24 hours post final challenge. Cytokines were assayed by standard ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05 and *** p ≤ 0.0001 comparing the LPS-reduced HDE to HDE the group by Student's t-test.\nIt has been suggested previously that the presence of TNFα can induce increased levels of Th2 cytokines in the BAL fluid of allergen challenged mice[19]. Further, asthma is largely characterized as a Th2 mediated disease, accompanied by high IL-4 and IL-5 production and low IFNγ[26]. To determine whether the reduction in TNFα seen in LPS-reduced HDE challenged mice affected cytokine production, we assayed both Th1 and Th2 cytokines 24 hours post final challenge. In conflict with the Th2 paradigm, IFNγ gamma was significantly increased in the LPS-reduced HDE mice (Figure 7A). Surprisingly, IL-5 and IL-13 were significantly increased in the BAL from LPS-reduced HDE challenged mice although there was no difference in IL-4 levels (Figure 7B, C). IL-17 has also been implicated as a key cytokine in the asthmatic response, and is also modulated by LPS levels [27,28]. However, we measured no significant differences in IL-17 levels between groups in the BAL fluid (data not shown).\nTh1 and Th2 cytokine production in the BAL fluid at 24 hours post final challenge. Cytokines were assayed by standard ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05 and *** p ≤ 0.0001 comparing the LPS-reduced HDE to HDE the group by Student's t-test.", "Dust was collected from the homes of asthmatic children as described previously[11]. This dust was shown to contain significantly higher concentrations of cockroach allergens Bla g1 and Bla g2 in comparison to other indoor and outdoor allergens[12]. This extract, when used in previous studies at a 1:10 dilution (containing a total of 61.9 ng CRA) successfully induced the hallmark features of asthma-like inflammation, including airways hyperresponsiveness, eosinophilia and plasma IgE[18-20]. In the present study, we specifically sought to determine whether removal of the potent innate immune activating agent LPS from the HDE at the time of allergen challenge would attenuate the severity of pulmonary inflammation in sensitized mice. To address this question, LPS in the HDE was reduced as described in materials and methods. Unfortunately, our previous attempts at LPS removal using polymyxin B columns (Pierce, Rockford, IL) also removed 99.9% of Bla g1 from the HDE, and these preparations were not appropriate for the present studies (Table 1). Although use of the EndoTrap column in this study slightly reduced CRA concentrations, through appropriate dilution, we were able to maintain the same total amount of 61.9 ng CRA for all treatments (Table 2).\nTotal amounts of CRA and LPS in the intact HDE and in the HDE treated with Polymyxin B or EndoTrap to deplete LPS.\nAmounts of CRA and LPS present in both the immunization and challenge mixtures.", "Sensitization to cockroach allergens plays a key role in the development of childhood asthma[21], and the level of LPS present in the home has also been shown to have a major effect on the severity of respiratory diseases[5,10,22]. Asthma-like inflammation was induced by i.p. immunization with 61.9 ng CRA, containing 11.6 ng of LPS. Mice were challenged on days 14 and 21 with either the intact HDE (182.1 ng of LPS) or LPS-reduced HDE (3.0 ng of LPS). Although different allergen mixtures are used for sensitization and challenge, we have previously shown that HDE immunization specifically induces an allergic response to CRA[12]. Consistent with the low levels of LPS present in the LPS-reduced HDE, TNFα production was significantly decreased in this group at 2 hours post final challenge (Figure 1). We have previously shown that this early timepoint reflects the point at which TNFα production peaks in the BAL fluid[19].\nTNFα production in bronchoalveolar lavage fluid. TNFα was measured by ELISA at the timepoints indicated. The 0 h sample was collected prior to the second pulmonary challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ***p ≤ 0.0001 comparing the LPS-reduced HDE group to the HDE group by Student's t-test.", "The presence or absence of TNFα has been shown to affect the antibody response to allergen[8]. We sought to determine whether the reduced levels of TNFα produced in response to challenge with the LPS-reduced HDE had an effect on antibody production in this model. Intact HDE challenge resulted in robust IgE production while challenge with LPS-reduced HDE did not induce IgE production beyond that seen in mice receiving only the CRA immunization (Figure 2A). Despite the global increase in IgE, measurement of CRA-specific IgE showed no difference between LPS-reduced HDE challenge and HDE challenge (Figure 2B). IgG1 production was similar to total IgE production with reduced levels in the mice challenged with the LPS-reduced HDE compared to immunized-only mice (Figure 2C). These changes in immunoglobulins were not found with all classes, since LPS-reduced HDE challenge induced significantly higher levels of plasma IgG2a compared to the immunized-only group (Figure 2D).\nPlasma antibody levels in immunized and challenged mice. Plasma levels of A) IgE, B) CRA-specific IgE, C) IgG1, and D) IgG2a were measured by standard ELISA from EDTA-plasma collected 24 hours post final challenge (day 22). The immunized-only group was sensitized on Day 0 and plasma was collected on day 22. Data are represented as the mean ± SEM. Statistical significance was assessed by one-way ANOVA followed by Turkey's post test as indicated in each panel. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ** p < 0.001.", "The composition of inflammatory cells in the BAL fluid was assessed at various timepoints after the final intratracheal challenge. Cytospin preparations demonstrate distinct inflammatory cell morphology. Eosinophils were characterized by a ring-shaped nucleus with pink cytoplasmic amyloid staining, while neutrophils show a lobular nucleus with no cytoplasmic staining. Similar numbers of eosinophils and neutrophils in the mice challenged with the intact HDE or the LPS-reduced HDE, in contrast to that of naïve mice containing predominantly macrophages (Figure 3). We observed a slight reduction in eosinophil numbers in the BAL fluid of the LPS-reduced HDE group, but this change did not reach statistical significance (Figure 4A). In the BAL fluid, the eosinophil-specific chemoattractant eotaxin-1 (CCL11) was significantly increased at 24 hours post final challenge in the LPS-reduced HDE mice, however no difference in eotaxin-2 (CCL24) was observed (Figure 4). We also assayed eosinophil-specific peroxidase activity (EPO) in the lungs of mice (after BAL) as a measure of eosinophils within the lung interstitium. No significant difference in EPO was measured between the two groups (Figure 4D). A statistically significant decrease in eotaxin-1 (CCL11) was measured in the lung homogenate of the LPS-reduced HDE group (Figure 4E), but no difference was measured in eotaxin-2 (CCL24) between groups (Figure 4F). While statistically significant differences in eotaxin-1 are observed in both the BAL and lung homogenate, it should be noted that eotaxin-2 is produced at much higher concentrations in both the alveolar compartment and the lung tissue, rendering an equivalent biological outcome (eosinophil recruitment) in both groups.\nCytospin preparations of cells recovered from BAL fluid. A) Naïve mice, B) HDE challenged mice and C) LPS-reduced HDE challenged mice at 24 hours post final challenge. Each is represented at 1000×. Representative eosinophils (Eo) and neutrophils (Ne) are indicated in the figure.\nEosinophil recruitment and production of eosinophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at the timepoints indicated were made and 300 cell differential counts performed to determine the absolute numbers of eosinophils (A). Concentrations of the chemokines B) CCL11 and C) CCL24 in the BAL fluid were measured by ELISA. D) Eosinophil peroxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CCL11 and F) CCL24 in the lung homogenate were assayed by ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment, in some data points the symbol is larger than the SEM. * p < 0.05 and *** p ≤ 0.0001 HDE vs. LPS-reduced HDE by Student's t-test.\nWe also assayed the number of neutrophils recruited in response to allergen challenge. Interestingly, comparable numbers of neutrophils were present in both groups despite reduced levels of LPS in the LPS-reduced HDE (Figure 5A). This suggests that other components of the HDE are able to activate the innate immune response in order to recruit neutrophils. The concentrations of the neutrophil chemoattractants CXCL1 and CXCL2 were assayed as the mechanism of neutrophil recruitment into the alveolar space and lung tissue. BAL levels of CXCL1 are significantly decreased at 2 hours in mice challenged with the LPS-reduced HDE, however no difference was seen in CXCL2 (Figure 5). Despite this slight decrease in chemokine production, we postulate that the concentrations at which they are being produced is adequate to recruit equivalent numbers of neutrophils shown in Figure 5A.\nNeutrophil recruitment and production of neutrophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at 24 hours post final challenge were made and 300 cell differential counts performed to determine the absolute numbers of A) neutrophils. Concentrations of chemokines B) CXCL1 and C) CXCL2 in the BAL fluid were measured by ELISA. D) Myeloperoxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CXCL1 and F) CXCL2 in the lung homogenate were assayed by standard ELISA. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05, ** p ≤ 0.001 and *** p ≤ 0.0001 comparing the LPS-reduced HDE group to HDE the group by Student's t-test.\nMyeloperoxidase activity (MPO) in the lung tissue was not changed between groups. (Figure 5D). Significant increases were measured in both CXCL1 and CXCL2 in the lung homogenate in the LPS-depleted HDE at all timepoints assayed (Figure 5E). This again suggests that although the chemokines were lower in the intact HDE, the concentrations were sufficient to sequester neutrophils within the lung interstitium. No differences in macrophage or lymphocyte numbers were seen in the BAL and no difference was seen in neutrophil numbers at 2 h post final challenge (data not shown).", "Several studies have shown that both LPS inhalation and pulmonary TNFα administration induce AHR, however little data exist regarding the synergistic role of allergen and endotoxin on AHR[23-25]. Therefore, we sought to determine whether the levels of LPS in the HDE would affect AHR severity. Challenge with increasing concentrations of aerosolized methacholine 4 hours post final challenge did not result in any difference in AHR between groups (Figure 6A). The 4 hour timepoint was assessed based on observations in our laboratory showing robust AHR induction at this timepoint. No increase in AHR was measured comparing normal mice and sensitized-only mice (data not shown). Given the controversy associated with non-invasive whole body plethysmography for assessing AHR, we confirmed our findings using invasive lung function tests. As shown in Figure 6B, invasive tests demonstrated no significant difference in airway resistance between groups at all methacholine concentrations tested. We also assessed BAL cysteinyl leukotriene (Cys-LT) concentrations at 2 hours as a mechanism for AHR, and found no significant change in BAL Cys-LT in mice challenged with LPS-reduced HDE compared to HDE challenge (Figure 6C).\nAirways hyperresponsiveness, airway resistance and cysteinyl leukotrienes post allergen challenge. A) For non-invasive measurement of AHR, mice were challenged with increasing concentrations of aerosolized methacholine at 4 hours after the final challenge and PenH values are represented as percent above baseline. B) For invasive measurements, mice were challenged with methacholine and pulmonary resistance is expressed as R cmH2O.s/mL. C) Cysteinyl leukotriene concentrations were measured in the BAL fluid 2 hours post final challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment.", "It has been suggested previously that the presence of TNFα can induce increased levels of Th2 cytokines in the BAL fluid of allergen challenged mice[19]. Further, asthma is largely characterized as a Th2 mediated disease, accompanied by high IL-4 and IL-5 production and low IFNγ[26]. To determine whether the reduction in TNFα seen in LPS-reduced HDE challenged mice affected cytokine production, we assayed both Th1 and Th2 cytokines 24 hours post final challenge. In conflict with the Th2 paradigm, IFNγ gamma was significantly increased in the LPS-reduced HDE mice (Figure 7A). Surprisingly, IL-5 and IL-13 were significantly increased in the BAL from LPS-reduced HDE challenged mice although there was no difference in IL-4 levels (Figure 7B, C). IL-17 has also been implicated as a key cytokine in the asthmatic response, and is also modulated by LPS levels [27,28]. However, we measured no significant differences in IL-17 levels between groups in the BAL fluid (data not shown).\nTh1 and Th2 cytokine production in the BAL fluid at 24 hours post final challenge. Cytokines were assayed by standard ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05 and *** p ≤ 0.0001 comparing the LPS-reduced HDE to HDE the group by Student's t-test.", "The effect of exposure to household dust on the development of atopy and asthma has recently become a focus of investigation; however this avenue of study has resulted in more controversy than consensus. Epidemiological evidence shows that the presence of cockroach allergens in inner city homes is a strong risk factor for the development of atopic disease [29,30]. Cockroach allergen sensitization was found to be responsible for exacerbations in a large proportion of asthmatic children in inner city Washington, D.C [21]. Further, asthma exacerbations were not the result of sensitization to other allergens such as cat dander and dust mite allergens, which were also present in these homes [1,21]. Exposure to innate immune activators such as TLR ligands, as well as environmental pollutants such as diesel particulates can also drive the adaptive immune response toward tolerance or sensitization [10,31]. This area of study is gaining significant interest in light of the increasing incidence of atopy and allergy in the developed world.\nThis study sought to examine whether the presence of LPS in the household environment modifies the allergic response after sensitization to cockroach allergens has been established. It is possible that our method of LPS reduction also removed other allergens from the HDE mixture which may contribute to the allergic response. While this is a real concern, it is a shortcoming of all column purification methods. We are confident that our methods resulted in physiologically relevant data given that the LPS-reduced HDE was still able to induce a robust adaptive immune response.\nRecent studies have employed house dust extracts to determine their roles in asthma severity, and have suggested that these extracts can induce allergic phenotypes or can skew towards tolerance, depending upon the dose and timing of the exposure [31,32]. However, these studies focused on the immunomodulatory effects of the house dust extracts on mice presensitized to ovalbumin, and thus study house dust as an exacerbating factor, rather than a causative agent. In contrast, the current study directly examines the synergistic effects of CRA and LPS contained in the HDE in mice presensitized to CRA, which is the major allergen in the HDE.\nLPS contamination in allergen preparations has been raised as an issue in many models of allergy, however we believe that approaching these studies using household components with naturally occurring LPS more closely and accurately reflects the environment in which one develops asthma. It is not appropriate to refer to the HDE as \"contaminated\" with LPS, similar to how one would not say it is contaminated with cockroach allergens. In an OVA study, Watanabe et.al. demonstrated that LPS contamination of commercially available OVA preparations significantly attenuated cellular influx, AHR and IgE production [33]. The differences observed between these models might be explained by the intrinsic protease activity of the cockroach allergen Bla g2, which may enhance allergen sensitization by increasing epithelial permeability, and allergen exposure to dendritic cells [34,35]. This highlights the need to study the effect of LPS in the development of asthma-like inflammation using a relevant allergen mixture.\nIn the lung, TNFα is primarily produced by alveolar macrophages and secreted within 1 hour of LPS challenge [36,37]. Our results show that mice challenged with the LPS-reduced HDE produce significantly less TNFα compared to those challenged with the HDE, verifying that removal of LPS from the HDE has the expected immunological effect. Several studies have implicated TNFα and TNFα signaling in the development of adaptive immunity, particularly in the context of class switch to IgE production [8,9]. Therefore, we would expect that we did not see differences in CRA-specific IgE, given that our study focused on the effect of LPS at challenge only. In contrast, the studies by Eisenbarth et. al. and Watanabe et.al. focused on LPS at the time of sensitization, which likely has more effect on the production of antigen-specific antibodies [8,33]. This was also demonstrated in a rat model where aerosol exposure to LPS before or shortly after sensitization reduced the development of OVA-specific IgE [22]. Levels of CRA-specific IgG1 and IgG2a are of significant interest, however reliable assays for these antibodies are not currently available.\nThe increase in total IgE and IgG1 in the HDE-challenged mice represents an increase in polyclonal, antigen non-specific IgE. LPS alone has been shown to induce expression of activation-induced cytidine deaminase, which is required for activation of downstream signals responsible for class switch recombination to both IgG1 and IgE in B cells [38]. Other models have shown polyclonal IgE to have a role in mast cell survival and cytokine release [39].\nOur data demonstrate that airways hyperresponsiveness was not diminished in the LPS-reduced HDE challenged groups despite significantly attenuated TNFα. We postulate that AHR is governed by other mediators and is not directly induced by TNFα. The role of IgE was explored in other studies that demonstrated that AHR may be modified independent of IgE alterations. Hamelmann et. al. showed robust AHR induction in both mast cell and B cell deficient mice and it has been shown that mast cell activation is not required for the induction of AHR[40,41]. Further, methacholine acts directly upon muscarinic receptors in order to induce smooth muscle contraction. We have recently shown that pulmonary exposure to high concentrations of LPS can induce modest differences in muscarinic receptor expression[42]. Therefore, it is possible that the concentrations of LPS used in this study were not sufficient to alter muscarinic receptor expression, and therefore did not alter AHR.\nThe role of LPS in driving allergic responses toward a Th2 phenotype has been demonstrated in several models and our data show this to be true in the context of total antibody production. However, our data show increases in both Th1 and Th2 cytokines in the LPS-reduced HDE group. One possible explanation for this is that the source of these cytokines is not only the T cells present in the lung. Other cell types such as eosinophils, epithelial cells and mast cells are contributing to the pool of cytokines present in the BAL fluid [43]. Therefore, without assessing the direct cellular source of these cytokines, it is not possible to conclude whether LPS-induced Th2 cell differentiation is impaired in this model.\nInterestingly, challenge with intact HDE results in decreased BAL cytokine production. We postulate that this is modulation of the innate immune response through induction of microbial cross-tolerance. We and others have shown that cross-tolerance to TLR ligands can be induced in the lung and in pulmonary macrophages [44,45]. It is likely that simultaneous exposure to various TLR ligands present in the HDE can induce cross-tolerance and therefore attenuate the cytokine response. Removal of the potent TLR4 agonist LPS, may reduce the tolerogenic response, driving increased cytokine production.\nLimited epidemiological data exist investigating whether removal of house dust can limit future asthma exacerbations. One such study carried out in the UK found no correlation between removal of house dust and amelioration of asthma symptoms in presensitized individuals. However, interventions to reduce dust in homes of children who were highly susceptible to developing atopy was effective in reducing asthma diagnosis [46,47]. Taken together our data show that in a clinically relevant model of house dust induced asthma-like inflammation, the removal of only LPS at challenge does not protect against the development of the inflammatory response. Further studies will be needed to clarify whether LPS in the home environment exacerbates asthma in presensitized individuals.", "Decreasing the endotoxin content endogenously present in cockroach allergens results in a complicated pulmonary inflammatory response. While there was differential regulation of cytokine, chemokines and plasma levels of immunoglobulins, there was no real change in AHR or pulmonary inflammatory cell recruitment.", "The authors declare that they have no competing interests.", "SN designed and executed the experiments and wrote the manuscript. JB and JK assisted with conducting the experiments, interpreting the data and reviewing the manuscript. WC assisted with the Flexivent measurements and reviewed the manuscript. DR assumed overall responsibility for the experimental design and manuscript preparation. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2466/11/12/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Animals", "Reduction of LPS content in House Dust Extract", "LPS Assay", "Asthma induction protocol", "Timepoints for data collection", "Airways Hyperresponsiveness and Airway Resistance", "Bronchoalveolar lavage and lung homogenate preparation", "Myeloperoxidase and Eosinophil Peroxidase Assays", "ELISA", "Cysteinyl-Leukotriene Immunoassay", "Statistical Analysis", "Results", "Depletion of LPS from the house dust extract", "TNFα production in BAL fluid post final challenge", "Role of LPS in antibody production", "Inflammatory cell recruitment in the BAL fluid post allergen challenge", "Induction of airways hyperresponsiveness after allergen challenge", "Th1 and Th2 cytokine production in the BAL post allergen challenge", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Asthma is a chronic allergic disorder characterized by IgE production, airway eosinophilia and bronchial hyperresponsiveness[2]. Numerous studies have shown that allergens are not solely responsible for the severity of the asthmatic response (reviewed in[3]). Microbial components, such as lipopolysaccharide (LPS) from Gram negative bacteria, are ubiquitously present in the environment, including the ambient air. In addition to being potent activators of innate immunity, these compounds have been shown to modulate asthma severity[4]. The presence of microbial components often is associated with environmental and household cleanliness, however the hygiene hypothesis postulates that the lack of exposure to these pathogens at a young age increases susceptibility to allergen sensitization, accounting for the dramatic increases in allergic diseases in the developed world [5,6].\nSeveral studies have been carried out investigating the role LPS plays in allergen sensitization. These studies suggest that in ovalbumin (OVA) models, increasing doses of LPS protect against eosinophilia and AHR[7,8]. Recently, more evidence has emerged implicating reduced tumor necrosis factor-α (TNFα) production as the mechanism responsible for this protection. Eisenbarth. et. al. showed that OVA sensitization in toll-like receptor-4 (TLR4) deficient mice did not result in IgE production or eosinophil recruitment, however these responses could be restored by exogenous administration of TNFα shortly after sensitization[8]. Additionally, mice deficient in TNF-receptor associated factor-1 do not develop eosinophilia or AHR in response to OVA sensitization and challenge[9].\nWhile LPS plays a key role in allergen sensitization, the synergistic effect of LPS and allergen during the effector phase of asthma remains unclear. We sought to determine whether the removal of LPS from house dust obtained from the home environment can protect against the severity of murine airways-inflammation. We employed a clinically relevant model of asthma-like pulmonary inflammation based on immunization with a CRA[1] extract containing LPS, followed by challenge with a house dust extract (HDE) collected from the home of an asthmatic child. This HDE contains high levels of both CRA and LPS, and can induce the cardinal features of asthma-like inflammation, such as airway eosinophilia, airways hyperresponsiveness and IgE production[10,11]. In order to investigate the role of LPS at the time of allergen challenge, LPS was removed from the HDE using a commercially available column. The current study shows that reduction of the LPS content in the HDE exacerbates cytokine production, reduces IgE, but does not alter AHR, airway eosinophilia or neutrophil recruitment.", "[SUBTITLE] Animals [SUBSECTION] Female BALB/c mice 9-12 weeks old were purchased from Jackson Laboratories (Bar Harbor, ME) and maintained under standard laboratory conditions. The mice were housed in a temperature and humidity controlled room with 12 hour light/dark cycles. Food and water were allowed ad libitum. All experiments were performed according to the National Institutes of Health guidelines and were approved by the Boston University and University of Michigan Institutional Animal Care and Use Committees.\nFemale BALB/c mice 9-12 weeks old were purchased from Jackson Laboratories (Bar Harbor, ME) and maintained under standard laboratory conditions. The mice were housed in a temperature and humidity controlled room with 12 hour light/dark cycles. Food and water were allowed ad libitum. All experiments were performed according to the National Institutes of Health guidelines and were approved by the Boston University and University of Michigan Institutional Animal Care and Use Committees.\n[SUBTITLE] Reduction of LPS content in House Dust Extract [SUBSECTION] LPS was removed from the house dust extract (HDE) using the EndoTrap Blue column (Profos AG, Regensburg, Germany), by modifying the manufacturer's protocol. HDE was diluted 1:1 in sterile PBS. The depletion column was equilibrated with the provided equilibration buffer modified to contain 400 mM NaCl. The HDE mixture was applied to the column and eluted using the provided elution buffer. The eluted mixture was immediately aliquoted and stored at -70°C until use. Concentrations of Bla g (Blatella germanica) 1 and Bla g2 and LPS were measured after application to the column, and appropriate dilutions prepared for in vivo administration.\nLPS was removed from the house dust extract (HDE) using the EndoTrap Blue column (Profos AG, Regensburg, Germany), by modifying the manufacturer's protocol. HDE was diluted 1:1 in sterile PBS. The depletion column was equilibrated with the provided equilibration buffer modified to contain 400 mM NaCl. The HDE mixture was applied to the column and eluted using the provided elution buffer. The eluted mixture was immediately aliquoted and stored at -70°C until use. Concentrations of Bla g (Blatella germanica) 1 and Bla g2 and LPS were measured after application to the column, and appropriate dilutions prepared for in vivo administration.\n[SUBTITLE] LPS Assay [SUBSECTION] LPS in the house dust extract and cockroach allergen preparations was assayed in pyrogen free water using the endpoint Limulus amoebocyte lysate (LAL) assay (Lonza, Basel Switzerland, 50-647 U). 96-well microplates and substrate solutions were warmed to 37°C. 50 μL of sample and standard were added to the plate in duplicate followed by 50 μL LAL regent. The plate was incubated at 37°C for 10 min. 100 μL of substrate solution was then added to each well and the plate was incubated at 37°C for 6 minutes. The reaction was stopped with 50 μL 25% glacial acetic acid and the absorbance was read at 405 nm.\nLPS in the house dust extract and cockroach allergen preparations was assayed in pyrogen free water using the endpoint Limulus amoebocyte lysate (LAL) assay (Lonza, Basel Switzerland, 50-647 U). 96-well microplates and substrate solutions were warmed to 37°C. 50 μL of sample and standard were added to the plate in duplicate followed by 50 μL LAL regent. The plate was incubated at 37°C for 10 min. 100 μL of substrate solution was then added to each well and the plate was incubated at 37°C for 6 minutes. The reaction was stopped with 50 μL 25% glacial acetic acid and the absorbance was read at 405 nm.\n[SUBTITLE] Asthma induction protocol [SUBSECTION] House dust collection and processing was performed as previously described[11]. We used a commercial CRA preparation for immunizations since there are limited amounts of HDE available. Additionally, our lab has demonstrated that asthma-like inflammation induced by the HDE is CRA specific[12] and CRA administration will produce asthma-like pulmonary inflammation[13,14]. For these reasons, the commercial German cockroach-allergen[1] (Greer Laboratories, LeNoir, NC, Item # B46) was used for immunization[12]. The CRA (61.9 ng) was diluted to a final volume of 50 μL in sterile phosphate buffered saline (PBS) immediately prior to use. This mixture was emulsified in 50 μL TiterMax Gold adjuvant (CytRx, Norcross, GA). Each mouse was injected i.p. with 100 μL of the adjuvant/allergen mixture on day 0. On day 14 and day 21, mice were challenged by direct intratracheal installation of 50 μL of the intact HDE or LPS-reduced HDE, each containing 61.9 ng CRA [15]. Briefly, mice were lightly anesthetized and suspended by their front incisors on a vertical board. Their tails were taped down to support the body weight. The tongue was gently extended and the liquid was placed at the base of the oropharynx so that it was inhaled.\nHouse dust collection and processing was performed as previously described[11]. We used a commercial CRA preparation for immunizations since there are limited amounts of HDE available. Additionally, our lab has demonstrated that asthma-like inflammation induced by the HDE is CRA specific[12] and CRA administration will produce asthma-like pulmonary inflammation[13,14]. For these reasons, the commercial German cockroach-allergen[1] (Greer Laboratories, LeNoir, NC, Item # B46) was used for immunization[12]. The CRA (61.9 ng) was diluted to a final volume of 50 μL in sterile phosphate buffered saline (PBS) immediately prior to use. This mixture was emulsified in 50 μL TiterMax Gold adjuvant (CytRx, Norcross, GA). Each mouse was injected i.p. with 100 μL of the adjuvant/allergen mixture on day 0. On day 14 and day 21, mice were challenged by direct intratracheal installation of 50 μL of the intact HDE or LPS-reduced HDE, each containing 61.9 ng CRA [15]. Briefly, mice were lightly anesthetized and suspended by their front incisors on a vertical board. Their tails were taped down to support the body weight. The tongue was gently extended and the liquid was placed at the base of the oropharynx so that it was inhaled.\n[SUBTITLE] Timepoints for data collection [SUBSECTION] Animals were euthanized by cervical dislocation following ketamine/xylazine anesthesia at 0, 2 or 24 hours post final allergen challenge. At the time of sacrifice, plasma, bronchoalveolar lavage, and lung homogenates were prepared. The 2 and 24 hour timepoints represent the early phase and late phase of the asthmatic response based on prior publications[11]. The 0 hour timepoint refers to animals sacrificed on day 21 without receiving the second pulmonary challenge. This timepoint was included to determine if sensitization and the first challenge caused any lasting inflammatory reaction in either group. Airway hyperresponsiveness [10] was measured 4 hours post final challenge, based on our observations of robust AHR induction at this timepoint (data not shown). These mice were then sacrificed at 24 hours.\nAnimals were euthanized by cervical dislocation following ketamine/xylazine anesthesia at 0, 2 or 24 hours post final allergen challenge. At the time of sacrifice, plasma, bronchoalveolar lavage, and lung homogenates were prepared. The 2 and 24 hour timepoints represent the early phase and late phase of the asthmatic response based on prior publications[11]. The 0 hour timepoint refers to animals sacrificed on day 21 without receiving the second pulmonary challenge. This timepoint was included to determine if sensitization and the first challenge caused any lasting inflammatory reaction in either group. Airway hyperresponsiveness [10] was measured 4 hours post final challenge, based on our observations of robust AHR induction at this timepoint (data not shown). These mice were then sacrificed at 24 hours.\n[SUBTITLE] Airways Hyperresponsiveness and Airway Resistance [SUBSECTION] Airways changes were measured either with direct invasive techniques (Flexivent, Scireq Scientific Respiratory Equipment, Montreal, Canada) or using unrestrained whole body plethysmography (Buxco Systems, Troy, NY). For whole body plethysmography, mice were placed in the instrument chamber and allowed to acclimate for at least 5 minutes. Baseline measurements were recorded for 5 minutes. Mice were then challenged for 2 minutes with aerosolized PBS and increasing doses of methacholine (Sigma, St. Louis, MO). Each aerosol challenge was followed by 5 minutes of monitoring and data collection. The partial pressure difference between the experimental and reference chambers represented the PenH parameter, and the data presented as the percent increase above baseline PenH measurements.\nThese data were further verified using invasive pulmonary function tests. For measurement of mouse airway resistance, mice were anesthetized with an i.p. injection of 1:5 diluted pentobarbital (Nembutal®, 0.016 ml/g body weight, Ovation Pharmaceutical, Deerfield, IL). The paralytic was pancuronium (Sigma-Aldrich, St. Louis, MO) at 0.5 micrograms per gram body weight. Once adequate surgical sedation was established, determined by a firm squeeze of the foot pad, a tracheotomy was performed by insertion of an 18 g polyethylene cannula into the distal trachea. The mouse was then placed on the FlexiVent mechanical ventilator (Scireq Scientific Respiratory Equipment, Montreal, Canada) and ventilated at 190 breaths per minute with positive-end expiratory pressure set at 3 cmH2O. Measurement of airway resistance in response to increasing concentrations of aerosolized methacholine was obtained through periodic computer-generated \"snapshot 150\" forced-maneuver interruptions in ventilation. Data are then presented as resistance change from baseline (cmH2O per milliliter per second).\nAirways changes were measured either with direct invasive techniques (Flexivent, Scireq Scientific Respiratory Equipment, Montreal, Canada) or using unrestrained whole body plethysmography (Buxco Systems, Troy, NY). For whole body plethysmography, mice were placed in the instrument chamber and allowed to acclimate for at least 5 minutes. Baseline measurements were recorded for 5 minutes. Mice were then challenged for 2 minutes with aerosolized PBS and increasing doses of methacholine (Sigma, St. Louis, MO). Each aerosol challenge was followed by 5 minutes of monitoring and data collection. The partial pressure difference between the experimental and reference chambers represented the PenH parameter, and the data presented as the percent increase above baseline PenH measurements.\nThese data were further verified using invasive pulmonary function tests. For measurement of mouse airway resistance, mice were anesthetized with an i.p. injection of 1:5 diluted pentobarbital (Nembutal®, 0.016 ml/g body weight, Ovation Pharmaceutical, Deerfield, IL). The paralytic was pancuronium (Sigma-Aldrich, St. Louis, MO) at 0.5 micrograms per gram body weight. Once adequate surgical sedation was established, determined by a firm squeeze of the foot pad, a tracheotomy was performed by insertion of an 18 g polyethylene cannula into the distal trachea. The mouse was then placed on the FlexiVent mechanical ventilator (Scireq Scientific Respiratory Equipment, Montreal, Canada) and ventilated at 190 breaths per minute with positive-end expiratory pressure set at 3 cmH2O. Measurement of airway resistance in response to increasing concentrations of aerosolized methacholine was obtained through periodic computer-generated \"snapshot 150\" forced-maneuver interruptions in ventilation. Data are then presented as resistance change from baseline (cmH2O per milliliter per second).\n[SUBTITLE] Bronchoalveolar lavage and lung homogenate preparation [SUBSECTION] Mice were exanguinated and BAL performed by cannulating the trachea. The lung was lavaged with 2, 1 mL aliquots of warm Hank's Buffered Salt Solution (HBSS, Gibco, Grand Island, NY). Both aliquots were centrifuged and the supernatant of the first wash removed and frozen at -20°C for cytokine analysis. The supernatant from the second aliquot was discarded and the cell pellets were resuspended and combined. Total cell counts were obtained using a Beckman-Coulter particle counter model ZF (Coulter Electronics Inc., Hialeah, FL). Cytospin preparations were stained with Diff-Quick and 300 cell differential counts were performed to determine the absolute numbers of inflammatory cells. The right lung was removed, placed in ice cold protease inhibitor cocktail (Roche, Indianapolis, IN) containing 0.00005% Triton X-100 in PBS, and homogenized with 3, 10 second passes in a Brinkmann Polytron PT3000 homogenizer. An aliquot was removed and sonicated in hexadecyltrimethylammonium bromide (HTAB) buffer for myeloperoxidase assay. A separate aliquot was removed and sonicated in 0.5% cetyltrimethylammoniumchloride (CTAC) (Sigma, St. Louis, MO) for eosinophil-specific peroxidase assay. The homogenized and sonicated mixtures were centrifuged at 15,000 g for 15 min. The homogenate supernatant was removed and stored at -20°C for cytokine analysis and the supernatant from the sonicated fractions was used immediately for peroxidase assays.\nMice were exanguinated and BAL performed by cannulating the trachea. The lung was lavaged with 2, 1 mL aliquots of warm Hank's Buffered Salt Solution (HBSS, Gibco, Grand Island, NY). Both aliquots were centrifuged and the supernatant of the first wash removed and frozen at -20°C for cytokine analysis. The supernatant from the second aliquot was discarded and the cell pellets were resuspended and combined. Total cell counts were obtained using a Beckman-Coulter particle counter model ZF (Coulter Electronics Inc., Hialeah, FL). Cytospin preparations were stained with Diff-Quick and 300 cell differential counts were performed to determine the absolute numbers of inflammatory cells. The right lung was removed, placed in ice cold protease inhibitor cocktail (Roche, Indianapolis, IN) containing 0.00005% Triton X-100 in PBS, and homogenized with 3, 10 second passes in a Brinkmann Polytron PT3000 homogenizer. An aliquot was removed and sonicated in hexadecyltrimethylammonium bromide (HTAB) buffer for myeloperoxidase assay. A separate aliquot was removed and sonicated in 0.5% cetyltrimethylammoniumchloride (CTAC) (Sigma, St. Louis, MO) for eosinophil-specific peroxidase assay. The homogenized and sonicated mixtures were centrifuged at 15,000 g for 15 min. The homogenate supernatant was removed and stored at -20°C for cytokine analysis and the supernatant from the sonicated fractions was used immediately for peroxidase assays.\n[SUBTITLE] Myeloperoxidase and Eosinophil Peroxidase Assays [SUBSECTION] Myeloperoxidase (MPO) and eosinophil peroxidase (EPO) assays were performed as described previously[16], with some modifications. EPO was performed by diluting the supernatant of the mixture sonicated in CTAC 1:3 in 10 mM HEPES, pH 8, in quadruplicate in a 96-well plate. 150 μL ice cold stop solution (4N H2SO4 + 2 mM resorcinol) was added to 2 of the sample wells. 75 μL of substrate solution containing 6 mM KBr, 1.5 mM o-phenylenediamine (Sigma, St. Louis, MO, P9029), and 0.3% H2O2 in 50 mM HEPES, pH8, was added to the two remaining sample wells and incubated in the dark for 30 seconds. 150 μL ice cold stop solution was added and the absorbance read at 490 nm.\nMPO was measured by diluting the supernatant of the mixture sonicated in HTAB 1:5 in 10 mM citrate buffer, pH 5 in quadruplicate. 150 μL ice cold stop solution (4N H2SO4) was added to 2 of the sample wells. 75 μL substrate solution containing 0.3 mM 3,3',5,5'-Tetramethylbenzidine, 120 mM resorcinol (an eosinophil peroxidase inhibitor), and 0.007% H2O2 in ddH2O was added to the two remaining sample wells and incubated in the dark for 2 minutes. 150 μL ice cold stop solution was added and the absorbance read at 450 nm. Data is expressed as ΔOD reflecting the difference in absorbance between the average of the sample and background wells.\nMyeloperoxidase (MPO) and eosinophil peroxidase (EPO) assays were performed as described previously[16], with some modifications. EPO was performed by diluting the supernatant of the mixture sonicated in CTAC 1:3 in 10 mM HEPES, pH 8, in quadruplicate in a 96-well plate. 150 μL ice cold stop solution (4N H2SO4 + 2 mM resorcinol) was added to 2 of the sample wells. 75 μL of substrate solution containing 6 mM KBr, 1.5 mM o-phenylenediamine (Sigma, St. Louis, MO, P9029), and 0.3% H2O2 in 50 mM HEPES, pH8, was added to the two remaining sample wells and incubated in the dark for 30 seconds. 150 μL ice cold stop solution was added and the absorbance read at 490 nm.\nMPO was measured by diluting the supernatant of the mixture sonicated in HTAB 1:5 in 10 mM citrate buffer, pH 5 in quadruplicate. 150 μL ice cold stop solution (4N H2SO4) was added to 2 of the sample wells. 75 μL substrate solution containing 0.3 mM 3,3',5,5'-Tetramethylbenzidine, 120 mM resorcinol (an eosinophil peroxidase inhibitor), and 0.007% H2O2 in ddH2O was added to the two remaining sample wells and incubated in the dark for 2 minutes. 150 μL ice cold stop solution was added and the absorbance read at 450 nm. Data is expressed as ΔOD reflecting the difference in absorbance between the average of the sample and background wells.\n[SUBTITLE] ELISA [SUBSECTION] Cytokines and chemokines were measured by ELISA as previously described[17]. Matched antibody pairs and recombinant standards were purchased from R&D Systems (Minneapolis, MN). Lung homogenate samples were assayed with the addition of 20% normal lung homogenate to the standards to adjust for the increased background caused by non-specific matrix effects. Antibody ELISA reagents (IgG and IgE) were purchased from Bethyl Laboratories (Montgomery, TX) and assays were performed by the same standard protocols as for cytokines and chemokines. Cockroach allergen ELISAs were performed as previously described[12] with antibody pairs and recombinant standards from Indoor Biotechnologies (Charlottesville, VA).\nCRA-specific IgE was measured by coating ELISA plates with CRA (Greer Labs) over-night at 4°C. Plates were washed and non-specific binding blocked for 2.5 hours at room temperature on an orbital shaker. Plates were washed an incubated with 1:10 dilutions of plasma overnight at 4°C. Plates were washed and incubated with HRP-goat anti mouse IgE for 1 hour at room temperature on an orbital shaker. Plates were developed using 3,3',5,5'-Tetramethylbenzidine as substrate and read as previously described[12]. Results are represented as ΔOD (OD465 - OD590). Since CRA-specific IgE standards are not currently available, plasma from a naïve mouse was run in parallel in each assay for baseline.\nCytokines and chemokines were measured by ELISA as previously described[17]. Matched antibody pairs and recombinant standards were purchased from R&D Systems (Minneapolis, MN). Lung homogenate samples were assayed with the addition of 20% normal lung homogenate to the standards to adjust for the increased background caused by non-specific matrix effects. Antibody ELISA reagents (IgG and IgE) were purchased from Bethyl Laboratories (Montgomery, TX) and assays were performed by the same standard protocols as for cytokines and chemokines. Cockroach allergen ELISAs were performed as previously described[12] with antibody pairs and recombinant standards from Indoor Biotechnologies (Charlottesville, VA).\nCRA-specific IgE was measured by coating ELISA plates with CRA (Greer Labs) over-night at 4°C. Plates were washed and non-specific binding blocked for 2.5 hours at room temperature on an orbital shaker. Plates were washed an incubated with 1:10 dilutions of plasma overnight at 4°C. Plates were washed and incubated with HRP-goat anti mouse IgE for 1 hour at room temperature on an orbital shaker. Plates were developed using 3,3',5,5'-Tetramethylbenzidine as substrate and read as previously described[12]. Results are represented as ΔOD (OD465 - OD590). Since CRA-specific IgE standards are not currently available, plasma from a naïve mouse was run in parallel in each assay for baseline.\n[SUBTITLE] Cysteinyl-Leukotriene Immunoassay [SUBSECTION] Cysteinyl leukotrienes in the BAL fluid were measured by an Enzyme-linked Immunoassay Kit (Cayman Chemicals, Ann Arbor, MI) according to the manufacturer instructions. All samples were run at two dilutions. %B/B0 values in the linear range of the standard curve were accepted. Sample values that did not fall in this range were appropriately diluted and rerun.\nCysteinyl leukotrienes in the BAL fluid were measured by an Enzyme-linked Immunoassay Kit (Cayman Chemicals, Ann Arbor, MI) according to the manufacturer instructions. All samples were run at two dilutions. %B/B0 values in the linear range of the standard curve were accepted. Sample values that did not fall in this range were appropriately diluted and rerun.\n[SUBTITLE] Statistical Analysis [SUBSECTION] All data are represented as mean ± SEM. Statistical significance was determined by unpaired Student's t-test or One-way ANOVA with Turkey's post test using GraphPad Prism version 4.0.3. (GraphPad Software, San Diego, CA). Statistical significance was achieved when p < 0.05.\nAll data are represented as mean ± SEM. Statistical significance was determined by unpaired Student's t-test or One-way ANOVA with Turkey's post test using GraphPad Prism version 4.0.3. (GraphPad Software, San Diego, CA). Statistical significance was achieved when p < 0.05.", "Female BALB/c mice 9-12 weeks old were purchased from Jackson Laboratories (Bar Harbor, ME) and maintained under standard laboratory conditions. The mice were housed in a temperature and humidity controlled room with 12 hour light/dark cycles. Food and water were allowed ad libitum. All experiments were performed according to the National Institutes of Health guidelines and were approved by the Boston University and University of Michigan Institutional Animal Care and Use Committees.", "LPS was removed from the house dust extract (HDE) using the EndoTrap Blue column (Profos AG, Regensburg, Germany), by modifying the manufacturer's protocol. HDE was diluted 1:1 in sterile PBS. The depletion column was equilibrated with the provided equilibration buffer modified to contain 400 mM NaCl. The HDE mixture was applied to the column and eluted using the provided elution buffer. The eluted mixture was immediately aliquoted and stored at -70°C until use. Concentrations of Bla g (Blatella germanica) 1 and Bla g2 and LPS were measured after application to the column, and appropriate dilutions prepared for in vivo administration.", "LPS in the house dust extract and cockroach allergen preparations was assayed in pyrogen free water using the endpoint Limulus amoebocyte lysate (LAL) assay (Lonza, Basel Switzerland, 50-647 U). 96-well microplates and substrate solutions were warmed to 37°C. 50 μL of sample and standard were added to the plate in duplicate followed by 50 μL LAL regent. The plate was incubated at 37°C for 10 min. 100 μL of substrate solution was then added to each well and the plate was incubated at 37°C for 6 minutes. The reaction was stopped with 50 μL 25% glacial acetic acid and the absorbance was read at 405 nm.", "House dust collection and processing was performed as previously described[11]. We used a commercial CRA preparation for immunizations since there are limited amounts of HDE available. Additionally, our lab has demonstrated that asthma-like inflammation induced by the HDE is CRA specific[12] and CRA administration will produce asthma-like pulmonary inflammation[13,14]. For these reasons, the commercial German cockroach-allergen[1] (Greer Laboratories, LeNoir, NC, Item # B46) was used for immunization[12]. The CRA (61.9 ng) was diluted to a final volume of 50 μL in sterile phosphate buffered saline (PBS) immediately prior to use. This mixture was emulsified in 50 μL TiterMax Gold adjuvant (CytRx, Norcross, GA). Each mouse was injected i.p. with 100 μL of the adjuvant/allergen mixture on day 0. On day 14 and day 21, mice were challenged by direct intratracheal installation of 50 μL of the intact HDE or LPS-reduced HDE, each containing 61.9 ng CRA [15]. Briefly, mice were lightly anesthetized and suspended by their front incisors on a vertical board. Their tails were taped down to support the body weight. The tongue was gently extended and the liquid was placed at the base of the oropharynx so that it was inhaled.", "Animals were euthanized by cervical dislocation following ketamine/xylazine anesthesia at 0, 2 or 24 hours post final allergen challenge. At the time of sacrifice, plasma, bronchoalveolar lavage, and lung homogenates were prepared. The 2 and 24 hour timepoints represent the early phase and late phase of the asthmatic response based on prior publications[11]. The 0 hour timepoint refers to animals sacrificed on day 21 without receiving the second pulmonary challenge. This timepoint was included to determine if sensitization and the first challenge caused any lasting inflammatory reaction in either group. Airway hyperresponsiveness [10] was measured 4 hours post final challenge, based on our observations of robust AHR induction at this timepoint (data not shown). These mice were then sacrificed at 24 hours.", "Airways changes were measured either with direct invasive techniques (Flexivent, Scireq Scientific Respiratory Equipment, Montreal, Canada) or using unrestrained whole body plethysmography (Buxco Systems, Troy, NY). For whole body plethysmography, mice were placed in the instrument chamber and allowed to acclimate for at least 5 minutes. Baseline measurements were recorded for 5 minutes. Mice were then challenged for 2 minutes with aerosolized PBS and increasing doses of methacholine (Sigma, St. Louis, MO). Each aerosol challenge was followed by 5 minutes of monitoring and data collection. The partial pressure difference between the experimental and reference chambers represented the PenH parameter, and the data presented as the percent increase above baseline PenH measurements.\nThese data were further verified using invasive pulmonary function tests. For measurement of mouse airway resistance, mice were anesthetized with an i.p. injection of 1:5 diluted pentobarbital (Nembutal®, 0.016 ml/g body weight, Ovation Pharmaceutical, Deerfield, IL). The paralytic was pancuronium (Sigma-Aldrich, St. Louis, MO) at 0.5 micrograms per gram body weight. Once adequate surgical sedation was established, determined by a firm squeeze of the foot pad, a tracheotomy was performed by insertion of an 18 g polyethylene cannula into the distal trachea. The mouse was then placed on the FlexiVent mechanical ventilator (Scireq Scientific Respiratory Equipment, Montreal, Canada) and ventilated at 190 breaths per minute with positive-end expiratory pressure set at 3 cmH2O. Measurement of airway resistance in response to increasing concentrations of aerosolized methacholine was obtained through periodic computer-generated \"snapshot 150\" forced-maneuver interruptions in ventilation. Data are then presented as resistance change from baseline (cmH2O per milliliter per second).", "Mice were exanguinated and BAL performed by cannulating the trachea. The lung was lavaged with 2, 1 mL aliquots of warm Hank's Buffered Salt Solution (HBSS, Gibco, Grand Island, NY). Both aliquots were centrifuged and the supernatant of the first wash removed and frozen at -20°C for cytokine analysis. The supernatant from the second aliquot was discarded and the cell pellets were resuspended and combined. Total cell counts were obtained using a Beckman-Coulter particle counter model ZF (Coulter Electronics Inc., Hialeah, FL). Cytospin preparations were stained with Diff-Quick and 300 cell differential counts were performed to determine the absolute numbers of inflammatory cells. The right lung was removed, placed in ice cold protease inhibitor cocktail (Roche, Indianapolis, IN) containing 0.00005% Triton X-100 in PBS, and homogenized with 3, 10 second passes in a Brinkmann Polytron PT3000 homogenizer. An aliquot was removed and sonicated in hexadecyltrimethylammonium bromide (HTAB) buffer for myeloperoxidase assay. A separate aliquot was removed and sonicated in 0.5% cetyltrimethylammoniumchloride (CTAC) (Sigma, St. Louis, MO) for eosinophil-specific peroxidase assay. The homogenized and sonicated mixtures were centrifuged at 15,000 g for 15 min. The homogenate supernatant was removed and stored at -20°C for cytokine analysis and the supernatant from the sonicated fractions was used immediately for peroxidase assays.", "Myeloperoxidase (MPO) and eosinophil peroxidase (EPO) assays were performed as described previously[16], with some modifications. EPO was performed by diluting the supernatant of the mixture sonicated in CTAC 1:3 in 10 mM HEPES, pH 8, in quadruplicate in a 96-well plate. 150 μL ice cold stop solution (4N H2SO4 + 2 mM resorcinol) was added to 2 of the sample wells. 75 μL of substrate solution containing 6 mM KBr, 1.5 mM o-phenylenediamine (Sigma, St. Louis, MO, P9029), and 0.3% H2O2 in 50 mM HEPES, pH8, was added to the two remaining sample wells and incubated in the dark for 30 seconds. 150 μL ice cold stop solution was added and the absorbance read at 490 nm.\nMPO was measured by diluting the supernatant of the mixture sonicated in HTAB 1:5 in 10 mM citrate buffer, pH 5 in quadruplicate. 150 μL ice cold stop solution (4N H2SO4) was added to 2 of the sample wells. 75 μL substrate solution containing 0.3 mM 3,3',5,5'-Tetramethylbenzidine, 120 mM resorcinol (an eosinophil peroxidase inhibitor), and 0.007% H2O2 in ddH2O was added to the two remaining sample wells and incubated in the dark for 2 minutes. 150 μL ice cold stop solution was added and the absorbance read at 450 nm. Data is expressed as ΔOD reflecting the difference in absorbance between the average of the sample and background wells.", "Cytokines and chemokines were measured by ELISA as previously described[17]. Matched antibody pairs and recombinant standards were purchased from R&D Systems (Minneapolis, MN). Lung homogenate samples were assayed with the addition of 20% normal lung homogenate to the standards to adjust for the increased background caused by non-specific matrix effects. Antibody ELISA reagents (IgG and IgE) were purchased from Bethyl Laboratories (Montgomery, TX) and assays were performed by the same standard protocols as for cytokines and chemokines. Cockroach allergen ELISAs were performed as previously described[12] with antibody pairs and recombinant standards from Indoor Biotechnologies (Charlottesville, VA).\nCRA-specific IgE was measured by coating ELISA plates with CRA (Greer Labs) over-night at 4°C. Plates were washed and non-specific binding blocked for 2.5 hours at room temperature on an orbital shaker. Plates were washed an incubated with 1:10 dilutions of plasma overnight at 4°C. Plates were washed and incubated with HRP-goat anti mouse IgE for 1 hour at room temperature on an orbital shaker. Plates were developed using 3,3',5,5'-Tetramethylbenzidine as substrate and read as previously described[12]. Results are represented as ΔOD (OD465 - OD590). Since CRA-specific IgE standards are not currently available, plasma from a naïve mouse was run in parallel in each assay for baseline.", "Cysteinyl leukotrienes in the BAL fluid were measured by an Enzyme-linked Immunoassay Kit (Cayman Chemicals, Ann Arbor, MI) according to the manufacturer instructions. All samples were run at two dilutions. %B/B0 values in the linear range of the standard curve were accepted. Sample values that did not fall in this range were appropriately diluted and rerun.", "All data are represented as mean ± SEM. Statistical significance was determined by unpaired Student's t-test or One-way ANOVA with Turkey's post test using GraphPad Prism version 4.0.3. (GraphPad Software, San Diego, CA). Statistical significance was achieved when p < 0.05.", "[SUBTITLE] Depletion of LPS from the house dust extract [SUBSECTION] Dust was collected from the homes of asthmatic children as described previously[11]. This dust was shown to contain significantly higher concentrations of cockroach allergens Bla g1 and Bla g2 in comparison to other indoor and outdoor allergens[12]. This extract, when used in previous studies at a 1:10 dilution (containing a total of 61.9 ng CRA) successfully induced the hallmark features of asthma-like inflammation, including airways hyperresponsiveness, eosinophilia and plasma IgE[18-20]. In the present study, we specifically sought to determine whether removal of the potent innate immune activating agent LPS from the HDE at the time of allergen challenge would attenuate the severity of pulmonary inflammation in sensitized mice. To address this question, LPS in the HDE was reduced as described in materials and methods. Unfortunately, our previous attempts at LPS removal using polymyxin B columns (Pierce, Rockford, IL) also removed 99.9% of Bla g1 from the HDE, and these preparations were not appropriate for the present studies (Table 1). Although use of the EndoTrap column in this study slightly reduced CRA concentrations, through appropriate dilution, we were able to maintain the same total amount of 61.9 ng CRA for all treatments (Table 2).\nTotal amounts of CRA and LPS in the intact HDE and in the HDE treated with Polymyxin B or EndoTrap to deplete LPS.\nAmounts of CRA and LPS present in both the immunization and challenge mixtures.\nDust was collected from the homes of asthmatic children as described previously[11]. This dust was shown to contain significantly higher concentrations of cockroach allergens Bla g1 and Bla g2 in comparison to other indoor and outdoor allergens[12]. This extract, when used in previous studies at a 1:10 dilution (containing a total of 61.9 ng CRA) successfully induced the hallmark features of asthma-like inflammation, including airways hyperresponsiveness, eosinophilia and plasma IgE[18-20]. In the present study, we specifically sought to determine whether removal of the potent innate immune activating agent LPS from the HDE at the time of allergen challenge would attenuate the severity of pulmonary inflammation in sensitized mice. To address this question, LPS in the HDE was reduced as described in materials and methods. Unfortunately, our previous attempts at LPS removal using polymyxin B columns (Pierce, Rockford, IL) also removed 99.9% of Bla g1 from the HDE, and these preparations were not appropriate for the present studies (Table 1). Although use of the EndoTrap column in this study slightly reduced CRA concentrations, through appropriate dilution, we were able to maintain the same total amount of 61.9 ng CRA for all treatments (Table 2).\nTotal amounts of CRA and LPS in the intact HDE and in the HDE treated with Polymyxin B or EndoTrap to deplete LPS.\nAmounts of CRA and LPS present in both the immunization and challenge mixtures.\n[SUBTITLE] TNFα production in BAL fluid post final challenge [SUBSECTION] Sensitization to cockroach allergens plays a key role in the development of childhood asthma[21], and the level of LPS present in the home has also been shown to have a major effect on the severity of respiratory diseases[5,10,22]. Asthma-like inflammation was induced by i.p. immunization with 61.9 ng CRA, containing 11.6 ng of LPS. Mice were challenged on days 14 and 21 with either the intact HDE (182.1 ng of LPS) or LPS-reduced HDE (3.0 ng of LPS). Although different allergen mixtures are used for sensitization and challenge, we have previously shown that HDE immunization specifically induces an allergic response to CRA[12]. Consistent with the low levels of LPS present in the LPS-reduced HDE, TNFα production was significantly decreased in this group at 2 hours post final challenge (Figure 1). We have previously shown that this early timepoint reflects the point at which TNFα production peaks in the BAL fluid[19].\nTNFα production in bronchoalveolar lavage fluid. TNFα was measured by ELISA at the timepoints indicated. The 0 h sample was collected prior to the second pulmonary challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ***p ≤ 0.0001 comparing the LPS-reduced HDE group to the HDE group by Student's t-test.\nSensitization to cockroach allergens plays a key role in the development of childhood asthma[21], and the level of LPS present in the home has also been shown to have a major effect on the severity of respiratory diseases[5,10,22]. Asthma-like inflammation was induced by i.p. immunization with 61.9 ng CRA, containing 11.6 ng of LPS. Mice were challenged on days 14 and 21 with either the intact HDE (182.1 ng of LPS) or LPS-reduced HDE (3.0 ng of LPS). Although different allergen mixtures are used for sensitization and challenge, we have previously shown that HDE immunization specifically induces an allergic response to CRA[12]. Consistent with the low levels of LPS present in the LPS-reduced HDE, TNFα production was significantly decreased in this group at 2 hours post final challenge (Figure 1). We have previously shown that this early timepoint reflects the point at which TNFα production peaks in the BAL fluid[19].\nTNFα production in bronchoalveolar lavage fluid. TNFα was measured by ELISA at the timepoints indicated. The 0 h sample was collected prior to the second pulmonary challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ***p ≤ 0.0001 comparing the LPS-reduced HDE group to the HDE group by Student's t-test.\n[SUBTITLE] Role of LPS in antibody production [SUBSECTION] The presence or absence of TNFα has been shown to affect the antibody response to allergen[8]. We sought to determine whether the reduced levels of TNFα produced in response to challenge with the LPS-reduced HDE had an effect on antibody production in this model. Intact HDE challenge resulted in robust IgE production while challenge with LPS-reduced HDE did not induce IgE production beyond that seen in mice receiving only the CRA immunization (Figure 2A). Despite the global increase in IgE, measurement of CRA-specific IgE showed no difference between LPS-reduced HDE challenge and HDE challenge (Figure 2B). IgG1 production was similar to total IgE production with reduced levels in the mice challenged with the LPS-reduced HDE compared to immunized-only mice (Figure 2C). These changes in immunoglobulins were not found with all classes, since LPS-reduced HDE challenge induced significantly higher levels of plasma IgG2a compared to the immunized-only group (Figure 2D).\nPlasma antibody levels in immunized and challenged mice. Plasma levels of A) IgE, B) CRA-specific IgE, C) IgG1, and D) IgG2a were measured by standard ELISA from EDTA-plasma collected 24 hours post final challenge (day 22). The immunized-only group was sensitized on Day 0 and plasma was collected on day 22. Data are represented as the mean ± SEM. Statistical significance was assessed by one-way ANOVA followed by Turkey's post test as indicated in each panel. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ** p < 0.001.\nThe presence or absence of TNFα has been shown to affect the antibody response to allergen[8]. We sought to determine whether the reduced levels of TNFα produced in response to challenge with the LPS-reduced HDE had an effect on antibody production in this model. Intact HDE challenge resulted in robust IgE production while challenge with LPS-reduced HDE did not induce IgE production beyond that seen in mice receiving only the CRA immunization (Figure 2A). Despite the global increase in IgE, measurement of CRA-specific IgE showed no difference between LPS-reduced HDE challenge and HDE challenge (Figure 2B). IgG1 production was similar to total IgE production with reduced levels in the mice challenged with the LPS-reduced HDE compared to immunized-only mice (Figure 2C). These changes in immunoglobulins were not found with all classes, since LPS-reduced HDE challenge induced significantly higher levels of plasma IgG2a compared to the immunized-only group (Figure 2D).\nPlasma antibody levels in immunized and challenged mice. Plasma levels of A) IgE, B) CRA-specific IgE, C) IgG1, and D) IgG2a were measured by standard ELISA from EDTA-plasma collected 24 hours post final challenge (day 22). The immunized-only group was sensitized on Day 0 and plasma was collected on day 22. Data are represented as the mean ± SEM. Statistical significance was assessed by one-way ANOVA followed by Turkey's post test as indicated in each panel. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ** p < 0.001.\n[SUBTITLE] Inflammatory cell recruitment in the BAL fluid post allergen challenge [SUBSECTION] The composition of inflammatory cells in the BAL fluid was assessed at various timepoints after the final intratracheal challenge. Cytospin preparations demonstrate distinct inflammatory cell morphology. Eosinophils were characterized by a ring-shaped nucleus with pink cytoplasmic amyloid staining, while neutrophils show a lobular nucleus with no cytoplasmic staining. Similar numbers of eosinophils and neutrophils in the mice challenged with the intact HDE or the LPS-reduced HDE, in contrast to that of naïve mice containing predominantly macrophages (Figure 3). We observed a slight reduction in eosinophil numbers in the BAL fluid of the LPS-reduced HDE group, but this change did not reach statistical significance (Figure 4A). In the BAL fluid, the eosinophil-specific chemoattractant eotaxin-1 (CCL11) was significantly increased at 24 hours post final challenge in the LPS-reduced HDE mice, however no difference in eotaxin-2 (CCL24) was observed (Figure 4). We also assayed eosinophil-specific peroxidase activity (EPO) in the lungs of mice (after BAL) as a measure of eosinophils within the lung interstitium. No significant difference in EPO was measured between the two groups (Figure 4D). A statistically significant decrease in eotaxin-1 (CCL11) was measured in the lung homogenate of the LPS-reduced HDE group (Figure 4E), but no difference was measured in eotaxin-2 (CCL24) between groups (Figure 4F). While statistically significant differences in eotaxin-1 are observed in both the BAL and lung homogenate, it should be noted that eotaxin-2 is produced at much higher concentrations in both the alveolar compartment and the lung tissue, rendering an equivalent biological outcome (eosinophil recruitment) in both groups.\nCytospin preparations of cells recovered from BAL fluid. A) Naïve mice, B) HDE challenged mice and C) LPS-reduced HDE challenged mice at 24 hours post final challenge. Each is represented at 1000×. Representative eosinophils (Eo) and neutrophils (Ne) are indicated in the figure.\nEosinophil recruitment and production of eosinophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at the timepoints indicated were made and 300 cell differential counts performed to determine the absolute numbers of eosinophils (A). Concentrations of the chemokines B) CCL11 and C) CCL24 in the BAL fluid were measured by ELISA. D) Eosinophil peroxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CCL11 and F) CCL24 in the lung homogenate were assayed by ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment, in some data points the symbol is larger than the SEM. * p < 0.05 and *** p ≤ 0.0001 HDE vs. LPS-reduced HDE by Student's t-test.\nWe also assayed the number of neutrophils recruited in response to allergen challenge. Interestingly, comparable numbers of neutrophils were present in both groups despite reduced levels of LPS in the LPS-reduced HDE (Figure 5A). This suggests that other components of the HDE are able to activate the innate immune response in order to recruit neutrophils. The concentrations of the neutrophil chemoattractants CXCL1 and CXCL2 were assayed as the mechanism of neutrophil recruitment into the alveolar space and lung tissue. BAL levels of CXCL1 are significantly decreased at 2 hours in mice challenged with the LPS-reduced HDE, however no difference was seen in CXCL2 (Figure 5). Despite this slight decrease in chemokine production, we postulate that the concentrations at which they are being produced is adequate to recruit equivalent numbers of neutrophils shown in Figure 5A.\nNeutrophil recruitment and production of neutrophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at 24 hours post final challenge were made and 300 cell differential counts performed to determine the absolute numbers of A) neutrophils. Concentrations of chemokines B) CXCL1 and C) CXCL2 in the BAL fluid were measured by ELISA. D) Myeloperoxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CXCL1 and F) CXCL2 in the lung homogenate were assayed by standard ELISA. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05, ** p ≤ 0.001 and *** p ≤ 0.0001 comparing the LPS-reduced HDE group to HDE the group by Student's t-test.\nMyeloperoxidase activity (MPO) in the lung tissue was not changed between groups. (Figure 5D). Significant increases were measured in both CXCL1 and CXCL2 in the lung homogenate in the LPS-depleted HDE at all timepoints assayed (Figure 5E). This again suggests that although the chemokines were lower in the intact HDE, the concentrations were sufficient to sequester neutrophils within the lung interstitium. No differences in macrophage or lymphocyte numbers were seen in the BAL and no difference was seen in neutrophil numbers at 2 h post final challenge (data not shown).\nThe composition of inflammatory cells in the BAL fluid was assessed at various timepoints after the final intratracheal challenge. Cytospin preparations demonstrate distinct inflammatory cell morphology. Eosinophils were characterized by a ring-shaped nucleus with pink cytoplasmic amyloid staining, while neutrophils show a lobular nucleus with no cytoplasmic staining. Similar numbers of eosinophils and neutrophils in the mice challenged with the intact HDE or the LPS-reduced HDE, in contrast to that of naïve mice containing predominantly macrophages (Figure 3). We observed a slight reduction in eosinophil numbers in the BAL fluid of the LPS-reduced HDE group, but this change did not reach statistical significance (Figure 4A). In the BAL fluid, the eosinophil-specific chemoattractant eotaxin-1 (CCL11) was significantly increased at 24 hours post final challenge in the LPS-reduced HDE mice, however no difference in eotaxin-2 (CCL24) was observed (Figure 4). We also assayed eosinophil-specific peroxidase activity (EPO) in the lungs of mice (after BAL) as a measure of eosinophils within the lung interstitium. No significant difference in EPO was measured between the two groups (Figure 4D). A statistically significant decrease in eotaxin-1 (CCL11) was measured in the lung homogenate of the LPS-reduced HDE group (Figure 4E), but no difference was measured in eotaxin-2 (CCL24) between groups (Figure 4F). While statistically significant differences in eotaxin-1 are observed in both the BAL and lung homogenate, it should be noted that eotaxin-2 is produced at much higher concentrations in both the alveolar compartment and the lung tissue, rendering an equivalent biological outcome (eosinophil recruitment) in both groups.\nCytospin preparations of cells recovered from BAL fluid. A) Naïve mice, B) HDE challenged mice and C) LPS-reduced HDE challenged mice at 24 hours post final challenge. Each is represented at 1000×. Representative eosinophils (Eo) and neutrophils (Ne) are indicated in the figure.\nEosinophil recruitment and production of eosinophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at the timepoints indicated were made and 300 cell differential counts performed to determine the absolute numbers of eosinophils (A). Concentrations of the chemokines B) CCL11 and C) CCL24 in the BAL fluid were measured by ELISA. D) Eosinophil peroxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CCL11 and F) CCL24 in the lung homogenate were assayed by ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment, in some data points the symbol is larger than the SEM. * p < 0.05 and *** p ≤ 0.0001 HDE vs. LPS-reduced HDE by Student's t-test.\nWe also assayed the number of neutrophils recruited in response to allergen challenge. Interestingly, comparable numbers of neutrophils were present in both groups despite reduced levels of LPS in the LPS-reduced HDE (Figure 5A). This suggests that other components of the HDE are able to activate the innate immune response in order to recruit neutrophils. The concentrations of the neutrophil chemoattractants CXCL1 and CXCL2 were assayed as the mechanism of neutrophil recruitment into the alveolar space and lung tissue. BAL levels of CXCL1 are significantly decreased at 2 hours in mice challenged with the LPS-reduced HDE, however no difference was seen in CXCL2 (Figure 5). Despite this slight decrease in chemokine production, we postulate that the concentrations at which they are being produced is adequate to recruit equivalent numbers of neutrophils shown in Figure 5A.\nNeutrophil recruitment and production of neutrophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at 24 hours post final challenge were made and 300 cell differential counts performed to determine the absolute numbers of A) neutrophils. Concentrations of chemokines B) CXCL1 and C) CXCL2 in the BAL fluid were measured by ELISA. D) Myeloperoxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CXCL1 and F) CXCL2 in the lung homogenate were assayed by standard ELISA. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05, ** p ≤ 0.001 and *** p ≤ 0.0001 comparing the LPS-reduced HDE group to HDE the group by Student's t-test.\nMyeloperoxidase activity (MPO) in the lung tissue was not changed between groups. (Figure 5D). Significant increases were measured in both CXCL1 and CXCL2 in the lung homogenate in the LPS-depleted HDE at all timepoints assayed (Figure 5E). This again suggests that although the chemokines were lower in the intact HDE, the concentrations were sufficient to sequester neutrophils within the lung interstitium. No differences in macrophage or lymphocyte numbers were seen in the BAL and no difference was seen in neutrophil numbers at 2 h post final challenge (data not shown).\n[SUBTITLE] Induction of airways hyperresponsiveness after allergen challenge [SUBSECTION] Several studies have shown that both LPS inhalation and pulmonary TNFα administration induce AHR, however little data exist regarding the synergistic role of allergen and endotoxin on AHR[23-25]. Therefore, we sought to determine whether the levels of LPS in the HDE would affect AHR severity. Challenge with increasing concentrations of aerosolized methacholine 4 hours post final challenge did not result in any difference in AHR between groups (Figure 6A). The 4 hour timepoint was assessed based on observations in our laboratory showing robust AHR induction at this timepoint. No increase in AHR was measured comparing normal mice and sensitized-only mice (data not shown). Given the controversy associated with non-invasive whole body plethysmography for assessing AHR, we confirmed our findings using invasive lung function tests. As shown in Figure 6B, invasive tests demonstrated no significant difference in airway resistance between groups at all methacholine concentrations tested. We also assessed BAL cysteinyl leukotriene (Cys-LT) concentrations at 2 hours as a mechanism for AHR, and found no significant change in BAL Cys-LT in mice challenged with LPS-reduced HDE compared to HDE challenge (Figure 6C).\nAirways hyperresponsiveness, airway resistance and cysteinyl leukotrienes post allergen challenge. A) For non-invasive measurement of AHR, mice were challenged with increasing concentrations of aerosolized methacholine at 4 hours after the final challenge and PenH values are represented as percent above baseline. B) For invasive measurements, mice were challenged with methacholine and pulmonary resistance is expressed as R cmH2O.s/mL. C) Cysteinyl leukotriene concentrations were measured in the BAL fluid 2 hours post final challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment.\nSeveral studies have shown that both LPS inhalation and pulmonary TNFα administration induce AHR, however little data exist regarding the synergistic role of allergen and endotoxin on AHR[23-25]. Therefore, we sought to determine whether the levels of LPS in the HDE would affect AHR severity. Challenge with increasing concentrations of aerosolized methacholine 4 hours post final challenge did not result in any difference in AHR between groups (Figure 6A). The 4 hour timepoint was assessed based on observations in our laboratory showing robust AHR induction at this timepoint. No increase in AHR was measured comparing normal mice and sensitized-only mice (data not shown). Given the controversy associated with non-invasive whole body plethysmography for assessing AHR, we confirmed our findings using invasive lung function tests. As shown in Figure 6B, invasive tests demonstrated no significant difference in airway resistance between groups at all methacholine concentrations tested. We also assessed BAL cysteinyl leukotriene (Cys-LT) concentrations at 2 hours as a mechanism for AHR, and found no significant change in BAL Cys-LT in mice challenged with LPS-reduced HDE compared to HDE challenge (Figure 6C).\nAirways hyperresponsiveness, airway resistance and cysteinyl leukotrienes post allergen challenge. A) For non-invasive measurement of AHR, mice were challenged with increasing concentrations of aerosolized methacholine at 4 hours after the final challenge and PenH values are represented as percent above baseline. B) For invasive measurements, mice were challenged with methacholine and pulmonary resistance is expressed as R cmH2O.s/mL. C) Cysteinyl leukotriene concentrations were measured in the BAL fluid 2 hours post final challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment.\n[SUBTITLE] Th1 and Th2 cytokine production in the BAL post allergen challenge [SUBSECTION] It has been suggested previously that the presence of TNFα can induce increased levels of Th2 cytokines in the BAL fluid of allergen challenged mice[19]. Further, asthma is largely characterized as a Th2 mediated disease, accompanied by high IL-4 and IL-5 production and low IFNγ[26]. To determine whether the reduction in TNFα seen in LPS-reduced HDE challenged mice affected cytokine production, we assayed both Th1 and Th2 cytokines 24 hours post final challenge. In conflict with the Th2 paradigm, IFNγ gamma was significantly increased in the LPS-reduced HDE mice (Figure 7A). Surprisingly, IL-5 and IL-13 were significantly increased in the BAL from LPS-reduced HDE challenged mice although there was no difference in IL-4 levels (Figure 7B, C). IL-17 has also been implicated as a key cytokine in the asthmatic response, and is also modulated by LPS levels [27,28]. However, we measured no significant differences in IL-17 levels between groups in the BAL fluid (data not shown).\nTh1 and Th2 cytokine production in the BAL fluid at 24 hours post final challenge. Cytokines were assayed by standard ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05 and *** p ≤ 0.0001 comparing the LPS-reduced HDE to HDE the group by Student's t-test.\nIt has been suggested previously that the presence of TNFα can induce increased levels of Th2 cytokines in the BAL fluid of allergen challenged mice[19]. Further, asthma is largely characterized as a Th2 mediated disease, accompanied by high IL-4 and IL-5 production and low IFNγ[26]. To determine whether the reduction in TNFα seen in LPS-reduced HDE challenged mice affected cytokine production, we assayed both Th1 and Th2 cytokines 24 hours post final challenge. In conflict with the Th2 paradigm, IFNγ gamma was significantly increased in the LPS-reduced HDE mice (Figure 7A). Surprisingly, IL-5 and IL-13 were significantly increased in the BAL from LPS-reduced HDE challenged mice although there was no difference in IL-4 levels (Figure 7B, C). IL-17 has also been implicated as a key cytokine in the asthmatic response, and is also modulated by LPS levels [27,28]. However, we measured no significant differences in IL-17 levels between groups in the BAL fluid (data not shown).\nTh1 and Th2 cytokine production in the BAL fluid at 24 hours post final challenge. Cytokines were assayed by standard ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05 and *** p ≤ 0.0001 comparing the LPS-reduced HDE to HDE the group by Student's t-test.", "Dust was collected from the homes of asthmatic children as described previously[11]. This dust was shown to contain significantly higher concentrations of cockroach allergens Bla g1 and Bla g2 in comparison to other indoor and outdoor allergens[12]. This extract, when used in previous studies at a 1:10 dilution (containing a total of 61.9 ng CRA) successfully induced the hallmark features of asthma-like inflammation, including airways hyperresponsiveness, eosinophilia and plasma IgE[18-20]. In the present study, we specifically sought to determine whether removal of the potent innate immune activating agent LPS from the HDE at the time of allergen challenge would attenuate the severity of pulmonary inflammation in sensitized mice. To address this question, LPS in the HDE was reduced as described in materials and methods. Unfortunately, our previous attempts at LPS removal using polymyxin B columns (Pierce, Rockford, IL) also removed 99.9% of Bla g1 from the HDE, and these preparations were not appropriate for the present studies (Table 1). Although use of the EndoTrap column in this study slightly reduced CRA concentrations, through appropriate dilution, we were able to maintain the same total amount of 61.9 ng CRA for all treatments (Table 2).\nTotal amounts of CRA and LPS in the intact HDE and in the HDE treated with Polymyxin B or EndoTrap to deplete LPS.\nAmounts of CRA and LPS present in both the immunization and challenge mixtures.", "Sensitization to cockroach allergens plays a key role in the development of childhood asthma[21], and the level of LPS present in the home has also been shown to have a major effect on the severity of respiratory diseases[5,10,22]. Asthma-like inflammation was induced by i.p. immunization with 61.9 ng CRA, containing 11.6 ng of LPS. Mice were challenged on days 14 and 21 with either the intact HDE (182.1 ng of LPS) or LPS-reduced HDE (3.0 ng of LPS). Although different allergen mixtures are used for sensitization and challenge, we have previously shown that HDE immunization specifically induces an allergic response to CRA[12]. Consistent with the low levels of LPS present in the LPS-reduced HDE, TNFα production was significantly decreased in this group at 2 hours post final challenge (Figure 1). We have previously shown that this early timepoint reflects the point at which TNFα production peaks in the BAL fluid[19].\nTNFα production in bronchoalveolar lavage fluid. TNFα was measured by ELISA at the timepoints indicated. The 0 h sample was collected prior to the second pulmonary challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ***p ≤ 0.0001 comparing the LPS-reduced HDE group to the HDE group by Student's t-test.", "The presence or absence of TNFα has been shown to affect the antibody response to allergen[8]. We sought to determine whether the reduced levels of TNFα produced in response to challenge with the LPS-reduced HDE had an effect on antibody production in this model. Intact HDE challenge resulted in robust IgE production while challenge with LPS-reduced HDE did not induce IgE production beyond that seen in mice receiving only the CRA immunization (Figure 2A). Despite the global increase in IgE, measurement of CRA-specific IgE showed no difference between LPS-reduced HDE challenge and HDE challenge (Figure 2B). IgG1 production was similar to total IgE production with reduced levels in the mice challenged with the LPS-reduced HDE compared to immunized-only mice (Figure 2C). These changes in immunoglobulins were not found with all classes, since LPS-reduced HDE challenge induced significantly higher levels of plasma IgG2a compared to the immunized-only group (Figure 2D).\nPlasma antibody levels in immunized and challenged mice. Plasma levels of A) IgE, B) CRA-specific IgE, C) IgG1, and D) IgG2a were measured by standard ELISA from EDTA-plasma collected 24 hours post final challenge (day 22). The immunized-only group was sensitized on Day 0 and plasma was collected on day 22. Data are represented as the mean ± SEM. Statistical significance was assessed by one-way ANOVA followed by Turkey's post test as indicated in each panel. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. ** p < 0.001.", "The composition of inflammatory cells in the BAL fluid was assessed at various timepoints after the final intratracheal challenge. Cytospin preparations demonstrate distinct inflammatory cell morphology. Eosinophils were characterized by a ring-shaped nucleus with pink cytoplasmic amyloid staining, while neutrophils show a lobular nucleus with no cytoplasmic staining. Similar numbers of eosinophils and neutrophils in the mice challenged with the intact HDE or the LPS-reduced HDE, in contrast to that of naïve mice containing predominantly macrophages (Figure 3). We observed a slight reduction in eosinophil numbers in the BAL fluid of the LPS-reduced HDE group, but this change did not reach statistical significance (Figure 4A). In the BAL fluid, the eosinophil-specific chemoattractant eotaxin-1 (CCL11) was significantly increased at 24 hours post final challenge in the LPS-reduced HDE mice, however no difference in eotaxin-2 (CCL24) was observed (Figure 4). We also assayed eosinophil-specific peroxidase activity (EPO) in the lungs of mice (after BAL) as a measure of eosinophils within the lung interstitium. No significant difference in EPO was measured between the two groups (Figure 4D). A statistically significant decrease in eotaxin-1 (CCL11) was measured in the lung homogenate of the LPS-reduced HDE group (Figure 4E), but no difference was measured in eotaxin-2 (CCL24) between groups (Figure 4F). While statistically significant differences in eotaxin-1 are observed in both the BAL and lung homogenate, it should be noted that eotaxin-2 is produced at much higher concentrations in both the alveolar compartment and the lung tissue, rendering an equivalent biological outcome (eosinophil recruitment) in both groups.\nCytospin preparations of cells recovered from BAL fluid. A) Naïve mice, B) HDE challenged mice and C) LPS-reduced HDE challenged mice at 24 hours post final challenge. Each is represented at 1000×. Representative eosinophils (Eo) and neutrophils (Ne) are indicated in the figure.\nEosinophil recruitment and production of eosinophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at the timepoints indicated were made and 300 cell differential counts performed to determine the absolute numbers of eosinophils (A). Concentrations of the chemokines B) CCL11 and C) CCL24 in the BAL fluid were measured by ELISA. D) Eosinophil peroxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CCL11 and F) CCL24 in the lung homogenate were assayed by ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment, in some data points the symbol is larger than the SEM. * p < 0.05 and *** p ≤ 0.0001 HDE vs. LPS-reduced HDE by Student's t-test.\nWe also assayed the number of neutrophils recruited in response to allergen challenge. Interestingly, comparable numbers of neutrophils were present in both groups despite reduced levels of LPS in the LPS-reduced HDE (Figure 5A). This suggests that other components of the HDE are able to activate the innate immune response in order to recruit neutrophils. The concentrations of the neutrophil chemoattractants CXCL1 and CXCL2 were assayed as the mechanism of neutrophil recruitment into the alveolar space and lung tissue. BAL levels of CXCL1 are significantly decreased at 2 hours in mice challenged with the LPS-reduced HDE, however no difference was seen in CXCL2 (Figure 5). Despite this slight decrease in chemokine production, we postulate that the concentrations at which they are being produced is adequate to recruit equivalent numbers of neutrophils shown in Figure 5A.\nNeutrophil recruitment and production of neutrophil-specific chemokines in response to allergen challenge. Cytospin preparations from cells collected in the BAL fluid at 24 hours post final challenge were made and 300 cell differential counts performed to determine the absolute numbers of A) neutrophils. Concentrations of chemokines B) CXCL1 and C) CXCL2 in the BAL fluid were measured by ELISA. D) Myeloperoxidase activity was measured in the lung tissue after BAL at 24 hours post final challenge. Concentrations of E) CXCL1 and F) CXCL2 in the lung homogenate were assayed by standard ELISA. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05, ** p ≤ 0.001 and *** p ≤ 0.0001 comparing the LPS-reduced HDE group to HDE the group by Student's t-test.\nMyeloperoxidase activity (MPO) in the lung tissue was not changed between groups. (Figure 5D). Significant increases were measured in both CXCL1 and CXCL2 in the lung homogenate in the LPS-depleted HDE at all timepoints assayed (Figure 5E). This again suggests that although the chemokines were lower in the intact HDE, the concentrations were sufficient to sequester neutrophils within the lung interstitium. No differences in macrophage or lymphocyte numbers were seen in the BAL and no difference was seen in neutrophil numbers at 2 h post final challenge (data not shown).", "Several studies have shown that both LPS inhalation and pulmonary TNFα administration induce AHR, however little data exist regarding the synergistic role of allergen and endotoxin on AHR[23-25]. Therefore, we sought to determine whether the levels of LPS in the HDE would affect AHR severity. Challenge with increasing concentrations of aerosolized methacholine 4 hours post final challenge did not result in any difference in AHR between groups (Figure 6A). The 4 hour timepoint was assessed based on observations in our laboratory showing robust AHR induction at this timepoint. No increase in AHR was measured comparing normal mice and sensitized-only mice (data not shown). Given the controversy associated with non-invasive whole body plethysmography for assessing AHR, we confirmed our findings using invasive lung function tests. As shown in Figure 6B, invasive tests demonstrated no significant difference in airway resistance between groups at all methacholine concentrations tested. We also assessed BAL cysteinyl leukotriene (Cys-LT) concentrations at 2 hours as a mechanism for AHR, and found no significant change in BAL Cys-LT in mice challenged with LPS-reduced HDE compared to HDE challenge (Figure 6C).\nAirways hyperresponsiveness, airway resistance and cysteinyl leukotrienes post allergen challenge. A) For non-invasive measurement of AHR, mice were challenged with increasing concentrations of aerosolized methacholine at 4 hours after the final challenge and PenH values are represented as percent above baseline. B) For invasive measurements, mice were challenged with methacholine and pulmonary resistance is expressed as R cmH2O.s/mL. C) Cysteinyl leukotriene concentrations were measured in the BAL fluid 2 hours post final challenge. Data are represented as the mean ± SEM, in some data points the symbol is larger than the SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment.", "It has been suggested previously that the presence of TNFα can induce increased levels of Th2 cytokines in the BAL fluid of allergen challenged mice[19]. Further, asthma is largely characterized as a Th2 mediated disease, accompanied by high IL-4 and IL-5 production and low IFNγ[26]. To determine whether the reduction in TNFα seen in LPS-reduced HDE challenged mice affected cytokine production, we assayed both Th1 and Th2 cytokines 24 hours post final challenge. In conflict with the Th2 paradigm, IFNγ gamma was significantly increased in the LPS-reduced HDE mice (Figure 7A). Surprisingly, IL-5 and IL-13 were significantly increased in the BAL from LPS-reduced HDE challenged mice although there was no difference in IL-4 levels (Figure 7B, C). IL-17 has also been implicated as a key cytokine in the asthmatic response, and is also modulated by LPS levels [27,28]. However, we measured no significant differences in IL-17 levels between groups in the BAL fluid (data not shown).\nTh1 and Th2 cytokine production in the BAL fluid at 24 hours post final challenge. Cytokines were assayed by standard ELISA. Data are represented as the mean ± SEM. Experiments were done a minimum of 3 times with 3-4 mice per group, per experiment. * p < 0.05 and *** p ≤ 0.0001 comparing the LPS-reduced HDE to HDE the group by Student's t-test.", "The effect of exposure to household dust on the development of atopy and asthma has recently become a focus of investigation; however this avenue of study has resulted in more controversy than consensus. Epidemiological evidence shows that the presence of cockroach allergens in inner city homes is a strong risk factor for the development of atopic disease [29,30]. Cockroach allergen sensitization was found to be responsible for exacerbations in a large proportion of asthmatic children in inner city Washington, D.C [21]. Further, asthma exacerbations were not the result of sensitization to other allergens such as cat dander and dust mite allergens, which were also present in these homes [1,21]. Exposure to innate immune activators such as TLR ligands, as well as environmental pollutants such as diesel particulates can also drive the adaptive immune response toward tolerance or sensitization [10,31]. This area of study is gaining significant interest in light of the increasing incidence of atopy and allergy in the developed world.\nThis study sought to examine whether the presence of LPS in the household environment modifies the allergic response after sensitization to cockroach allergens has been established. It is possible that our method of LPS reduction also removed other allergens from the HDE mixture which may contribute to the allergic response. While this is a real concern, it is a shortcoming of all column purification methods. We are confident that our methods resulted in physiologically relevant data given that the LPS-reduced HDE was still able to induce a robust adaptive immune response.\nRecent studies have employed house dust extracts to determine their roles in asthma severity, and have suggested that these extracts can induce allergic phenotypes or can skew towards tolerance, depending upon the dose and timing of the exposure [31,32]. However, these studies focused on the immunomodulatory effects of the house dust extracts on mice presensitized to ovalbumin, and thus study house dust as an exacerbating factor, rather than a causative agent. In contrast, the current study directly examines the synergistic effects of CRA and LPS contained in the HDE in mice presensitized to CRA, which is the major allergen in the HDE.\nLPS contamination in allergen preparations has been raised as an issue in many models of allergy, however we believe that approaching these studies using household components with naturally occurring LPS more closely and accurately reflects the environment in which one develops asthma. It is not appropriate to refer to the HDE as \"contaminated\" with LPS, similar to how one would not say it is contaminated with cockroach allergens. In an OVA study, Watanabe et.al. demonstrated that LPS contamination of commercially available OVA preparations significantly attenuated cellular influx, AHR and IgE production [33]. The differences observed between these models might be explained by the intrinsic protease activity of the cockroach allergen Bla g2, which may enhance allergen sensitization by increasing epithelial permeability, and allergen exposure to dendritic cells [34,35]. This highlights the need to study the effect of LPS in the development of asthma-like inflammation using a relevant allergen mixture.\nIn the lung, TNFα is primarily produced by alveolar macrophages and secreted within 1 hour of LPS challenge [36,37]. Our results show that mice challenged with the LPS-reduced HDE produce significantly less TNFα compared to those challenged with the HDE, verifying that removal of LPS from the HDE has the expected immunological effect. Several studies have implicated TNFα and TNFα signaling in the development of adaptive immunity, particularly in the context of class switch to IgE production [8,9]. Therefore, we would expect that we did not see differences in CRA-specific IgE, given that our study focused on the effect of LPS at challenge only. In contrast, the studies by Eisenbarth et. al. and Watanabe et.al. focused on LPS at the time of sensitization, which likely has more effect on the production of antigen-specific antibodies [8,33]. This was also demonstrated in a rat model where aerosol exposure to LPS before or shortly after sensitization reduced the development of OVA-specific IgE [22]. Levels of CRA-specific IgG1 and IgG2a are of significant interest, however reliable assays for these antibodies are not currently available.\nThe increase in total IgE and IgG1 in the HDE-challenged mice represents an increase in polyclonal, antigen non-specific IgE. LPS alone has been shown to induce expression of activation-induced cytidine deaminase, which is required for activation of downstream signals responsible for class switch recombination to both IgG1 and IgE in B cells [38]. Other models have shown polyclonal IgE to have a role in mast cell survival and cytokine release [39].\nOur data demonstrate that airways hyperresponsiveness was not diminished in the LPS-reduced HDE challenged groups despite significantly attenuated TNFα. We postulate that AHR is governed by other mediators and is not directly induced by TNFα. The role of IgE was explored in other studies that demonstrated that AHR may be modified independent of IgE alterations. Hamelmann et. al. showed robust AHR induction in both mast cell and B cell deficient mice and it has been shown that mast cell activation is not required for the induction of AHR[40,41]. Further, methacholine acts directly upon muscarinic receptors in order to induce smooth muscle contraction. We have recently shown that pulmonary exposure to high concentrations of LPS can induce modest differences in muscarinic receptor expression[42]. Therefore, it is possible that the concentrations of LPS used in this study were not sufficient to alter muscarinic receptor expression, and therefore did not alter AHR.\nThe role of LPS in driving allergic responses toward a Th2 phenotype has been demonstrated in several models and our data show this to be true in the context of total antibody production. However, our data show increases in both Th1 and Th2 cytokines in the LPS-reduced HDE group. One possible explanation for this is that the source of these cytokines is not only the T cells present in the lung. Other cell types such as eosinophils, epithelial cells and mast cells are contributing to the pool of cytokines present in the BAL fluid [43]. Therefore, without assessing the direct cellular source of these cytokines, it is not possible to conclude whether LPS-induced Th2 cell differentiation is impaired in this model.\nInterestingly, challenge with intact HDE results in decreased BAL cytokine production. We postulate that this is modulation of the innate immune response through induction of microbial cross-tolerance. We and others have shown that cross-tolerance to TLR ligands can be induced in the lung and in pulmonary macrophages [44,45]. It is likely that simultaneous exposure to various TLR ligands present in the HDE can induce cross-tolerance and therefore attenuate the cytokine response. Removal of the potent TLR4 agonist LPS, may reduce the tolerogenic response, driving increased cytokine production.\nLimited epidemiological data exist investigating whether removal of house dust can limit future asthma exacerbations. One such study carried out in the UK found no correlation between removal of house dust and amelioration of asthma symptoms in presensitized individuals. However, interventions to reduce dust in homes of children who were highly susceptible to developing atopy was effective in reducing asthma diagnosis [46,47]. Taken together our data show that in a clinically relevant model of house dust induced asthma-like inflammation, the removal of only LPS at challenge does not protect against the development of the inflammatory response. Further studies will be needed to clarify whether LPS in the home environment exacerbates asthma in presensitized individuals.", "Decreasing the endotoxin content endogenously present in cockroach allergens results in a complicated pulmonary inflammatory response. While there was differential regulation of cytokine, chemokines and plasma levels of immunoglobulins, there was no real change in AHR or pulmonary inflammatory cell recruitment.", "The authors declare that they have no competing interests.", "SN designed and executed the experiments and wrote the manuscript. JB and JK assisted with conducting the experiments, interpreting the data and reviewing the manuscript. WC assisted with the Flexivent measurements and reviewed the manuscript. DR assumed overall responsibility for the experimental design and manuscript preparation. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2466/11/12/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "ABO Blood-Group System", "Blood Cell Count", "Case-Control Studies", "Enzyme-Linked Immunosorbent Assay", "Female", "Ferritins", "Glucosephosphate Dehydrogenase", "Hemoglobin, Sickle", "Humans", "Iron Deficiencies", "Malaria, Falciparum", "Malawi", "Polymorphism, Genetic", "Pregnancy", "Pregnancy Complications, Infectious", "Receptors, Transferrin", "Risk Assessment" ]

[ "Background", "Study sample", "Study procedures", "Results", "Discussion", "Competing interests", "Authors' contributions" ]

[ "The influence of iron deficiency on infection risk has been a long-standing clinical concern [1], related to the concept that restriction of availability of iron for growth of pathogens will inhibit pathogen proliferation, and conversely making iron available would increase infection risk [2]. A recent Cochrane analysis of 68 trials of iron supplementation observed an increased risk of malaria with iron in trials that did not provide malaria surveillance and treatment [3]. The Cochrane review did not include studies of pregnant women, yet there are 85.3 million pregnancies in areas with Plasmodium falciparum transmission, 54.7 million of which occur in areas with stable transmission and 30.6 million in areas with unstable transmission [4].\nIron-malaria interactions in young non-pregnant women also have been little studied and it is unknown whether improving iron status of young women prior to pregnancy living in malaria endemic areas would increase their risk for pregnancy related malaria. Weekly iron and folic acid supplementation for all women of child-bearing age also has been recommended by the World Health Organization [5], although the safety of this strategy in terms of malaria risk has not been assessed.\nOnly five published studies on the effects of iron treatments or status during pregnancy on the prevalence of maternal malaria have been identified[6-10]. One showed an association of recent haematinic use with increased risk of Plasmodium vivax infection which was not related to parity [6], another no association of haematinics with placental P. falciparum infection in multigravidae with HbAS genotype [7]. Two uncontrolled studies reported parenteral iron given to severely anaemic women was associated with higher risk of P. falciparum malaria at delivery [8,9], one of which reported the association in primiparae, but not multiparae [9] As these women receiving parenteral iron all had severe antenatal anaemia, they would be the more likely to have malaria at the time of treatment and delivery independent of iron prescription. A single cross-sectional study at delivery reported iron deficiency was associated with decreased the risk of placental parasitaemia especially in the first pregnancy [10].\nMultigravidae have reduced prevalence of P. falciparum infection, which is attributed to parity specific immunity to malaria [11], although as they have higher risk of iron deficiency due to the cumulative iron needs of successive pregnancies, then their iron status could influence their immunity to malaria. Other factors, including ABO blood group phenotype [12] and adolescence, are also important for defining malaria risks in pregnancy [13]. The rationale for the present study was to determine the association of iron deficiency at delivery with placental malaria controlling for these factors, as well as using a histological classification of placental malaria which characterizes acute, chronic, and past placental infections.", "This was a case control study conducted between January and July 2005 at Montfort Hospital, in Southern Malawi amongst all pregnant women who attended the hospital for delivery and who provided informed consent. Malaria was highly endemic in this area and transmission occurred year round but intensifies during the rainy season from April to December. A total of 647 women were enrolled who attended the hospital for delivery. Women were approached during the first stage of labour and informed consent was requested for participation. A questionnaire was completed by a trained midwife requesting information on age, parity, literacy, obstetric history, bed net and anti-malarial use. Women with emergency obstetric problems (toxaemia, haemorrhage, obstructed labour or sepsis), or with recent blood transfusion were excluded. Maternal mid-upper-arm-circumference (MUAC) was measured at the mid-point of the left arm.\nFollowing delivery of placenta a tissue sample was obtained by cutting a one cm cube of tissue from both the central and peripheral cotyledons. These were fixed in 10% neutral buffered formalin. Biopsies were embedded in paraffin wax, sliced into thin sections and stained with Giemsa and/or haematoxylin and eosin. Slides were examined at the Histopathology Department, College of Medicine, Blantyre, by two observers, with discrepant findings assessed by a pathologist. Cases were women with placental malaria, and controls those with no evidence of infection. Placental malaria was histologically classified as acute, based on the presence of parasites, chronic, based on the presence of parasites and haemozoin, and past infection if haemozoin alone was present. Active infection included both acute and chronic categories [14].", "A maternal peripheral blood sample was obtained after delivery and placed in a EDTA tube. Thick and thin blood smears were stained with Giemsa and screened for malaria parasites. A full blood count was performed by an automated Beckman Coulter AcT 8 S/N. AA322060 (http://www.coulter.com). Ferritin and sTfR were quantified by enzyme-linked immunoassays (http://www.omegadiagnostics.co.uk;http://www.oriondiagnostica.fi). Iron deficiency was defined as a sTfR:log ferritin ratio > 1.6, as this cut-off has been shown to best predict iron deficiency based on bone marrow iron stores in Malawi, an area with high infection pressure [15]. DNA analysis was used to identify maternal genetic polymorphisms: ABO blood group, sickle cell polymorphism (HbS), and glucose -6-phosphate dehydrogenase deficiency (G6PD). Polymerase chain reactions (PCRs) were analysed by restriction fragment length polymorphism technique using restriction enzymes NlaIII and DdeI for G6PD and HbS respectively. All RFLPs were electrophoresed for 45 minutes on 2.5% agarose gel stained with ethidium bromide and visualized with a UVP GelDoc - It imaging system, USA.\nMaternal parity was included in the analysis as a known correlate of placental malaria risk. Adolescence was defined as < 20 years. Chi-square, Mann-Whitney and student t tests were used for comparisons using SPSS version 18. Multivariate logistic regression was used to describe factors associated with placental malaria.", "Between February-June 2004 and January-July 2005 a total of 112 infected cases were identified who were compared with the first 110 women identified with no evidence of placental infection. Over 95% of women had received one or more doses of sulphadoxine-pyrimethamine as intermittent treatment (IPTp-SP). There were no significant differences between cases and controls for the following parameters: mean age and parity, proportion with low mid-upper arm circumference (< 23 cm), marital or literacy status, adolescents, frequency of use of insecticide treated bed nets or uptake of IPTp-SP during pregnancy (all p > 0.1). 44.1% of cases and 42.5% of controls were anaemic (Hb < 11g/dl), There were no differences between cases and controls in prevalence of severe anaemia (Hb < 8 g/dl, mean 5.9%), G6PD deficiency (mean 18.7%), or sickle cell trait (mean 44.5%), (all p > 0.3).\nBlood group O phenotype was less frequent in multigravidae with active (35.4%, p = 0.037), or acute infection (34.8%, p = 0.03) than controls (54.8%), while the converse was true for primigravidae (61.5%, p = 0.06; and 63.2%, p = 0.09; versus 37.1%). Sera was available for iron biomarkers for 92 cases and 68 controls. Of cases 57.6% had acute, 9.8% chronic, and 32.6% past infection. Women with either acute (p = 0.001), or acute and chronic (p = 0.006) placental malaria were less likely to have iron deficiency than women without placental malaria infection (Table 1). For all infection categories (acute, chronic, past) the odds ratio for iron deficiency was 0.4 (95% CI 0.2-0.8, p = 0.01). Iron deficient multigravidae, but not primigravidae, were less likely to have placental malaria (all infection categories, p = 0.002) (Table 1).\nRisk of iron deficiency in cases and controls\nAll infected = acute, chronic and past infection\nOR: odds ratio for infected category versus non-infected\nThree models were used for the regression analysis, each for a placental malaria category (all categories, active, or past). Factors in the model were those with P < 0.1 in univariate analysis. These were iron deficiency, parity, ABO phenotype and adolescence. Risk of iron deficiency was reduced for combined infection categories (p = 0.008), active infection (p = 0.007), or acute infection (p = 0.001)(Table 2). In multigravidae there was reduced risk of malaria infection for all infection categories (p=0.03). Active infection was increased in adolescents (p=0.04).\nAdjusted odds ratios for factors related to placental malaria\nBrackets: 95% confidence interval\nReference: no placental infection (Controls)\nAll infected: acute, chronic and past infection", "Iron deficiency was less frequent in women with evidence of acute or chronic placental malarial infection indicating that women with placental infection have better iron status than those without. There was no difference in use of IPTp-SP during pregnancy between cases and controls, and the association remained highly significant controlling for maternal ABO phenotype.\nThe importance of these results relates to the influence which maternal iron status has on parity specific malaria immunity, especially as the results were most significant for multigravidae. The cumulative iron requirements of successive pregnancies would increase prevalence of iron deficiency leading to reduced risk for placental malaria in multigravidae. As there is a need to supplement these women with iron this is likely to increase malaria risk despite acquisition of parity specific malaria immunity. These results support the findings from a study in Tanzania based on intervillous blood smears although that study reported malaria was less prevalent among women with iron deficiency especially during the first pregnancy [9]. It will be important to establish the relative importance of differences in iron effects between parity groups in a much larger study in order to determine confounding effects on parity specific malarial immunity, and in relation to differential effects on malaria risk of iron supplementation in pregnancy.\nThe calculation of body iron store based on the sTfR:SF ratio provides the best estimate of iron status in adult men and non-pregnant women in populations where infectious disorders are not prevalent. Its utility where women are living under endemic conditions for malaria is less clear and in severely anaemic HIV infected adults had low sensitivity based on bone marrow iron stores [16]. It has good sensitivity and specificity (> 70%) for reduced bone marrow iron stores in Malawian children with severe anaemia (Hb < 5g/dl) irrespective of the presence of infection [15]. The prevalence of iron deficiency in all women was not high (47.8%), although similar to a previous estimate from this area [17]. Other causes for anaemia are likely to be important including vitamin A and B12 deficiencies.\nFour hypotheses might explain these effects of iron status in the context of malaria. These are: the availability of non-transferrin bound iron which might influence the growth of the malaria parasite which may be an acute effect related to recent iron intake or supplementation; an effect on the oxidant/anti-oxidant balance and oxidative free radical reactions in the placental intervillous blood which may alter cellular immune function and macrophage iron metabolism; altered expression of vascular endothelial molecules which may influence adhesion and sequestration of P.falciparum parasites; and altered erythropoiesis with down regulation of parasite invasion of young red cells.\nAll these women received haematinics of iron (60 mg) and folic acid (5 mg) in short daily courses mostly in the second half of pregnancy. Women were not routinely screened for anaemia at first antenatal attendance or during pregnancy and there was no preferential prescription of haematinics for selected women with anaemia. This is important because women with malarial anaemia could receive therapeutic iron supplements for anaemia treatment, reducing iron deficiency, and then independent of their iron status, these same women may be more likely to have placental malaria due to a higher malaria exposure risk. Although the possibility that some women received additional iron treatments cannot be excluded, the number was nevertheless likely to be small and the policy was to uniformly prescribe haematinics to all women.\nThis study needs to be replicated in women living under different levels of malaria transmission in order to determine the magnitude of these effects in subjects with different levels of malaria immunity. The small sample size and lack of available sera from all subjects limited detailed sub-analyses, for example, placental malaria iron interactions in younger adolescents less than 16 years. A large multi-centre case control study is required to adequately address this.\nAs many millions of pregnant women are exposed to malaria annually there is a priority to establish if reversing such iron deficiency states through routine supplementation programs either prior to or during pregnancy enhances malaria risk especially in areas where malaria control and surveillance is limited. It is particularly important to assess this in adolescents as their malaria immunity is lower than that in older women and many will experience pregnancy while still adolescent.", "The authors declare that they have no competing interests.", "ES carried out the field study, participated in the analysis and preparation of the manuscript; GH co-ordinated and participated in the laboratory analyses; GK supported the data preparation and analysis; PK participated in the study design and co-ordination and helped to draft the manuscript; BB conceived the study and participated in its design, co-ordination, analysis and helped draft the manuscript. All authors have read and approved the final manuscript." ]

[ null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study sample", "Study procedures", "Results", "Discussion", "Competing interests", "Authors' contributions" ]

[ "The influence of iron deficiency on infection risk has been a long-standing clinical concern [1], related to the concept that restriction of availability of iron for growth of pathogens will inhibit pathogen proliferation, and conversely making iron available would increase infection risk [2]. A recent Cochrane analysis of 68 trials of iron supplementation observed an increased risk of malaria with iron in trials that did not provide malaria surveillance and treatment [3]. The Cochrane review did not include studies of pregnant women, yet there are 85.3 million pregnancies in areas with Plasmodium falciparum transmission, 54.7 million of which occur in areas with stable transmission and 30.6 million in areas with unstable transmission [4].\nIron-malaria interactions in young non-pregnant women also have been little studied and it is unknown whether improving iron status of young women prior to pregnancy living in malaria endemic areas would increase their risk for pregnancy related malaria. Weekly iron and folic acid supplementation for all women of child-bearing age also has been recommended by the World Health Organization [5], although the safety of this strategy in terms of malaria risk has not been assessed.\nOnly five published studies on the effects of iron treatments or status during pregnancy on the prevalence of maternal malaria have been identified[6-10]. One showed an association of recent haematinic use with increased risk of Plasmodium vivax infection which was not related to parity [6], another no association of haematinics with placental P. falciparum infection in multigravidae with HbAS genotype [7]. Two uncontrolled studies reported parenteral iron given to severely anaemic women was associated with higher risk of P. falciparum malaria at delivery [8,9], one of which reported the association in primiparae, but not multiparae [9] As these women receiving parenteral iron all had severe antenatal anaemia, they would be the more likely to have malaria at the time of treatment and delivery independent of iron prescription. A single cross-sectional study at delivery reported iron deficiency was associated with decreased the risk of placental parasitaemia especially in the first pregnancy [10].\nMultigravidae have reduced prevalence of P. falciparum infection, which is attributed to parity specific immunity to malaria [11], although as they have higher risk of iron deficiency due to the cumulative iron needs of successive pregnancies, then their iron status could influence their immunity to malaria. Other factors, including ABO blood group phenotype [12] and adolescence, are also important for defining malaria risks in pregnancy [13]. The rationale for the present study was to determine the association of iron deficiency at delivery with placental malaria controlling for these factors, as well as using a histological classification of placental malaria which characterizes acute, chronic, and past placental infections.", "[SUBTITLE] Study sample [SUBSECTION] This was a case control study conducted between January and July 2005 at Montfort Hospital, in Southern Malawi amongst all pregnant women who attended the hospital for delivery and who provided informed consent. Malaria was highly endemic in this area and transmission occurred year round but intensifies during the rainy season from April to December. A total of 647 women were enrolled who attended the hospital for delivery. Women were approached during the first stage of labour and informed consent was requested for participation. A questionnaire was completed by a trained midwife requesting information on age, parity, literacy, obstetric history, bed net and anti-malarial use. Women with emergency obstetric problems (toxaemia, haemorrhage, obstructed labour or sepsis), or with recent blood transfusion were excluded. Maternal mid-upper-arm-circumference (MUAC) was measured at the mid-point of the left arm.\nFollowing delivery of placenta a tissue sample was obtained by cutting a one cm cube of tissue from both the central and peripheral cotyledons. These were fixed in 10% neutral buffered formalin. Biopsies were embedded in paraffin wax, sliced into thin sections and stained with Giemsa and/or haematoxylin and eosin. Slides were examined at the Histopathology Department, College of Medicine, Blantyre, by two observers, with discrepant findings assessed by a pathologist. Cases were women with placental malaria, and controls those with no evidence of infection. Placental malaria was histologically classified as acute, based on the presence of parasites, chronic, based on the presence of parasites and haemozoin, and past infection if haemozoin alone was present. Active infection included both acute and chronic categories [14].\nThis was a case control study conducted between January and July 2005 at Montfort Hospital, in Southern Malawi amongst all pregnant women who attended the hospital for delivery and who provided informed consent. Malaria was highly endemic in this area and transmission occurred year round but intensifies during the rainy season from April to December. A total of 647 women were enrolled who attended the hospital for delivery. Women were approached during the first stage of labour and informed consent was requested for participation. A questionnaire was completed by a trained midwife requesting information on age, parity, literacy, obstetric history, bed net and anti-malarial use. Women with emergency obstetric problems (toxaemia, haemorrhage, obstructed labour or sepsis), or with recent blood transfusion were excluded. Maternal mid-upper-arm-circumference (MUAC) was measured at the mid-point of the left arm.\nFollowing delivery of placenta a tissue sample was obtained by cutting a one cm cube of tissue from both the central and peripheral cotyledons. These were fixed in 10% neutral buffered formalin. Biopsies were embedded in paraffin wax, sliced into thin sections and stained with Giemsa and/or haematoxylin and eosin. Slides were examined at the Histopathology Department, College of Medicine, Blantyre, by two observers, with discrepant findings assessed by a pathologist. Cases were women with placental malaria, and controls those with no evidence of infection. Placental malaria was histologically classified as acute, based on the presence of parasites, chronic, based on the presence of parasites and haemozoin, and past infection if haemozoin alone was present. Active infection included both acute and chronic categories [14].\n[SUBTITLE] Study procedures [SUBSECTION] A maternal peripheral blood sample was obtained after delivery and placed in a EDTA tube. Thick and thin blood smears were stained with Giemsa and screened for malaria parasites. A full blood count was performed by an automated Beckman Coulter AcT 8 S/N. AA322060 (http://www.coulter.com). Ferritin and sTfR were quantified by enzyme-linked immunoassays (http://www.omegadiagnostics.co.uk;http://www.oriondiagnostica.fi). Iron deficiency was defined as a sTfR:log ferritin ratio > 1.6, as this cut-off has been shown to best predict iron deficiency based on bone marrow iron stores in Malawi, an area with high infection pressure [15]. DNA analysis was used to identify maternal genetic polymorphisms: ABO blood group, sickle cell polymorphism (HbS), and glucose -6-phosphate dehydrogenase deficiency (G6PD). Polymerase chain reactions (PCRs) were analysed by restriction fragment length polymorphism technique using restriction enzymes NlaIII and DdeI for G6PD and HbS respectively. All RFLPs were electrophoresed for 45 minutes on 2.5% agarose gel stained with ethidium bromide and visualized with a UVP GelDoc - It imaging system, USA.\nMaternal parity was included in the analysis as a known correlate of placental malaria risk. Adolescence was defined as < 20 years. Chi-square, Mann-Whitney and student t tests were used for comparisons using SPSS version 18. Multivariate logistic regression was used to describe factors associated with placental malaria.\nA maternal peripheral blood sample was obtained after delivery and placed in a EDTA tube. Thick and thin blood smears were stained with Giemsa and screened for malaria parasites. A full blood count was performed by an automated Beckman Coulter AcT 8 S/N. AA322060 (http://www.coulter.com). Ferritin and sTfR were quantified by enzyme-linked immunoassays (http://www.omegadiagnostics.co.uk;http://www.oriondiagnostica.fi). Iron deficiency was defined as a sTfR:log ferritin ratio > 1.6, as this cut-off has been shown to best predict iron deficiency based on bone marrow iron stores in Malawi, an area with high infection pressure [15]. DNA analysis was used to identify maternal genetic polymorphisms: ABO blood group, sickle cell polymorphism (HbS), and glucose -6-phosphate dehydrogenase deficiency (G6PD). Polymerase chain reactions (PCRs) were analysed by restriction fragment length polymorphism technique using restriction enzymes NlaIII and DdeI for G6PD and HbS respectively. All RFLPs were electrophoresed for 45 minutes on 2.5% agarose gel stained with ethidium bromide and visualized with a UVP GelDoc - It imaging system, USA.\nMaternal parity was included in the analysis as a known correlate of placental malaria risk. Adolescence was defined as < 20 years. Chi-square, Mann-Whitney and student t tests were used for comparisons using SPSS version 18. Multivariate logistic regression was used to describe factors associated with placental malaria.", "This was a case control study conducted between January and July 2005 at Montfort Hospital, in Southern Malawi amongst all pregnant women who attended the hospital for delivery and who provided informed consent. Malaria was highly endemic in this area and transmission occurred year round but intensifies during the rainy season from April to December. A total of 647 women were enrolled who attended the hospital for delivery. Women were approached during the first stage of labour and informed consent was requested for participation. A questionnaire was completed by a trained midwife requesting information on age, parity, literacy, obstetric history, bed net and anti-malarial use. Women with emergency obstetric problems (toxaemia, haemorrhage, obstructed labour or sepsis), or with recent blood transfusion were excluded. Maternal mid-upper-arm-circumference (MUAC) was measured at the mid-point of the left arm.\nFollowing delivery of placenta a tissue sample was obtained by cutting a one cm cube of tissue from both the central and peripheral cotyledons. These were fixed in 10% neutral buffered formalin. Biopsies were embedded in paraffin wax, sliced into thin sections and stained with Giemsa and/or haematoxylin and eosin. Slides were examined at the Histopathology Department, College of Medicine, Blantyre, by two observers, with discrepant findings assessed by a pathologist. Cases were women with placental malaria, and controls those with no evidence of infection. Placental malaria was histologically classified as acute, based on the presence of parasites, chronic, based on the presence of parasites and haemozoin, and past infection if haemozoin alone was present. Active infection included both acute and chronic categories [14].", "A maternal peripheral blood sample was obtained after delivery and placed in a EDTA tube. Thick and thin blood smears were stained with Giemsa and screened for malaria parasites. A full blood count was performed by an automated Beckman Coulter AcT 8 S/N. AA322060 (http://www.coulter.com). Ferritin and sTfR were quantified by enzyme-linked immunoassays (http://www.omegadiagnostics.co.uk;http://www.oriondiagnostica.fi). Iron deficiency was defined as a sTfR:log ferritin ratio > 1.6, as this cut-off has been shown to best predict iron deficiency based on bone marrow iron stores in Malawi, an area with high infection pressure [15]. DNA analysis was used to identify maternal genetic polymorphisms: ABO blood group, sickle cell polymorphism (HbS), and glucose -6-phosphate dehydrogenase deficiency (G6PD). Polymerase chain reactions (PCRs) were analysed by restriction fragment length polymorphism technique using restriction enzymes NlaIII and DdeI for G6PD and HbS respectively. All RFLPs were electrophoresed for 45 minutes on 2.5% agarose gel stained with ethidium bromide and visualized with a UVP GelDoc - It imaging system, USA.\nMaternal parity was included in the analysis as a known correlate of placental malaria risk. Adolescence was defined as < 20 years. Chi-square, Mann-Whitney and student t tests were used for comparisons using SPSS version 18. Multivariate logistic regression was used to describe factors associated with placental malaria.", "Between February-June 2004 and January-July 2005 a total of 112 infected cases were identified who were compared with the first 110 women identified with no evidence of placental infection. Over 95% of women had received one or more doses of sulphadoxine-pyrimethamine as intermittent treatment (IPTp-SP). There were no significant differences between cases and controls for the following parameters: mean age and parity, proportion with low mid-upper arm circumference (< 23 cm), marital or literacy status, adolescents, frequency of use of insecticide treated bed nets or uptake of IPTp-SP during pregnancy (all p > 0.1). 44.1% of cases and 42.5% of controls were anaemic (Hb < 11g/dl), There were no differences between cases and controls in prevalence of severe anaemia (Hb < 8 g/dl, mean 5.9%), G6PD deficiency (mean 18.7%), or sickle cell trait (mean 44.5%), (all p > 0.3).\nBlood group O phenotype was less frequent in multigravidae with active (35.4%, p = 0.037), or acute infection (34.8%, p = 0.03) than controls (54.8%), while the converse was true for primigravidae (61.5%, p = 0.06; and 63.2%, p = 0.09; versus 37.1%). Sera was available for iron biomarkers for 92 cases and 68 controls. Of cases 57.6% had acute, 9.8% chronic, and 32.6% past infection. Women with either acute (p = 0.001), or acute and chronic (p = 0.006) placental malaria were less likely to have iron deficiency than women without placental malaria infection (Table 1). For all infection categories (acute, chronic, past) the odds ratio for iron deficiency was 0.4 (95% CI 0.2-0.8, p = 0.01). Iron deficient multigravidae, but not primigravidae, were less likely to have placental malaria (all infection categories, p = 0.002) (Table 1).\nRisk of iron deficiency in cases and controls\nAll infected = acute, chronic and past infection\nOR: odds ratio for infected category versus non-infected\nThree models were used for the regression analysis, each for a placental malaria category (all categories, active, or past). Factors in the model were those with P < 0.1 in univariate analysis. These were iron deficiency, parity, ABO phenotype and adolescence. Risk of iron deficiency was reduced for combined infection categories (p = 0.008), active infection (p = 0.007), or acute infection (p = 0.001)(Table 2). In multigravidae there was reduced risk of malaria infection for all infection categories (p=0.03). Active infection was increased in adolescents (p=0.04).\nAdjusted odds ratios for factors related to placental malaria\nBrackets: 95% confidence interval\nReference: no placental infection (Controls)\nAll infected: acute, chronic and past infection", "Iron deficiency was less frequent in women with evidence of acute or chronic placental malarial infection indicating that women with placental infection have better iron status than those without. There was no difference in use of IPTp-SP during pregnancy between cases and controls, and the association remained highly significant controlling for maternal ABO phenotype.\nThe importance of these results relates to the influence which maternal iron status has on parity specific malaria immunity, especially as the results were most significant for multigravidae. The cumulative iron requirements of successive pregnancies would increase prevalence of iron deficiency leading to reduced risk for placental malaria in multigravidae. As there is a need to supplement these women with iron this is likely to increase malaria risk despite acquisition of parity specific malaria immunity. These results support the findings from a study in Tanzania based on intervillous blood smears although that study reported malaria was less prevalent among women with iron deficiency especially during the first pregnancy [9]. It will be important to establish the relative importance of differences in iron effects between parity groups in a much larger study in order to determine confounding effects on parity specific malarial immunity, and in relation to differential effects on malaria risk of iron supplementation in pregnancy.\nThe calculation of body iron store based on the sTfR:SF ratio provides the best estimate of iron status in adult men and non-pregnant women in populations where infectious disorders are not prevalent. Its utility where women are living under endemic conditions for malaria is less clear and in severely anaemic HIV infected adults had low sensitivity based on bone marrow iron stores [16]. It has good sensitivity and specificity (> 70%) for reduced bone marrow iron stores in Malawian children with severe anaemia (Hb < 5g/dl) irrespective of the presence of infection [15]. The prevalence of iron deficiency in all women was not high (47.8%), although similar to a previous estimate from this area [17]. Other causes for anaemia are likely to be important including vitamin A and B12 deficiencies.\nFour hypotheses might explain these effects of iron status in the context of malaria. These are: the availability of non-transferrin bound iron which might influence the growth of the malaria parasite which may be an acute effect related to recent iron intake or supplementation; an effect on the oxidant/anti-oxidant balance and oxidative free radical reactions in the placental intervillous blood which may alter cellular immune function and macrophage iron metabolism; altered expression of vascular endothelial molecules which may influence adhesion and sequestration of P.falciparum parasites; and altered erythropoiesis with down regulation of parasite invasion of young red cells.\nAll these women received haematinics of iron (60 mg) and folic acid (5 mg) in short daily courses mostly in the second half of pregnancy. Women were not routinely screened for anaemia at first antenatal attendance or during pregnancy and there was no preferential prescription of haematinics for selected women with anaemia. This is important because women with malarial anaemia could receive therapeutic iron supplements for anaemia treatment, reducing iron deficiency, and then independent of their iron status, these same women may be more likely to have placental malaria due to a higher malaria exposure risk. Although the possibility that some women received additional iron treatments cannot be excluded, the number was nevertheless likely to be small and the policy was to uniformly prescribe haematinics to all women.\nThis study needs to be replicated in women living under different levels of malaria transmission in order to determine the magnitude of these effects in subjects with different levels of malaria immunity. The small sample size and lack of available sera from all subjects limited detailed sub-analyses, for example, placental malaria iron interactions in younger adolescents less than 16 years. A large multi-centre case control study is required to adequately address this.\nAs many millions of pregnant women are exposed to malaria annually there is a priority to establish if reversing such iron deficiency states through routine supplementation programs either prior to or during pregnancy enhances malaria risk especially in areas where malaria control and surveillance is limited. It is particularly important to assess this in adolescents as their malaria immunity is lower than that in older women and many will experience pregnancy while still adolescent.", "The authors declare that they have no competing interests.", "ES carried out the field study, participated in the analysis and preparation of the manuscript; GH co-ordinated and participated in the laboratory analyses; GK supported the data preparation and analysis; PK participated in the study design and co-ordination and helped to draft the manuscript; BB conceived the study and participated in its design, co-ordination, analysis and helped draft the manuscript. All authors have read and approved the final manuscript." ]

[ null, "methods", null, null, null, null, null, null ]

[]

[ "Adolescent", "Adult", "Aged", "Anti-Bacterial Agents", "Bacterial Typing Techniques", "Child", "Child, Preschool", "Deoxyribonucleases, Type II Site-Specific", "Electrophoresis, Gel, Pulsed-Field", "Female", "Humans", "Infant", "Male", "Microbial Sensitivity Tests", "Middle Aged", "Molecular Typing", "Salmonella Infections", "Salmonella enterica", "Serotyping", "Switzerland", "Young Adult" ]

[ "Background", "Strains", "Antimicrobial susceptibility testing", "Genotyping", "Results", "Clinical and epidemiological data", "Antimicrobial susceptibility", "Genetic relationship among Swiss S. Virchow isolates", "Discussion", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Non-typhoidal salmonellae are worldwide responsible for numerous foodborne infections and are a major threat to public health [1,2]. Based on the lipopolysaccharide (O antigen) and the flagellar structures (H antigen) Salmonella spp. are divided into greater than 2500 serotypes, of which only few obtain substantial relevance to humans [3]. In Europe, as well as in Switzerland, Salmonella serotype Virchow ranks among the five most frequent serovars and has been reported to be poultry associated [4-9]. As with other enteritic salmonellae, infections due to S. Virchow commonly manifest themselves as self-limiting gastroenteritis, but severe invasive infections can also occur which then need antibiotic treatment [10]. Whereas fluoroquinolones are used to treat extraintestinal infections in adults, third generation cephalosporins are drugs of choice in children. Alarmingly, in the last decade numerous European countries reported emergence of S. Virchow strains resistant to the above mentioned antimicrobials, leading to treatment failures [9,11,12]. Resistance to quinolones is either mediated by chromosomal mutations in genes encoding DNA gyrase or topoisomerase IV or in genes affecting the uptake or efflux of drugs [13]. Plasmid associated quinolone corrupting genes such as aac(6')-Ib-cr or qnr etc., are also known to cause reduced susceptibility. Resistance to third generation cephalosporins is mediated by extended spectrum β-lactamases (ESBL), which are plasmid encoded or integron associated [14].\nIn this study, a total of 153 S. Virchow strains isolated from different patients from 2004 through 2009 in Switzerland were further characterized by (i) assessing phenotypic antibiotic resistance profiles using the disk diffusion method and (ii) by genotyping using pulsed-field gel electrophoresis (PFGE) after macrorestriction with XbaI in order to evaluate strain relationship.", "Human salmonellosis is a reportable disease in Switzerland and isolates have to be submitted to the National Centre for Enteropathogenic Bacteria (NENT). The NENT - in cooperation with the Swiss Federal Office of Public Health - collects clinical data and carries out final serological identification by slide agglutination with commercial antisera according to the Kauffmann-White scheme. Over the years 2004 through 2009, 10395 Salmonella strains were submitted to the NENT. From this collection of strains 153 S. Virchow strains (1,5%) from different patients (multiple isolates from same patient excluded, but family members not excluded) were selected and integrated in this study. All samples had been collected by hospitals or family doctors.", "The strains were tested for antimicrobial resistance by the disk diffusion method according to the Clinical and Laboratory Standards Institute (CLSI) [15]. The panel of antibiotic disks (Becton, Dickinson and Company, Maryland, USA) consisted of ampicillin (AM10), amoxicillin/clavulanic acid (AMC30), cephalothin (CF30), cefotaxime (CTX30), ciprofloxacin (CIP5), gentamicin (GM10), tetracycline (Te30), streptomycin (S10), chloramphenicol (C30), kanamycin (K30), nalidixic acid (NA30), sulfamethoxazole (SMZ), and trimethoprim (TMP5). The strains were classified as resistant, intermediate or susceptible to each antibiotic agent according to the CLSI criteria [15].\nFor presumptive ESBL producers - strains that produced synergistic inhibition zone enlargements between adjacent discs containing an oxyimino cephalosporin and clavulanic acid, respectively - Etest ESBL (bioMérieux, Marcy l'Etoile, France) tests, containing cefotaxime (CT/CTL), ceftazidime (TZ/TZL), or cefepime (PM/PML) alone and in combination with clavulanic acid, were performed for confirmation according the manufacturer's guidelines.", "Pulsed-field gel electrophoresis (PFGE) was performed by following the CDC PulseNet protocol http://www.cdc.gov/pulsenet/protocols.htm with minor modifications. In brief, strains were grown on blood agar at 37°C over night. Colonies from blood agar were resuspended in cell suspension buffer (OD600 = 1). The bacterial cell suspension was mixed with 400 μl of 1.4% Pulsed Field Certified Agarose (BIO-RAD, Munich, Germany) and cells were lysed by proteinase K treatment over night. After lysis the plugs were washed twice for 30 min in ultrapure water and 4 times for an hour in Tris-EDTA (TE) buffer. After washing with TE buffer, DNA agarose plugs were incubated over night in the presence of XbaI (Roche, Mannheim, Germany) following the manufacturer's instructions. Restricted DNA in plug slices was separated in a 1% SeaKem Gold (BioConcept, Allschwil, Switzerland) agarose gel at 6 V/cm in 0.5 M Tris-Borate-EDTA buffer cooled to 12°C in a CHEF-DR III system (BIO-RAD, Munich, Germany). The pulse times were ramped from 2 to 64 sec for 20 h at an angle of 120°. Gels were stained with ethidium bromide and visualized under UV light transillumination with Gel Doc (BIO-RAD, Munich, Germany) and analysed with BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium). Pair wise similarities between the XbaI pulsed-field patterns were calculated by the DICE's similarity coefficient. Clustering was based on the unweighted pair-group method with averages (UPGMA), setting tolerance and optimization each at 1.5%. We used Salmonella Braenderup strain H9812 (ATCC BAA 664) as a reference strain.", "[SUBTITLE] Clinical and epidemiological data [SUBSECTION] The 153 S. Virchow strains were isolated from 124 faecal samples (81%), 17 blood samples (11.1%), two urine samples (1.3%), two swabs (1.3%), and one synovial puncture (0.7%). For 7 strains (4.6%) the origin was unknown. The relative frequency of S. Virchow among other Salmonella serovars varied between 4th to 8th rank. The annual incidence ranged from 0.45/100'000 in 2004 to 0.40/100'000 in 2009 (lowest incidence: 0.39/100'000 in 2006, highest incidence: 0.47/100'000 in 2007).\nThe age of the patients was known for 99% of the isolates submitted. Fourteen percent of the S. Virchow strains were submitted from patients aged 5 years and younger (incidence: 0.9/100'000), 3.5% from patients aged 6 to 14 years (incidence: 0.13/100'000), 74.5% from patients aged 15 to 65 years (incidence: 0.4/100'000) and 7% from patients older than 65 years (incidence: 0.17/100'000). The incidence among infants of 5 years and younger was seven fold higher than that among children aged 6 to 14 years. The incidence of people between 15 and 65 years was three fold higher than that of children aged 6 to 14 years. Of the patients infected with S. Virchow 52% were female and 48% were male.\nOf all patients, 35 (23%) had travelled abroad, the travelling status was unknown for 117 patients (76%), and only one person confirmed not having visited a foreign country within two weeks prior to infection. Eleven persons (7.2%) had visited Egypt, 7 (4.6%) Thailand, 3 (2%) India, 2 (1.3%) Kenya, 2 (1.3%) Asia. The following countries had been visited by one person each (0.7%): Guinea, Jordan, Senegal/Gambia, China, Sri Lanka and Tanzania. Four patients (2.6%) affirmed having visited a foreign country within two weeks prior to infection start, but gave no further information about the destination. As many as 82.5% who had travelled to a foreign country within two weeks before disease outbreak, were between 15 and 65 years old, 7.5% were younger than 6 years, another 7.5% were aged from 6 to 14 years, and 2.5% were older than 65. Subtracting the number of patients with travel background from the total of the patients lowered the average incidence from 0.42/100'000 to 0.34/100'000. Turning the attention on the 6 to 14 year old children, the incidence reduced even from 0.13/100'000 to 0.07/100'000 by subtracting those who stayed abroad. Regarding the 15 to 65 year old patients, the incidence showed similar behaviour and decreased from 0.41/100'000 to 0.3/100'000 while the incidence of other age-groups exhibited only minor changes.\nThe 153 S. Virchow strains were isolated from 124 faecal samples (81%), 17 blood samples (11.1%), two urine samples (1.3%), two swabs (1.3%), and one synovial puncture (0.7%). For 7 strains (4.6%) the origin was unknown. The relative frequency of S. Virchow among other Salmonella serovars varied between 4th to 8th rank. The annual incidence ranged from 0.45/100'000 in 2004 to 0.40/100'000 in 2009 (lowest incidence: 0.39/100'000 in 2006, highest incidence: 0.47/100'000 in 2007).\nThe age of the patients was known for 99% of the isolates submitted. Fourteen percent of the S. Virchow strains were submitted from patients aged 5 years and younger (incidence: 0.9/100'000), 3.5% from patients aged 6 to 14 years (incidence: 0.13/100'000), 74.5% from patients aged 15 to 65 years (incidence: 0.4/100'000) and 7% from patients older than 65 years (incidence: 0.17/100'000). The incidence among infants of 5 years and younger was seven fold higher than that among children aged 6 to 14 years. The incidence of people between 15 and 65 years was three fold higher than that of children aged 6 to 14 years. Of the patients infected with S. Virchow 52% were female and 48% were male.\nOf all patients, 35 (23%) had travelled abroad, the travelling status was unknown for 117 patients (76%), and only one person confirmed not having visited a foreign country within two weeks prior to infection. Eleven persons (7.2%) had visited Egypt, 7 (4.6%) Thailand, 3 (2%) India, 2 (1.3%) Kenya, 2 (1.3%) Asia. The following countries had been visited by one person each (0.7%): Guinea, Jordan, Senegal/Gambia, China, Sri Lanka and Tanzania. Four patients (2.6%) affirmed having visited a foreign country within two weeks prior to infection start, but gave no further information about the destination. As many as 82.5% who had travelled to a foreign country within two weeks before disease outbreak, were between 15 and 65 years old, 7.5% were younger than 6 years, another 7.5% were aged from 6 to 14 years, and 2.5% were older than 65. Subtracting the number of patients with travel background from the total of the patients lowered the average incidence from 0.42/100'000 to 0.34/100'000. Turning the attention on the 6 to 14 year old children, the incidence reduced even from 0.13/100'000 to 0.07/100'000 by subtracting those who stayed abroad. Regarding the 15 to 65 year old patients, the incidence showed similar behaviour and decreased from 0.41/100'000 to 0.3/100'000 while the incidence of other age-groups exhibited only minor changes.\n[SUBTITLE] Antimicrobial susceptibility [SUBSECTION] A high prevalence of resistant (resistant to 1 to 3 antimicrobials) and multi drug resistant (MDR = resistant to more than 3 antibiotics) S. Virchow strains was detected. For calculation of percentages, intermediate susceptibility was counted as \"susceptible\". A total of 48 strains (32%) were resistant and 54 strains (36%) displayed MDR, whereas only 50 strains (33%) showed full susceptibility. There is no clear trend identifiable over the years 2004 to 2009. Antimicrobial susceptibility and resistance profiles varied considerably (Figure 1). Thus in 2008, 67% of all isolates exhibited MDR, and only 19% were susceptible to all antibiotic agents tested. However, in 2009 only 25% of all strains displayed MDR but 57% showed no resistances.\nAssigned to the number of antibiotic resistances (1 to 10), for each year, the number of resistant strains are constituted as dark blue, intermediate strains as light blue columns, respectively.\nWe found 62% of all S. Virchow strains resistant to nalidixic acid, while 98% were susceptible to ciprofloxacin, and only 2% (3 isolates) were intermediate according to CLSI criteria [15]. As many as 49% of the isolates were resistant to sulfamethoxazole, 41% to trimethoprim and 36% to tetracycline. All strains were susceptible to cefotaxime, and only 1.3% were resistant to amoxicillin/clavulanic acid. During the years 2004 to 2007 the resistance situation remained consistent. In 2008, for several antimicrobials the percentage of resistant strains was significantly higher (p < 0.05) than in the preceding years. Interestingly, multiple percentages during 2009 lay beneath the average values of the years 2004 to 2009 (p < 0.05) (Figure 2). Hence, there were no significant trends in resistance development observable for any antibiotic.\nResistance of S. Virchow strains isolated over the years 2004 trough 2009 in Switzerland. Each box shows the percentage fraction of the resistant, intermediate and susceptible sub-population over time for one antimicrobial. Values significantly (p < 0.05) deviating from average are marked with * (more resistant) or + (less resistant), respectively.\nTwo S. Virchow strains (05N2379 and 06N1956) produced ESBL. Both strains tested positive with the confirmatory Etest ESBL (ratio CT/CTL = 62,5; ratio TZ/TZL >340; ratio PM/PML >6).\nA high prevalence of resistant (resistant to 1 to 3 antimicrobials) and multi drug resistant (MDR = resistant to more than 3 antibiotics) S. Virchow strains was detected. For calculation of percentages, intermediate susceptibility was counted as \"susceptible\". A total of 48 strains (32%) were resistant and 54 strains (36%) displayed MDR, whereas only 50 strains (33%) showed full susceptibility. There is no clear trend identifiable over the years 2004 to 2009. Antimicrobial susceptibility and resistance profiles varied considerably (Figure 1). Thus in 2008, 67% of all isolates exhibited MDR, and only 19% were susceptible to all antibiotic agents tested. However, in 2009 only 25% of all strains displayed MDR but 57% showed no resistances.\nAssigned to the number of antibiotic resistances (1 to 10), for each year, the number of resistant strains are constituted as dark blue, intermediate strains as light blue columns, respectively.\nWe found 62% of all S. Virchow strains resistant to nalidixic acid, while 98% were susceptible to ciprofloxacin, and only 2% (3 isolates) were intermediate according to CLSI criteria [15]. As many as 49% of the isolates were resistant to sulfamethoxazole, 41% to trimethoprim and 36% to tetracycline. All strains were susceptible to cefotaxime, and only 1.3% were resistant to amoxicillin/clavulanic acid. During the years 2004 to 2007 the resistance situation remained consistent. In 2008, for several antimicrobials the percentage of resistant strains was significantly higher (p < 0.05) than in the preceding years. Interestingly, multiple percentages during 2009 lay beneath the average values of the years 2004 to 2009 (p < 0.05) (Figure 2). Hence, there were no significant trends in resistance development observable for any antibiotic.\nResistance of S. Virchow strains isolated over the years 2004 trough 2009 in Switzerland. Each box shows the percentage fraction of the resistant, intermediate and susceptible sub-population over time for one antimicrobial. Values significantly (p < 0.05) deviating from average are marked with * (more resistant) or + (less resistant), respectively.\nTwo S. Virchow strains (05N2379 and 06N1956) produced ESBL. Both strains tested positive with the confirmatory Etest ESBL (ratio CT/CTL = 62,5; ratio TZ/TZL >340; ratio PM/PML >6).\n[SUBTITLE] Genetic relationship among Swiss S. Virchow isolates [SUBSECTION] PFGE analysis of the 153 XbaI-digested strains revealed 104 pulsotypes with 14 to 22 DNA fragments ranging from 20,5 kb to 1135 kb. Five and more related strains defined the formation of a cluster. We discovered four pattern clusters (A-D, Figure 3), when strains were considered related if their similarity coefficient exceeded 80%. There were no correlations regarding gender, year of isolation, age-group or stay-abroad, and no other common properties were found for isolates within a certain cluster. A number of pulsotypes incorporated several indistinguishable isolates. Their content ranged from 2 to 8 indistinguishable isolates (Figure 3).\nAnamnestic data, resistance profiles and PFGE patterns of the 153 S. Virchow strains. PFGE clusters are faintly coloured and marked A-D. Strains resistant, intermediate or susceptible to a specific antibiotic are highlighted in red, white, or green, respectively.\nPFGE analysis of the 153 XbaI-digested strains revealed 104 pulsotypes with 14 to 22 DNA fragments ranging from 20,5 kb to 1135 kb. Five and more related strains defined the formation of a cluster. We discovered four pattern clusters (A-D, Figure 3), when strains were considered related if their similarity coefficient exceeded 80%. There were no correlations regarding gender, year of isolation, age-group or stay-abroad, and no other common properties were found for isolates within a certain cluster. A number of pulsotypes incorporated several indistinguishable isolates. Their content ranged from 2 to 8 indistinguishable isolates (Figure 3).\nAnamnestic data, resistance profiles and PFGE patterns of the 153 S. Virchow strains. PFGE clusters are faintly coloured and marked A-D. Strains resistant, intermediate or susceptible to a specific antibiotic are highlighted in red, white, or green, respectively.", "The 153 S. Virchow strains were isolated from 124 faecal samples (81%), 17 blood samples (11.1%), two urine samples (1.3%), two swabs (1.3%), and one synovial puncture (0.7%). For 7 strains (4.6%) the origin was unknown. The relative frequency of S. Virchow among other Salmonella serovars varied between 4th to 8th rank. The annual incidence ranged from 0.45/100'000 in 2004 to 0.40/100'000 in 2009 (lowest incidence: 0.39/100'000 in 2006, highest incidence: 0.47/100'000 in 2007).\nThe age of the patients was known for 99% of the isolates submitted. Fourteen percent of the S. Virchow strains were submitted from patients aged 5 years and younger (incidence: 0.9/100'000), 3.5% from patients aged 6 to 14 years (incidence: 0.13/100'000), 74.5% from patients aged 15 to 65 years (incidence: 0.4/100'000) and 7% from patients older than 65 years (incidence: 0.17/100'000). The incidence among infants of 5 years and younger was seven fold higher than that among children aged 6 to 14 years. The incidence of people between 15 and 65 years was three fold higher than that of children aged 6 to 14 years. Of the patients infected with S. Virchow 52% were female and 48% were male.\nOf all patients, 35 (23%) had travelled abroad, the travelling status was unknown for 117 patients (76%), and only one person confirmed not having visited a foreign country within two weeks prior to infection. Eleven persons (7.2%) had visited Egypt, 7 (4.6%) Thailand, 3 (2%) India, 2 (1.3%) Kenya, 2 (1.3%) Asia. The following countries had been visited by one person each (0.7%): Guinea, Jordan, Senegal/Gambia, China, Sri Lanka and Tanzania. Four patients (2.6%) affirmed having visited a foreign country within two weeks prior to infection start, but gave no further information about the destination. As many as 82.5% who had travelled to a foreign country within two weeks before disease outbreak, were between 15 and 65 years old, 7.5% were younger than 6 years, another 7.5% were aged from 6 to 14 years, and 2.5% were older than 65. Subtracting the number of patients with travel background from the total of the patients lowered the average incidence from 0.42/100'000 to 0.34/100'000. Turning the attention on the 6 to 14 year old children, the incidence reduced even from 0.13/100'000 to 0.07/100'000 by subtracting those who stayed abroad. Regarding the 15 to 65 year old patients, the incidence showed similar behaviour and decreased from 0.41/100'000 to 0.3/100'000 while the incidence of other age-groups exhibited only minor changes.", "A high prevalence of resistant (resistant to 1 to 3 antimicrobials) and multi drug resistant (MDR = resistant to more than 3 antibiotics) S. Virchow strains was detected. For calculation of percentages, intermediate susceptibility was counted as \"susceptible\". A total of 48 strains (32%) were resistant and 54 strains (36%) displayed MDR, whereas only 50 strains (33%) showed full susceptibility. There is no clear trend identifiable over the years 2004 to 2009. Antimicrobial susceptibility and resistance profiles varied considerably (Figure 1). Thus in 2008, 67% of all isolates exhibited MDR, and only 19% were susceptible to all antibiotic agents tested. However, in 2009 only 25% of all strains displayed MDR but 57% showed no resistances.\nAssigned to the number of antibiotic resistances (1 to 10), for each year, the number of resistant strains are constituted as dark blue, intermediate strains as light blue columns, respectively.\nWe found 62% of all S. Virchow strains resistant to nalidixic acid, while 98% were susceptible to ciprofloxacin, and only 2% (3 isolates) were intermediate according to CLSI criteria [15]. As many as 49% of the isolates were resistant to sulfamethoxazole, 41% to trimethoprim and 36% to tetracycline. All strains were susceptible to cefotaxime, and only 1.3% were resistant to amoxicillin/clavulanic acid. During the years 2004 to 2007 the resistance situation remained consistent. In 2008, for several antimicrobials the percentage of resistant strains was significantly higher (p < 0.05) than in the preceding years. Interestingly, multiple percentages during 2009 lay beneath the average values of the years 2004 to 2009 (p < 0.05) (Figure 2). Hence, there were no significant trends in resistance development observable for any antibiotic.\nResistance of S. Virchow strains isolated over the years 2004 trough 2009 in Switzerland. Each box shows the percentage fraction of the resistant, intermediate and susceptible sub-population over time for one antimicrobial. Values significantly (p < 0.05) deviating from average are marked with * (more resistant) or + (less resistant), respectively.\nTwo S. Virchow strains (05N2379 and 06N1956) produced ESBL. Both strains tested positive with the confirmatory Etest ESBL (ratio CT/CTL = 62,5; ratio TZ/TZL >340; ratio PM/PML >6).", "PFGE analysis of the 153 XbaI-digested strains revealed 104 pulsotypes with 14 to 22 DNA fragments ranging from 20,5 kb to 1135 kb. Five and more related strains defined the formation of a cluster. We discovered four pattern clusters (A-D, Figure 3), when strains were considered related if their similarity coefficient exceeded 80%. There were no correlations regarding gender, year of isolation, age-group or stay-abroad, and no other common properties were found for isolates within a certain cluster. A number of pulsotypes incorporated several indistinguishable isolates. Their content ranged from 2 to 8 indistinguishable isolates (Figure 3).\nAnamnestic data, resistance profiles and PFGE patterns of the 153 S. Virchow strains. PFGE clusters are faintly coloured and marked A-D. Strains resistant, intermediate or susceptible to a specific antibiotic are highlighted in red, white, or green, respectively.", "During the study period, 1.5% of all Salmonella strains submitted to the NENT belonged to the serotype Virchow. This amounted to a prevalence of 0.45 (2004) to 0.40 (2009) cases per 100'000 population, and was within the average in Europe [4,5]. For a comparison, the prevalence of total salmonellosis in Switzerland was 26 (2004) to 25 (2009) cases per 100'000 population. Whereas in France the annual number as well as the corresponding rank decreased during 2005 to 2008 [5], in Switzerland the number of cases associated with S. Virchow infections and its rank remained constant from 2004 through 2009.\nConspicuously, the incidence of infected children aged 0 to 5 years was seven fold higher than that in children older than 5 years. Reasons may be an insufficient immune defence that lowers the minimal infective dose, as the immune system has not yet fully developed. Regarding the age-group from 15 to 65, we discovered a three fold higher incidence compared to the incidence of persons aged older than 65. This could be linked to considerable differences in travelling activity. Thus, 25% of all patients between 15 and 65 years of age had been visiting a foreign country within 2 weeks before infection. Even 50% of the infections among 6 to 14 years old children were associated with travelling, whereas other age-groups could hardly be linked to a stay-abroad. Upon the present investigation, no case control study could be performed, and, unfortunately, exact data about travel destinations were available only with 31 of the 153 cases. Nevertheless, 16 and 15 out of the 31 cases could be traced back to African and Asian countries, respectively.\nIn this study, S. Virchow strains isolated from blood samples, and presumably associated with invasive manifestation, were enclosed. Interestingly, we did not receive any blood samples derived from children up to 5 years. In contrast, 16% blood samples were submitted from adult patients of the 15 to 65 age group. Similar findings were published in 2005 from a university hospital in Crete (Greece). Galanakis and his team found 41% of all patients infected with non-thyphoidal salmonellae aged under 15 years, but only 4.06% suffered from an invasive disease. Adult patients (> 15 years) suffered twice as much from an invasive manifestation [16]. However, Parry reported about sub-Saharan Africa, where invasive infections caused by enteritic serotypes are particularly common in children under 5 years [17]. Reasons for such divergent findings are unknown, but differences between developed and developing countries seem to exist.\nGenotyping revealed 104 strains with a high degree (> 82%) of relationship, summarised in cluster A, which constituted the major part of all isolates. Additionally, several groups of samples were discovered, representing indistinguishable pulsotypes. Strains belonging to cluster A were collected within the years 2004 to 2009 and originated from different countries. This indicates, that these closely related isolates are highly established and widespread.\nIndistinguishable patterns give evidence to small outbreaks or point sources, that spread a certain strain over several years. This becomes particularly apparent regarding non-discriminable isolates derived from patients that stayed in the same foreign country. As already mentioned, most of the strains from patients with a travelling background had returned from the North-African or Asian region. Close clonal relationship between strains from those regions is exemplified by two PFGE sub-clusters within the large Cluster A. One sub-cluster is associated with Egypt (Figure 3, strains no. 8-N1063, 9-N2543, 7-N2080, 8-N2820, 8-N2492, 8-N2393), the other with Thailand (Figure 3, strains no. 6-N0978, 6-N1949, 7-N0090, 7-N0368).\nThe number of resistant (31.5%) and multiresistant (35.5%) strains is considerable. Similar rates from European countries have already been published. In 2004 a study performed by Meakins and co-authors revealed 73% of all S. Virchow strains resistant to at least one antimicrobial agent [18]. Threlfall, in a European multi-centre study with over 27000 cases of salmonellosis in 2000, discovered 36% multi drug resistant isolates [4]. Strikingly, in our study almost all MDR isolates lay within cluster A (Figure 3). Similarities in the resistance situation between the European Union and Switzerland indicate a close relationship of strains belonging to cluster A and strains responsible for the high level of resistance in Europe.\nA high prevalence of nalidixic acid resistant strains was found within cluster A, whereas nearly no resistance to nalidixic acid could be observed among the remaining isolates. A similar behaviour for resistances to tetracycline, sulfonamide and trimetoprim was noted.\nFluoroquinolones and third-generation cephalosporins are drugs-of-choice for invasive Salmonella infections. During the past years numerous studies reported the emergence of fluoroquinolone or third-generation cephalosporine resistant strains [9,11,12]. We found a high prevalence of isolates resistant to nalidixic acid but no full resistance against ciprofloxacin, and only 2% of isolates being intermediate over the whole study period. Because of the intersecting resistance mechanisms against quinolones and fluoroquinolones, the behaviour against ciprofloxacin must be discussed in association with that to nalidixic acid [19]. Because of repeated reports of fluoroquinolone treatment failure following disease caused by Salmonella strains that were susceptible to fluoroquinolones using CLSI criteria, but were simultaneously resistant to nalidixic acid, several authors called for re-evaluation of the CLSI breakpoints for fluoroquinolones involving salmonellae [20-22]. Consequently, comments no. 2, 18, and 20 were added to Table Two A in CLSI documents, as can be seen, for instance, in the 2008 edition [15]. They serve as a warning, that fluoroquinolone-susceptible but nalidixic acid-resistant extraintestinal salmonellae must be considered reduced-susceptible to fluoroquinolones. By the comments, the reader is also urged to perform a nalidixic acid susceptibility test, in order to become aware of such strains. Based on these arguments, the 62% naldixic acid-resistant S. Virchow isolates of the present study have to be considered reduced-susceptible to ciprofloxacin, and thus prone to cause a significant rate of treatment failure. The observation of a low rate of \"obvious\" resistance to fluoroquinolones is in accordance with data from other European countries [4,5]. As we discovered only two ESBL producing isolates in the years 2005 and 2006, the occurrence of such strains in Switzerland are rather rare events compared to other European countries [5,11,23].\nIn many countries, S. Virchow infections are poultry associated [6-9]. In Switzerland, however, there is a favourable situation in poultry flocks with respect to Salmonella in general and S. Virchow in particular [24]. However, 52% of all poultry meat is imported and 71% of the imports originate from Brazil, Germany and France. From all these countries, detection of S. Virchow in poultry products has been reported [7,25,26]. On this basis, one might speculate, that a majority of the S. Virchow infections in Switzerland may be linked to travelling activities and/or consumption of imported poultry meat. This, however, would have to be further investigated by additional case control- and epidemiological studies.", "The worldwide increase of antibiotic resistances in Salmonella is an emerging public health problem. For Switzerland, no clear trend is identifiable over the years 2004 to 2009 for S. Virchow. Antimicrobial susceptibility and resistance profiles varied considerably within this period. Nevertheless, the situation in Switzerland coincided with findings in other European countries. Genotyping results of this strain collection revealed no evidence for an undetected outbreak within this time period", "The authors declare that they have no competing interests.", "RS, LA and HH designed the study. MB, NC, UK have done the phenotypic and genotypic characterization of the strains. MB and RS drafted the manuscript. All authors read, commented on and approved of the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2334/11/49/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Strains", "Antimicrobial susceptibility testing", "Genotyping", "Results", "Clinical and epidemiological data", "Antimicrobial susceptibility", "Genetic relationship among Swiss S. Virchow isolates", "Discussion", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Non-typhoidal salmonellae are worldwide responsible for numerous foodborne infections and are a major threat to public health [1,2]. Based on the lipopolysaccharide (O antigen) and the flagellar structures (H antigen) Salmonella spp. are divided into greater than 2500 serotypes, of which only few obtain substantial relevance to humans [3]. In Europe, as well as in Switzerland, Salmonella serotype Virchow ranks among the five most frequent serovars and has been reported to be poultry associated [4-9]. As with other enteritic salmonellae, infections due to S. Virchow commonly manifest themselves as self-limiting gastroenteritis, but severe invasive infections can also occur which then need antibiotic treatment [10]. Whereas fluoroquinolones are used to treat extraintestinal infections in adults, third generation cephalosporins are drugs of choice in children. Alarmingly, in the last decade numerous European countries reported emergence of S. Virchow strains resistant to the above mentioned antimicrobials, leading to treatment failures [9,11,12]. Resistance to quinolones is either mediated by chromosomal mutations in genes encoding DNA gyrase or topoisomerase IV or in genes affecting the uptake or efflux of drugs [13]. Plasmid associated quinolone corrupting genes such as aac(6')-Ib-cr or qnr etc., are also known to cause reduced susceptibility. Resistance to third generation cephalosporins is mediated by extended spectrum β-lactamases (ESBL), which are plasmid encoded or integron associated [14].\nIn this study, a total of 153 S. Virchow strains isolated from different patients from 2004 through 2009 in Switzerland were further characterized by (i) assessing phenotypic antibiotic resistance profiles using the disk diffusion method and (ii) by genotyping using pulsed-field gel electrophoresis (PFGE) after macrorestriction with XbaI in order to evaluate strain relationship.", "[SUBTITLE] Strains [SUBSECTION] Human salmonellosis is a reportable disease in Switzerland and isolates have to be submitted to the National Centre for Enteropathogenic Bacteria (NENT). The NENT - in cooperation with the Swiss Federal Office of Public Health - collects clinical data and carries out final serological identification by slide agglutination with commercial antisera according to the Kauffmann-White scheme. Over the years 2004 through 2009, 10395 Salmonella strains were submitted to the NENT. From this collection of strains 153 S. Virchow strains (1,5%) from different patients (multiple isolates from same patient excluded, but family members not excluded) were selected and integrated in this study. All samples had been collected by hospitals or family doctors.\nHuman salmonellosis is a reportable disease in Switzerland and isolates have to be submitted to the National Centre for Enteropathogenic Bacteria (NENT). The NENT - in cooperation with the Swiss Federal Office of Public Health - collects clinical data and carries out final serological identification by slide agglutination with commercial antisera according to the Kauffmann-White scheme. Over the years 2004 through 2009, 10395 Salmonella strains were submitted to the NENT. From this collection of strains 153 S. Virchow strains (1,5%) from different patients (multiple isolates from same patient excluded, but family members not excluded) were selected and integrated in this study. All samples had been collected by hospitals or family doctors.\n[SUBTITLE] Antimicrobial susceptibility testing [SUBSECTION] The strains were tested for antimicrobial resistance by the disk diffusion method according to the Clinical and Laboratory Standards Institute (CLSI) [15]. The panel of antibiotic disks (Becton, Dickinson and Company, Maryland, USA) consisted of ampicillin (AM10), amoxicillin/clavulanic acid (AMC30), cephalothin (CF30), cefotaxime (CTX30), ciprofloxacin (CIP5), gentamicin (GM10), tetracycline (Te30), streptomycin (S10), chloramphenicol (C30), kanamycin (K30), nalidixic acid (NA30), sulfamethoxazole (SMZ), and trimethoprim (TMP5). The strains were classified as resistant, intermediate or susceptible to each antibiotic agent according to the CLSI criteria [15].\nFor presumptive ESBL producers - strains that produced synergistic inhibition zone enlargements between adjacent discs containing an oxyimino cephalosporin and clavulanic acid, respectively - Etest ESBL (bioMérieux, Marcy l'Etoile, France) tests, containing cefotaxime (CT/CTL), ceftazidime (TZ/TZL), or cefepime (PM/PML) alone and in combination with clavulanic acid, were performed for confirmation according the manufacturer's guidelines.\nThe strains were tested for antimicrobial resistance by the disk diffusion method according to the Clinical and Laboratory Standards Institute (CLSI) [15]. The panel of antibiotic disks (Becton, Dickinson and Company, Maryland, USA) consisted of ampicillin (AM10), amoxicillin/clavulanic acid (AMC30), cephalothin (CF30), cefotaxime (CTX30), ciprofloxacin (CIP5), gentamicin (GM10), tetracycline (Te30), streptomycin (S10), chloramphenicol (C30), kanamycin (K30), nalidixic acid (NA30), sulfamethoxazole (SMZ), and trimethoprim (TMP5). The strains were classified as resistant, intermediate or susceptible to each antibiotic agent according to the CLSI criteria [15].\nFor presumptive ESBL producers - strains that produced synergistic inhibition zone enlargements between adjacent discs containing an oxyimino cephalosporin and clavulanic acid, respectively - Etest ESBL (bioMérieux, Marcy l'Etoile, France) tests, containing cefotaxime (CT/CTL), ceftazidime (TZ/TZL), or cefepime (PM/PML) alone and in combination with clavulanic acid, were performed for confirmation according the manufacturer's guidelines.\n[SUBTITLE] Genotyping [SUBSECTION] Pulsed-field gel electrophoresis (PFGE) was performed by following the CDC PulseNet protocol http://www.cdc.gov/pulsenet/protocols.htm with minor modifications. In brief, strains were grown on blood agar at 37°C over night. Colonies from blood agar were resuspended in cell suspension buffer (OD600 = 1). The bacterial cell suspension was mixed with 400 μl of 1.4% Pulsed Field Certified Agarose (BIO-RAD, Munich, Germany) and cells were lysed by proteinase K treatment over night. After lysis the plugs were washed twice for 30 min in ultrapure water and 4 times for an hour in Tris-EDTA (TE) buffer. After washing with TE buffer, DNA agarose plugs were incubated over night in the presence of XbaI (Roche, Mannheim, Germany) following the manufacturer's instructions. Restricted DNA in plug slices was separated in a 1% SeaKem Gold (BioConcept, Allschwil, Switzerland) agarose gel at 6 V/cm in 0.5 M Tris-Borate-EDTA buffer cooled to 12°C in a CHEF-DR III system (BIO-RAD, Munich, Germany). The pulse times were ramped from 2 to 64 sec for 20 h at an angle of 120°. Gels were stained with ethidium bromide and visualized under UV light transillumination with Gel Doc (BIO-RAD, Munich, Germany) and analysed with BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium). Pair wise similarities between the XbaI pulsed-field patterns were calculated by the DICE's similarity coefficient. Clustering was based on the unweighted pair-group method with averages (UPGMA), setting tolerance and optimization each at 1.5%. We used Salmonella Braenderup strain H9812 (ATCC BAA 664) as a reference strain.\nPulsed-field gel electrophoresis (PFGE) was performed by following the CDC PulseNet protocol http://www.cdc.gov/pulsenet/protocols.htm with minor modifications. In brief, strains were grown on blood agar at 37°C over night. Colonies from blood agar were resuspended in cell suspension buffer (OD600 = 1). The bacterial cell suspension was mixed with 400 μl of 1.4% Pulsed Field Certified Agarose (BIO-RAD, Munich, Germany) and cells were lysed by proteinase K treatment over night. After lysis the plugs were washed twice for 30 min in ultrapure water and 4 times for an hour in Tris-EDTA (TE) buffer. After washing with TE buffer, DNA agarose plugs were incubated over night in the presence of XbaI (Roche, Mannheim, Germany) following the manufacturer's instructions. Restricted DNA in plug slices was separated in a 1% SeaKem Gold (BioConcept, Allschwil, Switzerland) agarose gel at 6 V/cm in 0.5 M Tris-Borate-EDTA buffer cooled to 12°C in a CHEF-DR III system (BIO-RAD, Munich, Germany). The pulse times were ramped from 2 to 64 sec for 20 h at an angle of 120°. Gels were stained with ethidium bromide and visualized under UV light transillumination with Gel Doc (BIO-RAD, Munich, Germany) and analysed with BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium). Pair wise similarities between the XbaI pulsed-field patterns were calculated by the DICE's similarity coefficient. Clustering was based on the unweighted pair-group method with averages (UPGMA), setting tolerance and optimization each at 1.5%. We used Salmonella Braenderup strain H9812 (ATCC BAA 664) as a reference strain.", "Human salmonellosis is a reportable disease in Switzerland and isolates have to be submitted to the National Centre for Enteropathogenic Bacteria (NENT). The NENT - in cooperation with the Swiss Federal Office of Public Health - collects clinical data and carries out final serological identification by slide agglutination with commercial antisera according to the Kauffmann-White scheme. Over the years 2004 through 2009, 10395 Salmonella strains were submitted to the NENT. From this collection of strains 153 S. Virchow strains (1,5%) from different patients (multiple isolates from same patient excluded, but family members not excluded) were selected and integrated in this study. All samples had been collected by hospitals or family doctors.", "The strains were tested for antimicrobial resistance by the disk diffusion method according to the Clinical and Laboratory Standards Institute (CLSI) [15]. The panel of antibiotic disks (Becton, Dickinson and Company, Maryland, USA) consisted of ampicillin (AM10), amoxicillin/clavulanic acid (AMC30), cephalothin (CF30), cefotaxime (CTX30), ciprofloxacin (CIP5), gentamicin (GM10), tetracycline (Te30), streptomycin (S10), chloramphenicol (C30), kanamycin (K30), nalidixic acid (NA30), sulfamethoxazole (SMZ), and trimethoprim (TMP5). The strains were classified as resistant, intermediate or susceptible to each antibiotic agent according to the CLSI criteria [15].\nFor presumptive ESBL producers - strains that produced synergistic inhibition zone enlargements between adjacent discs containing an oxyimino cephalosporin and clavulanic acid, respectively - Etest ESBL (bioMérieux, Marcy l'Etoile, France) tests, containing cefotaxime (CT/CTL), ceftazidime (TZ/TZL), or cefepime (PM/PML) alone and in combination with clavulanic acid, were performed for confirmation according the manufacturer's guidelines.", "Pulsed-field gel electrophoresis (PFGE) was performed by following the CDC PulseNet protocol http://www.cdc.gov/pulsenet/protocols.htm with minor modifications. In brief, strains were grown on blood agar at 37°C over night. Colonies from blood agar were resuspended in cell suspension buffer (OD600 = 1). The bacterial cell suspension was mixed with 400 μl of 1.4% Pulsed Field Certified Agarose (BIO-RAD, Munich, Germany) and cells were lysed by proteinase K treatment over night. After lysis the plugs were washed twice for 30 min in ultrapure water and 4 times for an hour in Tris-EDTA (TE) buffer. After washing with TE buffer, DNA agarose plugs were incubated over night in the presence of XbaI (Roche, Mannheim, Germany) following the manufacturer's instructions. Restricted DNA in plug slices was separated in a 1% SeaKem Gold (BioConcept, Allschwil, Switzerland) agarose gel at 6 V/cm in 0.5 M Tris-Borate-EDTA buffer cooled to 12°C in a CHEF-DR III system (BIO-RAD, Munich, Germany). The pulse times were ramped from 2 to 64 sec for 20 h at an angle of 120°. Gels were stained with ethidium bromide and visualized under UV light transillumination with Gel Doc (BIO-RAD, Munich, Germany) and analysed with BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium). Pair wise similarities between the XbaI pulsed-field patterns were calculated by the DICE's similarity coefficient. Clustering was based on the unweighted pair-group method with averages (UPGMA), setting tolerance and optimization each at 1.5%. We used Salmonella Braenderup strain H9812 (ATCC BAA 664) as a reference strain.", "[SUBTITLE] Clinical and epidemiological data [SUBSECTION] The 153 S. Virchow strains were isolated from 124 faecal samples (81%), 17 blood samples (11.1%), two urine samples (1.3%), two swabs (1.3%), and one synovial puncture (0.7%). For 7 strains (4.6%) the origin was unknown. The relative frequency of S. Virchow among other Salmonella serovars varied between 4th to 8th rank. The annual incidence ranged from 0.45/100'000 in 2004 to 0.40/100'000 in 2009 (lowest incidence: 0.39/100'000 in 2006, highest incidence: 0.47/100'000 in 2007).\nThe age of the patients was known for 99% of the isolates submitted. Fourteen percent of the S. Virchow strains were submitted from patients aged 5 years and younger (incidence: 0.9/100'000), 3.5% from patients aged 6 to 14 years (incidence: 0.13/100'000), 74.5% from patients aged 15 to 65 years (incidence: 0.4/100'000) and 7% from patients older than 65 years (incidence: 0.17/100'000). The incidence among infants of 5 years and younger was seven fold higher than that among children aged 6 to 14 years. The incidence of people between 15 and 65 years was three fold higher than that of children aged 6 to 14 years. Of the patients infected with S. Virchow 52% were female and 48% were male.\nOf all patients, 35 (23%) had travelled abroad, the travelling status was unknown for 117 patients (76%), and only one person confirmed not having visited a foreign country within two weeks prior to infection. Eleven persons (7.2%) had visited Egypt, 7 (4.6%) Thailand, 3 (2%) India, 2 (1.3%) Kenya, 2 (1.3%) Asia. The following countries had been visited by one person each (0.7%): Guinea, Jordan, Senegal/Gambia, China, Sri Lanka and Tanzania. Four patients (2.6%) affirmed having visited a foreign country within two weeks prior to infection start, but gave no further information about the destination. As many as 82.5% who had travelled to a foreign country within two weeks before disease outbreak, were between 15 and 65 years old, 7.5% were younger than 6 years, another 7.5% were aged from 6 to 14 years, and 2.5% were older than 65. Subtracting the number of patients with travel background from the total of the patients lowered the average incidence from 0.42/100'000 to 0.34/100'000. Turning the attention on the 6 to 14 year old children, the incidence reduced even from 0.13/100'000 to 0.07/100'000 by subtracting those who stayed abroad. Regarding the 15 to 65 year old patients, the incidence showed similar behaviour and decreased from 0.41/100'000 to 0.3/100'000 while the incidence of other age-groups exhibited only minor changes.\nThe 153 S. Virchow strains were isolated from 124 faecal samples (81%), 17 blood samples (11.1%), two urine samples (1.3%), two swabs (1.3%), and one synovial puncture (0.7%). For 7 strains (4.6%) the origin was unknown. The relative frequency of S. Virchow among other Salmonella serovars varied between 4th to 8th rank. The annual incidence ranged from 0.45/100'000 in 2004 to 0.40/100'000 in 2009 (lowest incidence: 0.39/100'000 in 2006, highest incidence: 0.47/100'000 in 2007).\nThe age of the patients was known for 99% of the isolates submitted. Fourteen percent of the S. Virchow strains were submitted from patients aged 5 years and younger (incidence: 0.9/100'000), 3.5% from patients aged 6 to 14 years (incidence: 0.13/100'000), 74.5% from patients aged 15 to 65 years (incidence: 0.4/100'000) and 7% from patients older than 65 years (incidence: 0.17/100'000). The incidence among infants of 5 years and younger was seven fold higher than that among children aged 6 to 14 years. The incidence of people between 15 and 65 years was three fold higher than that of children aged 6 to 14 years. Of the patients infected with S. Virchow 52% were female and 48% were male.\nOf all patients, 35 (23%) had travelled abroad, the travelling status was unknown for 117 patients (76%), and only one person confirmed not having visited a foreign country within two weeks prior to infection. Eleven persons (7.2%) had visited Egypt, 7 (4.6%) Thailand, 3 (2%) India, 2 (1.3%) Kenya, 2 (1.3%) Asia. The following countries had been visited by one person each (0.7%): Guinea, Jordan, Senegal/Gambia, China, Sri Lanka and Tanzania. Four patients (2.6%) affirmed having visited a foreign country within two weeks prior to infection start, but gave no further information about the destination. As many as 82.5% who had travelled to a foreign country within two weeks before disease outbreak, were between 15 and 65 years old, 7.5% were younger than 6 years, another 7.5% were aged from 6 to 14 years, and 2.5% were older than 65. Subtracting the number of patients with travel background from the total of the patients lowered the average incidence from 0.42/100'000 to 0.34/100'000. Turning the attention on the 6 to 14 year old children, the incidence reduced even from 0.13/100'000 to 0.07/100'000 by subtracting those who stayed abroad. Regarding the 15 to 65 year old patients, the incidence showed similar behaviour and decreased from 0.41/100'000 to 0.3/100'000 while the incidence of other age-groups exhibited only minor changes.\n[SUBTITLE] Antimicrobial susceptibility [SUBSECTION] A high prevalence of resistant (resistant to 1 to 3 antimicrobials) and multi drug resistant (MDR = resistant to more than 3 antibiotics) S. Virchow strains was detected. For calculation of percentages, intermediate susceptibility was counted as \"susceptible\". A total of 48 strains (32%) were resistant and 54 strains (36%) displayed MDR, whereas only 50 strains (33%) showed full susceptibility. There is no clear trend identifiable over the years 2004 to 2009. Antimicrobial susceptibility and resistance profiles varied considerably (Figure 1). Thus in 2008, 67% of all isolates exhibited MDR, and only 19% were susceptible to all antibiotic agents tested. However, in 2009 only 25% of all strains displayed MDR but 57% showed no resistances.\nAssigned to the number of antibiotic resistances (1 to 10), for each year, the number of resistant strains are constituted as dark blue, intermediate strains as light blue columns, respectively.\nWe found 62% of all S. Virchow strains resistant to nalidixic acid, while 98% were susceptible to ciprofloxacin, and only 2% (3 isolates) were intermediate according to CLSI criteria [15]. As many as 49% of the isolates were resistant to sulfamethoxazole, 41% to trimethoprim and 36% to tetracycline. All strains were susceptible to cefotaxime, and only 1.3% were resistant to amoxicillin/clavulanic acid. During the years 2004 to 2007 the resistance situation remained consistent. In 2008, for several antimicrobials the percentage of resistant strains was significantly higher (p < 0.05) than in the preceding years. Interestingly, multiple percentages during 2009 lay beneath the average values of the years 2004 to 2009 (p < 0.05) (Figure 2). Hence, there were no significant trends in resistance development observable for any antibiotic.\nResistance of S. Virchow strains isolated over the years 2004 trough 2009 in Switzerland. Each box shows the percentage fraction of the resistant, intermediate and susceptible sub-population over time for one antimicrobial. Values significantly (p < 0.05) deviating from average are marked with * (more resistant) or + (less resistant), respectively.\nTwo S. Virchow strains (05N2379 and 06N1956) produced ESBL. Both strains tested positive with the confirmatory Etest ESBL (ratio CT/CTL = 62,5; ratio TZ/TZL >340; ratio PM/PML >6).\nA high prevalence of resistant (resistant to 1 to 3 antimicrobials) and multi drug resistant (MDR = resistant to more than 3 antibiotics) S. Virchow strains was detected. For calculation of percentages, intermediate susceptibility was counted as \"susceptible\". A total of 48 strains (32%) were resistant and 54 strains (36%) displayed MDR, whereas only 50 strains (33%) showed full susceptibility. There is no clear trend identifiable over the years 2004 to 2009. Antimicrobial susceptibility and resistance profiles varied considerably (Figure 1). Thus in 2008, 67% of all isolates exhibited MDR, and only 19% were susceptible to all antibiotic agents tested. However, in 2009 only 25% of all strains displayed MDR but 57% showed no resistances.\nAssigned to the number of antibiotic resistances (1 to 10), for each year, the number of resistant strains are constituted as dark blue, intermediate strains as light blue columns, respectively.\nWe found 62% of all S. Virchow strains resistant to nalidixic acid, while 98% were susceptible to ciprofloxacin, and only 2% (3 isolates) were intermediate according to CLSI criteria [15]. As many as 49% of the isolates were resistant to sulfamethoxazole, 41% to trimethoprim and 36% to tetracycline. All strains were susceptible to cefotaxime, and only 1.3% were resistant to amoxicillin/clavulanic acid. During the years 2004 to 2007 the resistance situation remained consistent. In 2008, for several antimicrobials the percentage of resistant strains was significantly higher (p < 0.05) than in the preceding years. Interestingly, multiple percentages during 2009 lay beneath the average values of the years 2004 to 2009 (p < 0.05) (Figure 2). Hence, there were no significant trends in resistance development observable for any antibiotic.\nResistance of S. Virchow strains isolated over the years 2004 trough 2009 in Switzerland. Each box shows the percentage fraction of the resistant, intermediate and susceptible sub-population over time for one antimicrobial. Values significantly (p < 0.05) deviating from average are marked with * (more resistant) or + (less resistant), respectively.\nTwo S. Virchow strains (05N2379 and 06N1956) produced ESBL. Both strains tested positive with the confirmatory Etest ESBL (ratio CT/CTL = 62,5; ratio TZ/TZL >340; ratio PM/PML >6).\n[SUBTITLE] Genetic relationship among Swiss S. Virchow isolates [SUBSECTION] PFGE analysis of the 153 XbaI-digested strains revealed 104 pulsotypes with 14 to 22 DNA fragments ranging from 20,5 kb to 1135 kb. Five and more related strains defined the formation of a cluster. We discovered four pattern clusters (A-D, Figure 3), when strains were considered related if their similarity coefficient exceeded 80%. There were no correlations regarding gender, year of isolation, age-group or stay-abroad, and no other common properties were found for isolates within a certain cluster. A number of pulsotypes incorporated several indistinguishable isolates. Their content ranged from 2 to 8 indistinguishable isolates (Figure 3).\nAnamnestic data, resistance profiles and PFGE patterns of the 153 S. Virchow strains. PFGE clusters are faintly coloured and marked A-D. Strains resistant, intermediate or susceptible to a specific antibiotic are highlighted in red, white, or green, respectively.\nPFGE analysis of the 153 XbaI-digested strains revealed 104 pulsotypes with 14 to 22 DNA fragments ranging from 20,5 kb to 1135 kb. Five and more related strains defined the formation of a cluster. We discovered four pattern clusters (A-D, Figure 3), when strains were considered related if their similarity coefficient exceeded 80%. There were no correlations regarding gender, year of isolation, age-group or stay-abroad, and no other common properties were found for isolates within a certain cluster. A number of pulsotypes incorporated several indistinguishable isolates. Their content ranged from 2 to 8 indistinguishable isolates (Figure 3).\nAnamnestic data, resistance profiles and PFGE patterns of the 153 S. Virchow strains. PFGE clusters are faintly coloured and marked A-D. Strains resistant, intermediate or susceptible to a specific antibiotic are highlighted in red, white, or green, respectively.", "The 153 S. Virchow strains were isolated from 124 faecal samples (81%), 17 blood samples (11.1%), two urine samples (1.3%), two swabs (1.3%), and one synovial puncture (0.7%). For 7 strains (4.6%) the origin was unknown. The relative frequency of S. Virchow among other Salmonella serovars varied between 4th to 8th rank. The annual incidence ranged from 0.45/100'000 in 2004 to 0.40/100'000 in 2009 (lowest incidence: 0.39/100'000 in 2006, highest incidence: 0.47/100'000 in 2007).\nThe age of the patients was known for 99% of the isolates submitted. Fourteen percent of the S. Virchow strains were submitted from patients aged 5 years and younger (incidence: 0.9/100'000), 3.5% from patients aged 6 to 14 years (incidence: 0.13/100'000), 74.5% from patients aged 15 to 65 years (incidence: 0.4/100'000) and 7% from patients older than 65 years (incidence: 0.17/100'000). The incidence among infants of 5 years and younger was seven fold higher than that among children aged 6 to 14 years. The incidence of people between 15 and 65 years was three fold higher than that of children aged 6 to 14 years. Of the patients infected with S. Virchow 52% were female and 48% were male.\nOf all patients, 35 (23%) had travelled abroad, the travelling status was unknown for 117 patients (76%), and only one person confirmed not having visited a foreign country within two weeks prior to infection. Eleven persons (7.2%) had visited Egypt, 7 (4.6%) Thailand, 3 (2%) India, 2 (1.3%) Kenya, 2 (1.3%) Asia. The following countries had been visited by one person each (0.7%): Guinea, Jordan, Senegal/Gambia, China, Sri Lanka and Tanzania. Four patients (2.6%) affirmed having visited a foreign country within two weeks prior to infection start, but gave no further information about the destination. As many as 82.5% who had travelled to a foreign country within two weeks before disease outbreak, were between 15 and 65 years old, 7.5% were younger than 6 years, another 7.5% were aged from 6 to 14 years, and 2.5% were older than 65. Subtracting the number of patients with travel background from the total of the patients lowered the average incidence from 0.42/100'000 to 0.34/100'000. Turning the attention on the 6 to 14 year old children, the incidence reduced even from 0.13/100'000 to 0.07/100'000 by subtracting those who stayed abroad. Regarding the 15 to 65 year old patients, the incidence showed similar behaviour and decreased from 0.41/100'000 to 0.3/100'000 while the incidence of other age-groups exhibited only minor changes.", "A high prevalence of resistant (resistant to 1 to 3 antimicrobials) and multi drug resistant (MDR = resistant to more than 3 antibiotics) S. Virchow strains was detected. For calculation of percentages, intermediate susceptibility was counted as \"susceptible\". A total of 48 strains (32%) were resistant and 54 strains (36%) displayed MDR, whereas only 50 strains (33%) showed full susceptibility. There is no clear trend identifiable over the years 2004 to 2009. Antimicrobial susceptibility and resistance profiles varied considerably (Figure 1). Thus in 2008, 67% of all isolates exhibited MDR, and only 19% were susceptible to all antibiotic agents tested. However, in 2009 only 25% of all strains displayed MDR but 57% showed no resistances.\nAssigned to the number of antibiotic resistances (1 to 10), for each year, the number of resistant strains are constituted as dark blue, intermediate strains as light blue columns, respectively.\nWe found 62% of all S. Virchow strains resistant to nalidixic acid, while 98% were susceptible to ciprofloxacin, and only 2% (3 isolates) were intermediate according to CLSI criteria [15]. As many as 49% of the isolates were resistant to sulfamethoxazole, 41% to trimethoprim and 36% to tetracycline. All strains were susceptible to cefotaxime, and only 1.3% were resistant to amoxicillin/clavulanic acid. During the years 2004 to 2007 the resistance situation remained consistent. In 2008, for several antimicrobials the percentage of resistant strains was significantly higher (p < 0.05) than in the preceding years. Interestingly, multiple percentages during 2009 lay beneath the average values of the years 2004 to 2009 (p < 0.05) (Figure 2). Hence, there were no significant trends in resistance development observable for any antibiotic.\nResistance of S. Virchow strains isolated over the years 2004 trough 2009 in Switzerland. Each box shows the percentage fraction of the resistant, intermediate and susceptible sub-population over time for one antimicrobial. Values significantly (p < 0.05) deviating from average are marked with * (more resistant) or + (less resistant), respectively.\nTwo S. Virchow strains (05N2379 and 06N1956) produced ESBL. Both strains tested positive with the confirmatory Etest ESBL (ratio CT/CTL = 62,5; ratio TZ/TZL >340; ratio PM/PML >6).", "PFGE analysis of the 153 XbaI-digested strains revealed 104 pulsotypes with 14 to 22 DNA fragments ranging from 20,5 kb to 1135 kb. Five and more related strains defined the formation of a cluster. We discovered four pattern clusters (A-D, Figure 3), when strains were considered related if their similarity coefficient exceeded 80%. There were no correlations regarding gender, year of isolation, age-group or stay-abroad, and no other common properties were found for isolates within a certain cluster. A number of pulsotypes incorporated several indistinguishable isolates. Their content ranged from 2 to 8 indistinguishable isolates (Figure 3).\nAnamnestic data, resistance profiles and PFGE patterns of the 153 S. Virchow strains. PFGE clusters are faintly coloured and marked A-D. Strains resistant, intermediate or susceptible to a specific antibiotic are highlighted in red, white, or green, respectively.", "During the study period, 1.5% of all Salmonella strains submitted to the NENT belonged to the serotype Virchow. This amounted to a prevalence of 0.45 (2004) to 0.40 (2009) cases per 100'000 population, and was within the average in Europe [4,5]. For a comparison, the prevalence of total salmonellosis in Switzerland was 26 (2004) to 25 (2009) cases per 100'000 population. Whereas in France the annual number as well as the corresponding rank decreased during 2005 to 2008 [5], in Switzerland the number of cases associated with S. Virchow infections and its rank remained constant from 2004 through 2009.\nConspicuously, the incidence of infected children aged 0 to 5 years was seven fold higher than that in children older than 5 years. Reasons may be an insufficient immune defence that lowers the minimal infective dose, as the immune system has not yet fully developed. Regarding the age-group from 15 to 65, we discovered a three fold higher incidence compared to the incidence of persons aged older than 65. This could be linked to considerable differences in travelling activity. Thus, 25% of all patients between 15 and 65 years of age had been visiting a foreign country within 2 weeks before infection. Even 50% of the infections among 6 to 14 years old children were associated with travelling, whereas other age-groups could hardly be linked to a stay-abroad. Upon the present investigation, no case control study could be performed, and, unfortunately, exact data about travel destinations were available only with 31 of the 153 cases. Nevertheless, 16 and 15 out of the 31 cases could be traced back to African and Asian countries, respectively.\nIn this study, S. Virchow strains isolated from blood samples, and presumably associated with invasive manifestation, were enclosed. Interestingly, we did not receive any blood samples derived from children up to 5 years. In contrast, 16% blood samples were submitted from adult patients of the 15 to 65 age group. Similar findings were published in 2005 from a university hospital in Crete (Greece). Galanakis and his team found 41% of all patients infected with non-thyphoidal salmonellae aged under 15 years, but only 4.06% suffered from an invasive disease. Adult patients (> 15 years) suffered twice as much from an invasive manifestation [16]. However, Parry reported about sub-Saharan Africa, where invasive infections caused by enteritic serotypes are particularly common in children under 5 years [17]. Reasons for such divergent findings are unknown, but differences between developed and developing countries seem to exist.\nGenotyping revealed 104 strains with a high degree (> 82%) of relationship, summarised in cluster A, which constituted the major part of all isolates. Additionally, several groups of samples were discovered, representing indistinguishable pulsotypes. Strains belonging to cluster A were collected within the years 2004 to 2009 and originated from different countries. This indicates, that these closely related isolates are highly established and widespread.\nIndistinguishable patterns give evidence to small outbreaks or point sources, that spread a certain strain over several years. This becomes particularly apparent regarding non-discriminable isolates derived from patients that stayed in the same foreign country. As already mentioned, most of the strains from patients with a travelling background had returned from the North-African or Asian region. Close clonal relationship between strains from those regions is exemplified by two PFGE sub-clusters within the large Cluster A. One sub-cluster is associated with Egypt (Figure 3, strains no. 8-N1063, 9-N2543, 7-N2080, 8-N2820, 8-N2492, 8-N2393), the other with Thailand (Figure 3, strains no. 6-N0978, 6-N1949, 7-N0090, 7-N0368).\nThe number of resistant (31.5%) and multiresistant (35.5%) strains is considerable. Similar rates from European countries have already been published. In 2004 a study performed by Meakins and co-authors revealed 73% of all S. Virchow strains resistant to at least one antimicrobial agent [18]. Threlfall, in a European multi-centre study with over 27000 cases of salmonellosis in 2000, discovered 36% multi drug resistant isolates [4]. Strikingly, in our study almost all MDR isolates lay within cluster A (Figure 3). Similarities in the resistance situation between the European Union and Switzerland indicate a close relationship of strains belonging to cluster A and strains responsible for the high level of resistance in Europe.\nA high prevalence of nalidixic acid resistant strains was found within cluster A, whereas nearly no resistance to nalidixic acid could be observed among the remaining isolates. A similar behaviour for resistances to tetracycline, sulfonamide and trimetoprim was noted.\nFluoroquinolones and third-generation cephalosporins are drugs-of-choice for invasive Salmonella infections. During the past years numerous studies reported the emergence of fluoroquinolone or third-generation cephalosporine resistant strains [9,11,12]. We found a high prevalence of isolates resistant to nalidixic acid but no full resistance against ciprofloxacin, and only 2% of isolates being intermediate over the whole study period. Because of the intersecting resistance mechanisms against quinolones and fluoroquinolones, the behaviour against ciprofloxacin must be discussed in association with that to nalidixic acid [19]. Because of repeated reports of fluoroquinolone treatment failure following disease caused by Salmonella strains that were susceptible to fluoroquinolones using CLSI criteria, but were simultaneously resistant to nalidixic acid, several authors called for re-evaluation of the CLSI breakpoints for fluoroquinolones involving salmonellae [20-22]. Consequently, comments no. 2, 18, and 20 were added to Table Two A in CLSI documents, as can be seen, for instance, in the 2008 edition [15]. They serve as a warning, that fluoroquinolone-susceptible but nalidixic acid-resistant extraintestinal salmonellae must be considered reduced-susceptible to fluoroquinolones. By the comments, the reader is also urged to perform a nalidixic acid susceptibility test, in order to become aware of such strains. Based on these arguments, the 62% naldixic acid-resistant S. Virchow isolates of the present study have to be considered reduced-susceptible to ciprofloxacin, and thus prone to cause a significant rate of treatment failure. The observation of a low rate of \"obvious\" resistance to fluoroquinolones is in accordance with data from other European countries [4,5]. As we discovered only two ESBL producing isolates in the years 2005 and 2006, the occurrence of such strains in Switzerland are rather rare events compared to other European countries [5,11,23].\nIn many countries, S. Virchow infections are poultry associated [6-9]. In Switzerland, however, there is a favourable situation in poultry flocks with respect to Salmonella in general and S. Virchow in particular [24]. However, 52% of all poultry meat is imported and 71% of the imports originate from Brazil, Germany and France. From all these countries, detection of S. Virchow in poultry products has been reported [7,25,26]. On this basis, one might speculate, that a majority of the S. Virchow infections in Switzerland may be linked to travelling activities and/or consumption of imported poultry meat. This, however, would have to be further investigated by additional case control- and epidemiological studies.", "The worldwide increase of antibiotic resistances in Salmonella is an emerging public health problem. For Switzerland, no clear trend is identifiable over the years 2004 to 2009 for S. Virchow. Antimicrobial susceptibility and resistance profiles varied considerably within this period. Nevertheless, the situation in Switzerland coincided with findings in other European countries. Genotyping results of this strain collection revealed no evidence for an undetected outbreak within this time period", "The authors declare that they have no competing interests.", "RS, LA and HH designed the study. MB, NC, UK have done the phenotypic and genotypic characterization of the strains. MB and RS drafted the manuscript. All authors read, commented on and approved of the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2334/11/49/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Arthritis, Rheumatoid", "Bone Density", "Disease Progression", "Female", "Follow-Up Studies", "Humans", "Image Processing, Computer-Assisted", "Male", "Metacarpal Bones", "Middle Aged", "Radiography" ]

[ "Introduction", "Patients", "Larsen score", "Statistics", "Patients and characteristics according to early BMD-DXR", "Early DXR-BMD association with and prediction of Larsen score", "Comparing DXR-BMD for Larsen scores for hands and feet", "DXR-BMD progression rate over time", "Association with and prediction of early Larsen changes on later Larsen scores", "Baseline BMD or Larsen scores and late Larsen scores", "Abbreviations", "Competing interests", "Authors' contributions" ]

[ "Rheumatoid arthritis (RA) is an inflammatory disease characterized by chronic synovial inflammation commonly associated with destruction in cartilage and bone tissue. Changes in bone metabolism during the course of RA are usually divided into periarticular bone loss with or without focal articular bone erosion and generalized osteopenia manifested by loss of both trabecular and cortical bone. Periarticular bone loss occurs early and often before erosion is apparent. This is linked to the inflammatory process and may be driven by locally released inflammatory mediators such as the receptor activator of nuclear factor (NF)-κB and its ligand RANKL and the decoy receptor of RANKL, osteoprotegerin. These findings have been extensively investigated and reviewed [1-9]. In RA, periarticular bone loss is seen predominantly in the hands and feet, which are affected early in the disease course. Thus, bone loss may be a marker of disease activity, and it has been suggested that it may also relate to future joint damage [4,5,8].\nEarly loss of bone mineral density (BMD) in the hand has been shown to covariate with radiographic joint damage as measured by the Sharp-van der Heide (SvdH) or Larsen score in RA patients [4-11]. However, outcome measures do not necessarily agree. The two more established radiological scores in RA, the Larsen and the SvdH, demonstrate only a modest correlation. This may partly be explained by the fact that the Larsen score is more global, while the SvdH score measures specific joint regions [12-14].\nThe bone mineral density measured by digital X-ray radiogrammetry (DXR-BMD) determination has the advantage of being well standardized and not subject to the interpretation and measurement errors that are inherent in both the Larsen and the SvdH scoring systems. Furthermore, DXR-BMD is reliable in quantifying demineralization and/or osteoporosis, which have been shown to be only imprecisely verified with conventional radiography. The radiographic scoring methods and DXR-BMD can be performed on historical radiographs, provided that these are of sufficient quality and taken in a standardized fashion. This is in contrast to other methods of BMD measurements such as ultrasound, dual-emission X-ray absorptiometry and peripheral quantitative computed tomography, all of which have as a focus the measurement of BMD, are not retrospectively applicable to standard radiographs of the hands and sometimes focus on generalized osteopenia [6-8,15-18].\nIn Lund, Sweden, we have prospectively monitored a cohort of early RA patients since 1985 and have more than 20 years of follow-up information on radiographic and clinical outcomes [19,20]. In the present study, we have examined the relationship between early DXR-BMD changes and short- and long-term outcomes as measured radiographically by the Larsen score of the hands and feet. Furthermore, we have examined the relationship between baseline BMD, Larsen score and early Larsen progression and long-term radiographic outcome.", "A total of 183 patients (115 women and 68 men) with early definite RA who had a mean symptom duration of 11 months (SD ± 7; range, 0 to 24 months) included between 1985 and 1989 were followed prospectively (the Lund early RA cohort). Clinical and functional measures were assessed, as previously reported, at least once yearly [19]. The validated Swedish version of the Health Assessment Questionnaire (HAQ), which includes the use of aids, was used [21]. Approval from the Ethical Review Board at Lund University (LU 525-02) and informed consent from each patient were obtained for this study. Throughout the study all patients with active disease were offered treatment with disease-modifying antirheumatic drugs (DMARDs) according to current clinical practice. About 50% of patients from the original cohort started treatment with DMARDs within 1 year after diagnosis. D-penicillamine and antimalarial drugs were most commonly used in the early years, and methotrexate became most frequently used during the 1990s.\nIn total, 48 patients from the original cohort were excluded from the analysis. In three cases, only hand and no feet radiographs were available as the first radiographs; in eight cases, the first DXR measurements were made on prediagnosis inclusion radiographs which were not Larsen scored; and in an additional 37 patients, radiographs were taken using a magnifying analogous technique precluding DXR measurements. There were no statistically significant differences regarding patient characteristics between included and excluded patients.\nDXR was developed as a computerized method of radiogrammetry to measure cortical bone thickness in diaphysis of the second to fourth metacarpal bones using standard hand radiographs. Thus, DXR-BMD quantifies only the cortical bone tissue where the bone metabolism is minor compared to trabecular bone tissue [9]. The method has been described in more detail elsewhere [9,15-17]. Briefly, the early DXR-BMD change rate (bone loss) per year, expressed in grams per square centimeter, was calculated on the basis of the first two available conventional radiographs of the hands. Exact dates of the radiographs were used to calculate the yearly change rate. The same radiographs were used for both radiographic scoring and measurement of hand bone density. The radiographs were digitized to 300 dpi, 12-bit gray scale format with DICOM software using a Vidar Diagnostic Pro Plus digitizer (VIDAR Systems Corp., Herndon, VA, USA), and the resulting digital radiographs were analyzed using dxr-online software (Sectra, Linköping, Sweden). The dxr-online software recognizes regions of interest around the narrowest part of the second, third and fourth metacarpal bone diaphysis and automatically measures the BMD through a combination of radiogrammetry and textural analysis. Mean values of the right and left hand were used. The smallest detectable difference in elevated early DXR-BMD loss has been shown to be 0.0048 g/cm2 [7].", "Joint damage was evaluated on radiographs of the hands and feet according to the Larsen method as described previously [13]. In short, 32 joints in the hands and feet were evaluated. Each joint was compared to a standard reference film, and changes were graded from 0 to 5, with 0 being normal; 1 being joint space narrowing, soft tissue swelling or periarticular osteoporosis; and 2 to 5 representing a progressively increasing degree of erosion and destruction. A joint damage score was calculated by adding all scores, with the scores for joints in the wrists multiplied by 5, resulting in a range of 0 to 200. Radiographs were obtained at inclusion in the study as well as after 1, 2, 3, 4, 5, 10, 15 and 20 years. Early progression of Larsen scores per year was measured as the difference between the first two available sets of both hand and feet radiograms using the exact dates of the radiograms to calculate the progression rate per year (Larsen units/year).", "Early DXR-BMD changes stratified according to the median and upper and lower halves were compared to each other. No imputation of missing data was performed. Differences between the groups were calculated using the χ2 test for ordinal variables and the Mann-Whitney U test for numerical variables. Influences of age and sex on baseline BMD, early DXR-BMD change and baseline Larsen score were estimated by performing analysis of covariance (ANCOVA). To maximize statistical power, differences between Larsen scores at 10 years were analyzed using a generalized linear regression model adjusted for age, sex and baseline values of the variable tested. This was done because 69 patients died during follow-up (the majority after 10 years), and thus only contributed information for a more limited follow-up period. Correlations were calculated using Spearman's rank correlation coefficient. Values are given as means with 95% confidence intervals (95% CIs) unless stated otherwise. A P value < 0.05 was considered statistically significant.", "Complete data were available for 135 patients. Mean (± SD) DXR-BMD at baseline was 0.593 ± 0.08 g/cm2. Mean (± SD) DXR-BMD progression rate (bone loss), defined as change in DXR-BMD between the first two existing radiographs was -0.023 g/cm2 ± 0.025.\nPatient characteristics stratified according to the median value (-0.019) of early DXR-BMD loss are presented in Table 1. This stratification identified several differences in baseline variables. Patients with high bone loss were older, had longer symptom duration at diagnosis and had higher erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), HAQ scores and Disease Activity Score using 44 joint counts (DAS44), while sex and Larsen score were not significantly different. ANCOVA showed that older patients had both significantly lower baseline BMD (P < 0.001) and higher early DXR-BMD changes (P < 0.001), while females only had lower baseline BMD (P < 0.001), but not significantly higher early DXR-BMD changes (P = 0.690).\nPatient characteristics at diagnosis stratified according to the median of the early DXR-BMD lossa\naSD, standard deviation; DXR-BMD, bone mineral density (BMD) measured by digital X-ray radiogrammetry; HAQ, Health Assessment Questionnaire; DAS44, Disease Activity Score using 44 joint counts; ESR, erythrocyte sedimentation rate; CRP, C-reactive protein; RF, rheumatoid factor; ACPA, anticitrullinated peptide antibodies; DMARDs, disease-modifying antirheumatic drugs; MTX, methotrexate; ns, not significant. Medication refers to treatment for the whole observation period.", "On the basis of univariate linear regression analysis, early bone loss was significantly associated with higher Larsen score at year 10 as were ESR and CRP levels, HAQ score and Larsen score at baseline. ANCOVA showed that the adjusted 10-year Larsen score in the group with high early bone loss (mean, 95% CI) was 30.2 (95% CI, 15.7 to 44.7) above the group with low early bone loss (P < 0.001).\nIn multivariate regression analysis, high early DXR-BMD loss (patients with early bone loss levels above the median), baseline Larsen scores and ESR levels remained predictive for the Larsen score at year 10 (Table 2). In patients with erosive disease at baseline (that is, Larsen score at baseline ≥2) who had early DXR-BMD, the median value was still associated with a higher Larsen score at 10-year follow-up (P = 0.031). Very few patients (n = 2) in this cohort had already been treated with prednisolone at RA diagnosis (baseline), and therefore this variable was not entered into the regression model.\nImpact of high early DXR-BMD loss and baseline demographic and disease characteristics on Larsen score at year 10 (linear regression model)a\na B, the estimated regression coefficient; ESR, erythrocyte sedimentation rate; ACPA, anticitrullinated peptide antibodies; high early DXR-BMD loss is defined as bone loss above median value.\nLarsen scores at years 0, 1, 3, 4, 5, 10, 15 and 20 stratified according to the median of early DXR-BMD loss are given in Figure 1. There is a diverging trend during the whole follow-up period, which is most marked in the first 5 years. The increase in 95% CI at the later time points illustrates the increasing number of deceased patients. When using the smallest detectable difference, 0.0048 g/cm2/year [7], as a cutoff for elevated early DXR-BMD loss, a total of 97 patients (72%) were in the higher bone loss group, while 38 patients (28%) were in the lower bone loss group. Plotting the patients with early bone loss >0.0048 compared to those with lower early bone loss yielded results similar to those of median stratification (data not shown). During the first 5 years, the Larsen score deteriorated at a faster pace in patients with elevated early bone loss. However at time points 10, 15 and 20 years, these differences were not significant between the groups. The Spearman correlations between early bone loss and Larsen scores at years 1, 2, 3, 4, 5, 10 and 20 were all significant, with Spearman's ρ values between 0.359 and 0.223.\nRadiological Larsen score progression over time after stratification according to the median (-0.0185 g/cm2 per year) of early bone mineral density measured by digital X-ray radiogrammetry (DXR-BMD) change. High loss (black solid line) versus low loss (red dotted line) of DXR-BMD. Values are given as means with 95% confidence intervals (95% CIs). The larger 95% CIs at 15 and 20 years illustrate the decreasing number of patients at these time points.", "When looking at Larsen scores for hands and feet separately, the clear differences between groups stratified according to early DXR-BMD loss remained for the hands. For the Larsen scores of the feet, the picture was similar, with the group with high early DXR-BMD loss always having higher mean Larsen scores, but at most time points the differences were nonsignificant (data not shown).", "To study the progression rate of DXR-BMD loss over time, we calculated the data for the yearly progression rates between time points 0 and 1, 2, 3 and 4 years, which were (means and 95% CIs): 2.2 (1.3 to 3.0), 2.2 (1.5 to 2.9), 2.7 (2.0 to 3.3) and 2.3 (1.8 to 2.8), respectively. There was an overall trend of decreasing DXR-BMD progression rate over time, but the limited number of patients precludes firm statistical confirmation.", "After stratifying for the median of early changes in Larsen scores, a very similar picture to that observed for early DXR-BMD changes was found (Figure 2). Also, here there is a diverging trend during the whole follow-up period, and ANCOVA showed that the adjusted 10-year Larsen scores (95% CI) in the group with more elevated early Larsen score progression was 32.3 (20.6 to 44.0) above the group with low early Larsen score progression. The association between early Larsen score changes with 10-year Larsen scores was 0.540 (Spearman's ρ).\nRadiological Larsen score progression over time after stratification according to the median (5.5 units) of early change in Larsen score. High (black solid line) versus low (red dotted line) early ΔLarsen scores. Values are given as means with 95% confidence intervals (95% CIs). The larger 95% CIs at 15 and 20 years illustrate the decreasing number of patients at these time points.", "Median stratification of baseline BMD or Larsen score did not identify any trends of predictability for long-term radiological progression as measured by the Larsen score (Figures 3 and 4). The Spearman's ρ correlation between 10-year Larsen score and baseline BMD was -0.056, and between 10-year Larsen score and baseline Larsen score it was 0.199.\nRadiological Larsen score progression over time after stratification according to median (0.6 g/cm2) baseline BMD. High BMD (black solid line) versus low BMD (red dotted line) values at baseline are shown.\nRadiological Larsen score progression over time after stratification according to median (5.5 units) of baseline Larsen score. High Larsen score (black solid line) versus low Larsen score (red dotted line) at baseline is shown.", "BMD: bone mineral density; DXR: digital X-ray radiogrammetry; DXR-BMD: bone mineral density measured by digital X-ray radiogrammetry; RA: rheumatoid arthritis.", "JA is employed by Sectra but is not a shareholder in the company. All other authors have declared no competing interests. The radiographs were digitized using a Vidar Diagnostic Pro Plus digitizer (VIDAR Systems Corp., Herndon, VA, USA), and the resulting digital radiographs were analyzed using DXR online (Sectra, Linköping, Sweden), but any financial or other support was obtained from the company.", "All authors have made substantial contributions to the study's conception and design, acquisition of data or analysis and interpretation of data and have been involved in drafting the manuscript or revising it critically for important intellectual content. MCK, EL and PG wrote the manuscript. EL and PG helped plan the study. JA developed the DXR method. KJL scored the radiograms. EL, TS and KE conceived the study. MCK helped in performing the statistical analysis. KE conceived the original Lund early RA cohort study. PG handled the database. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Materials and methods", "Patients", "Larsen score", "Statistics", "Results", "Patients and characteristics according to early BMD-DXR", "Early DXR-BMD association with and prediction of Larsen score", "Comparing DXR-BMD for Larsen scores for hands and feet", "DXR-BMD progression rate over time", "Association with and prediction of early Larsen changes on later Larsen scores", "Baseline BMD or Larsen scores and late Larsen scores", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions" ]

[ "Rheumatoid arthritis (RA) is an inflammatory disease characterized by chronic synovial inflammation commonly associated with destruction in cartilage and bone tissue. Changes in bone metabolism during the course of RA are usually divided into periarticular bone loss with or without focal articular bone erosion and generalized osteopenia manifested by loss of both trabecular and cortical bone. Periarticular bone loss occurs early and often before erosion is apparent. This is linked to the inflammatory process and may be driven by locally released inflammatory mediators such as the receptor activator of nuclear factor (NF)-κB and its ligand RANKL and the decoy receptor of RANKL, osteoprotegerin. These findings have been extensively investigated and reviewed [1-9]. In RA, periarticular bone loss is seen predominantly in the hands and feet, which are affected early in the disease course. Thus, bone loss may be a marker of disease activity, and it has been suggested that it may also relate to future joint damage [4,5,8].\nEarly loss of bone mineral density (BMD) in the hand has been shown to covariate with radiographic joint damage as measured by the Sharp-van der Heide (SvdH) or Larsen score in RA patients [4-11]. However, outcome measures do not necessarily agree. The two more established radiological scores in RA, the Larsen and the SvdH, demonstrate only a modest correlation. This may partly be explained by the fact that the Larsen score is more global, while the SvdH score measures specific joint regions [12-14].\nThe bone mineral density measured by digital X-ray radiogrammetry (DXR-BMD) determination has the advantage of being well standardized and not subject to the interpretation and measurement errors that are inherent in both the Larsen and the SvdH scoring systems. Furthermore, DXR-BMD is reliable in quantifying demineralization and/or osteoporosis, which have been shown to be only imprecisely verified with conventional radiography. The radiographic scoring methods and DXR-BMD can be performed on historical radiographs, provided that these are of sufficient quality and taken in a standardized fashion. This is in contrast to other methods of BMD measurements such as ultrasound, dual-emission X-ray absorptiometry and peripheral quantitative computed tomography, all of which have as a focus the measurement of BMD, are not retrospectively applicable to standard radiographs of the hands and sometimes focus on generalized osteopenia [6-8,15-18].\nIn Lund, Sweden, we have prospectively monitored a cohort of early RA patients since 1985 and have more than 20 years of follow-up information on radiographic and clinical outcomes [19,20]. In the present study, we have examined the relationship between early DXR-BMD changes and short- and long-term outcomes as measured radiographically by the Larsen score of the hands and feet. Furthermore, we have examined the relationship between baseline BMD, Larsen score and early Larsen progression and long-term radiographic outcome.", "[SUBTITLE] Patients [SUBSECTION] A total of 183 patients (115 women and 68 men) with early definite RA who had a mean symptom duration of 11 months (SD ± 7; range, 0 to 24 months) included between 1985 and 1989 were followed prospectively (the Lund early RA cohort). Clinical and functional measures were assessed, as previously reported, at least once yearly [19]. The validated Swedish version of the Health Assessment Questionnaire (HAQ), which includes the use of aids, was used [21]. Approval from the Ethical Review Board at Lund University (LU 525-02) and informed consent from each patient were obtained for this study. Throughout the study all patients with active disease were offered treatment with disease-modifying antirheumatic drugs (DMARDs) according to current clinical practice. About 50% of patients from the original cohort started treatment with DMARDs within 1 year after diagnosis. D-penicillamine and antimalarial drugs were most commonly used in the early years, and methotrexate became most frequently used during the 1990s.\nIn total, 48 patients from the original cohort were excluded from the analysis. In three cases, only hand and no feet radiographs were available as the first radiographs; in eight cases, the first DXR measurements were made on prediagnosis inclusion radiographs which were not Larsen scored; and in an additional 37 patients, radiographs were taken using a magnifying analogous technique precluding DXR measurements. There were no statistically significant differences regarding patient characteristics between included and excluded patients.\nDXR was developed as a computerized method of radiogrammetry to measure cortical bone thickness in diaphysis of the second to fourth metacarpal bones using standard hand radiographs. Thus, DXR-BMD quantifies only the cortical bone tissue where the bone metabolism is minor compared to trabecular bone tissue [9]. The method has been described in more detail elsewhere [9,15-17]. Briefly, the early DXR-BMD change rate (bone loss) per year, expressed in grams per square centimeter, was calculated on the basis of the first two available conventional radiographs of the hands. Exact dates of the radiographs were used to calculate the yearly change rate. The same radiographs were used for both radiographic scoring and measurement of hand bone density. The radiographs were digitized to 300 dpi, 12-bit gray scale format with DICOM software using a Vidar Diagnostic Pro Plus digitizer (VIDAR Systems Corp., Herndon, VA, USA), and the resulting digital radiographs were analyzed using dxr-online software (Sectra, Linköping, Sweden). The dxr-online software recognizes regions of interest around the narrowest part of the second, third and fourth metacarpal bone diaphysis and automatically measures the BMD through a combination of radiogrammetry and textural analysis. Mean values of the right and left hand were used. The smallest detectable difference in elevated early DXR-BMD loss has been shown to be 0.0048 g/cm2 [7].\nA total of 183 patients (115 women and 68 men) with early definite RA who had a mean symptom duration of 11 months (SD ± 7; range, 0 to 24 months) included between 1985 and 1989 were followed prospectively (the Lund early RA cohort). Clinical and functional measures were assessed, as previously reported, at least once yearly [19]. The validated Swedish version of the Health Assessment Questionnaire (HAQ), which includes the use of aids, was used [21]. Approval from the Ethical Review Board at Lund University (LU 525-02) and informed consent from each patient were obtained for this study. Throughout the study all patients with active disease were offered treatment with disease-modifying antirheumatic drugs (DMARDs) according to current clinical practice. About 50% of patients from the original cohort started treatment with DMARDs within 1 year after diagnosis. D-penicillamine and antimalarial drugs were most commonly used in the early years, and methotrexate became most frequently used during the 1990s.\nIn total, 48 patients from the original cohort were excluded from the analysis. In three cases, only hand and no feet radiographs were available as the first radiographs; in eight cases, the first DXR measurements were made on prediagnosis inclusion radiographs which were not Larsen scored; and in an additional 37 patients, radiographs were taken using a magnifying analogous technique precluding DXR measurements. There were no statistically significant differences regarding patient characteristics between included and excluded patients.\nDXR was developed as a computerized method of radiogrammetry to measure cortical bone thickness in diaphysis of the second to fourth metacarpal bones using standard hand radiographs. Thus, DXR-BMD quantifies only the cortical bone tissue where the bone metabolism is minor compared to trabecular bone tissue [9]. The method has been described in more detail elsewhere [9,15-17]. Briefly, the early DXR-BMD change rate (bone loss) per year, expressed in grams per square centimeter, was calculated on the basis of the first two available conventional radiographs of the hands. Exact dates of the radiographs were used to calculate the yearly change rate. The same radiographs were used for both radiographic scoring and measurement of hand bone density. The radiographs were digitized to 300 dpi, 12-bit gray scale format with DICOM software using a Vidar Diagnostic Pro Plus digitizer (VIDAR Systems Corp., Herndon, VA, USA), and the resulting digital radiographs were analyzed using dxr-online software (Sectra, Linköping, Sweden). The dxr-online software recognizes regions of interest around the narrowest part of the second, third and fourth metacarpal bone diaphysis and automatically measures the BMD through a combination of radiogrammetry and textural analysis. Mean values of the right and left hand were used. The smallest detectable difference in elevated early DXR-BMD loss has been shown to be 0.0048 g/cm2 [7].\n[SUBTITLE] Larsen score [SUBSECTION] Joint damage was evaluated on radiographs of the hands and feet according to the Larsen method as described previously [13]. In short, 32 joints in the hands and feet were evaluated. Each joint was compared to a standard reference film, and changes were graded from 0 to 5, with 0 being normal; 1 being joint space narrowing, soft tissue swelling or periarticular osteoporosis; and 2 to 5 representing a progressively increasing degree of erosion and destruction. A joint damage score was calculated by adding all scores, with the scores for joints in the wrists multiplied by 5, resulting in a range of 0 to 200. Radiographs were obtained at inclusion in the study as well as after 1, 2, 3, 4, 5, 10, 15 and 20 years. Early progression of Larsen scores per year was measured as the difference between the first two available sets of both hand and feet radiograms using the exact dates of the radiograms to calculate the progression rate per year (Larsen units/year).\nJoint damage was evaluated on radiographs of the hands and feet according to the Larsen method as described previously [13]. In short, 32 joints in the hands and feet were evaluated. Each joint was compared to a standard reference film, and changes were graded from 0 to 5, with 0 being normal; 1 being joint space narrowing, soft tissue swelling or periarticular osteoporosis; and 2 to 5 representing a progressively increasing degree of erosion and destruction. A joint damage score was calculated by adding all scores, with the scores for joints in the wrists multiplied by 5, resulting in a range of 0 to 200. Radiographs were obtained at inclusion in the study as well as after 1, 2, 3, 4, 5, 10, 15 and 20 years. Early progression of Larsen scores per year was measured as the difference between the first two available sets of both hand and feet radiograms using the exact dates of the radiograms to calculate the progression rate per year (Larsen units/year).\n[SUBTITLE] Statistics [SUBSECTION] Early DXR-BMD changes stratified according to the median and upper and lower halves were compared to each other. No imputation of missing data was performed. Differences between the groups were calculated using the χ2 test for ordinal variables and the Mann-Whitney U test for numerical variables. Influences of age and sex on baseline BMD, early DXR-BMD change and baseline Larsen score were estimated by performing analysis of covariance (ANCOVA). To maximize statistical power, differences between Larsen scores at 10 years were analyzed using a generalized linear regression model adjusted for age, sex and baseline values of the variable tested. This was done because 69 patients died during follow-up (the majority after 10 years), and thus only contributed information for a more limited follow-up period. Correlations were calculated using Spearman's rank correlation coefficient. Values are given as means with 95% confidence intervals (95% CIs) unless stated otherwise. A P value < 0.05 was considered statistically significant.\nEarly DXR-BMD changes stratified according to the median and upper and lower halves were compared to each other. No imputation of missing data was performed. Differences between the groups were calculated using the χ2 test for ordinal variables and the Mann-Whitney U test for numerical variables. Influences of age and sex on baseline BMD, early DXR-BMD change and baseline Larsen score were estimated by performing analysis of covariance (ANCOVA). To maximize statistical power, differences between Larsen scores at 10 years were analyzed using a generalized linear regression model adjusted for age, sex and baseline values of the variable tested. This was done because 69 patients died during follow-up (the majority after 10 years), and thus only contributed information for a more limited follow-up period. Correlations were calculated using Spearman's rank correlation coefficient. Values are given as means with 95% confidence intervals (95% CIs) unless stated otherwise. A P value < 0.05 was considered statistically significant.", "A total of 183 patients (115 women and 68 men) with early definite RA who had a mean symptom duration of 11 months (SD ± 7; range, 0 to 24 months) included between 1985 and 1989 were followed prospectively (the Lund early RA cohort). Clinical and functional measures were assessed, as previously reported, at least once yearly [19]. The validated Swedish version of the Health Assessment Questionnaire (HAQ), which includes the use of aids, was used [21]. Approval from the Ethical Review Board at Lund University (LU 525-02) and informed consent from each patient were obtained for this study. Throughout the study all patients with active disease were offered treatment with disease-modifying antirheumatic drugs (DMARDs) according to current clinical practice. About 50% of patients from the original cohort started treatment with DMARDs within 1 year after diagnosis. D-penicillamine and antimalarial drugs were most commonly used in the early years, and methotrexate became most frequently used during the 1990s.\nIn total, 48 patients from the original cohort were excluded from the analysis. In three cases, only hand and no feet radiographs were available as the first radiographs; in eight cases, the first DXR measurements were made on prediagnosis inclusion radiographs which were not Larsen scored; and in an additional 37 patients, radiographs were taken using a magnifying analogous technique precluding DXR measurements. There were no statistically significant differences regarding patient characteristics between included and excluded patients.\nDXR was developed as a computerized method of radiogrammetry to measure cortical bone thickness in diaphysis of the second to fourth metacarpal bones using standard hand radiographs. Thus, DXR-BMD quantifies only the cortical bone tissue where the bone metabolism is minor compared to trabecular bone tissue [9]. The method has been described in more detail elsewhere [9,15-17]. Briefly, the early DXR-BMD change rate (bone loss) per year, expressed in grams per square centimeter, was calculated on the basis of the first two available conventional radiographs of the hands. Exact dates of the radiographs were used to calculate the yearly change rate. The same radiographs were used for both radiographic scoring and measurement of hand bone density. The radiographs were digitized to 300 dpi, 12-bit gray scale format with DICOM software using a Vidar Diagnostic Pro Plus digitizer (VIDAR Systems Corp., Herndon, VA, USA), and the resulting digital radiographs were analyzed using dxr-online software (Sectra, Linköping, Sweden). The dxr-online software recognizes regions of interest around the narrowest part of the second, third and fourth metacarpal bone diaphysis and automatically measures the BMD through a combination of radiogrammetry and textural analysis. Mean values of the right and left hand were used. The smallest detectable difference in elevated early DXR-BMD loss has been shown to be 0.0048 g/cm2 [7].", "Joint damage was evaluated on radiographs of the hands and feet according to the Larsen method as described previously [13]. In short, 32 joints in the hands and feet were evaluated. Each joint was compared to a standard reference film, and changes were graded from 0 to 5, with 0 being normal; 1 being joint space narrowing, soft tissue swelling or periarticular osteoporosis; and 2 to 5 representing a progressively increasing degree of erosion and destruction. A joint damage score was calculated by adding all scores, with the scores for joints in the wrists multiplied by 5, resulting in a range of 0 to 200. Radiographs were obtained at inclusion in the study as well as after 1, 2, 3, 4, 5, 10, 15 and 20 years. Early progression of Larsen scores per year was measured as the difference between the first two available sets of both hand and feet radiograms using the exact dates of the radiograms to calculate the progression rate per year (Larsen units/year).", "Early DXR-BMD changes stratified according to the median and upper and lower halves were compared to each other. No imputation of missing data was performed. Differences between the groups were calculated using the χ2 test for ordinal variables and the Mann-Whitney U test for numerical variables. Influences of age and sex on baseline BMD, early DXR-BMD change and baseline Larsen score were estimated by performing analysis of covariance (ANCOVA). To maximize statistical power, differences between Larsen scores at 10 years were analyzed using a generalized linear regression model adjusted for age, sex and baseline values of the variable tested. This was done because 69 patients died during follow-up (the majority after 10 years), and thus only contributed information for a more limited follow-up period. Correlations were calculated using Spearman's rank correlation coefficient. Values are given as means with 95% confidence intervals (95% CIs) unless stated otherwise. A P value < 0.05 was considered statistically significant.", "[SUBTITLE] Patients and characteristics according to early BMD-DXR [SUBSECTION] Complete data were available for 135 patients. Mean (± SD) DXR-BMD at baseline was 0.593 ± 0.08 g/cm2. Mean (± SD) DXR-BMD progression rate (bone loss), defined as change in DXR-BMD between the first two existing radiographs was -0.023 g/cm2 ± 0.025.\nPatient characteristics stratified according to the median value (-0.019) of early DXR-BMD loss are presented in Table 1. This stratification identified several differences in baseline variables. Patients with high bone loss were older, had longer symptom duration at diagnosis and had higher erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), HAQ scores and Disease Activity Score using 44 joint counts (DAS44), while sex and Larsen score were not significantly different. ANCOVA showed that older patients had both significantly lower baseline BMD (P < 0.001) and higher early DXR-BMD changes (P < 0.001), while females only had lower baseline BMD (P < 0.001), but not significantly higher early DXR-BMD changes (P = 0.690).\nPatient characteristics at diagnosis stratified according to the median of the early DXR-BMD lossa\naSD, standard deviation; DXR-BMD, bone mineral density (BMD) measured by digital X-ray radiogrammetry; HAQ, Health Assessment Questionnaire; DAS44, Disease Activity Score using 44 joint counts; ESR, erythrocyte sedimentation rate; CRP, C-reactive protein; RF, rheumatoid factor; ACPA, anticitrullinated peptide antibodies; DMARDs, disease-modifying antirheumatic drugs; MTX, methotrexate; ns, not significant. Medication refers to treatment for the whole observation period.\nComplete data were available for 135 patients. Mean (± SD) DXR-BMD at baseline was 0.593 ± 0.08 g/cm2. Mean (± SD) DXR-BMD progression rate (bone loss), defined as change in DXR-BMD between the first two existing radiographs was -0.023 g/cm2 ± 0.025.\nPatient characteristics stratified according to the median value (-0.019) of early DXR-BMD loss are presented in Table 1. This stratification identified several differences in baseline variables. Patients with high bone loss were older, had longer symptom duration at diagnosis and had higher erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), HAQ scores and Disease Activity Score using 44 joint counts (DAS44), while sex and Larsen score were not significantly different. ANCOVA showed that older patients had both significantly lower baseline BMD (P < 0.001) and higher early DXR-BMD changes (P < 0.001), while females only had lower baseline BMD (P < 0.001), but not significantly higher early DXR-BMD changes (P = 0.690).\nPatient characteristics at diagnosis stratified according to the median of the early DXR-BMD lossa\naSD, standard deviation; DXR-BMD, bone mineral density (BMD) measured by digital X-ray radiogrammetry; HAQ, Health Assessment Questionnaire; DAS44, Disease Activity Score using 44 joint counts; ESR, erythrocyte sedimentation rate; CRP, C-reactive protein; RF, rheumatoid factor; ACPA, anticitrullinated peptide antibodies; DMARDs, disease-modifying antirheumatic drugs; MTX, methotrexate; ns, not significant. Medication refers to treatment for the whole observation period.\n[SUBTITLE] Early DXR-BMD association with and prediction of Larsen score [SUBSECTION] On the basis of univariate linear regression analysis, early bone loss was significantly associated with higher Larsen score at year 10 as were ESR and CRP levels, HAQ score and Larsen score at baseline. ANCOVA showed that the adjusted 10-year Larsen score in the group with high early bone loss (mean, 95% CI) was 30.2 (95% CI, 15.7 to 44.7) above the group with low early bone loss (P < 0.001).\nIn multivariate regression analysis, high early DXR-BMD loss (patients with early bone loss levels above the median), baseline Larsen scores and ESR levels remained predictive for the Larsen score at year 10 (Table 2). In patients with erosive disease at baseline (that is, Larsen score at baseline ≥2) who had early DXR-BMD, the median value was still associated with a higher Larsen score at 10-year follow-up (P = 0.031). Very few patients (n = 2) in this cohort had already been treated with prednisolone at RA diagnosis (baseline), and therefore this variable was not entered into the regression model.\nImpact of high early DXR-BMD loss and baseline demographic and disease characteristics on Larsen score at year 10 (linear regression model)a\na B, the estimated regression coefficient; ESR, erythrocyte sedimentation rate; ACPA, anticitrullinated peptide antibodies; high early DXR-BMD loss is defined as bone loss above median value.\nLarsen scores at years 0, 1, 3, 4, 5, 10, 15 and 20 stratified according to the median of early DXR-BMD loss are given in Figure 1. There is a diverging trend during the whole follow-up period, which is most marked in the first 5 years. The increase in 95% CI at the later time points illustrates the increasing number of deceased patients. When using the smallest detectable difference, 0.0048 g/cm2/year [7], as a cutoff for elevated early DXR-BMD loss, a total of 97 patients (72%) were in the higher bone loss group, while 38 patients (28%) were in the lower bone loss group. Plotting the patients with early bone loss >0.0048 compared to those with lower early bone loss yielded results similar to those of median stratification (data not shown). During the first 5 years, the Larsen score deteriorated at a faster pace in patients with elevated early bone loss. However at time points 10, 15 and 20 years, these differences were not significant between the groups. The Spearman correlations between early bone loss and Larsen scores at years 1, 2, 3, 4, 5, 10 and 20 were all significant, with Spearman's ρ values between 0.359 and 0.223.\nRadiological Larsen score progression over time after stratification according to the median (-0.0185 g/cm2 per year) of early bone mineral density measured by digital X-ray radiogrammetry (DXR-BMD) change. High loss (black solid line) versus low loss (red dotted line) of DXR-BMD. Values are given as means with 95% confidence intervals (95% CIs). The larger 95% CIs at 15 and 20 years illustrate the decreasing number of patients at these time points.\nOn the basis of univariate linear regression analysis, early bone loss was significantly associated with higher Larsen score at year 10 as were ESR and CRP levels, HAQ score and Larsen score at baseline. ANCOVA showed that the adjusted 10-year Larsen score in the group with high early bone loss (mean, 95% CI) was 30.2 (95% CI, 15.7 to 44.7) above the group with low early bone loss (P < 0.001).\nIn multivariate regression analysis, high early DXR-BMD loss (patients with early bone loss levels above the median), baseline Larsen scores and ESR levels remained predictive for the Larsen score at year 10 (Table 2). In patients with erosive disease at baseline (that is, Larsen score at baseline ≥2) who had early DXR-BMD, the median value was still associated with a higher Larsen score at 10-year follow-up (P = 0.031). Very few patients (n = 2) in this cohort had already been treated with prednisolone at RA diagnosis (baseline), and therefore this variable was not entered into the regression model.\nImpact of high early DXR-BMD loss and baseline demographic and disease characteristics on Larsen score at year 10 (linear regression model)a\na B, the estimated regression coefficient; ESR, erythrocyte sedimentation rate; ACPA, anticitrullinated peptide antibodies; high early DXR-BMD loss is defined as bone loss above median value.\nLarsen scores at years 0, 1, 3, 4, 5, 10, 15 and 20 stratified according to the median of early DXR-BMD loss are given in Figure 1. There is a diverging trend during the whole follow-up period, which is most marked in the first 5 years. The increase in 95% CI at the later time points illustrates the increasing number of deceased patients. When using the smallest detectable difference, 0.0048 g/cm2/year [7], as a cutoff for elevated early DXR-BMD loss, a total of 97 patients (72%) were in the higher bone loss group, while 38 patients (28%) were in the lower bone loss group. Plotting the patients with early bone loss >0.0048 compared to those with lower early bone loss yielded results similar to those of median stratification (data not shown). During the first 5 years, the Larsen score deteriorated at a faster pace in patients with elevated early bone loss. However at time points 10, 15 and 20 years, these differences were not significant between the groups. The Spearman correlations between early bone loss and Larsen scores at years 1, 2, 3, 4, 5, 10 and 20 were all significant, with Spearman's ρ values between 0.359 and 0.223.\nRadiological Larsen score progression over time after stratification according to the median (-0.0185 g/cm2 per year) of early bone mineral density measured by digital X-ray radiogrammetry (DXR-BMD) change. High loss (black solid line) versus low loss (red dotted line) of DXR-BMD. Values are given as means with 95% confidence intervals (95% CIs). The larger 95% CIs at 15 and 20 years illustrate the decreasing number of patients at these time points.\n[SUBTITLE] Comparing DXR-BMD for Larsen scores for hands and feet [SUBSECTION] When looking at Larsen scores for hands and feet separately, the clear differences between groups stratified according to early DXR-BMD loss remained for the hands. For the Larsen scores of the feet, the picture was similar, with the group with high early DXR-BMD loss always having higher mean Larsen scores, but at most time points the differences were nonsignificant (data not shown).\nWhen looking at Larsen scores for hands and feet separately, the clear differences between groups stratified according to early DXR-BMD loss remained for the hands. For the Larsen scores of the feet, the picture was similar, with the group with high early DXR-BMD loss always having higher mean Larsen scores, but at most time points the differences were nonsignificant (data not shown).\n[SUBTITLE] DXR-BMD progression rate over time [SUBSECTION] To study the progression rate of DXR-BMD loss over time, we calculated the data for the yearly progression rates between time points 0 and 1, 2, 3 and 4 years, which were (means and 95% CIs): 2.2 (1.3 to 3.0), 2.2 (1.5 to 2.9), 2.7 (2.0 to 3.3) and 2.3 (1.8 to 2.8), respectively. There was an overall trend of decreasing DXR-BMD progression rate over time, but the limited number of patients precludes firm statistical confirmation.\nTo study the progression rate of DXR-BMD loss over time, we calculated the data for the yearly progression rates between time points 0 and 1, 2, 3 and 4 years, which were (means and 95% CIs): 2.2 (1.3 to 3.0), 2.2 (1.5 to 2.9), 2.7 (2.0 to 3.3) and 2.3 (1.8 to 2.8), respectively. There was an overall trend of decreasing DXR-BMD progression rate over time, but the limited number of patients precludes firm statistical confirmation.\n[SUBTITLE] Association with and prediction of early Larsen changes on later Larsen scores [SUBSECTION] After stratifying for the median of early changes in Larsen scores, a very similar picture to that observed for early DXR-BMD changes was found (Figure 2). Also, here there is a diverging trend during the whole follow-up period, and ANCOVA showed that the adjusted 10-year Larsen scores (95% CI) in the group with more elevated early Larsen score progression was 32.3 (20.6 to 44.0) above the group with low early Larsen score progression. The association between early Larsen score changes with 10-year Larsen scores was 0.540 (Spearman's ρ).\nRadiological Larsen score progression over time after stratification according to the median (5.5 units) of early change in Larsen score. High (black solid line) versus low (red dotted line) early ΔLarsen scores. Values are given as means with 95% confidence intervals (95% CIs). The larger 95% CIs at 15 and 20 years illustrate the decreasing number of patients at these time points.\nAfter stratifying for the median of early changes in Larsen scores, a very similar picture to that observed for early DXR-BMD changes was found (Figure 2). Also, here there is a diverging trend during the whole follow-up period, and ANCOVA showed that the adjusted 10-year Larsen scores (95% CI) in the group with more elevated early Larsen score progression was 32.3 (20.6 to 44.0) above the group with low early Larsen score progression. The association between early Larsen score changes with 10-year Larsen scores was 0.540 (Spearman's ρ).\nRadiological Larsen score progression over time after stratification according to the median (5.5 units) of early change in Larsen score. High (black solid line) versus low (red dotted line) early ΔLarsen scores. Values are given as means with 95% confidence intervals (95% CIs). The larger 95% CIs at 15 and 20 years illustrate the decreasing number of patients at these time points.\n[SUBTITLE] Baseline BMD or Larsen scores and late Larsen scores [SUBSECTION] Median stratification of baseline BMD or Larsen score did not identify any trends of predictability for long-term radiological progression as measured by the Larsen score (Figures 3 and 4). The Spearman's ρ correlation between 10-year Larsen score and baseline BMD was -0.056, and between 10-year Larsen score and baseline Larsen score it was 0.199.\nRadiological Larsen score progression over time after stratification according to median (0.6 g/cm2) baseline BMD. High BMD (black solid line) versus low BMD (red dotted line) values at baseline are shown.\nRadiological Larsen score progression over time after stratification according to median (5.5 units) of baseline Larsen score. High Larsen score (black solid line) versus low Larsen score (red dotted line) at baseline is shown.\nMedian stratification of baseline BMD or Larsen score did not identify any trends of predictability for long-term radiological progression as measured by the Larsen score (Figures 3 and 4). The Spearman's ρ correlation between 10-year Larsen score and baseline BMD was -0.056, and between 10-year Larsen score and baseline Larsen score it was 0.199.\nRadiological Larsen score progression over time after stratification according to median (0.6 g/cm2) baseline BMD. High BMD (black solid line) versus low BMD (red dotted line) values at baseline are shown.\nRadiological Larsen score progression over time after stratification according to median (5.5 units) of baseline Larsen score. High Larsen score (black solid line) versus low Larsen score (red dotted line) at baseline is shown.", "Complete data were available for 135 patients. Mean (± SD) DXR-BMD at baseline was 0.593 ± 0.08 g/cm2. Mean (± SD) DXR-BMD progression rate (bone loss), defined as change in DXR-BMD between the first two existing radiographs was -0.023 g/cm2 ± 0.025.\nPatient characteristics stratified according to the median value (-0.019) of early DXR-BMD loss are presented in Table 1. This stratification identified several differences in baseline variables. Patients with high bone loss were older, had longer symptom duration at diagnosis and had higher erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), HAQ scores and Disease Activity Score using 44 joint counts (DAS44), while sex and Larsen score were not significantly different. ANCOVA showed that older patients had both significantly lower baseline BMD (P < 0.001) and higher early DXR-BMD changes (P < 0.001), while females only had lower baseline BMD (P < 0.001), but not significantly higher early DXR-BMD changes (P = 0.690).\nPatient characteristics at diagnosis stratified according to the median of the early DXR-BMD lossa\naSD, standard deviation; DXR-BMD, bone mineral density (BMD) measured by digital X-ray radiogrammetry; HAQ, Health Assessment Questionnaire; DAS44, Disease Activity Score using 44 joint counts; ESR, erythrocyte sedimentation rate; CRP, C-reactive protein; RF, rheumatoid factor; ACPA, anticitrullinated peptide antibodies; DMARDs, disease-modifying antirheumatic drugs; MTX, methotrexate; ns, not significant. Medication refers to treatment for the whole observation period.", "On the basis of univariate linear regression analysis, early bone loss was significantly associated with higher Larsen score at year 10 as were ESR and CRP levels, HAQ score and Larsen score at baseline. ANCOVA showed that the adjusted 10-year Larsen score in the group with high early bone loss (mean, 95% CI) was 30.2 (95% CI, 15.7 to 44.7) above the group with low early bone loss (P < 0.001).\nIn multivariate regression analysis, high early DXR-BMD loss (patients with early bone loss levels above the median), baseline Larsen scores and ESR levels remained predictive for the Larsen score at year 10 (Table 2). In patients with erosive disease at baseline (that is, Larsen score at baseline ≥2) who had early DXR-BMD, the median value was still associated with a higher Larsen score at 10-year follow-up (P = 0.031). Very few patients (n = 2) in this cohort had already been treated with prednisolone at RA diagnosis (baseline), and therefore this variable was not entered into the regression model.\nImpact of high early DXR-BMD loss and baseline demographic and disease characteristics on Larsen score at year 10 (linear regression model)a\na B, the estimated regression coefficient; ESR, erythrocyte sedimentation rate; ACPA, anticitrullinated peptide antibodies; high early DXR-BMD loss is defined as bone loss above median value.\nLarsen scores at years 0, 1, 3, 4, 5, 10, 15 and 20 stratified according to the median of early DXR-BMD loss are given in Figure 1. There is a diverging trend during the whole follow-up period, which is most marked in the first 5 years. The increase in 95% CI at the later time points illustrates the increasing number of deceased patients. When using the smallest detectable difference, 0.0048 g/cm2/year [7], as a cutoff for elevated early DXR-BMD loss, a total of 97 patients (72%) were in the higher bone loss group, while 38 patients (28%) were in the lower bone loss group. Plotting the patients with early bone loss >0.0048 compared to those with lower early bone loss yielded results similar to those of median stratification (data not shown). During the first 5 years, the Larsen score deteriorated at a faster pace in patients with elevated early bone loss. However at time points 10, 15 and 20 years, these differences were not significant between the groups. The Spearman correlations between early bone loss and Larsen scores at years 1, 2, 3, 4, 5, 10 and 20 were all significant, with Spearman's ρ values between 0.359 and 0.223.\nRadiological Larsen score progression over time after stratification according to the median (-0.0185 g/cm2 per year) of early bone mineral density measured by digital X-ray radiogrammetry (DXR-BMD) change. High loss (black solid line) versus low loss (red dotted line) of DXR-BMD. Values are given as means with 95% confidence intervals (95% CIs). The larger 95% CIs at 15 and 20 years illustrate the decreasing number of patients at these time points.", "When looking at Larsen scores for hands and feet separately, the clear differences between groups stratified according to early DXR-BMD loss remained for the hands. For the Larsen scores of the feet, the picture was similar, with the group with high early DXR-BMD loss always having higher mean Larsen scores, but at most time points the differences were nonsignificant (data not shown).", "To study the progression rate of DXR-BMD loss over time, we calculated the data for the yearly progression rates between time points 0 and 1, 2, 3 and 4 years, which were (means and 95% CIs): 2.2 (1.3 to 3.0), 2.2 (1.5 to 2.9), 2.7 (2.0 to 3.3) and 2.3 (1.8 to 2.8), respectively. There was an overall trend of decreasing DXR-BMD progression rate over time, but the limited number of patients precludes firm statistical confirmation.", "After stratifying for the median of early changes in Larsen scores, a very similar picture to that observed for early DXR-BMD changes was found (Figure 2). Also, here there is a diverging trend during the whole follow-up period, and ANCOVA showed that the adjusted 10-year Larsen scores (95% CI) in the group with more elevated early Larsen score progression was 32.3 (20.6 to 44.0) above the group with low early Larsen score progression. The association between early Larsen score changes with 10-year Larsen scores was 0.540 (Spearman's ρ).\nRadiological Larsen score progression over time after stratification according to the median (5.5 units) of early change in Larsen score. High (black solid line) versus low (red dotted line) early ΔLarsen scores. Values are given as means with 95% confidence intervals (95% CIs). The larger 95% CIs at 15 and 20 years illustrate the decreasing number of patients at these time points.", "Median stratification of baseline BMD or Larsen score did not identify any trends of predictability for long-term radiological progression as measured by the Larsen score (Figures 3 and 4). The Spearman's ρ correlation between 10-year Larsen score and baseline BMD was -0.056, and between 10-year Larsen score and baseline Larsen score it was 0.199.\nRadiological Larsen score progression over time after stratification according to median (0.6 g/cm2) baseline BMD. High BMD (black solid line) versus low BMD (red dotted line) values at baseline are shown.\nRadiological Larsen score progression over time after stratification according to median (5.5 units) of baseline Larsen score. High Larsen score (black solid line) versus low Larsen score (red dotted line) at baseline is shown.", "In the present report, we have demonstrated that early bone mineral loss in the hands of patients with RA predict joint damage as measured by the radiographic Larsen score over a 20-year observation period. The difference was seen as early as 1 year after baseline all the way to 20-year follow-up.\nWe could not verify that DXR-BMD loss is more marked in the early phase of the disease, which has been reported by others, but our study had more limited power [4-11]. Overall our results are consistent with those of other studies. A longitudinal observational study including early RA patients with disease duration <1 year showed that changes in DXR-BMD of the second to fourth metacarpal bones at year 1 were specific and sensitive in identifying both patients who developed erosions scored using both the Larsen and SvdH methods and those whose existing erosion had progressed at year 4 [4]. Similarly, in RA patients with disease duration up to 4 years, hand bone loss measured by DXR-BMD at year 1 was predictive of subsequent radiographic damage scored by the SvdH method at 5 and 10 years [7].\nWe did not attempt to study the implication of late DXR-BMD loss in this study, since we aimed to find early predictors of outcome. On the other hand, the clinical value of early bone mineral loss as a predictor of future joint damage as measured by Larsen score should not be overemphasized. The relationship is well validated on the group level, but cannot be recommended for use as the only predictor as illustrated by the relatively low correlations. On the other hand, early progression in Larsen score also shows a clear predictive value for later joint damage, but the low correlations revealed that this information also must be confined to the group level. Early bone loss is a risk factor, in addition to other known risk factors. Thus, although statistically significant predictive value for RA prognosis can be found for several early findings such as the presence of rheumatoid factor, anticyclic citrullinated peptide antibody, elevated ESR level and cartilage oligomeric matrix protein level, their value in clinical decisions regarding treatment of the individual patient must not be overemphasized [22-29]. Treatment decisions rely heavily on several aspects not covered by conventional predictive analyses, such as perceived pain, working situation, deteriorating function and so on.\nThe findings that early DXR-BMD loss in the cortical shaft of metacarpal bones is predictive of later progression of joint-related hand Larsen score was not unexpected because of the close vicinity of these structures. We were not able to demonstrate an unequivocal relationship between early DXR-BMD loss in the metacarpal bones and Larsen score in the foot, even though there was a quite clear trend for such a relationship as has been reported for SvdH [11]. However, in the Larsen scoring system, as in the SvdH scoring system, the feet do not have the same number of joints as the hands, thus giving a ceiling effect with reduced possibility of finding differences [12-14]. This could at least partly explain the nonsignificant differences found in the Larsen scores of the feet. Methods of using DXR-BMD have not been developed for the metatarsal bones, thus precluding direct comparisons of hand and foot bone loss as well as a possible relationship between more generalized bone loss and multiple joint damage as measured by the total Larsen score.\nSeveral factors are known to promote osteopenia, such as female sex, older age and severe RA [30,31]. It was therefore not unexpected to find that the degree of early DXR-BMD loss identified patients with different baseline characteristics. Thus early markers of more severe disease, such as high HAQ score, high DAS44 score and high ESR and CRP levels, were all more common among the patients with higher DXR-BMD loss. Also older age contributed to high early bone loss, while rheumatoid factor status or anticitrullinated peptide antibody positivity and sex did not. The low numbers of patients taking systemic glucocorticoids represent the therapeutic tradition in the 1980s and prevent any meaningful analysis of this cause of osteopenia.\nThe strength of the DXR-BMD technique is that it is well standardized, can be used to perform a large number of objective measurements and can be done on existing radiographs. However, the radiographs must be taken in a standardized fashion if the results are to be given in grams per square centimeter. DXR-BMD can be somewhat more generalizable if the results are given as percentages, which are sufficient to measure relative changes. Both the Larsen and SvdH scores require experienced and licensed readers and still have a sizable reading error [12], a shortcoming that DXR-BMD has minimized, although the radiographic equipment used can influence the results to a minor degree. DXR-BMD is able to quantify only cortical and not trabecular bone mass, thus limiting its use as a general osteopenia measure. However, periarticular cortical bone loss, in some cases measured by DXR-BMD, has been shown to predict erosive disease in RA [4-11].\nThe strengths of the current study include its community-based recruitment, its prospective gathering of a large amount of laboratory and clinical information and its almost complete follow-up. The exclusion of the sizable proportion of RA patients who died during follow-up could be argued for. However, they did represent a group of RA patients that presumably had several severity markers in common, and the exclusion of these patients could seriously hamper the results of the study. We also refrained from extrapolating and imputing missing Larsen scores, since this also leads to several assumptions and limitations in this restricted sample size.\nThe limitations of the present study include the retrospective approach of DXR-BMD measurements on preexisting radiographs. This accounted for the loss of 48 patients. However, their baseline characteristics were largely similar to those of the total group of patients. Also, the total sample size was not very large, but this was one of the very first early RA cohorts recruited, and it represents what one center could bear during a total of >25 years of follow-up. Thus, although some of the finer details may not always be statistically verified, the cohort size permits firm conclusions on several robust sets of data such as those in the present report. It must be remembered that in comparing data with more recent early RA studies, in which new and often more liberal criteria sets are being used, we used the 1958 American Rheumatism Association criteria to establish definite diagnoses of RA in the present study [32]. Also, the introduction of new and potentially bone loss-protective treatment cohorts must be accounted for in more recent early RA studies when comparing them with the Lund early RA cohort treated according to the clinical standards used over 25 years ago.", "Early DXR-BMD loss in the present study predicted future joint damage as measured by Larsen score both in the short-term perspective (1 year), which confirms previous studies that used the SvdH score, and for the first time in the very long-term perspective (20 years).", "BMD: bone mineral density; DXR: digital X-ray radiogrammetry; DXR-BMD: bone mineral density measured by digital X-ray radiogrammetry; RA: rheumatoid arthritis.", "JA is employed by Sectra but is not a shareholder in the company. All other authors have declared no competing interests. The radiographs were digitized using a Vidar Diagnostic Pro Plus digitizer (VIDAR Systems Corp., Herndon, VA, USA), and the resulting digital radiographs were analyzed using DXR online (Sectra, Linköping, Sweden), but any financial or other support was obtained from the company.", "All authors have made substantial contributions to the study's conception and design, acquisition of data or analysis and interpretation of data and have been involved in drafting the manuscript or revising it critically for important intellectual content. MCK, EL and PG wrote the manuscript. EL and PG helped plan the study. JA developed the DXR method. KJL scored the radiograms. EL, TS and KE conceived the study. MCK helped in performing the statistical analysis. KE conceived the original Lund early RA cohort study. PG handled the database. All authors read and approved the final manuscript." ]

[ null, "materials|methods", null, null, null, "results", null, null, null, null, null, null, "discussion", "conclusions", null, null, null ]

[]

[ "Adult", "Carcinoma, Squamous Cell", "Female", "HSP70 Heat-Shock Proteins", "Humans", "Immunohistochemistry", "Lymphatic Metastasis", "Male", "Middle Aged", "Mouth Neoplasms", "Multivariate Analysis", "Neoplasm Invasiveness", "Neoplasm Staging", "Prognosis", "Survival Rate" ]

[ "Background", "Patients", "Immunohistochemistry", "Evaluation of HSP70 Expression", "Evaluation of Clinicopathological Parameters", "Statistical analysis", "Results", "Clinical and pathological features", "Clinicopathological findings and survival", "Immunohistochemistry of HSP70", "Prognostic significance of HSP expression", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Oral squamous cell carcinoma (OSCC), a frequently occurring cancer in the head and neck region, is characterized by an aggressive growth pattern, local invasiveness, and spread to cervical lymph nodes. Patient outcome depends on the conventional prognostic factors used in clinical practice. Advances in surgical and nonsurgical treatments have led to increased local tumor control in recent years. However, overall survival rates have not improved because of the prevalence of locoregional tumor recurrence and distant metastasis. Although there is general agreement that tumor infiltration of the resection margin is one of the most relevant predictive factors for the development of a local recurrent carcinoma, the presence of tumor-free margins does not guarantee against recurrence because carcinoma can develop following discontinued expansion of tumor cells in the vicinity [1,2]. To improve survival periods of these patients, molecular and histological markers must be identified to target tumors with a high likelihood of metastatic spread. To date, no reliable or clinically applicable marker of tumor aggressiveness has been identified for OSCC. Different markers/marker complexes have been identified as active in tumor suppression or antitumor defense and display potential as prognostic factors. Furthermore, molecular biology investigations of resection margins have shown that detection of mutant p53 genes is linked with increased incidence of recurrent tumors. The p53 molecule is a 53-kD polypeptide. It acts as a transcription factor that controls the cell cycle by either arresting cells in the G1 phase through activation of the p21 gene or triggering apoptosis by activating genes. Another more recent approach to oral carcinogenesis focuses on the escape of malignant cells from apoptotic signals. Extensive research has been carried out on p53 in this respect, and there is broad evidence for its role in the manifestation of oral carcinoma. However, published data indicates that p53 alone is not particularly valuable in predicting prognosis. Additional markers of apoptosis such as Fas, Fas ligand (FasL), and Bax, as well as anti-apoptotic molecules such as bcl2/BAG-1, are reported to be relevant to prognosis in a smaller number of publications. All of these have shown a significant correlation with prognosis, but it is difficult to draw conclusions on the prognostic validity of these markers on the basis of the present data [3-5].\nAnother approach in predicting prognosis in cancer is expression of heat shock proteins (HSPs). HSPs are found in all organisms and all cell types. They are the most phylogenetically conserved proteins known with respect to both structure and function [6]. Usually, HSPs are expressed at low levels, and under normal physiological conditions, many members of the HSP family are involved in protein synthesis. When a cell is stressed, oligomeric complexes disassemble and polypeptides unfold. Under these conditions, the role of HSPs is to reverse such changes and, if refolding becomes impossible, to potentially speed up the removal of such denatured proteins. Expression of HSPs is induced even under nonstress conditions, including those of the cell cycle, development, and differentiation [7,8]. During carcinogenesis, HSPs have been reported to alter their expression levels, showing either an increase or a decrease [9,10]). HSP70 expression in colorectal carcinoma and breast carcinoma has been significantly correlated with low differentiation and poor prognosis [11,12], whereas in renal cell carcinoma it has been reported to be associated with good prognosis [13]. Five main families of HSPs are known: low molecular weight, HSP65, HSP70, HSP90, and HSP100. Over the last few years, HSPs have also been shown to play a role in the antigenicity of tumors. Expression of HSPs on the surface of tumor cells, instead of their normal intracellular location, suggests they play a role in inducing an immune response against cancer. In a chemically induced mouse sarcoma, Ullrich et al. identified a tumor-specific transplantation antigen that appears to be an HSP, which is expressed on the cell surface and induces protective immunity [14]. Furthermore, a protein related to the HSP70 family has been shown to be immunogenic in oncogene-transformed rat fibroblasts [15], and HSP70 derived from MethA sarcoma (but not from normal tissue) has been demonstrated to be immunogenic, not in itself but in association with tumor peptides. The prognostic significance of HSP70 in esophageal squamous cell carcinoma has been reported [16]. The aim of the present study was to investigate the expression of highly stress-inducible HSP70 in OSCC and its use in predicting prognosis.", "Surgical specimens were obtained from 61 patients (49 men and 12 women) with OSCC who underwent potentially curative surgery at the Department of Oral and Maxillofacial Surgery, Hanover Medical School, Hanover, Germany, between 1996 and 2001. Tumor stage and disease grade were classified according to the fifth edition of the TNM classification of the International Union Against Cancer (UICC). None of the patients had received irradiation or chemotherapy prior to surgery or had distant metastases at the time of surgery. Patients who underwent no curative surgery and/or inadequate follow-up were not included in the study.", "Serial 3-mm sections were deparaffinized, rehydrated, washed and, treated with a solution of 2% horse serum, 0.1% bovine serum albumin (Sigma Corporation, Steinheim, Germany), and 0.1% sodium acid in 150 mmol/l phosphate-buffered saline (PBS; pH 7.2) for 15 min to block nonspecific antibody-binding. A polyclonal rabbit anti-HSP70 antibody (Dako, Carpinteria, CA, USA), specific to HSP from Escherichia coli, which shares more than 48% sequence homology with mammalian HSP70 was the first layer. The optimal dilution of anti-HSP antibody (1:250) was determined by titration. The selected sections were incubated with this antibody for 120 min at room temperature (RT). The second layer, a biotin-conjugated goat anti-rabbit immunoglobulin (Oncogene, San Diego, CA USA) diluted 1:200 in PBS was incubated for 30 min at RT. The third layer was an avidin-biotin-horseradish peroxidase complex (Dako) diluted 1:50 in PBS. Incubation was, as before, 30 min at RT. Sections were washed for 10 min in 2 changes of PBS between each layer. The color reaction was developed with a solution consisting of 0.05% 3,30-diaminobenzidine tetrahydrochloride (Sigma, St Louis, MO, USA), 0.03% nickel chloride (Sigma), and 0.01% hydrogen peroxide in 48 mmol/l Tris-HCL, pH 7.6 (Sigma). Counterstaining was carried out with Mayer's hematoxylin [6].", "Light microscopy and analysis 3.1® (Soft Imaging System, Münster, Germany), an image processing and analysis program, were used for evaluating HSP70 expression. The tumor region was defined as the region of interest (ROI) and HSP70-positive staining was analyzed. When 20% or more of the tumor cells in a given specimen were positively stained, the sample was graded as HSP70 positive; it was graded negative when fewer than 20% of the tumor cells were stained [16].", "Histological evaluation was performed using hematoxylin and eosin staining. TNM-classification was applied with regard to the clinical stage, depth of tumor invasion, lymph node status, and histological typing using the World Health Organization classification.", "The statistical significance of the data was analyzed using the chi-square test. Patients' postoperative status was assessed on December 31, 2008. No patient with incomplete follow-up was included in the evaluation. The cumulative survival rate was calculated by the Kaplan-Meier method [17], and statistical significance was analyzed by the log-rank test. To confirm the statistical significance of HSP70 expression as a prognostic indicator, multivariate analysis was performed using the Cox proportional-hazards regression model [18].", "[SUBTITLE] Clinical and pathological features [SUBSECTION] Of the 61 patients, 49 were men and 12 women. The age of the patients ranged from 35 to 60 years with a mean age of 50.7 years. T2 tumors were present in 28 patients, T3 tumors in 3 patients, and T4 tumors in 30 patients. Lymph node metastases were found in 25 patients.\nOf the 61 patients, 49 were men and 12 women. The age of the patients ranged from 35 to 60 years with a mean age of 50.7 years. T2 tumors were present in 28 patients, T3 tumors in 3 patients, and T4 tumors in 30 patients. Lymph node metastases were found in 25 patients.\n[SUBTITLE] Clinicopathological findings and survival [SUBSECTION] There was no statistical influence on prognosis of grading, T- stadium, age, and sex of the patients. Presence of positive lymph nodes (N+) was identified as a prognostic factor with significant influence in univariate and multivariate analysis, as well in all patients suffering from T2 tumors (Table 1, Figure 1A, and Figure 1B). The mortality risk of these patients is 13 times than that of patients with negative nodes.\nExpression of HSP70 and other tumor characteristics multivariate analyses for disease-free survival prediction in 28 patients with T2 oral squamous cell carcinoma.\n(Cox's partially nonparametric regression model was used to evaluate the predictive power of various combinations and interactions of prognostic factors. Cl = confidence interval.)\nInfluence of nodal stage on overall survival following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Considering all patients (A), 25 of 35 had a positive lymph node stage. Regarding the patients suffering from T2 tumors (B), 6 of 28 showed a positive lymph node stage. Survival of patients with negative lymph nodes (green) and positive lymph nodes (blue) (p = 0.002).\nThere was no statistical influence on prognosis of grading, T- stadium, age, and sex of the patients. Presence of positive lymph nodes (N+) was identified as a prognostic factor with significant influence in univariate and multivariate analysis, as well in all patients suffering from T2 tumors (Table 1, Figure 1A, and Figure 1B). The mortality risk of these patients is 13 times than that of patients with negative nodes.\nExpression of HSP70 and other tumor characteristics multivariate analyses for disease-free survival prediction in 28 patients with T2 oral squamous cell carcinoma.\n(Cox's partially nonparametric regression model was used to evaluate the predictive power of various combinations and interactions of prognostic factors. Cl = confidence interval.)\nInfluence of nodal stage on overall survival following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Considering all patients (A), 25 of 35 had a positive lymph node stage. Regarding the patients suffering from T2 tumors (B), 6 of 28 showed a positive lymph node stage. Survival of patients with negative lymph nodes (green) and positive lymph nodes (blue) (p = 0.002).\n[SUBTITLE] Immunohistochemistry of HSP70 [SUBSECTION] Immunoreactivity for HSP70 was positive in tumor cells of 38 of all patients (63.3%). Positive immunoreactivity of tumor cells could be detected in 17 of 28 patients with T2 tumors (60.7%) (Table 2, Figure 2).\nProportional distribution of HSP70-positive and HSP70-negative tumors separated into T2 and T3/T4 tumors.\nPhotographs from tissue sections of oral squamous cell carcinoma immunostained for HSP70. Expression pattern of HSP70-negative (A) and HSP70-positive (brown color, B) specimens.\nImmunoreactivity for HSP70 was positive in tumor cells of 38 of all patients (63.3%). Positive immunoreactivity of tumor cells could be detected in 17 of 28 patients with T2 tumors (60.7%) (Table 2, Figure 2).\nProportional distribution of HSP70-positive and HSP70-negative tumors separated into T2 and T3/T4 tumors.\nPhotographs from tissue sections of oral squamous cell carcinoma immunostained for HSP70. Expression pattern of HSP70-negative (A) and HSP70-positive (brown color, B) specimens.\n[SUBTITLE] Prognostic significance of HSP expression [SUBSECTION] When all patients were considered, no statistical influence on survival of HSP70 could be detected. After classifying the samples as T2 and T3/T4 tumors, prognostic significance of HSP70 expression in tumor cells could be detected in patients suffering from T2 tumors. Figure 3 shows cumulative survival curves for patients with T2 tumors with positive and negative HSP70 expression (Table 1 and Figure 3).\nInfluence of HSP70 expression on overall survival of patients suffering from T2 tumors following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Of 28 patients, 17 patients showed HSP70 expression greater or equal to 20% (blue curve). This predicts significantly improved survival compared to patients with less then 20% expression of HSP70 (red curve). [p = 0.009]\nThe survival of patients suffering from T2 tumors with positive HSP70 expression was 8 times higher than that for patients with negative HSP70 expression.\nWhen all patients were considered, no statistical influence on survival of HSP70 could be detected. After classifying the samples as T2 and T3/T4 tumors, prognostic significance of HSP70 expression in tumor cells could be detected in patients suffering from T2 tumors. Figure 3 shows cumulative survival curves for patients with T2 tumors with positive and negative HSP70 expression (Table 1 and Figure 3).\nInfluence of HSP70 expression on overall survival of patients suffering from T2 tumors following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Of 28 patients, 17 patients showed HSP70 expression greater or equal to 20% (blue curve). This predicts significantly improved survival compared to patients with less then 20% expression of HSP70 (red curve). [p = 0.009]\nThe survival of patients suffering from T2 tumors with positive HSP70 expression was 8 times higher than that for patients with negative HSP70 expression.", "Of the 61 patients, 49 were men and 12 women. The age of the patients ranged from 35 to 60 years with a mean age of 50.7 years. T2 tumors were present in 28 patients, T3 tumors in 3 patients, and T4 tumors in 30 patients. Lymph node metastases were found in 25 patients.", "There was no statistical influence on prognosis of grading, T- stadium, age, and sex of the patients. Presence of positive lymph nodes (N+) was identified as a prognostic factor with significant influence in univariate and multivariate analysis, as well in all patients suffering from T2 tumors (Table 1, Figure 1A, and Figure 1B). The mortality risk of these patients is 13 times than that of patients with negative nodes.\nExpression of HSP70 and other tumor characteristics multivariate analyses for disease-free survival prediction in 28 patients with T2 oral squamous cell carcinoma.\n(Cox's partially nonparametric regression model was used to evaluate the predictive power of various combinations and interactions of prognostic factors. Cl = confidence interval.)\nInfluence of nodal stage on overall survival following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Considering all patients (A), 25 of 35 had a positive lymph node stage. Regarding the patients suffering from T2 tumors (B), 6 of 28 showed a positive lymph node stage. Survival of patients with negative lymph nodes (green) and positive lymph nodes (blue) (p = 0.002).", "Immunoreactivity for HSP70 was positive in tumor cells of 38 of all patients (63.3%). Positive immunoreactivity of tumor cells could be detected in 17 of 28 patients with T2 tumors (60.7%) (Table 2, Figure 2).\nProportional distribution of HSP70-positive and HSP70-negative tumors separated into T2 and T3/T4 tumors.\nPhotographs from tissue sections of oral squamous cell carcinoma immunostained for HSP70. Expression pattern of HSP70-negative (A) and HSP70-positive (brown color, B) specimens.", "When all patients were considered, no statistical influence on survival of HSP70 could be detected. After classifying the samples as T2 and T3/T4 tumors, prognostic significance of HSP70 expression in tumor cells could be detected in patients suffering from T2 tumors. Figure 3 shows cumulative survival curves for patients with T2 tumors with positive and negative HSP70 expression (Table 1 and Figure 3).\nInfluence of HSP70 expression on overall survival of patients suffering from T2 tumors following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Of 28 patients, 17 patients showed HSP70 expression greater or equal to 20% (blue curve). This predicts significantly improved survival compared to patients with less then 20% expression of HSP70 (red curve). [p = 0.009]\nThe survival of patients suffering from T2 tumors with positive HSP70 expression was 8 times higher than that for patients with negative HSP70 expression.", "Many studies in cell biology have addressed the roles of HSPs as molecular chaperones, including protein folding and unfolding, translocation, and prevention of inappropriate protein aggregation [7,19]. These findings show an association between HSPs and cell proliferation and the prevention of apoptosis [20]. While HSP expression has been recognized as a factor of prognostic value in certain tumors, the data are limited and the results often contradictory. For example, inducible HSP72 has been shown to be a negative prognostic factor for disease-free survival (DFS) in patients with lymph node-negative breast cancer [21], whereas other authors have shown that HSP72 positively correlates with estrogen receptors and inversely with expression of mutant p53 [22]. Santarosa et al. demonstrated prognostic implications of expression of HSP70 in patients suffering from renal cancer [23]. Although some evidence indicates that HSPs are involved in various aspects of cell transformation and immune response against cancer [24], their biological role and its implications for the clinical course in cancer patients are not clear. In OSCC, the number of positive nodes, macroscopic extracapsular spread, and tumor infiltration of the resection margins have been described as significant prognostic factors [1,2]. Studies concerning HSP70 as a prognostic factor in esophageal carcinoma suggest that reduction of HSP70 expression is significantly correlated with poor prognosis [16,25]. Although several studies have been performed to elucidate the relationship between HSPs and tumors in various organs, to our knowledge, there are comparatively few reports related to OSCC. Sugarman et al. reported that HSP70 expression is not a definitive marker of oral malignancy or malignant potential [26]. In terms of prognostic significance, Ito et al. examined 24 specimens of patients suffering from OSCC. Although HSP immunohistochemistry revealed changes in HSP expression during tumorigenesis of squamous epithelium of the tongue, there was no correlation between HSP staining and survival period, stage, lymph node metastasis, histological grade, or p53 immunostaining [27]. These results are in line with those of Gandour-Edwards et al, who found that HSPs were expressed in normal upper respiratory tract squamous mucosa, and their role in carcinoma thus remains unclear. None of the markers (p53, HSP27, or HSP70) demonstrated prognostic significance for 5-year survival. We confirm the previously identified association of cervical lymph node metastases with decreased survival [5].\nThe results of the previously published data are not in line with the findings of the present study. In this study, we demonstrated that for patients suffering from T2-tumors, positive expression of HSP70 results in a significantly lower mortality risk compared with patients with negative expression. The main difference regarding the study design may be that, in contrast to previous studies, our patient population was divided according to tumor stages in the different T-stadium of the tumor. The purpose of this division was to analyze the prognostic significance of each T-stadium separately. Considering the entire patient group, our findings would suggest the same result: HSP70 has no prognostic implications for our patients. Also in the seperated T3- and T4-tumors, HSP 70 did not show any prognostic significance. Only after focusing on patients suffering from T2-tumors could a significant difference be detected. The survival of patients suffering from T2 tumors with positive HSP70 expression was 8 times higher than that for patients with negative HSP70 expression.", "The results of our study suggests that expression of HSP70 affects survival only in the early stage of the disease and that HSP70 membrane expression, in particular, is a target for natural killer cells [28]. Therefore in T2-tumors, the increased level of HSP70 may result in an extended tumor control by the natural killer cells. By implication, T1-T2 tumors of OSCC with low expression of Hsp70 could require more radical treatment.", "The authors declare that they have no competing interests.", "FT, HK, NCG and AE conceived of the study and participated in its design and coordination. FT drafted the manuscript. CMT helped to draft the manuscript. OFS carried out the immunohistochemistry and the HSP 70 Expression. GW performed the statistical analysis. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Patients", "Immunohistochemistry", "Evaluation of HSP70 Expression", "Evaluation of Clinicopathological Parameters", "Statistical analysis", "Results", "Clinical and pathological features", "Clinicopathological findings and survival", "Immunohistochemistry of HSP70", "Prognostic significance of HSP expression", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Oral squamous cell carcinoma (OSCC), a frequently occurring cancer in the head and neck region, is characterized by an aggressive growth pattern, local invasiveness, and spread to cervical lymph nodes. Patient outcome depends on the conventional prognostic factors used in clinical practice. Advances in surgical and nonsurgical treatments have led to increased local tumor control in recent years. However, overall survival rates have not improved because of the prevalence of locoregional tumor recurrence and distant metastasis. Although there is general agreement that tumor infiltration of the resection margin is one of the most relevant predictive factors for the development of a local recurrent carcinoma, the presence of tumor-free margins does not guarantee against recurrence because carcinoma can develop following discontinued expansion of tumor cells in the vicinity [1,2]. To improve survival periods of these patients, molecular and histological markers must be identified to target tumors with a high likelihood of metastatic spread. To date, no reliable or clinically applicable marker of tumor aggressiveness has been identified for OSCC. Different markers/marker complexes have been identified as active in tumor suppression or antitumor defense and display potential as prognostic factors. Furthermore, molecular biology investigations of resection margins have shown that detection of mutant p53 genes is linked with increased incidence of recurrent tumors. The p53 molecule is a 53-kD polypeptide. It acts as a transcription factor that controls the cell cycle by either arresting cells in the G1 phase through activation of the p21 gene or triggering apoptosis by activating genes. Another more recent approach to oral carcinogenesis focuses on the escape of malignant cells from apoptotic signals. Extensive research has been carried out on p53 in this respect, and there is broad evidence for its role in the manifestation of oral carcinoma. However, published data indicates that p53 alone is not particularly valuable in predicting prognosis. Additional markers of apoptosis such as Fas, Fas ligand (FasL), and Bax, as well as anti-apoptotic molecules such as bcl2/BAG-1, are reported to be relevant to prognosis in a smaller number of publications. All of these have shown a significant correlation with prognosis, but it is difficult to draw conclusions on the prognostic validity of these markers on the basis of the present data [3-5].\nAnother approach in predicting prognosis in cancer is expression of heat shock proteins (HSPs). HSPs are found in all organisms and all cell types. They are the most phylogenetically conserved proteins known with respect to both structure and function [6]. Usually, HSPs are expressed at low levels, and under normal physiological conditions, many members of the HSP family are involved in protein synthesis. When a cell is stressed, oligomeric complexes disassemble and polypeptides unfold. Under these conditions, the role of HSPs is to reverse such changes and, if refolding becomes impossible, to potentially speed up the removal of such denatured proteins. Expression of HSPs is induced even under nonstress conditions, including those of the cell cycle, development, and differentiation [7,8]. During carcinogenesis, HSPs have been reported to alter their expression levels, showing either an increase or a decrease [9,10]). HSP70 expression in colorectal carcinoma and breast carcinoma has been significantly correlated with low differentiation and poor prognosis [11,12], whereas in renal cell carcinoma it has been reported to be associated with good prognosis [13]. Five main families of HSPs are known: low molecular weight, HSP65, HSP70, HSP90, and HSP100. Over the last few years, HSPs have also been shown to play a role in the antigenicity of tumors. Expression of HSPs on the surface of tumor cells, instead of their normal intracellular location, suggests they play a role in inducing an immune response against cancer. In a chemically induced mouse sarcoma, Ullrich et al. identified a tumor-specific transplantation antigen that appears to be an HSP, which is expressed on the cell surface and induces protective immunity [14]. Furthermore, a protein related to the HSP70 family has been shown to be immunogenic in oncogene-transformed rat fibroblasts [15], and HSP70 derived from MethA sarcoma (but not from normal tissue) has been demonstrated to be immunogenic, not in itself but in association with tumor peptides. The prognostic significance of HSP70 in esophageal squamous cell carcinoma has been reported [16]. The aim of the present study was to investigate the expression of highly stress-inducible HSP70 in OSCC and its use in predicting prognosis.", "[SUBTITLE] Patients [SUBSECTION] Surgical specimens were obtained from 61 patients (49 men and 12 women) with OSCC who underwent potentially curative surgery at the Department of Oral and Maxillofacial Surgery, Hanover Medical School, Hanover, Germany, between 1996 and 2001. Tumor stage and disease grade were classified according to the fifth edition of the TNM classification of the International Union Against Cancer (UICC). None of the patients had received irradiation or chemotherapy prior to surgery or had distant metastases at the time of surgery. Patients who underwent no curative surgery and/or inadequate follow-up were not included in the study.\nSurgical specimens were obtained from 61 patients (49 men and 12 women) with OSCC who underwent potentially curative surgery at the Department of Oral and Maxillofacial Surgery, Hanover Medical School, Hanover, Germany, between 1996 and 2001. Tumor stage and disease grade were classified according to the fifth edition of the TNM classification of the International Union Against Cancer (UICC). None of the patients had received irradiation or chemotherapy prior to surgery or had distant metastases at the time of surgery. Patients who underwent no curative surgery and/or inadequate follow-up were not included in the study.\n[SUBTITLE] Immunohistochemistry [SUBSECTION] Serial 3-mm sections were deparaffinized, rehydrated, washed and, treated with a solution of 2% horse serum, 0.1% bovine serum albumin (Sigma Corporation, Steinheim, Germany), and 0.1% sodium acid in 150 mmol/l phosphate-buffered saline (PBS; pH 7.2) for 15 min to block nonspecific antibody-binding. A polyclonal rabbit anti-HSP70 antibody (Dako, Carpinteria, CA, USA), specific to HSP from Escherichia coli, which shares more than 48% sequence homology with mammalian HSP70 was the first layer. The optimal dilution of anti-HSP antibody (1:250) was determined by titration. The selected sections were incubated with this antibody for 120 min at room temperature (RT). The second layer, a biotin-conjugated goat anti-rabbit immunoglobulin (Oncogene, San Diego, CA USA) diluted 1:200 in PBS was incubated for 30 min at RT. The third layer was an avidin-biotin-horseradish peroxidase complex (Dako) diluted 1:50 in PBS. Incubation was, as before, 30 min at RT. Sections were washed for 10 min in 2 changes of PBS between each layer. The color reaction was developed with a solution consisting of 0.05% 3,30-diaminobenzidine tetrahydrochloride (Sigma, St Louis, MO, USA), 0.03% nickel chloride (Sigma), and 0.01% hydrogen peroxide in 48 mmol/l Tris-HCL, pH 7.6 (Sigma). Counterstaining was carried out with Mayer's hematoxylin [6].\nSerial 3-mm sections were deparaffinized, rehydrated, washed and, treated with a solution of 2% horse serum, 0.1% bovine serum albumin (Sigma Corporation, Steinheim, Germany), and 0.1% sodium acid in 150 mmol/l phosphate-buffered saline (PBS; pH 7.2) for 15 min to block nonspecific antibody-binding. A polyclonal rabbit anti-HSP70 antibody (Dako, Carpinteria, CA, USA), specific to HSP from Escherichia coli, which shares more than 48% sequence homology with mammalian HSP70 was the first layer. The optimal dilution of anti-HSP antibody (1:250) was determined by titration. The selected sections were incubated with this antibody for 120 min at room temperature (RT). The second layer, a biotin-conjugated goat anti-rabbit immunoglobulin (Oncogene, San Diego, CA USA) diluted 1:200 in PBS was incubated for 30 min at RT. The third layer was an avidin-biotin-horseradish peroxidase complex (Dako) diluted 1:50 in PBS. Incubation was, as before, 30 min at RT. Sections were washed for 10 min in 2 changes of PBS between each layer. The color reaction was developed with a solution consisting of 0.05% 3,30-diaminobenzidine tetrahydrochloride (Sigma, St Louis, MO, USA), 0.03% nickel chloride (Sigma), and 0.01% hydrogen peroxide in 48 mmol/l Tris-HCL, pH 7.6 (Sigma). Counterstaining was carried out with Mayer's hematoxylin [6].\n[SUBTITLE] Evaluation of HSP70 Expression [SUBSECTION] Light microscopy and analysis 3.1® (Soft Imaging System, Münster, Germany), an image processing and analysis program, were used for evaluating HSP70 expression. The tumor region was defined as the region of interest (ROI) and HSP70-positive staining was analyzed. When 20% or more of the tumor cells in a given specimen were positively stained, the sample was graded as HSP70 positive; it was graded negative when fewer than 20% of the tumor cells were stained [16].\nLight microscopy and analysis 3.1® (Soft Imaging System, Münster, Germany), an image processing and analysis program, were used for evaluating HSP70 expression. The tumor region was defined as the region of interest (ROI) and HSP70-positive staining was analyzed. When 20% or more of the tumor cells in a given specimen were positively stained, the sample was graded as HSP70 positive; it was graded negative when fewer than 20% of the tumor cells were stained [16].\n[SUBTITLE] Evaluation of Clinicopathological Parameters [SUBSECTION] Histological evaluation was performed using hematoxylin and eosin staining. TNM-classification was applied with regard to the clinical stage, depth of tumor invasion, lymph node status, and histological typing using the World Health Organization classification.\nHistological evaluation was performed using hematoxylin and eosin staining. TNM-classification was applied with regard to the clinical stage, depth of tumor invasion, lymph node status, and histological typing using the World Health Organization classification.\n[SUBTITLE] Statistical analysis [SUBSECTION] The statistical significance of the data was analyzed using the chi-square test. Patients' postoperative status was assessed on December 31, 2008. No patient with incomplete follow-up was included in the evaluation. The cumulative survival rate was calculated by the Kaplan-Meier method [17], and statistical significance was analyzed by the log-rank test. To confirm the statistical significance of HSP70 expression as a prognostic indicator, multivariate analysis was performed using the Cox proportional-hazards regression model [18].\nThe statistical significance of the data was analyzed using the chi-square test. Patients' postoperative status was assessed on December 31, 2008. No patient with incomplete follow-up was included in the evaluation. The cumulative survival rate was calculated by the Kaplan-Meier method [17], and statistical significance was analyzed by the log-rank test. To confirm the statistical significance of HSP70 expression as a prognostic indicator, multivariate analysis was performed using the Cox proportional-hazards regression model [18].", "Surgical specimens were obtained from 61 patients (49 men and 12 women) with OSCC who underwent potentially curative surgery at the Department of Oral and Maxillofacial Surgery, Hanover Medical School, Hanover, Germany, between 1996 and 2001. Tumor stage and disease grade were classified according to the fifth edition of the TNM classification of the International Union Against Cancer (UICC). None of the patients had received irradiation or chemotherapy prior to surgery or had distant metastases at the time of surgery. Patients who underwent no curative surgery and/or inadequate follow-up were not included in the study.", "Serial 3-mm sections were deparaffinized, rehydrated, washed and, treated with a solution of 2% horse serum, 0.1% bovine serum albumin (Sigma Corporation, Steinheim, Germany), and 0.1% sodium acid in 150 mmol/l phosphate-buffered saline (PBS; pH 7.2) for 15 min to block nonspecific antibody-binding. A polyclonal rabbit anti-HSP70 antibody (Dako, Carpinteria, CA, USA), specific to HSP from Escherichia coli, which shares more than 48% sequence homology with mammalian HSP70 was the first layer. The optimal dilution of anti-HSP antibody (1:250) was determined by titration. The selected sections were incubated with this antibody for 120 min at room temperature (RT). The second layer, a biotin-conjugated goat anti-rabbit immunoglobulin (Oncogene, San Diego, CA USA) diluted 1:200 in PBS was incubated for 30 min at RT. The third layer was an avidin-biotin-horseradish peroxidase complex (Dako) diluted 1:50 in PBS. Incubation was, as before, 30 min at RT. Sections were washed for 10 min in 2 changes of PBS between each layer. The color reaction was developed with a solution consisting of 0.05% 3,30-diaminobenzidine tetrahydrochloride (Sigma, St Louis, MO, USA), 0.03% nickel chloride (Sigma), and 0.01% hydrogen peroxide in 48 mmol/l Tris-HCL, pH 7.6 (Sigma). Counterstaining was carried out with Mayer's hematoxylin [6].", "Light microscopy and analysis 3.1® (Soft Imaging System, Münster, Germany), an image processing and analysis program, were used for evaluating HSP70 expression. The tumor region was defined as the region of interest (ROI) and HSP70-positive staining was analyzed. When 20% or more of the tumor cells in a given specimen were positively stained, the sample was graded as HSP70 positive; it was graded negative when fewer than 20% of the tumor cells were stained [16].", "Histological evaluation was performed using hematoxylin and eosin staining. TNM-classification was applied with regard to the clinical stage, depth of tumor invasion, lymph node status, and histological typing using the World Health Organization classification.", "The statistical significance of the data was analyzed using the chi-square test. Patients' postoperative status was assessed on December 31, 2008. No patient with incomplete follow-up was included in the evaluation. The cumulative survival rate was calculated by the Kaplan-Meier method [17], and statistical significance was analyzed by the log-rank test. To confirm the statistical significance of HSP70 expression as a prognostic indicator, multivariate analysis was performed using the Cox proportional-hazards regression model [18].", "[SUBTITLE] Clinical and pathological features [SUBSECTION] Of the 61 patients, 49 were men and 12 women. The age of the patients ranged from 35 to 60 years with a mean age of 50.7 years. T2 tumors were present in 28 patients, T3 tumors in 3 patients, and T4 tumors in 30 patients. Lymph node metastases were found in 25 patients.\nOf the 61 patients, 49 were men and 12 women. The age of the patients ranged from 35 to 60 years with a mean age of 50.7 years. T2 tumors were present in 28 patients, T3 tumors in 3 patients, and T4 tumors in 30 patients. Lymph node metastases were found in 25 patients.\n[SUBTITLE] Clinicopathological findings and survival [SUBSECTION] There was no statistical influence on prognosis of grading, T- stadium, age, and sex of the patients. Presence of positive lymph nodes (N+) was identified as a prognostic factor with significant influence in univariate and multivariate analysis, as well in all patients suffering from T2 tumors (Table 1, Figure 1A, and Figure 1B). The mortality risk of these patients is 13 times than that of patients with negative nodes.\nExpression of HSP70 and other tumor characteristics multivariate analyses for disease-free survival prediction in 28 patients with T2 oral squamous cell carcinoma.\n(Cox's partially nonparametric regression model was used to evaluate the predictive power of various combinations and interactions of prognostic factors. Cl = confidence interval.)\nInfluence of nodal stage on overall survival following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Considering all patients (A), 25 of 35 had a positive lymph node stage. Regarding the patients suffering from T2 tumors (B), 6 of 28 showed a positive lymph node stage. Survival of patients with negative lymph nodes (green) and positive lymph nodes (blue) (p = 0.002).\nThere was no statistical influence on prognosis of grading, T- stadium, age, and sex of the patients. Presence of positive lymph nodes (N+) was identified as a prognostic factor with significant influence in univariate and multivariate analysis, as well in all patients suffering from T2 tumors (Table 1, Figure 1A, and Figure 1B). The mortality risk of these patients is 13 times than that of patients with negative nodes.\nExpression of HSP70 and other tumor characteristics multivariate analyses for disease-free survival prediction in 28 patients with T2 oral squamous cell carcinoma.\n(Cox's partially nonparametric regression model was used to evaluate the predictive power of various combinations and interactions of prognostic factors. Cl = confidence interval.)\nInfluence of nodal stage on overall survival following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Considering all patients (A), 25 of 35 had a positive lymph node stage. Regarding the patients suffering from T2 tumors (B), 6 of 28 showed a positive lymph node stage. Survival of patients with negative lymph nodes (green) and positive lymph nodes (blue) (p = 0.002).\n[SUBTITLE] Immunohistochemistry of HSP70 [SUBSECTION] Immunoreactivity for HSP70 was positive in tumor cells of 38 of all patients (63.3%). Positive immunoreactivity of tumor cells could be detected in 17 of 28 patients with T2 tumors (60.7%) (Table 2, Figure 2).\nProportional distribution of HSP70-positive and HSP70-negative tumors separated into T2 and T3/T4 tumors.\nPhotographs from tissue sections of oral squamous cell carcinoma immunostained for HSP70. Expression pattern of HSP70-negative (A) and HSP70-positive (brown color, B) specimens.\nImmunoreactivity for HSP70 was positive in tumor cells of 38 of all patients (63.3%). Positive immunoreactivity of tumor cells could be detected in 17 of 28 patients with T2 tumors (60.7%) (Table 2, Figure 2).\nProportional distribution of HSP70-positive and HSP70-negative tumors separated into T2 and T3/T4 tumors.\nPhotographs from tissue sections of oral squamous cell carcinoma immunostained for HSP70. Expression pattern of HSP70-negative (A) and HSP70-positive (brown color, B) specimens.\n[SUBTITLE] Prognostic significance of HSP expression [SUBSECTION] When all patients were considered, no statistical influence on survival of HSP70 could be detected. After classifying the samples as T2 and T3/T4 tumors, prognostic significance of HSP70 expression in tumor cells could be detected in patients suffering from T2 tumors. Figure 3 shows cumulative survival curves for patients with T2 tumors with positive and negative HSP70 expression (Table 1 and Figure 3).\nInfluence of HSP70 expression on overall survival of patients suffering from T2 tumors following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Of 28 patients, 17 patients showed HSP70 expression greater or equal to 20% (blue curve). This predicts significantly improved survival compared to patients with less then 20% expression of HSP70 (red curve). [p = 0.009]\nThe survival of patients suffering from T2 tumors with positive HSP70 expression was 8 times higher than that for patients with negative HSP70 expression.\nWhen all patients were considered, no statistical influence on survival of HSP70 could be detected. After classifying the samples as T2 and T3/T4 tumors, prognostic significance of HSP70 expression in tumor cells could be detected in patients suffering from T2 tumors. Figure 3 shows cumulative survival curves for patients with T2 tumors with positive and negative HSP70 expression (Table 1 and Figure 3).\nInfluence of HSP70 expression on overall survival of patients suffering from T2 tumors following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Of 28 patients, 17 patients showed HSP70 expression greater or equal to 20% (blue curve). This predicts significantly improved survival compared to patients with less then 20% expression of HSP70 (red curve). [p = 0.009]\nThe survival of patients suffering from T2 tumors with positive HSP70 expression was 8 times higher than that for patients with negative HSP70 expression.", "Of the 61 patients, 49 were men and 12 women. The age of the patients ranged from 35 to 60 years with a mean age of 50.7 years. T2 tumors were present in 28 patients, T3 tumors in 3 patients, and T4 tumors in 30 patients. Lymph node metastases were found in 25 patients.", "There was no statistical influence on prognosis of grading, T- stadium, age, and sex of the patients. Presence of positive lymph nodes (N+) was identified as a prognostic factor with significant influence in univariate and multivariate analysis, as well in all patients suffering from T2 tumors (Table 1, Figure 1A, and Figure 1B). The mortality risk of these patients is 13 times than that of patients with negative nodes.\nExpression of HSP70 and other tumor characteristics multivariate analyses for disease-free survival prediction in 28 patients with T2 oral squamous cell carcinoma.\n(Cox's partially nonparametric regression model was used to evaluate the predictive power of various combinations and interactions of prognostic factors. Cl = confidence interval.)\nInfluence of nodal stage on overall survival following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Considering all patients (A), 25 of 35 had a positive lymph node stage. Regarding the patients suffering from T2 tumors (B), 6 of 28 showed a positive lymph node stage. Survival of patients with negative lymph nodes (green) and positive lymph nodes (blue) (p = 0.002).", "Immunoreactivity for HSP70 was positive in tumor cells of 38 of all patients (63.3%). Positive immunoreactivity of tumor cells could be detected in 17 of 28 patients with T2 tumors (60.7%) (Table 2, Figure 2).\nProportional distribution of HSP70-positive and HSP70-negative tumors separated into T2 and T3/T4 tumors.\nPhotographs from tissue sections of oral squamous cell carcinoma immunostained for HSP70. Expression pattern of HSP70-negative (A) and HSP70-positive (brown color, B) specimens.", "When all patients were considered, no statistical influence on survival of HSP70 could be detected. After classifying the samples as T2 and T3/T4 tumors, prognostic significance of HSP70 expression in tumor cells could be detected in patients suffering from T2 tumors. Figure 3 shows cumulative survival curves for patients with T2 tumors with positive and negative HSP70 expression (Table 1 and Figure 3).\nInfluence of HSP70 expression on overall survival of patients suffering from T2 tumors following diagnosis of oral cancer, using the product-limit-method of Kaplan and Meier. Of 28 patients, 17 patients showed HSP70 expression greater or equal to 20% (blue curve). This predicts significantly improved survival compared to patients with less then 20% expression of HSP70 (red curve). [p = 0.009]\nThe survival of patients suffering from T2 tumors with positive HSP70 expression was 8 times higher than that for patients with negative HSP70 expression.", "Many studies in cell biology have addressed the roles of HSPs as molecular chaperones, including protein folding and unfolding, translocation, and prevention of inappropriate protein aggregation [7,19]. These findings show an association between HSPs and cell proliferation and the prevention of apoptosis [20]. While HSP expression has been recognized as a factor of prognostic value in certain tumors, the data are limited and the results often contradictory. For example, inducible HSP72 has been shown to be a negative prognostic factor for disease-free survival (DFS) in patients with lymph node-negative breast cancer [21], whereas other authors have shown that HSP72 positively correlates with estrogen receptors and inversely with expression of mutant p53 [22]. Santarosa et al. demonstrated prognostic implications of expression of HSP70 in patients suffering from renal cancer [23]. Although some evidence indicates that HSPs are involved in various aspects of cell transformation and immune response against cancer [24], their biological role and its implications for the clinical course in cancer patients are not clear. In OSCC, the number of positive nodes, macroscopic extracapsular spread, and tumor infiltration of the resection margins have been described as significant prognostic factors [1,2]. Studies concerning HSP70 as a prognostic factor in esophageal carcinoma suggest that reduction of HSP70 expression is significantly correlated with poor prognosis [16,25]. Although several studies have been performed to elucidate the relationship between HSPs and tumors in various organs, to our knowledge, there are comparatively few reports related to OSCC. Sugarman et al. reported that HSP70 expression is not a definitive marker of oral malignancy or malignant potential [26]. In terms of prognostic significance, Ito et al. examined 24 specimens of patients suffering from OSCC. Although HSP immunohistochemistry revealed changes in HSP expression during tumorigenesis of squamous epithelium of the tongue, there was no correlation between HSP staining and survival period, stage, lymph node metastasis, histological grade, or p53 immunostaining [27]. These results are in line with those of Gandour-Edwards et al, who found that HSPs were expressed in normal upper respiratory tract squamous mucosa, and their role in carcinoma thus remains unclear. None of the markers (p53, HSP27, or HSP70) demonstrated prognostic significance for 5-year survival. We confirm the previously identified association of cervical lymph node metastases with decreased survival [5].\nThe results of the previously published data are not in line with the findings of the present study. In this study, we demonstrated that for patients suffering from T2-tumors, positive expression of HSP70 results in a significantly lower mortality risk compared with patients with negative expression. The main difference regarding the study design may be that, in contrast to previous studies, our patient population was divided according to tumor stages in the different T-stadium of the tumor. The purpose of this division was to analyze the prognostic significance of each T-stadium separately. Considering the entire patient group, our findings would suggest the same result: HSP70 has no prognostic implications for our patients. Also in the seperated T3- and T4-tumors, HSP 70 did not show any prognostic significance. Only after focusing on patients suffering from T2-tumors could a significant difference be detected. The survival of patients suffering from T2 tumors with positive HSP70 expression was 8 times higher than that for patients with negative HSP70 expression.", "The results of our study suggests that expression of HSP70 affects survival only in the early stage of the disease and that HSP70 membrane expression, in particular, is a target for natural killer cells [28]. Therefore in T2-tumors, the increased level of HSP70 may result in an extended tumor control by the natural killer cells. By implication, T1-T2 tumors of OSCC with low expression of Hsp70 could require more radical treatment.", "The authors declare that they have no competing interests.", "FT, HK, NCG and AE conceived of the study and participated in its design and coordination. FT drafted the manuscript. CMT helped to draft the manuscript. OFS carried out the immunohistochemistry and the HSP 70 Expression. GW performed the statistical analysis. All authors read and approved the final manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Gene Expression", "Genetic Variation", "Genetics, Population", "Humans", "Leukocytes", "Mutation", "Phenotype", "RNA, Messenger", "Tuberous Sclerosis", "Tuberous Sclerosis Complex 1 Protein", "Tuberous Sclerosis Complex 2 Protein", "Tumor Suppressor Proteins" ]

[ "Background", "IRB Approval", "Isolation of DNA and RNA from blood samples", "cDNA Synthesis from RNA samples", "Selection of Coding Region SNPs in TSC1 and TSC2", "Genotyping using the SNaPshot assay", "Allelic expression imbalance assay", "Results", "SNP frequencies in the sample population", "AEI in the TSC1 gene", "AEI in the TSC2 gene", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Tuberous sclerosis complex (TSC) is an autosomal dominant neurogenetic disease caused by a mutation in either the TSC1 or TSC2 gene [1-3]. Roughly two-thirds of TSC cases reported in mutational and epidemiological studies are sporadic (simplex), while the remaining cases are familial [4-9]. Neurological symptoms include seizures, cognitive delay, impulsivity, attention deficit, and learning disabilities. TSC patients often present with characteristic brain lesions, including cortical tubers, subependymal nodules (SENs), and subependymal giant cell astrocytomas (SEGAs). The severity of neurological symptoms is variable, although mental retardation and intractable epilepsy are fairly common and are frequently the most debilitating symptoms [2,10,11].\nLesions outside of the nervous system, including renal angiomyolipomas (AMLs), renal cysts, cardiac rhabdomyomas, facial angiofibromas, periungual fibromas, retinal hamartomas, and pulmonary lymphangioleiomyomas (LAM), are also characteristic of TSC [2,11]. Some of these lesions may result in life threatening events, such as hemorrhage into a large AML [12,13] or spontaneous pneumothorax or chylothorax from a ruptured LAM [14].\nMany of the hamartomatous growths associated with TSC are likely to be caused by loss of heterozygosity (LOH) due to a \"second-hit\" mutation that compromises the remaining normal TSC allele. This has been demonstrated in renal AMLs, cardiac rhabdomyomas, SEGAs and SENs [13,15-18]. By contrast, LOH has only rarely been demonstrated in cortical tubers [19,20]. While the lesions of TSC are generally associated with LOH, cognitive symptoms, including mental retardation, hyperactivity, impulsivity and attention deficit, may occur by a different mechanism, likely involving haploinsufficiency of TSC proteins in brain cells. In fact, the pathway in which hamartin and tuberin function has been shown to influence both neuronal structure and function [21]. It is therefore plausible that dysregulation of this pathway (a quantitative effect) produces cognitive deficits.\nStudies of coding and splice region mutations of the TSC1 and TSC2 genes have not yet produced a clear understanding of the relationship between genotype and phenotype, as people with the same primary mutation often have very different phenotypic outcomes [22,23]. It is generally accepted, however, that mutations in the TSC1 gene produce milder symptoms compared to mutations in the TSC2 gene [4,5,9,24]. Although most studies have failed to consistently link specific mutations to distinct phenotypes, there are exceptions such as the TSC2 R905Q mutation, which produces a mild form of the disease, and the TSC2 R905W and R905G mutations, which are associated with more severe forms of TSC [25].\nOur research is aimed at understanding why individuals carrying identical TSC gene mutations often have widely varying clinical outcomes. It has been repeatedly noted in the literature that phenotypic variation of TSC disease is very common within families [2,26-31]. The reason for this intra-familial variability in phenotype is currently unknown, although potential explanations include the modifying effects of unlinked genes, epigenetic factors [32,33], or mosaicism [34,35].\nIn many simple genetic disorders, pathogenic mutations inactivate the encoded protein or reduce its quantity or stability, thereby leading to an inadequate level of functional protein in the cell. TSC is an autosomal dominant genetic disease and, consequently, affected individuals are heterozygous for mutations in TSC1 or TSC2, i.e., one mutant and one normal allele is present in each cell [1-3]. We hypothesize that the differential expression of normal and mutant alleles may account for some proportion of the observed phenotypic variation. For example, it is possible that at the cellular level, relatively high levels of expression of the normal allele may compensate for the abnormal protein produced by the mutant allele. Conversely, high expression of a \"gain of function\" mutant protein, such as a mutant with dominant-negative properties, may be particularly deleterious. Based upon these considerations, it is plausible that allele-specific cis-acting elements that regulate mRNA expression [36-39] contribute to differences in disease severity in TSC.\nIf common regulatory elements within the TSC loci play a role in modulating disease phenotype in individuals carrying a mutation at one of the TSC genes, we would expect to be able to detect the same regulatory elements in subjects selected from the normal population. To test this hypothesis we used a PCR/primer extension-based assay to measure allele-specific differences in expression of TSC1 and TSC2 mRNAs in leukocytes isolated from normal volunteers. The use of this assay allows highly accurate measurements of \"allelic expression imbalance\" (AEI) for individuals who are heterozygous for a \"marker\" single nucleotide polymorphism (SNP) located within the mRNA. Based on these measurements, we estimate that about 19% of the population (our sample group was of mixed races with a predominance of Caucasian individuals) is heterozygous for high- and low-expression alleles at the TSC1 locus and 10% of the population is heterozygous for high- and low-expression alleles at the TSC2 locus.", "This research was approved by the St. Joseph's Hospital and Medical Center Institutional Review Board (IRB) for Human Research. Informed consent was obtained from all study participants. Participants were healthy volunteers who denied any personal or familial history of Tuberous Sclerosis Complex.", "DNA was extracted from blood leukocytes using Gentra Puregene Blood Kits (Qiagen, Valencia, California) and stored at 4°C. RNA was extracted from blood leukocytes using PAXgene Blood RNA Kits (Qiagen, Valencia, California) and stored at -80°C.", "TSC1 (NM_000368) and TSC2 (NM_000548) mRNAs were reverse-transcribed to cDNA using gene-specific primers and the SuperScript III First-Strand synthesis system for RT-PCR, according to the manufacturer's protocol (Invitrogen, Carlsbad, California). The cDNA synthesis primer sequence for TSC1 mRNA was 5'-GGGCCTGTGCTGACTCTGGTTAGTG-3'. The sequence of the cDNA synthesis primer for TSC2 mRNA was 5'-TTTCACTGACAGGCAATACC-3'. cDNAs were stored at -20°C.", "To distinguish TSC gene alleles we chose marker SNPs with relatively high rates of heterozygosity, as indicated in the NCBI Human Genome Resource SNP database http://www.ncbi.nlm.nih.gov, the SNPper resource (CHIP Bioinformatics resource - http://snpper.chip.org) and by our own genotyping data. Due to the need to distinguish the alleles, only samples heterozygous at marker SNPs were analyzed. Two SNPs were chosen as markers for TSC1 alleles: rs739442 (C/T) and rs2809243 (C/T), both located in the 3'-untranslated region (UTR) of TSC1 mRNA. The DNA samples were genotyped at several exonic SNPs in the TSC2 gene. One synonymous SNP located within exon 40, rs1748 (C/T), proved to have the highest rate of heterozygosity among the tested SNPs (~24%) and was therefore used for AEI analysis. Together, these three marker SNPs tag all known TSC1 and TSC2 mRNA splice variants.", "All samples were genotyped using the SNaPshot assay. This method of genotyping relies on the presence of heterozygous marker SNPs to distinguish between two alleles of a gene. In homozygous samples, where the gene alleles have the same nucleotide at the SNP locus, electropherograms will show a single peak using the forward primer and a single peak using the reverse primer. In heterozygous samples, the presence of different nucleotides at the SNP locus on each allele will result in the production of two peaks in both forward and reverse reactions.\nPCR primer pairs were designed for amplifying genomic DNA segments that included each SNP of interest. The amplimer segments were used in a SNaPshot assay (ABI Prism SNaPshot Multiplex Kit) to establish the genotype (homozygous versus heterozygous) of individuals at each of the marker SNPs. The primers for amplifying the 3'UTR genomic DNA segment containing the TSC1 SNPs rs2809243 and rs739442 (amplimer size = 587 bp) were: 5'-TAGTAATGGCAGAGCAGTCTAAACA-3' (forward) and 5'-TCCAGGTCTCATTCTCCCAACCGTA-3' (reverse). The primers for amplifying a genomic DNA segment surrounding TSC2 exon 40 were: 5'-CTGGGCAACGACTTTGTGTCCATTGTCTAC-3' (forward) and 5'-CTGACAGGCAATACCGTCCAA-3' (reverse). This primer pair produces an 1857 bp amplimer when used with genomic DNA. The PCR program consisted of an initial denaturation at 94°C for 40 seconds. This was followed by 40 cycles of 94°C for 20 seconds, 55°C for 1 minute, 72°C for 1 1/2 minutes, and a final extension step at 72°C for 5 minutes. PCR products were gel purified from 1-1.5% low melt agarose gels (IBI Scientific, Peosta, Iowa) using the Wizard PCR Preps DNA Purification System (Promega, Madison, Wisconsin).\nGenotyping was done with primers designed for SNaPshot analysis. The SNaPshot assay was performed according to the manufacturer's protocol (Applied Biosciences, Inc.). Briefly, a PCR reaction was run in which a single fluorescently-tagged dideoxynucleotide was added at the 3'-end of an annealed primer that was designed to terminate exactly one nucleotide before the SNP of interest (different fluorophores for ddA, ddG, ddC, and ddT). This allowed the identity of the SNP nucleotide to be determined using a capillary sequencer (Applied Biosystems Inc. Prism 310 Genetic Analyzer). This assay was used both for genotyping individuals at various SNPs and for AEI determination (as described in the following section). While both forward and reverse primers can be used for this analysis, the forward primers for each of the SNPs analyzed were found to give cleaner, more reliable results and were thus used in this assay.\nThe primers for SNaPshot analysis of the TSC1 gene alleles were: rs2809243 5'-AAACTCAACAAGTGCTCCTGAAAGAAA-3' (forward) and rs739442 5'-TACGAAATCTTAGTGCC-3' (forward). The primer for SNaPshot analysis of the TSC2 gene allele was rs1748 (forward): 5'-GCATCATAGCCGCTCCAACCCCACCGA-3'. The PCR program consisted of 25 cycles of a 96°C denaturation step for 10 seconds, 50°C for 5 seconds and 60°C for 30 seconds. Subsequently, samples were treated for 45 minutes at 37°C with 5 units of antarctic phosphatase (New England Biolabs, Ipswich, Massachusetts). The phosphatase was then inactivated by incubating at 65°C for 10 minutes. Samples were run on the capillary sequencer and the genotype determined from the electropherogram generated during the run.", "Samples heterozygous at marker SNPs were tested for AEI using the SNaPshot assay. Genomic and cDNA fragments flanking each SNP (as described above) were amplified in triplicate and gel purified. The same primers previously described for use in the amplification of genomic DNA segments were used in this assay to amplify TSC1 and TSC2 cDNA gene segments. The primer pair amplifying the TSC2 cDNA segment produces a 553 bp fragment when used with cDNA rather than the 1857 bp fragment produced with genomic DNA as the template. This is due to inclusion of intronic sequence in the PCR product from genomic DNA. The PCR reactions amplify both alleles, preserving the existing allele ratios in genomic DNA and in cDNA. The concentrations of gel purified samples (purification performed as indicated above) were measured using a Nanodrop 2000c (Thermo Scientific, Waltham, Massachusetts). Equal concentrations of the amplified fragments were then used in SNaPshot assays. All genomic and cDNA samples were analyzed in triplicate using the ABI capillary sequencer.\nGenomic DNA has a theoretical allele ratio of 1.0, but due to differences in the detection efficiency of various fluorophores, this ratio often deviates from 1.0. Therefore, genomic DNA was used as an internal control to correct for the differences in detection. In order to calculate the necessary correction factor, the genomic DNA allelic ratios for a specific SNP from each capillary sequencer run were averaged and the correction factor was calculated as the inverse of this average genomic allelic ratio. A diagram of the method used for calculating and applying the correction factor is shown in Figure 1. The allele ratios for genomic and cDNA samples were calculated as the ratio of heterozygous peak heights (analysis done using Gene Mapper 3.0 software from Applied Biosystems, Inc.). The experimental values for both the genomic DNA and the cDNA were then multiplied by the correction factor and average values were calculated for each sample analyzed in triplicate (see Results section for additional details). Standard error of the mean (SEM) was calculated for each sample using Excel software (Microsoft, Inc.) and error bars indicating 2X SEM were used in graphing the results.\nMethod for correcting genomic and cDNA allele ratios (AR). Genomic DNA segments containing marker SNPs are amplified by PCR and used as templates in SNaPshot primer-extension assays. Extended primers containing one of two different fluorescently labeled nucleotides at their 3'-ends are resolved by capillary electrophoresis and the ratio of peak heights calculated. An average genomic AR for a specific SNP is determined from all the genomic samples (each analyzed in triplicate). A correction factor (CF) is then calculated as the inverse of the average genomic AR. The genomic samples analyzed in triplicate are then individually multiplied by the CF to normalize the data to approximately 1.0, which is the theoretical ratio of two gene alleles in genomic DNA. The corrected average genomic AR is then determined. For each RNA sample, cDNA is synthesized and heterozygous SNP containing segments are amplified by PCR in triplicate and subjected to a SNaPshot PCR reaction. Samples are run on a capillary sequencer and the ratio of heterozygous peak heights is determined. Individual cDNA ARs are calculated and corrected by multiplying by the CF. The average corrected cDNA AR for each sample is then calculated. A sample is designated as showing AEI if the average corrected genomic AR and average corrected cDNA AR differ by greater than 2X the standard error of the mean and by at least 10%.", "AEI was examined in the TSC1 and TSC2 genes by quantifying the relative amounts of mRNA derived from each of the two alleles of each gene in leukocyte RNA samples isolated from normal individuals heterozygous for mRNA marker SNPs. To avoid the confounding effects of alternative splicing, SNPs located within the 3'-UTR of the TSC1 gene and in exon 40 of the TSC2 gene were selected as markers. These regions are included in all known mRNA forms generated from the TSC1 and TSC2 genes. The rs numbers (NCBI SNP data base, http://www.ncbi.nlm.nih.gov/snp) and locations of the three TSC1 and TSC2 marker SNPs used in this study are shown in Figure 2.\nThis diagram shows the exon/intron structure of the TSC1 and TSC2 genes. Exons are represented by numbered boxes. Exons subject to alternative splicing are indicated by brackets. The locations of the SNPs used for analysis of AEI are indicated by stars.\nAs described in detail in Methods, our AEI assays involve PCR amplification of short segments of TSC1 or TSC2 cDNA containing a marker SNP, followed by annealing of a synthetic oligonucleotide primer to a site immediately upstream from the SNP and primer extension in the presence of fluorescently tagged dideoxynucleotide triphosphates (ddNTPs). Because each ddNTP is tagged with a different fluorophore, the identity of the added base can be determined by resolving the fluorescently labeled primers by capillary electrophoresis and identifying the \"color\" of each extended primer [40].\nDifferences in expression between two alleles can be quantified by calculating the ratio of the peak heights of the traces corresponding to each fluorescently labeled primer. To correct for artifactual imbalances related to technical aspects of the assay, AEI assays were also carried out using genomic DNA, which in the absence of local chromosome deletions or duplications, would be expected to contain equal numbers of each allele. A correction factor derived from these measurements was used to correct AEI measurements obtained using cDNA templates.\n[SUBTITLE] SNP frequencies in the sample population [SUBSECTION] Our assay uses SNPs located within protein coding exons or the 3'-UTR to distinguish between the mRNA species that are transcribed from the two alleles of a gene in each individual. Because only subjects who are heterozygous at marker SNPs are informative in our assays, we first genotyped our subjects to identify individuals who are heterozygous for one or more of the TSC1 and TSC2 marker SNPs described above. The heterozygosities of the TSC1 markers rs2809243 and rs739442 were approximately 49% (41/83) and 45% (37/83), respectively, in our sample. Heterozygosity of the TSC2 marker SNP rs1748 was approximately 22% (18/82). These data are similar to average population heterozygosities for subjects of all races reported for these SNPs on the SNPper website (CHIP Bioinformatics resource - http://snpper.chip.org) and the NCBI SNP database. Approximately 36% (30/83) of subjects were heterozygous at both TSC1 SNPs.\nOur assay uses SNPs located within protein coding exons or the 3'-UTR to distinguish between the mRNA species that are transcribed from the two alleles of a gene in each individual. Because only subjects who are heterozygous at marker SNPs are informative in our assays, we first genotyped our subjects to identify individuals who are heterozygous for one or more of the TSC1 and TSC2 marker SNPs described above. The heterozygosities of the TSC1 markers rs2809243 and rs739442 were approximately 49% (41/83) and 45% (37/83), respectively, in our sample. Heterozygosity of the TSC2 marker SNP rs1748 was approximately 22% (18/82). These data are similar to average population heterozygosities for subjects of all races reported for these SNPs on the SNPper website (CHIP Bioinformatics resource - http://snpper.chip.org) and the NCBI SNP database. Approximately 36% (30/83) of subjects were heterozygous at both TSC1 SNPs.\n[SUBTITLE] AEI in the TSC1 gene [SUBSECTION] AEI analysis of TSC1 mRNA expression was performed independently using the marker SNPs rs2809243 and rs739442. As indicated above, 41 individuals were heterozygous at rs2809243 and 37 were heterozygous at rs739442. 30 individuals were heterozygous at both of the marker SNPs. Data from these doubly heterozygous individuals was used for independent validation of the results from each SNP. As outlined in Figure 1, AEI measurements using genomic DNA as template were carried out to permit the calculation of a correction factor for AEI measurements using cDNA as template. AEIs were considered to be significant if the corrected cDNA allelic expression ratio differed from the corrected genomic allele ratio by greater than 10%, and if the error bars (defined here as 2x the standard error of the mean) for the average genomic and cDNA allele ratios did not overlap.\nFigure 3 displays the corrected genomic and cDNA AEI ratios for each individual in our sample. Shown to the left in each graph is the data for individuals heterozygous at both marker SNPs. Shown to the right in each graph is additional data for individuals heterozygous at a single marker SNP. 8/41 individuals show AEI using rs2809243 while 7/37 individuals show AEI using rs739442. Of the doubly heterozygous individuals, rs2809243 revealed 6 individuals with AEI reaching our defined level of significance while rs739442 showed 5 individuals demonstrating AEI. The 5 individuals with AEI by rs739442 were the same as those with AEI by rs2809243. A 6th individual's sample (#11) reached AEI significance by a small margin using rs2809243 but did not reach significance using rs739442 as the marker SNP, thus emphasizing the importance of using a second SNP to validate data. We were able to consistently score 5 out of 6 individuals as demonstrating significant AEI in blood leukocytes using two different SNPs. The AEI of TSC1 mRNA expression in this control group ranged from 10% to greater than 30%. While this degree of imbalance is relatively small, it could be sufficient to modulate the phenotype in a TSC patient heterozygous for a mutation in the TSC1 gene. Based on this small cohort of control subjects we estimate that the population frequency of AEI at the TSC1 locus may be as high as 15-20%.\nAEI analysis of TSC1 mRNA expression in leukocytes isolated from 30 individuals doubly heterozygous for the marker SNPs rs2809243 and rs739442 and additional individuals (11 and 7 individuals respectively) singly heterozygous for one of the two SNPs. For doubly heterozygous individuals (data at the left side of graphs 3A and 3B), blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate the corrected cDNA ARs. Data for singly heterozygous individuals is located to the right side of the 3A and 3B graphs using red and green bars to indicate genomic ARs and shaded bars to indicate cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the genomic AR by more than 10% and had error bars that did not overlap with those of the genomic DNA.\nIt should be noted that the cDNA allelic expression ratios measured using the marker SNP rs280943 range from greater that 1 (samples 29 and 36) to less than 1 (samples 3 and 27). This implies that the regulatory variant or variants resulting in this AEI are not tightly linked to the marker SNP. Thus, if these individuals were heterozygous for a remote regulatory variant comprising one high-expression allele and one low-expression allele, the results of our AEI measurements imply that the high-expression allele is \"in phase\" (ie., located on the same chromosome) with the rs280943 C-allele in individuals 29 and 36, but is \"in-phase\" with the rs280943 T-allele in individuals 3 and 27.\nSimilar arguments hold for the AEI measurement obtained using rs739442 as the marker SNP. The fact that the directions of the measured AEI differ for individual #27, depending upon the choice of marker SNP, implies that the \"phase\" of the marker SNPs with respect to the functional variant is different in this individual. That is, in this individual the rs739442 C-allele is located on the same chromosome as the high-expression allele of the remote regulatory variant. Although the two SNPs used for these analyses (rs280943 and rs739442) are separated by only 166 bp, our data indicate that these SNPs are not tightly linked. As previously indicated, only around 29% of our sample population is heterozygous at both TSC1 marker SNPs despite their close proximity. This is apparent in the data of individual #27 which shows the marker SNPs to be on different chromosomes. Using the Hapmap database http://hapmap.ncbi.nlm.nih.gov the linkage disequilibrium D' value for these SNPs is 0.671, confirming that these SNPs are not tightly linked despite the small separation distance. As these two SNPs are both located in the 3'UTR of TSC1, it is not surprising to see this level of variation as mutations in this area are less likely to affect the protein function.\nAEI analysis of TSC1 mRNA expression was performed independently using the marker SNPs rs2809243 and rs739442. As indicated above, 41 individuals were heterozygous at rs2809243 and 37 were heterozygous at rs739442. 30 individuals were heterozygous at both of the marker SNPs. Data from these doubly heterozygous individuals was used for independent validation of the results from each SNP. As outlined in Figure 1, AEI measurements using genomic DNA as template were carried out to permit the calculation of a correction factor for AEI measurements using cDNA as template. AEIs were considered to be significant if the corrected cDNA allelic expression ratio differed from the corrected genomic allele ratio by greater than 10%, and if the error bars (defined here as 2x the standard error of the mean) for the average genomic and cDNA allele ratios did not overlap.\nFigure 3 displays the corrected genomic and cDNA AEI ratios for each individual in our sample. Shown to the left in each graph is the data for individuals heterozygous at both marker SNPs. Shown to the right in each graph is additional data for individuals heterozygous at a single marker SNP. 8/41 individuals show AEI using rs2809243 while 7/37 individuals show AEI using rs739442. Of the doubly heterozygous individuals, rs2809243 revealed 6 individuals with AEI reaching our defined level of significance while rs739442 showed 5 individuals demonstrating AEI. The 5 individuals with AEI by rs739442 were the same as those with AEI by rs2809243. A 6th individual's sample (#11) reached AEI significance by a small margin using rs2809243 but did not reach significance using rs739442 as the marker SNP, thus emphasizing the importance of using a second SNP to validate data. We were able to consistently score 5 out of 6 individuals as demonstrating significant AEI in blood leukocytes using two different SNPs. The AEI of TSC1 mRNA expression in this control group ranged from 10% to greater than 30%. While this degree of imbalance is relatively small, it could be sufficient to modulate the phenotype in a TSC patient heterozygous for a mutation in the TSC1 gene. Based on this small cohort of control subjects we estimate that the population frequency of AEI at the TSC1 locus may be as high as 15-20%.\nAEI analysis of TSC1 mRNA expression in leukocytes isolated from 30 individuals doubly heterozygous for the marker SNPs rs2809243 and rs739442 and additional individuals (11 and 7 individuals respectively) singly heterozygous for one of the two SNPs. For doubly heterozygous individuals (data at the left side of graphs 3A and 3B), blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate the corrected cDNA ARs. Data for singly heterozygous individuals is located to the right side of the 3A and 3B graphs using red and green bars to indicate genomic ARs and shaded bars to indicate cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the genomic AR by more than 10% and had error bars that did not overlap with those of the genomic DNA.\nIt should be noted that the cDNA allelic expression ratios measured using the marker SNP rs280943 range from greater that 1 (samples 29 and 36) to less than 1 (samples 3 and 27). This implies that the regulatory variant or variants resulting in this AEI are not tightly linked to the marker SNP. Thus, if these individuals were heterozygous for a remote regulatory variant comprising one high-expression allele and one low-expression allele, the results of our AEI measurements imply that the high-expression allele is \"in phase\" (ie., located on the same chromosome) with the rs280943 C-allele in individuals 29 and 36, but is \"in-phase\" with the rs280943 T-allele in individuals 3 and 27.\nSimilar arguments hold for the AEI measurement obtained using rs739442 as the marker SNP. The fact that the directions of the measured AEI differ for individual #27, depending upon the choice of marker SNP, implies that the \"phase\" of the marker SNPs with respect to the functional variant is different in this individual. That is, in this individual the rs739442 C-allele is located on the same chromosome as the high-expression allele of the remote regulatory variant. Although the two SNPs used for these analyses (rs280943 and rs739442) are separated by only 166 bp, our data indicate that these SNPs are not tightly linked. As previously indicated, only around 29% of our sample population is heterozygous at both TSC1 marker SNPs despite their close proximity. This is apparent in the data of individual #27 which shows the marker SNPs to be on different chromosomes. Using the Hapmap database http://hapmap.ncbi.nlm.nih.gov the linkage disequilibrium D' value for these SNPs is 0.671, confirming that these SNPs are not tightly linked despite the small separation distance. As these two SNPs are both located in the 3'UTR of TSC1, it is not surprising to see this level of variation as mutations in this area are less likely to affect the protein function.\n[SUBTITLE] AEI in the TSC2 gene [SUBSECTION] Twenty out of 83 individuals in our sample were heterozygous for the TSC2 mRNA marker SNP rs1748, As shown in Figure 4, 10% (2/20) of these individuals demonstrated AEI above the 10% cut-off, with a difference of more than 2x the SEM. An independent marker SNP was not available for verification; however, AEI measurements were highly reproducible. Our data demonstrates that AEI is relatively common in both the TSC1 and TSC2 genes.\nAEI analysis of TSC2 mRNA expression in leukocytes isolated from 20 individuals heterozygous for the marker SNP rs1748. Blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate corrected cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the average corrected genomic AR by more than 10% and had error bars (2X SEM) that did not overlap with those of the genomic DNA.\nTwenty out of 83 individuals in our sample were heterozygous for the TSC2 mRNA marker SNP rs1748, As shown in Figure 4, 10% (2/20) of these individuals demonstrated AEI above the 10% cut-off, with a difference of more than 2x the SEM. An independent marker SNP was not available for verification; however, AEI measurements were highly reproducible. Our data demonstrates that AEI is relatively common in both the TSC1 and TSC2 genes.\nAEI analysis of TSC2 mRNA expression in leukocytes isolated from 20 individuals heterozygous for the marker SNP rs1748. Blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate corrected cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the average corrected genomic AR by more than 10% and had error bars (2X SEM) that did not overlap with those of the genomic DNA.", "Our assay uses SNPs located within protein coding exons or the 3'-UTR to distinguish between the mRNA species that are transcribed from the two alleles of a gene in each individual. Because only subjects who are heterozygous at marker SNPs are informative in our assays, we first genotyped our subjects to identify individuals who are heterozygous for one or more of the TSC1 and TSC2 marker SNPs described above. The heterozygosities of the TSC1 markers rs2809243 and rs739442 were approximately 49% (41/83) and 45% (37/83), respectively, in our sample. Heterozygosity of the TSC2 marker SNP rs1748 was approximately 22% (18/82). These data are similar to average population heterozygosities for subjects of all races reported for these SNPs on the SNPper website (CHIP Bioinformatics resource - http://snpper.chip.org) and the NCBI SNP database. Approximately 36% (30/83) of subjects were heterozygous at both TSC1 SNPs.", "AEI analysis of TSC1 mRNA expression was performed independently using the marker SNPs rs2809243 and rs739442. As indicated above, 41 individuals were heterozygous at rs2809243 and 37 were heterozygous at rs739442. 30 individuals were heterozygous at both of the marker SNPs. Data from these doubly heterozygous individuals was used for independent validation of the results from each SNP. As outlined in Figure 1, AEI measurements using genomic DNA as template were carried out to permit the calculation of a correction factor for AEI measurements using cDNA as template. AEIs were considered to be significant if the corrected cDNA allelic expression ratio differed from the corrected genomic allele ratio by greater than 10%, and if the error bars (defined here as 2x the standard error of the mean) for the average genomic and cDNA allele ratios did not overlap.\nFigure 3 displays the corrected genomic and cDNA AEI ratios for each individual in our sample. Shown to the left in each graph is the data for individuals heterozygous at both marker SNPs. Shown to the right in each graph is additional data for individuals heterozygous at a single marker SNP. 8/41 individuals show AEI using rs2809243 while 7/37 individuals show AEI using rs739442. Of the doubly heterozygous individuals, rs2809243 revealed 6 individuals with AEI reaching our defined level of significance while rs739442 showed 5 individuals demonstrating AEI. The 5 individuals with AEI by rs739442 were the same as those with AEI by rs2809243. A 6th individual's sample (#11) reached AEI significance by a small margin using rs2809243 but did not reach significance using rs739442 as the marker SNP, thus emphasizing the importance of using a second SNP to validate data. We were able to consistently score 5 out of 6 individuals as demonstrating significant AEI in blood leukocytes using two different SNPs. The AEI of TSC1 mRNA expression in this control group ranged from 10% to greater than 30%. While this degree of imbalance is relatively small, it could be sufficient to modulate the phenotype in a TSC patient heterozygous for a mutation in the TSC1 gene. Based on this small cohort of control subjects we estimate that the population frequency of AEI at the TSC1 locus may be as high as 15-20%.\nAEI analysis of TSC1 mRNA expression in leukocytes isolated from 30 individuals doubly heterozygous for the marker SNPs rs2809243 and rs739442 and additional individuals (11 and 7 individuals respectively) singly heterozygous for one of the two SNPs. For doubly heterozygous individuals (data at the left side of graphs 3A and 3B), blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate the corrected cDNA ARs. Data for singly heterozygous individuals is located to the right side of the 3A and 3B graphs using red and green bars to indicate genomic ARs and shaded bars to indicate cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the genomic AR by more than 10% and had error bars that did not overlap with those of the genomic DNA.\nIt should be noted that the cDNA allelic expression ratios measured using the marker SNP rs280943 range from greater that 1 (samples 29 and 36) to less than 1 (samples 3 and 27). This implies that the regulatory variant or variants resulting in this AEI are not tightly linked to the marker SNP. Thus, if these individuals were heterozygous for a remote regulatory variant comprising one high-expression allele and one low-expression allele, the results of our AEI measurements imply that the high-expression allele is \"in phase\" (ie., located on the same chromosome) with the rs280943 C-allele in individuals 29 and 36, but is \"in-phase\" with the rs280943 T-allele in individuals 3 and 27.\nSimilar arguments hold for the AEI measurement obtained using rs739442 as the marker SNP. The fact that the directions of the measured AEI differ for individual #27, depending upon the choice of marker SNP, implies that the \"phase\" of the marker SNPs with respect to the functional variant is different in this individual. That is, in this individual the rs739442 C-allele is located on the same chromosome as the high-expression allele of the remote regulatory variant. Although the two SNPs used for these analyses (rs280943 and rs739442) are separated by only 166 bp, our data indicate that these SNPs are not tightly linked. As previously indicated, only around 29% of our sample population is heterozygous at both TSC1 marker SNPs despite their close proximity. This is apparent in the data of individual #27 which shows the marker SNPs to be on different chromosomes. Using the Hapmap database http://hapmap.ncbi.nlm.nih.gov the linkage disequilibrium D' value for these SNPs is 0.671, confirming that these SNPs are not tightly linked despite the small separation distance. As these two SNPs are both located in the 3'UTR of TSC1, it is not surprising to see this level of variation as mutations in this area are less likely to affect the protein function.", "Twenty out of 83 individuals in our sample were heterozygous for the TSC2 mRNA marker SNP rs1748, As shown in Figure 4, 10% (2/20) of these individuals demonstrated AEI above the 10% cut-off, with a difference of more than 2x the SEM. An independent marker SNP was not available for verification; however, AEI measurements were highly reproducible. Our data demonstrates that AEI is relatively common in both the TSC1 and TSC2 genes.\nAEI analysis of TSC2 mRNA expression in leukocytes isolated from 20 individuals heterozygous for the marker SNP rs1748. Blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate corrected cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the average corrected genomic AR by more than 10% and had error bars (2X SEM) that did not overlap with those of the genomic DNA.", "There is a growing consensus that cis-acting genetic variants significantly contribute to phenotypic differences among individuals, including disease risk [36,39,41-45]. Regulatory polymorphisms are one of the predominant mechanisms by which cis-acting gene regulation has been found to occur. These polymorphisms, located in regulatory regions, influence the expression of genes by affecting transcriptional activation or repression, generally by altering the DNA binding sites for transcription factors [36,39,46,47]. In addition, splicing errors, changes in mRNA stablility, epigenetic modifications and polymorphic (ACn) microsatellites have also been implicated in cis-acting gene regulation [36,37,39,46-49].\nThe best known examples of allele-specific differences in gene expression have been associated with X-inactivation [50] or genomic imprinting [51]. However, allelic variation in expression has also been demonstrated in non-imprinted genes and this allelic variation itself can be regulated by cis-acting elements [36-39]. Variations in allele expression have been previously linked to disease. For example, allelic variation in APC (adenomatous polyposis coli) expression plays a role in predisposition to colon cancer [41]. Allelic expression imbalance has also been studied in the cancer associated genes BRCA1/2 and CDH1 and used to identify polymorphisms, mutations and other defects that alter allelic expression and influence disease state [45,48]. As many genes are active within networks, variation in the expression of specific gene alleles may ultimately result in multiple downstream effects within a network and between related gene networks. This creates an avenue by which even small differences in the expression of specific genes can ultimately result in substantial phenotypic changes [46,47].\nIt has been noted frequently that disease causing mutations in families with TSC may produce very few problems in certain individuals while having catastrophic effects in others [2,26-31]. Clearly, there are additional factors outside of the mutation itself that affect disease severity. Differences in allele specific mRNA expression could potentially be one of these disease modifying factors. It is possible that the overall amount of normal TSC protein in cells may determine the severity of the disease phenotype in patients. As TSC is a disease carried in a heterozygous state [1-3], some amount of normal TSC protein should be present in most cells since one normal allele of each TSC gene is present (the exception being abnormal tissue growths exhibiting LOH [15-18]). It is possible that higher relative expression of protein from the normal allele may be protective, while higher relative expression of abnormal protein from the mutant allele may have deleterious effects.\nWe began to study this issue by determining the frequency of occurrence of AEI of the TSC1 and TSC2 genes in a control population. The intent was to establish if mRNA expression variation might be common enough to be a mechanism by which phenotypes are modified in patients with TSC gene mutations. In a cohort of normal volunteers we were able to quantify allele specific expression of the TSC genes in blood RNA and estimate the frequency of allelic skewing of expression for these two genes. In our studies we found that there was significant skewing of allelic expression of the TSC1 gene in about 19% of our sample population and of the TSC2 gene in 10% of our population. This estimate is based on a small sample of informative individuals (48 individuals who were heterozygous for a TSC1 marker SNP and 20 who were heterozygous for a TSC2 marker SNP). This was a sample of convenience, but individuals were recruited without bias and should be representative of the general population. If we assume a binomial distribution for the occurrence of AEI in the general population, we can use the exact test to calculate 95% confidence intervals for the actual population frequencies of AEI at the TSC1 and TSC2 genes. Based on such calculations, the 95% confidence interval for prevalence of AEI at TSC1 is 9% to 33% and for AEI at the TSC2 gene is 1.2% to 32%. These confidence intervals for the estimates for the actual population frequencies of AEI at TSC1 and TSC2 can of course be sharpened with larger sample sizes.\nWhile these are not large proportions, AEI may be occurring frequently enough to be a potential contributor to the phenotypic differences in TSC patients. In any given individual patient within a particular family, the phenotype could be determined not just by the mutation, but also by SNPs located within regulatory regions of the TSC genes. In such familial cases of TSC, the implication is that regulatory SNPs inherited from the parents in various combinations with the normal and mutant gene alleles can affect the phenotype of the child.\nThe TSC1 and TSC2 gene products, hamartin and tuberin, function together as a protein complex. Therefore, mutation of either of the TSC genes results in the same disease [1,3,52]. The hamartin-tuberin complex is a modulator of the mTOR signaling pathway, which is important in the regulation of cell growth. We know that haploinsufficiency due to mutation of a single TSC gene allele is sufficient to cause TSC and represents an approximately 50% loss of the total expression of that TSC gene [2]. Loss of a single TSC gene allele is sufficient to disrupt neuronal morphology and function in mouse models [21]. Loss of both alleles of a TSC gene can result in the formation of hamartomas common to TSC as is demonstrated by LOH studies [13,15-18]. These points clearly suggest that pathways modulated by hamartin and tuberin are sensitive to gene dosage effects. If a 50% reduction in expression of a TSC gene is sufficient to cause disease, it is plausible that smaller variations in expression, such as the 10-30% that we found in our experiments, might be sufficient to influence phenotype. In our control sample group, this level of variation in allelic expression of TSC1 or TSC2 does not result in a phenotype, as both alleles encode normal proteins. This degree of variation in mRNA expression combined with mutation of a TSC gene allele may be sufficient to influence phenotype either positively or negatively.\nIt has previously been reported that a 50% decrease in the expression of a single allele of the adenomatous polyposis coli tumor suppressor gene (APC), representing an overall 25% decrease of APC mRNA expression, is sufficient to cause the development of familial adenomatous polyposis[41]. An additional study of a gene associated with osteoarthritis (GDF5) discovered that a promoter polymorphism which created a small reduction of the expression of the T allele (less than 27%), significantly increased individuals susceptibility to developing osteoarthritis [53]. These reports indicate that even small variations in allelic expression are important to disease outcomes. This supports our hypothesis that variation in expression of the TSC gene alleles, particularly in the presence of an existing genetic mutation, may influence disease severity in TSC patients.\nTissue specific expression of genes is an important consideration when assessing the effects of variation in allelic expression. A sequence variant in the regulatory region of a gene might be relevant in some tissues and not in others, leading to conflicting results in different tissues [54]. Our study was performed in blood samples as this is a readily available tissue specimen. It is important to determine if allelic expression ratios measured in peripheral blood correlate with ratios measured in brain tissue, something that may be done using banked tissue samples. A difficulty we've encountered is the availability of good-quality matched blood and brain tissue samples from which intact RNA and DNA can be extracted. Establishing a correlation between blood and brain expression levels is especially important as we try to relate expression of the TSC alleles in blood to severity of cognitive impairment in patients with TSC. The goal of our research is to determine if the levels of expression of mutant and wild-type alleles in patients with TSC, as measured in peripheral blood, correlates with phenotypic severity. To this end, we plan to next study familial cases of TSC, where multiple affected individuals have the same identical gene mutation, but are discordant in terms of disease severity. We shall determine if, in these multiplex families, disease severity is correlated with skewing of allele specific expression. Ultimately, we hope to use the combination of mutation detection and measures of AEI in blood samples to predict disease severity (at least in relation to cognitive impairment). Early identification of patients who are at risk for developing severe disease may allow for aggressive preventive interventions, and may protect the patient from additional damaging effects of the disease.", "We have concluded from our research that variation in TSC1 and TSC2 gene allele expression is common in normal individuals, as it was easily detected in a relatively small sample population. It is likely that this variation in allele expression will also be seen in some patients carrying TSC gene mutations and may therefore help to explain the intra-familial variation in disease severity frequently observed in TSC. These ideas can be tested in multiplex families that include patients with TSC that are discordant in disease severity (particularly cognitive symptoms). After such validation, we might be able to develop a simple blood test (ratio of wild-type to mutant TSC mRNA levels) that predicts disease severity in simplex cases of TSC.", "The authors declare that they have no competing interests.", "VN conceived of the study and VN, GJ and SR participated in its design and coordination. SR performed the DNA/RNA isolation and the AEI analysis. SO performed DNA sequencing and genotyping and provided theoretical advice. GJ and VN drafted the manuscript and SR and SO edited the manuscript. DS provided key technical advice and critical review of the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2350/12/29/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "IRB Approval", "Isolation of DNA and RNA from blood samples", "cDNA Synthesis from RNA samples", "Selection of Coding Region SNPs in TSC1 and TSC2", "Genotyping using the SNaPshot assay", "Allelic expression imbalance assay", "Results", "SNP frequencies in the sample population", "AEI in the TSC1 gene", "AEI in the TSC2 gene", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Tuberous sclerosis complex (TSC) is an autosomal dominant neurogenetic disease caused by a mutation in either the TSC1 or TSC2 gene [1-3]. Roughly two-thirds of TSC cases reported in mutational and epidemiological studies are sporadic (simplex), while the remaining cases are familial [4-9]. Neurological symptoms include seizures, cognitive delay, impulsivity, attention deficit, and learning disabilities. TSC patients often present with characteristic brain lesions, including cortical tubers, subependymal nodules (SENs), and subependymal giant cell astrocytomas (SEGAs). The severity of neurological symptoms is variable, although mental retardation and intractable epilepsy are fairly common and are frequently the most debilitating symptoms [2,10,11].\nLesions outside of the nervous system, including renal angiomyolipomas (AMLs), renal cysts, cardiac rhabdomyomas, facial angiofibromas, periungual fibromas, retinal hamartomas, and pulmonary lymphangioleiomyomas (LAM), are also characteristic of TSC [2,11]. Some of these lesions may result in life threatening events, such as hemorrhage into a large AML [12,13] or spontaneous pneumothorax or chylothorax from a ruptured LAM [14].\nMany of the hamartomatous growths associated with TSC are likely to be caused by loss of heterozygosity (LOH) due to a \"second-hit\" mutation that compromises the remaining normal TSC allele. This has been demonstrated in renal AMLs, cardiac rhabdomyomas, SEGAs and SENs [13,15-18]. By contrast, LOH has only rarely been demonstrated in cortical tubers [19,20]. While the lesions of TSC are generally associated with LOH, cognitive symptoms, including mental retardation, hyperactivity, impulsivity and attention deficit, may occur by a different mechanism, likely involving haploinsufficiency of TSC proteins in brain cells. In fact, the pathway in which hamartin and tuberin function has been shown to influence both neuronal structure and function [21]. It is therefore plausible that dysregulation of this pathway (a quantitative effect) produces cognitive deficits.\nStudies of coding and splice region mutations of the TSC1 and TSC2 genes have not yet produced a clear understanding of the relationship between genotype and phenotype, as people with the same primary mutation often have very different phenotypic outcomes [22,23]. It is generally accepted, however, that mutations in the TSC1 gene produce milder symptoms compared to mutations in the TSC2 gene [4,5,9,24]. Although most studies have failed to consistently link specific mutations to distinct phenotypes, there are exceptions such as the TSC2 R905Q mutation, which produces a mild form of the disease, and the TSC2 R905W and R905G mutations, which are associated with more severe forms of TSC [25].\nOur research is aimed at understanding why individuals carrying identical TSC gene mutations often have widely varying clinical outcomes. It has been repeatedly noted in the literature that phenotypic variation of TSC disease is very common within families [2,26-31]. The reason for this intra-familial variability in phenotype is currently unknown, although potential explanations include the modifying effects of unlinked genes, epigenetic factors [32,33], or mosaicism [34,35].\nIn many simple genetic disorders, pathogenic mutations inactivate the encoded protein or reduce its quantity or stability, thereby leading to an inadequate level of functional protein in the cell. TSC is an autosomal dominant genetic disease and, consequently, affected individuals are heterozygous for mutations in TSC1 or TSC2, i.e., one mutant and one normal allele is present in each cell [1-3]. We hypothesize that the differential expression of normal and mutant alleles may account for some proportion of the observed phenotypic variation. For example, it is possible that at the cellular level, relatively high levels of expression of the normal allele may compensate for the abnormal protein produced by the mutant allele. Conversely, high expression of a \"gain of function\" mutant protein, such as a mutant with dominant-negative properties, may be particularly deleterious. Based upon these considerations, it is plausible that allele-specific cis-acting elements that regulate mRNA expression [36-39] contribute to differences in disease severity in TSC.\nIf common regulatory elements within the TSC loci play a role in modulating disease phenotype in individuals carrying a mutation at one of the TSC genes, we would expect to be able to detect the same regulatory elements in subjects selected from the normal population. To test this hypothesis we used a PCR/primer extension-based assay to measure allele-specific differences in expression of TSC1 and TSC2 mRNAs in leukocytes isolated from normal volunteers. The use of this assay allows highly accurate measurements of \"allelic expression imbalance\" (AEI) for individuals who are heterozygous for a \"marker\" single nucleotide polymorphism (SNP) located within the mRNA. Based on these measurements, we estimate that about 19% of the population (our sample group was of mixed races with a predominance of Caucasian individuals) is heterozygous for high- and low-expression alleles at the TSC1 locus and 10% of the population is heterozygous for high- and low-expression alleles at the TSC2 locus.", "[SUBTITLE] IRB Approval [SUBSECTION] This research was approved by the St. Joseph's Hospital and Medical Center Institutional Review Board (IRB) for Human Research. Informed consent was obtained from all study participants. Participants were healthy volunteers who denied any personal or familial history of Tuberous Sclerosis Complex.\nThis research was approved by the St. Joseph's Hospital and Medical Center Institutional Review Board (IRB) for Human Research. Informed consent was obtained from all study participants. Participants were healthy volunteers who denied any personal or familial history of Tuberous Sclerosis Complex.\n[SUBTITLE] Isolation of DNA and RNA from blood samples [SUBSECTION] DNA was extracted from blood leukocytes using Gentra Puregene Blood Kits (Qiagen, Valencia, California) and stored at 4°C. RNA was extracted from blood leukocytes using PAXgene Blood RNA Kits (Qiagen, Valencia, California) and stored at -80°C.\nDNA was extracted from blood leukocytes using Gentra Puregene Blood Kits (Qiagen, Valencia, California) and stored at 4°C. RNA was extracted from blood leukocytes using PAXgene Blood RNA Kits (Qiagen, Valencia, California) and stored at -80°C.\n[SUBTITLE] cDNA Synthesis from RNA samples [SUBSECTION] TSC1 (NM_000368) and TSC2 (NM_000548) mRNAs were reverse-transcribed to cDNA using gene-specific primers and the SuperScript III First-Strand synthesis system for RT-PCR, according to the manufacturer's protocol (Invitrogen, Carlsbad, California). The cDNA synthesis primer sequence for TSC1 mRNA was 5'-GGGCCTGTGCTGACTCTGGTTAGTG-3'. The sequence of the cDNA synthesis primer for TSC2 mRNA was 5'-TTTCACTGACAGGCAATACC-3'. cDNAs were stored at -20°C.\nTSC1 (NM_000368) and TSC2 (NM_000548) mRNAs were reverse-transcribed to cDNA using gene-specific primers and the SuperScript III First-Strand synthesis system for RT-PCR, according to the manufacturer's protocol (Invitrogen, Carlsbad, California). The cDNA synthesis primer sequence for TSC1 mRNA was 5'-GGGCCTGTGCTGACTCTGGTTAGTG-3'. The sequence of the cDNA synthesis primer for TSC2 mRNA was 5'-TTTCACTGACAGGCAATACC-3'. cDNAs were stored at -20°C.\n[SUBTITLE] Selection of Coding Region SNPs in TSC1 and TSC2 [SUBSECTION] To distinguish TSC gene alleles we chose marker SNPs with relatively high rates of heterozygosity, as indicated in the NCBI Human Genome Resource SNP database http://www.ncbi.nlm.nih.gov, the SNPper resource (CHIP Bioinformatics resource - http://snpper.chip.org) and by our own genotyping data. Due to the need to distinguish the alleles, only samples heterozygous at marker SNPs were analyzed. Two SNPs were chosen as markers for TSC1 alleles: rs739442 (C/T) and rs2809243 (C/T), both located in the 3'-untranslated region (UTR) of TSC1 mRNA. The DNA samples were genotyped at several exonic SNPs in the TSC2 gene. One synonymous SNP located within exon 40, rs1748 (C/T), proved to have the highest rate of heterozygosity among the tested SNPs (~24%) and was therefore used for AEI analysis. Together, these three marker SNPs tag all known TSC1 and TSC2 mRNA splice variants.\nTo distinguish TSC gene alleles we chose marker SNPs with relatively high rates of heterozygosity, as indicated in the NCBI Human Genome Resource SNP database http://www.ncbi.nlm.nih.gov, the SNPper resource (CHIP Bioinformatics resource - http://snpper.chip.org) and by our own genotyping data. Due to the need to distinguish the alleles, only samples heterozygous at marker SNPs were analyzed. Two SNPs were chosen as markers for TSC1 alleles: rs739442 (C/T) and rs2809243 (C/T), both located in the 3'-untranslated region (UTR) of TSC1 mRNA. The DNA samples were genotyped at several exonic SNPs in the TSC2 gene. One synonymous SNP located within exon 40, rs1748 (C/T), proved to have the highest rate of heterozygosity among the tested SNPs (~24%) and was therefore used for AEI analysis. Together, these three marker SNPs tag all known TSC1 and TSC2 mRNA splice variants.\n[SUBTITLE] Genotyping using the SNaPshot assay [SUBSECTION] All samples were genotyped using the SNaPshot assay. This method of genotyping relies on the presence of heterozygous marker SNPs to distinguish between two alleles of a gene. In homozygous samples, where the gene alleles have the same nucleotide at the SNP locus, electropherograms will show a single peak using the forward primer and a single peak using the reverse primer. In heterozygous samples, the presence of different nucleotides at the SNP locus on each allele will result in the production of two peaks in both forward and reverse reactions.\nPCR primer pairs were designed for amplifying genomic DNA segments that included each SNP of interest. The amplimer segments were used in a SNaPshot assay (ABI Prism SNaPshot Multiplex Kit) to establish the genotype (homozygous versus heterozygous) of individuals at each of the marker SNPs. The primers for amplifying the 3'UTR genomic DNA segment containing the TSC1 SNPs rs2809243 and rs739442 (amplimer size = 587 bp) were: 5'-TAGTAATGGCAGAGCAGTCTAAACA-3' (forward) and 5'-TCCAGGTCTCATTCTCCCAACCGTA-3' (reverse). The primers for amplifying a genomic DNA segment surrounding TSC2 exon 40 were: 5'-CTGGGCAACGACTTTGTGTCCATTGTCTAC-3' (forward) and 5'-CTGACAGGCAATACCGTCCAA-3' (reverse). This primer pair produces an 1857 bp amplimer when used with genomic DNA. The PCR program consisted of an initial denaturation at 94°C for 40 seconds. This was followed by 40 cycles of 94°C for 20 seconds, 55°C for 1 minute, 72°C for 1 1/2 minutes, and a final extension step at 72°C for 5 minutes. PCR products were gel purified from 1-1.5% low melt agarose gels (IBI Scientific, Peosta, Iowa) using the Wizard PCR Preps DNA Purification System (Promega, Madison, Wisconsin).\nGenotyping was done with primers designed for SNaPshot analysis. The SNaPshot assay was performed according to the manufacturer's protocol (Applied Biosciences, Inc.). Briefly, a PCR reaction was run in which a single fluorescently-tagged dideoxynucleotide was added at the 3'-end of an annealed primer that was designed to terminate exactly one nucleotide before the SNP of interest (different fluorophores for ddA, ddG, ddC, and ddT). This allowed the identity of the SNP nucleotide to be determined using a capillary sequencer (Applied Biosystems Inc. Prism 310 Genetic Analyzer). This assay was used both for genotyping individuals at various SNPs and for AEI determination (as described in the following section). While both forward and reverse primers can be used for this analysis, the forward primers for each of the SNPs analyzed were found to give cleaner, more reliable results and were thus used in this assay.\nThe primers for SNaPshot analysis of the TSC1 gene alleles were: rs2809243 5'-AAACTCAACAAGTGCTCCTGAAAGAAA-3' (forward) and rs739442 5'-TACGAAATCTTAGTGCC-3' (forward). The primer for SNaPshot analysis of the TSC2 gene allele was rs1748 (forward): 5'-GCATCATAGCCGCTCCAACCCCACCGA-3'. The PCR program consisted of 25 cycles of a 96°C denaturation step for 10 seconds, 50°C for 5 seconds and 60°C for 30 seconds. Subsequently, samples were treated for 45 minutes at 37°C with 5 units of antarctic phosphatase (New England Biolabs, Ipswich, Massachusetts). The phosphatase was then inactivated by incubating at 65°C for 10 minutes. Samples were run on the capillary sequencer and the genotype determined from the electropherogram generated during the run.\nAll samples were genotyped using the SNaPshot assay. This method of genotyping relies on the presence of heterozygous marker SNPs to distinguish between two alleles of a gene. In homozygous samples, where the gene alleles have the same nucleotide at the SNP locus, electropherograms will show a single peak using the forward primer and a single peak using the reverse primer. In heterozygous samples, the presence of different nucleotides at the SNP locus on each allele will result in the production of two peaks in both forward and reverse reactions.\nPCR primer pairs were designed for amplifying genomic DNA segments that included each SNP of interest. The amplimer segments were used in a SNaPshot assay (ABI Prism SNaPshot Multiplex Kit) to establish the genotype (homozygous versus heterozygous) of individuals at each of the marker SNPs. The primers for amplifying the 3'UTR genomic DNA segment containing the TSC1 SNPs rs2809243 and rs739442 (amplimer size = 587 bp) were: 5'-TAGTAATGGCAGAGCAGTCTAAACA-3' (forward) and 5'-TCCAGGTCTCATTCTCCCAACCGTA-3' (reverse). The primers for amplifying a genomic DNA segment surrounding TSC2 exon 40 were: 5'-CTGGGCAACGACTTTGTGTCCATTGTCTAC-3' (forward) and 5'-CTGACAGGCAATACCGTCCAA-3' (reverse). This primer pair produces an 1857 bp amplimer when used with genomic DNA. The PCR program consisted of an initial denaturation at 94°C for 40 seconds. This was followed by 40 cycles of 94°C for 20 seconds, 55°C for 1 minute, 72°C for 1 1/2 minutes, and a final extension step at 72°C for 5 minutes. PCR products were gel purified from 1-1.5% low melt agarose gels (IBI Scientific, Peosta, Iowa) using the Wizard PCR Preps DNA Purification System (Promega, Madison, Wisconsin).\nGenotyping was done with primers designed for SNaPshot analysis. The SNaPshot assay was performed according to the manufacturer's protocol (Applied Biosciences, Inc.). Briefly, a PCR reaction was run in which a single fluorescently-tagged dideoxynucleotide was added at the 3'-end of an annealed primer that was designed to terminate exactly one nucleotide before the SNP of interest (different fluorophores for ddA, ddG, ddC, and ddT). This allowed the identity of the SNP nucleotide to be determined using a capillary sequencer (Applied Biosystems Inc. Prism 310 Genetic Analyzer). This assay was used both for genotyping individuals at various SNPs and for AEI determination (as described in the following section). While both forward and reverse primers can be used for this analysis, the forward primers for each of the SNPs analyzed were found to give cleaner, more reliable results and were thus used in this assay.\nThe primers for SNaPshot analysis of the TSC1 gene alleles were: rs2809243 5'-AAACTCAACAAGTGCTCCTGAAAGAAA-3' (forward) and rs739442 5'-TACGAAATCTTAGTGCC-3' (forward). The primer for SNaPshot analysis of the TSC2 gene allele was rs1748 (forward): 5'-GCATCATAGCCGCTCCAACCCCACCGA-3'. The PCR program consisted of 25 cycles of a 96°C denaturation step for 10 seconds, 50°C for 5 seconds and 60°C for 30 seconds. Subsequently, samples were treated for 45 minutes at 37°C with 5 units of antarctic phosphatase (New England Biolabs, Ipswich, Massachusetts). The phosphatase was then inactivated by incubating at 65°C for 10 minutes. Samples were run on the capillary sequencer and the genotype determined from the electropherogram generated during the run.\n[SUBTITLE] Allelic expression imbalance assay [SUBSECTION] Samples heterozygous at marker SNPs were tested for AEI using the SNaPshot assay. Genomic and cDNA fragments flanking each SNP (as described above) were amplified in triplicate and gel purified. The same primers previously described for use in the amplification of genomic DNA segments were used in this assay to amplify TSC1 and TSC2 cDNA gene segments. The primer pair amplifying the TSC2 cDNA segment produces a 553 bp fragment when used with cDNA rather than the 1857 bp fragment produced with genomic DNA as the template. This is due to inclusion of intronic sequence in the PCR product from genomic DNA. The PCR reactions amplify both alleles, preserving the existing allele ratios in genomic DNA and in cDNA. The concentrations of gel purified samples (purification performed as indicated above) were measured using a Nanodrop 2000c (Thermo Scientific, Waltham, Massachusetts). Equal concentrations of the amplified fragments were then used in SNaPshot assays. All genomic and cDNA samples were analyzed in triplicate using the ABI capillary sequencer.\nGenomic DNA has a theoretical allele ratio of 1.0, but due to differences in the detection efficiency of various fluorophores, this ratio often deviates from 1.0. Therefore, genomic DNA was used as an internal control to correct for the differences in detection. In order to calculate the necessary correction factor, the genomic DNA allelic ratios for a specific SNP from each capillary sequencer run were averaged and the correction factor was calculated as the inverse of this average genomic allelic ratio. A diagram of the method used for calculating and applying the correction factor is shown in Figure 1. The allele ratios for genomic and cDNA samples were calculated as the ratio of heterozygous peak heights (analysis done using Gene Mapper 3.0 software from Applied Biosystems, Inc.). The experimental values for both the genomic DNA and the cDNA were then multiplied by the correction factor and average values were calculated for each sample analyzed in triplicate (see Results section for additional details). Standard error of the mean (SEM) was calculated for each sample using Excel software (Microsoft, Inc.) and error bars indicating 2X SEM were used in graphing the results.\nMethod for correcting genomic and cDNA allele ratios (AR). Genomic DNA segments containing marker SNPs are amplified by PCR and used as templates in SNaPshot primer-extension assays. Extended primers containing one of two different fluorescently labeled nucleotides at their 3'-ends are resolved by capillary electrophoresis and the ratio of peak heights calculated. An average genomic AR for a specific SNP is determined from all the genomic samples (each analyzed in triplicate). A correction factor (CF) is then calculated as the inverse of the average genomic AR. The genomic samples analyzed in triplicate are then individually multiplied by the CF to normalize the data to approximately 1.0, which is the theoretical ratio of two gene alleles in genomic DNA. The corrected average genomic AR is then determined. For each RNA sample, cDNA is synthesized and heterozygous SNP containing segments are amplified by PCR in triplicate and subjected to a SNaPshot PCR reaction. Samples are run on a capillary sequencer and the ratio of heterozygous peak heights is determined. Individual cDNA ARs are calculated and corrected by multiplying by the CF. The average corrected cDNA AR for each sample is then calculated. A sample is designated as showing AEI if the average corrected genomic AR and average corrected cDNA AR differ by greater than 2X the standard error of the mean and by at least 10%.\nSamples heterozygous at marker SNPs were tested for AEI using the SNaPshot assay. Genomic and cDNA fragments flanking each SNP (as described above) were amplified in triplicate and gel purified. The same primers previously described for use in the amplification of genomic DNA segments were used in this assay to amplify TSC1 and TSC2 cDNA gene segments. The primer pair amplifying the TSC2 cDNA segment produces a 553 bp fragment when used with cDNA rather than the 1857 bp fragment produced with genomic DNA as the template. This is due to inclusion of intronic sequence in the PCR product from genomic DNA. The PCR reactions amplify both alleles, preserving the existing allele ratios in genomic DNA and in cDNA. The concentrations of gel purified samples (purification performed as indicated above) were measured using a Nanodrop 2000c (Thermo Scientific, Waltham, Massachusetts). Equal concentrations of the amplified fragments were then used in SNaPshot assays. All genomic and cDNA samples were analyzed in triplicate using the ABI capillary sequencer.\nGenomic DNA has a theoretical allele ratio of 1.0, but due to differences in the detection efficiency of various fluorophores, this ratio often deviates from 1.0. Therefore, genomic DNA was used as an internal control to correct for the differences in detection. In order to calculate the necessary correction factor, the genomic DNA allelic ratios for a specific SNP from each capillary sequencer run were averaged and the correction factor was calculated as the inverse of this average genomic allelic ratio. A diagram of the method used for calculating and applying the correction factor is shown in Figure 1. The allele ratios for genomic and cDNA samples were calculated as the ratio of heterozygous peak heights (analysis done using Gene Mapper 3.0 software from Applied Biosystems, Inc.). The experimental values for both the genomic DNA and the cDNA were then multiplied by the correction factor and average values were calculated for each sample analyzed in triplicate (see Results section for additional details). Standard error of the mean (SEM) was calculated for each sample using Excel software (Microsoft, Inc.) and error bars indicating 2X SEM were used in graphing the results.\nMethod for correcting genomic and cDNA allele ratios (AR). Genomic DNA segments containing marker SNPs are amplified by PCR and used as templates in SNaPshot primer-extension assays. Extended primers containing one of two different fluorescently labeled nucleotides at their 3'-ends are resolved by capillary electrophoresis and the ratio of peak heights calculated. An average genomic AR for a specific SNP is determined from all the genomic samples (each analyzed in triplicate). A correction factor (CF) is then calculated as the inverse of the average genomic AR. The genomic samples analyzed in triplicate are then individually multiplied by the CF to normalize the data to approximately 1.0, which is the theoretical ratio of two gene alleles in genomic DNA. The corrected average genomic AR is then determined. For each RNA sample, cDNA is synthesized and heterozygous SNP containing segments are amplified by PCR in triplicate and subjected to a SNaPshot PCR reaction. Samples are run on a capillary sequencer and the ratio of heterozygous peak heights is determined. Individual cDNA ARs are calculated and corrected by multiplying by the CF. The average corrected cDNA AR for each sample is then calculated. A sample is designated as showing AEI if the average corrected genomic AR and average corrected cDNA AR differ by greater than 2X the standard error of the mean and by at least 10%.", "This research was approved by the St. Joseph's Hospital and Medical Center Institutional Review Board (IRB) for Human Research. Informed consent was obtained from all study participants. Participants were healthy volunteers who denied any personal or familial history of Tuberous Sclerosis Complex.", "DNA was extracted from blood leukocytes using Gentra Puregene Blood Kits (Qiagen, Valencia, California) and stored at 4°C. RNA was extracted from blood leukocytes using PAXgene Blood RNA Kits (Qiagen, Valencia, California) and stored at -80°C.", "TSC1 (NM_000368) and TSC2 (NM_000548) mRNAs were reverse-transcribed to cDNA using gene-specific primers and the SuperScript III First-Strand synthesis system for RT-PCR, according to the manufacturer's protocol (Invitrogen, Carlsbad, California). The cDNA synthesis primer sequence for TSC1 mRNA was 5'-GGGCCTGTGCTGACTCTGGTTAGTG-3'. The sequence of the cDNA synthesis primer for TSC2 mRNA was 5'-TTTCACTGACAGGCAATACC-3'. cDNAs were stored at -20°C.", "To distinguish TSC gene alleles we chose marker SNPs with relatively high rates of heterozygosity, as indicated in the NCBI Human Genome Resource SNP database http://www.ncbi.nlm.nih.gov, the SNPper resource (CHIP Bioinformatics resource - http://snpper.chip.org) and by our own genotyping data. Due to the need to distinguish the alleles, only samples heterozygous at marker SNPs were analyzed. Two SNPs were chosen as markers for TSC1 alleles: rs739442 (C/T) and rs2809243 (C/T), both located in the 3'-untranslated region (UTR) of TSC1 mRNA. The DNA samples were genotyped at several exonic SNPs in the TSC2 gene. One synonymous SNP located within exon 40, rs1748 (C/T), proved to have the highest rate of heterozygosity among the tested SNPs (~24%) and was therefore used for AEI analysis. Together, these three marker SNPs tag all known TSC1 and TSC2 mRNA splice variants.", "All samples were genotyped using the SNaPshot assay. This method of genotyping relies on the presence of heterozygous marker SNPs to distinguish between two alleles of a gene. In homozygous samples, where the gene alleles have the same nucleotide at the SNP locus, electropherograms will show a single peak using the forward primer and a single peak using the reverse primer. In heterozygous samples, the presence of different nucleotides at the SNP locus on each allele will result in the production of two peaks in both forward and reverse reactions.\nPCR primer pairs were designed for amplifying genomic DNA segments that included each SNP of interest. The amplimer segments were used in a SNaPshot assay (ABI Prism SNaPshot Multiplex Kit) to establish the genotype (homozygous versus heterozygous) of individuals at each of the marker SNPs. The primers for amplifying the 3'UTR genomic DNA segment containing the TSC1 SNPs rs2809243 and rs739442 (amplimer size = 587 bp) were: 5'-TAGTAATGGCAGAGCAGTCTAAACA-3' (forward) and 5'-TCCAGGTCTCATTCTCCCAACCGTA-3' (reverse). The primers for amplifying a genomic DNA segment surrounding TSC2 exon 40 were: 5'-CTGGGCAACGACTTTGTGTCCATTGTCTAC-3' (forward) and 5'-CTGACAGGCAATACCGTCCAA-3' (reverse). This primer pair produces an 1857 bp amplimer when used with genomic DNA. The PCR program consisted of an initial denaturation at 94°C for 40 seconds. This was followed by 40 cycles of 94°C for 20 seconds, 55°C for 1 minute, 72°C for 1 1/2 minutes, and a final extension step at 72°C for 5 minutes. PCR products were gel purified from 1-1.5% low melt agarose gels (IBI Scientific, Peosta, Iowa) using the Wizard PCR Preps DNA Purification System (Promega, Madison, Wisconsin).\nGenotyping was done with primers designed for SNaPshot analysis. The SNaPshot assay was performed according to the manufacturer's protocol (Applied Biosciences, Inc.). Briefly, a PCR reaction was run in which a single fluorescently-tagged dideoxynucleotide was added at the 3'-end of an annealed primer that was designed to terminate exactly one nucleotide before the SNP of interest (different fluorophores for ddA, ddG, ddC, and ddT). This allowed the identity of the SNP nucleotide to be determined using a capillary sequencer (Applied Biosystems Inc. Prism 310 Genetic Analyzer). This assay was used both for genotyping individuals at various SNPs and for AEI determination (as described in the following section). While both forward and reverse primers can be used for this analysis, the forward primers for each of the SNPs analyzed were found to give cleaner, more reliable results and were thus used in this assay.\nThe primers for SNaPshot analysis of the TSC1 gene alleles were: rs2809243 5'-AAACTCAACAAGTGCTCCTGAAAGAAA-3' (forward) and rs739442 5'-TACGAAATCTTAGTGCC-3' (forward). The primer for SNaPshot analysis of the TSC2 gene allele was rs1748 (forward): 5'-GCATCATAGCCGCTCCAACCCCACCGA-3'. The PCR program consisted of 25 cycles of a 96°C denaturation step for 10 seconds, 50°C for 5 seconds and 60°C for 30 seconds. Subsequently, samples were treated for 45 minutes at 37°C with 5 units of antarctic phosphatase (New England Biolabs, Ipswich, Massachusetts). The phosphatase was then inactivated by incubating at 65°C for 10 minutes. Samples were run on the capillary sequencer and the genotype determined from the electropherogram generated during the run.", "Samples heterozygous at marker SNPs were tested for AEI using the SNaPshot assay. Genomic and cDNA fragments flanking each SNP (as described above) were amplified in triplicate and gel purified. The same primers previously described for use in the amplification of genomic DNA segments were used in this assay to amplify TSC1 and TSC2 cDNA gene segments. The primer pair amplifying the TSC2 cDNA segment produces a 553 bp fragment when used with cDNA rather than the 1857 bp fragment produced with genomic DNA as the template. This is due to inclusion of intronic sequence in the PCR product from genomic DNA. The PCR reactions amplify both alleles, preserving the existing allele ratios in genomic DNA and in cDNA. The concentrations of gel purified samples (purification performed as indicated above) were measured using a Nanodrop 2000c (Thermo Scientific, Waltham, Massachusetts). Equal concentrations of the amplified fragments were then used in SNaPshot assays. All genomic and cDNA samples were analyzed in triplicate using the ABI capillary sequencer.\nGenomic DNA has a theoretical allele ratio of 1.0, but due to differences in the detection efficiency of various fluorophores, this ratio often deviates from 1.0. Therefore, genomic DNA was used as an internal control to correct for the differences in detection. In order to calculate the necessary correction factor, the genomic DNA allelic ratios for a specific SNP from each capillary sequencer run were averaged and the correction factor was calculated as the inverse of this average genomic allelic ratio. A diagram of the method used for calculating and applying the correction factor is shown in Figure 1. The allele ratios for genomic and cDNA samples were calculated as the ratio of heterozygous peak heights (analysis done using Gene Mapper 3.0 software from Applied Biosystems, Inc.). The experimental values for both the genomic DNA and the cDNA were then multiplied by the correction factor and average values were calculated for each sample analyzed in triplicate (see Results section for additional details). Standard error of the mean (SEM) was calculated for each sample using Excel software (Microsoft, Inc.) and error bars indicating 2X SEM were used in graphing the results.\nMethod for correcting genomic and cDNA allele ratios (AR). Genomic DNA segments containing marker SNPs are amplified by PCR and used as templates in SNaPshot primer-extension assays. Extended primers containing one of two different fluorescently labeled nucleotides at their 3'-ends are resolved by capillary electrophoresis and the ratio of peak heights calculated. An average genomic AR for a specific SNP is determined from all the genomic samples (each analyzed in triplicate). A correction factor (CF) is then calculated as the inverse of the average genomic AR. The genomic samples analyzed in triplicate are then individually multiplied by the CF to normalize the data to approximately 1.0, which is the theoretical ratio of two gene alleles in genomic DNA. The corrected average genomic AR is then determined. For each RNA sample, cDNA is synthesized and heterozygous SNP containing segments are amplified by PCR in triplicate and subjected to a SNaPshot PCR reaction. Samples are run on a capillary sequencer and the ratio of heterozygous peak heights is determined. Individual cDNA ARs are calculated and corrected by multiplying by the CF. The average corrected cDNA AR for each sample is then calculated. A sample is designated as showing AEI if the average corrected genomic AR and average corrected cDNA AR differ by greater than 2X the standard error of the mean and by at least 10%.", "AEI was examined in the TSC1 and TSC2 genes by quantifying the relative amounts of mRNA derived from each of the two alleles of each gene in leukocyte RNA samples isolated from normal individuals heterozygous for mRNA marker SNPs. To avoid the confounding effects of alternative splicing, SNPs located within the 3'-UTR of the TSC1 gene and in exon 40 of the TSC2 gene were selected as markers. These regions are included in all known mRNA forms generated from the TSC1 and TSC2 genes. The rs numbers (NCBI SNP data base, http://www.ncbi.nlm.nih.gov/snp) and locations of the three TSC1 and TSC2 marker SNPs used in this study are shown in Figure 2.\nThis diagram shows the exon/intron structure of the TSC1 and TSC2 genes. Exons are represented by numbered boxes. Exons subject to alternative splicing are indicated by brackets. The locations of the SNPs used for analysis of AEI are indicated by stars.\nAs described in detail in Methods, our AEI assays involve PCR amplification of short segments of TSC1 or TSC2 cDNA containing a marker SNP, followed by annealing of a synthetic oligonucleotide primer to a site immediately upstream from the SNP and primer extension in the presence of fluorescently tagged dideoxynucleotide triphosphates (ddNTPs). Because each ddNTP is tagged with a different fluorophore, the identity of the added base can be determined by resolving the fluorescently labeled primers by capillary electrophoresis and identifying the \"color\" of each extended primer [40].\nDifferences in expression between two alleles can be quantified by calculating the ratio of the peak heights of the traces corresponding to each fluorescently labeled primer. To correct for artifactual imbalances related to technical aspects of the assay, AEI assays were also carried out using genomic DNA, which in the absence of local chromosome deletions or duplications, would be expected to contain equal numbers of each allele. A correction factor derived from these measurements was used to correct AEI measurements obtained using cDNA templates.\n[SUBTITLE] SNP frequencies in the sample population [SUBSECTION] Our assay uses SNPs located within protein coding exons or the 3'-UTR to distinguish between the mRNA species that are transcribed from the two alleles of a gene in each individual. Because only subjects who are heterozygous at marker SNPs are informative in our assays, we first genotyped our subjects to identify individuals who are heterozygous for one or more of the TSC1 and TSC2 marker SNPs described above. The heterozygosities of the TSC1 markers rs2809243 and rs739442 were approximately 49% (41/83) and 45% (37/83), respectively, in our sample. Heterozygosity of the TSC2 marker SNP rs1748 was approximately 22% (18/82). These data are similar to average population heterozygosities for subjects of all races reported for these SNPs on the SNPper website (CHIP Bioinformatics resource - http://snpper.chip.org) and the NCBI SNP database. Approximately 36% (30/83) of subjects were heterozygous at both TSC1 SNPs.\nOur assay uses SNPs located within protein coding exons or the 3'-UTR to distinguish between the mRNA species that are transcribed from the two alleles of a gene in each individual. Because only subjects who are heterozygous at marker SNPs are informative in our assays, we first genotyped our subjects to identify individuals who are heterozygous for one or more of the TSC1 and TSC2 marker SNPs described above. The heterozygosities of the TSC1 markers rs2809243 and rs739442 were approximately 49% (41/83) and 45% (37/83), respectively, in our sample. Heterozygosity of the TSC2 marker SNP rs1748 was approximately 22% (18/82). These data are similar to average population heterozygosities for subjects of all races reported for these SNPs on the SNPper website (CHIP Bioinformatics resource - http://snpper.chip.org) and the NCBI SNP database. Approximately 36% (30/83) of subjects were heterozygous at both TSC1 SNPs.\n[SUBTITLE] AEI in the TSC1 gene [SUBSECTION] AEI analysis of TSC1 mRNA expression was performed independently using the marker SNPs rs2809243 and rs739442. As indicated above, 41 individuals were heterozygous at rs2809243 and 37 were heterozygous at rs739442. 30 individuals were heterozygous at both of the marker SNPs. Data from these doubly heterozygous individuals was used for independent validation of the results from each SNP. As outlined in Figure 1, AEI measurements using genomic DNA as template were carried out to permit the calculation of a correction factor for AEI measurements using cDNA as template. AEIs were considered to be significant if the corrected cDNA allelic expression ratio differed from the corrected genomic allele ratio by greater than 10%, and if the error bars (defined here as 2x the standard error of the mean) for the average genomic and cDNA allele ratios did not overlap.\nFigure 3 displays the corrected genomic and cDNA AEI ratios for each individual in our sample. Shown to the left in each graph is the data for individuals heterozygous at both marker SNPs. Shown to the right in each graph is additional data for individuals heterozygous at a single marker SNP. 8/41 individuals show AEI using rs2809243 while 7/37 individuals show AEI using rs739442. Of the doubly heterozygous individuals, rs2809243 revealed 6 individuals with AEI reaching our defined level of significance while rs739442 showed 5 individuals demonstrating AEI. The 5 individuals with AEI by rs739442 were the same as those with AEI by rs2809243. A 6th individual's sample (#11) reached AEI significance by a small margin using rs2809243 but did not reach significance using rs739442 as the marker SNP, thus emphasizing the importance of using a second SNP to validate data. We were able to consistently score 5 out of 6 individuals as demonstrating significant AEI in blood leukocytes using two different SNPs. The AEI of TSC1 mRNA expression in this control group ranged from 10% to greater than 30%. While this degree of imbalance is relatively small, it could be sufficient to modulate the phenotype in a TSC patient heterozygous for a mutation in the TSC1 gene. Based on this small cohort of control subjects we estimate that the population frequency of AEI at the TSC1 locus may be as high as 15-20%.\nAEI analysis of TSC1 mRNA expression in leukocytes isolated from 30 individuals doubly heterozygous for the marker SNPs rs2809243 and rs739442 and additional individuals (11 and 7 individuals respectively) singly heterozygous for one of the two SNPs. For doubly heterozygous individuals (data at the left side of graphs 3A and 3B), blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate the corrected cDNA ARs. Data for singly heterozygous individuals is located to the right side of the 3A and 3B graphs using red and green bars to indicate genomic ARs and shaded bars to indicate cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the genomic AR by more than 10% and had error bars that did not overlap with those of the genomic DNA.\nIt should be noted that the cDNA allelic expression ratios measured using the marker SNP rs280943 range from greater that 1 (samples 29 and 36) to less than 1 (samples 3 and 27). This implies that the regulatory variant or variants resulting in this AEI are not tightly linked to the marker SNP. Thus, if these individuals were heterozygous for a remote regulatory variant comprising one high-expression allele and one low-expression allele, the results of our AEI measurements imply that the high-expression allele is \"in phase\" (ie., located on the same chromosome) with the rs280943 C-allele in individuals 29 and 36, but is \"in-phase\" with the rs280943 T-allele in individuals 3 and 27.\nSimilar arguments hold for the AEI measurement obtained using rs739442 as the marker SNP. The fact that the directions of the measured AEI differ for individual #27, depending upon the choice of marker SNP, implies that the \"phase\" of the marker SNPs with respect to the functional variant is different in this individual. That is, in this individual the rs739442 C-allele is located on the same chromosome as the high-expression allele of the remote regulatory variant. Although the two SNPs used for these analyses (rs280943 and rs739442) are separated by only 166 bp, our data indicate that these SNPs are not tightly linked. As previously indicated, only around 29% of our sample population is heterozygous at both TSC1 marker SNPs despite their close proximity. This is apparent in the data of individual #27 which shows the marker SNPs to be on different chromosomes. Using the Hapmap database http://hapmap.ncbi.nlm.nih.gov the linkage disequilibrium D' value for these SNPs is 0.671, confirming that these SNPs are not tightly linked despite the small separation distance. As these two SNPs are both located in the 3'UTR of TSC1, it is not surprising to see this level of variation as mutations in this area are less likely to affect the protein function.\nAEI analysis of TSC1 mRNA expression was performed independently using the marker SNPs rs2809243 and rs739442. As indicated above, 41 individuals were heterozygous at rs2809243 and 37 were heterozygous at rs739442. 30 individuals were heterozygous at both of the marker SNPs. Data from these doubly heterozygous individuals was used for independent validation of the results from each SNP. As outlined in Figure 1, AEI measurements using genomic DNA as template were carried out to permit the calculation of a correction factor for AEI measurements using cDNA as template. AEIs were considered to be significant if the corrected cDNA allelic expression ratio differed from the corrected genomic allele ratio by greater than 10%, and if the error bars (defined here as 2x the standard error of the mean) for the average genomic and cDNA allele ratios did not overlap.\nFigure 3 displays the corrected genomic and cDNA AEI ratios for each individual in our sample. Shown to the left in each graph is the data for individuals heterozygous at both marker SNPs. Shown to the right in each graph is additional data for individuals heterozygous at a single marker SNP. 8/41 individuals show AEI using rs2809243 while 7/37 individuals show AEI using rs739442. Of the doubly heterozygous individuals, rs2809243 revealed 6 individuals with AEI reaching our defined level of significance while rs739442 showed 5 individuals demonstrating AEI. The 5 individuals with AEI by rs739442 were the same as those with AEI by rs2809243. A 6th individual's sample (#11) reached AEI significance by a small margin using rs2809243 but did not reach significance using rs739442 as the marker SNP, thus emphasizing the importance of using a second SNP to validate data. We were able to consistently score 5 out of 6 individuals as demonstrating significant AEI in blood leukocytes using two different SNPs. The AEI of TSC1 mRNA expression in this control group ranged from 10% to greater than 30%. While this degree of imbalance is relatively small, it could be sufficient to modulate the phenotype in a TSC patient heterozygous for a mutation in the TSC1 gene. Based on this small cohort of control subjects we estimate that the population frequency of AEI at the TSC1 locus may be as high as 15-20%.\nAEI analysis of TSC1 mRNA expression in leukocytes isolated from 30 individuals doubly heterozygous for the marker SNPs rs2809243 and rs739442 and additional individuals (11 and 7 individuals respectively) singly heterozygous for one of the two SNPs. For doubly heterozygous individuals (data at the left side of graphs 3A and 3B), blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate the corrected cDNA ARs. Data for singly heterozygous individuals is located to the right side of the 3A and 3B graphs using red and green bars to indicate genomic ARs and shaded bars to indicate cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the genomic AR by more than 10% and had error bars that did not overlap with those of the genomic DNA.\nIt should be noted that the cDNA allelic expression ratios measured using the marker SNP rs280943 range from greater that 1 (samples 29 and 36) to less than 1 (samples 3 and 27). This implies that the regulatory variant or variants resulting in this AEI are not tightly linked to the marker SNP. Thus, if these individuals were heterozygous for a remote regulatory variant comprising one high-expression allele and one low-expression allele, the results of our AEI measurements imply that the high-expression allele is \"in phase\" (ie., located on the same chromosome) with the rs280943 C-allele in individuals 29 and 36, but is \"in-phase\" with the rs280943 T-allele in individuals 3 and 27.\nSimilar arguments hold for the AEI measurement obtained using rs739442 as the marker SNP. The fact that the directions of the measured AEI differ for individual #27, depending upon the choice of marker SNP, implies that the \"phase\" of the marker SNPs with respect to the functional variant is different in this individual. That is, in this individual the rs739442 C-allele is located on the same chromosome as the high-expression allele of the remote regulatory variant. Although the two SNPs used for these analyses (rs280943 and rs739442) are separated by only 166 bp, our data indicate that these SNPs are not tightly linked. As previously indicated, only around 29% of our sample population is heterozygous at both TSC1 marker SNPs despite their close proximity. This is apparent in the data of individual #27 which shows the marker SNPs to be on different chromosomes. Using the Hapmap database http://hapmap.ncbi.nlm.nih.gov the linkage disequilibrium D' value for these SNPs is 0.671, confirming that these SNPs are not tightly linked despite the small separation distance. As these two SNPs are both located in the 3'UTR of TSC1, it is not surprising to see this level of variation as mutations in this area are less likely to affect the protein function.\n[SUBTITLE] AEI in the TSC2 gene [SUBSECTION] Twenty out of 83 individuals in our sample were heterozygous for the TSC2 mRNA marker SNP rs1748, As shown in Figure 4, 10% (2/20) of these individuals demonstrated AEI above the 10% cut-off, with a difference of more than 2x the SEM. An independent marker SNP was not available for verification; however, AEI measurements were highly reproducible. Our data demonstrates that AEI is relatively common in both the TSC1 and TSC2 genes.\nAEI analysis of TSC2 mRNA expression in leukocytes isolated from 20 individuals heterozygous for the marker SNP rs1748. Blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate corrected cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the average corrected genomic AR by more than 10% and had error bars (2X SEM) that did not overlap with those of the genomic DNA.\nTwenty out of 83 individuals in our sample were heterozygous for the TSC2 mRNA marker SNP rs1748, As shown in Figure 4, 10% (2/20) of these individuals demonstrated AEI above the 10% cut-off, with a difference of more than 2x the SEM. An independent marker SNP was not available for verification; however, AEI measurements were highly reproducible. Our data demonstrates that AEI is relatively common in both the TSC1 and TSC2 genes.\nAEI analysis of TSC2 mRNA expression in leukocytes isolated from 20 individuals heterozygous for the marker SNP rs1748. Blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate corrected cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the average corrected genomic AR by more than 10% and had error bars (2X SEM) that did not overlap with those of the genomic DNA.", "Our assay uses SNPs located within protein coding exons or the 3'-UTR to distinguish between the mRNA species that are transcribed from the two alleles of a gene in each individual. Because only subjects who are heterozygous at marker SNPs are informative in our assays, we first genotyped our subjects to identify individuals who are heterozygous for one or more of the TSC1 and TSC2 marker SNPs described above. The heterozygosities of the TSC1 markers rs2809243 and rs739442 were approximately 49% (41/83) and 45% (37/83), respectively, in our sample. Heterozygosity of the TSC2 marker SNP rs1748 was approximately 22% (18/82). These data are similar to average population heterozygosities for subjects of all races reported for these SNPs on the SNPper website (CHIP Bioinformatics resource - http://snpper.chip.org) and the NCBI SNP database. Approximately 36% (30/83) of subjects were heterozygous at both TSC1 SNPs.", "AEI analysis of TSC1 mRNA expression was performed independently using the marker SNPs rs2809243 and rs739442. As indicated above, 41 individuals were heterozygous at rs2809243 and 37 were heterozygous at rs739442. 30 individuals were heterozygous at both of the marker SNPs. Data from these doubly heterozygous individuals was used for independent validation of the results from each SNP. As outlined in Figure 1, AEI measurements using genomic DNA as template were carried out to permit the calculation of a correction factor for AEI measurements using cDNA as template. AEIs were considered to be significant if the corrected cDNA allelic expression ratio differed from the corrected genomic allele ratio by greater than 10%, and if the error bars (defined here as 2x the standard error of the mean) for the average genomic and cDNA allele ratios did not overlap.\nFigure 3 displays the corrected genomic and cDNA AEI ratios for each individual in our sample. Shown to the left in each graph is the data for individuals heterozygous at both marker SNPs. Shown to the right in each graph is additional data for individuals heterozygous at a single marker SNP. 8/41 individuals show AEI using rs2809243 while 7/37 individuals show AEI using rs739442. Of the doubly heterozygous individuals, rs2809243 revealed 6 individuals with AEI reaching our defined level of significance while rs739442 showed 5 individuals demonstrating AEI. The 5 individuals with AEI by rs739442 were the same as those with AEI by rs2809243. A 6th individual's sample (#11) reached AEI significance by a small margin using rs2809243 but did not reach significance using rs739442 as the marker SNP, thus emphasizing the importance of using a second SNP to validate data. We were able to consistently score 5 out of 6 individuals as demonstrating significant AEI in blood leukocytes using two different SNPs. The AEI of TSC1 mRNA expression in this control group ranged from 10% to greater than 30%. While this degree of imbalance is relatively small, it could be sufficient to modulate the phenotype in a TSC patient heterozygous for a mutation in the TSC1 gene. Based on this small cohort of control subjects we estimate that the population frequency of AEI at the TSC1 locus may be as high as 15-20%.\nAEI analysis of TSC1 mRNA expression in leukocytes isolated from 30 individuals doubly heterozygous for the marker SNPs rs2809243 and rs739442 and additional individuals (11 and 7 individuals respectively) singly heterozygous for one of the two SNPs. For doubly heterozygous individuals (data at the left side of graphs 3A and 3B), blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate the corrected cDNA ARs. Data for singly heterozygous individuals is located to the right side of the 3A and 3B graphs using red and green bars to indicate genomic ARs and shaded bars to indicate cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the genomic AR by more than 10% and had error bars that did not overlap with those of the genomic DNA.\nIt should be noted that the cDNA allelic expression ratios measured using the marker SNP rs280943 range from greater that 1 (samples 29 and 36) to less than 1 (samples 3 and 27). This implies that the regulatory variant or variants resulting in this AEI are not tightly linked to the marker SNP. Thus, if these individuals were heterozygous for a remote regulatory variant comprising one high-expression allele and one low-expression allele, the results of our AEI measurements imply that the high-expression allele is \"in phase\" (ie., located on the same chromosome) with the rs280943 C-allele in individuals 29 and 36, but is \"in-phase\" with the rs280943 T-allele in individuals 3 and 27.\nSimilar arguments hold for the AEI measurement obtained using rs739442 as the marker SNP. The fact that the directions of the measured AEI differ for individual #27, depending upon the choice of marker SNP, implies that the \"phase\" of the marker SNPs with respect to the functional variant is different in this individual. That is, in this individual the rs739442 C-allele is located on the same chromosome as the high-expression allele of the remote regulatory variant. Although the two SNPs used for these analyses (rs280943 and rs739442) are separated by only 166 bp, our data indicate that these SNPs are not tightly linked. As previously indicated, only around 29% of our sample population is heterozygous at both TSC1 marker SNPs despite their close proximity. This is apparent in the data of individual #27 which shows the marker SNPs to be on different chromosomes. Using the Hapmap database http://hapmap.ncbi.nlm.nih.gov the linkage disequilibrium D' value for these SNPs is 0.671, confirming that these SNPs are not tightly linked despite the small separation distance. As these two SNPs are both located in the 3'UTR of TSC1, it is not surprising to see this level of variation as mutations in this area are less likely to affect the protein function.", "Twenty out of 83 individuals in our sample were heterozygous for the TSC2 mRNA marker SNP rs1748, As shown in Figure 4, 10% (2/20) of these individuals demonstrated AEI above the 10% cut-off, with a difference of more than 2x the SEM. An independent marker SNP was not available for verification; however, AEI measurements were highly reproducible. Our data demonstrates that AEI is relatively common in both the TSC1 and TSC2 genes.\nAEI analysis of TSC2 mRNA expression in leukocytes isolated from 20 individuals heterozygous for the marker SNP rs1748. Blue bars indicate corrected genomic allelic ratios (AR) and grey bars indicate corrected cDNA ARs. Error bars indicate 2X the standard error of the mean (SEM). Stars indicate samples for which the average corrected cDNA AR differed from the average corrected genomic AR by more than 10% and had error bars (2X SEM) that did not overlap with those of the genomic DNA.", "There is a growing consensus that cis-acting genetic variants significantly contribute to phenotypic differences among individuals, including disease risk [36,39,41-45]. Regulatory polymorphisms are one of the predominant mechanisms by which cis-acting gene regulation has been found to occur. These polymorphisms, located in regulatory regions, influence the expression of genes by affecting transcriptional activation or repression, generally by altering the DNA binding sites for transcription factors [36,39,46,47]. In addition, splicing errors, changes in mRNA stablility, epigenetic modifications and polymorphic (ACn) microsatellites have also been implicated in cis-acting gene regulation [36,37,39,46-49].\nThe best known examples of allele-specific differences in gene expression have been associated with X-inactivation [50] or genomic imprinting [51]. However, allelic variation in expression has also been demonstrated in non-imprinted genes and this allelic variation itself can be regulated by cis-acting elements [36-39]. Variations in allele expression have been previously linked to disease. For example, allelic variation in APC (adenomatous polyposis coli) expression plays a role in predisposition to colon cancer [41]. Allelic expression imbalance has also been studied in the cancer associated genes BRCA1/2 and CDH1 and used to identify polymorphisms, mutations and other defects that alter allelic expression and influence disease state [45,48]. As many genes are active within networks, variation in the expression of specific gene alleles may ultimately result in multiple downstream effects within a network and between related gene networks. This creates an avenue by which even small differences in the expression of specific genes can ultimately result in substantial phenotypic changes [46,47].\nIt has been noted frequently that disease causing mutations in families with TSC may produce very few problems in certain individuals while having catastrophic effects in others [2,26-31]. Clearly, there are additional factors outside of the mutation itself that affect disease severity. Differences in allele specific mRNA expression could potentially be one of these disease modifying factors. It is possible that the overall amount of normal TSC protein in cells may determine the severity of the disease phenotype in patients. As TSC is a disease carried in a heterozygous state [1-3], some amount of normal TSC protein should be present in most cells since one normal allele of each TSC gene is present (the exception being abnormal tissue growths exhibiting LOH [15-18]). It is possible that higher relative expression of protein from the normal allele may be protective, while higher relative expression of abnormal protein from the mutant allele may have deleterious effects.\nWe began to study this issue by determining the frequency of occurrence of AEI of the TSC1 and TSC2 genes in a control population. The intent was to establish if mRNA expression variation might be common enough to be a mechanism by which phenotypes are modified in patients with TSC gene mutations. In a cohort of normal volunteers we were able to quantify allele specific expression of the TSC genes in blood RNA and estimate the frequency of allelic skewing of expression for these two genes. In our studies we found that there was significant skewing of allelic expression of the TSC1 gene in about 19% of our sample population and of the TSC2 gene in 10% of our population. This estimate is based on a small sample of informative individuals (48 individuals who were heterozygous for a TSC1 marker SNP and 20 who were heterozygous for a TSC2 marker SNP). This was a sample of convenience, but individuals were recruited without bias and should be representative of the general population. If we assume a binomial distribution for the occurrence of AEI in the general population, we can use the exact test to calculate 95% confidence intervals for the actual population frequencies of AEI at the TSC1 and TSC2 genes. Based on such calculations, the 95% confidence interval for prevalence of AEI at TSC1 is 9% to 33% and for AEI at the TSC2 gene is 1.2% to 32%. These confidence intervals for the estimates for the actual population frequencies of AEI at TSC1 and TSC2 can of course be sharpened with larger sample sizes.\nWhile these are not large proportions, AEI may be occurring frequently enough to be a potential contributor to the phenotypic differences in TSC patients. In any given individual patient within a particular family, the phenotype could be determined not just by the mutation, but also by SNPs located within regulatory regions of the TSC genes. In such familial cases of TSC, the implication is that regulatory SNPs inherited from the parents in various combinations with the normal and mutant gene alleles can affect the phenotype of the child.\nThe TSC1 and TSC2 gene products, hamartin and tuberin, function together as a protein complex. Therefore, mutation of either of the TSC genes results in the same disease [1,3,52]. The hamartin-tuberin complex is a modulator of the mTOR signaling pathway, which is important in the regulation of cell growth. We know that haploinsufficiency due to mutation of a single TSC gene allele is sufficient to cause TSC and represents an approximately 50% loss of the total expression of that TSC gene [2]. Loss of a single TSC gene allele is sufficient to disrupt neuronal morphology and function in mouse models [21]. Loss of both alleles of a TSC gene can result in the formation of hamartomas common to TSC as is demonstrated by LOH studies [13,15-18]. These points clearly suggest that pathways modulated by hamartin and tuberin are sensitive to gene dosage effects. If a 50% reduction in expression of a TSC gene is sufficient to cause disease, it is plausible that smaller variations in expression, such as the 10-30% that we found in our experiments, might be sufficient to influence phenotype. In our control sample group, this level of variation in allelic expression of TSC1 or TSC2 does not result in a phenotype, as both alleles encode normal proteins. This degree of variation in mRNA expression combined with mutation of a TSC gene allele may be sufficient to influence phenotype either positively or negatively.\nIt has previously been reported that a 50% decrease in the expression of a single allele of the adenomatous polyposis coli tumor suppressor gene (APC), representing an overall 25% decrease of APC mRNA expression, is sufficient to cause the development of familial adenomatous polyposis[41]. An additional study of a gene associated with osteoarthritis (GDF5) discovered that a promoter polymorphism which created a small reduction of the expression of the T allele (less than 27%), significantly increased individuals susceptibility to developing osteoarthritis [53]. These reports indicate that even small variations in allelic expression are important to disease outcomes. This supports our hypothesis that variation in expression of the TSC gene alleles, particularly in the presence of an existing genetic mutation, may influence disease severity in TSC patients.\nTissue specific expression of genes is an important consideration when assessing the effects of variation in allelic expression. A sequence variant in the regulatory region of a gene might be relevant in some tissues and not in others, leading to conflicting results in different tissues [54]. Our study was performed in blood samples as this is a readily available tissue specimen. It is important to determine if allelic expression ratios measured in peripheral blood correlate with ratios measured in brain tissue, something that may be done using banked tissue samples. A difficulty we've encountered is the availability of good-quality matched blood and brain tissue samples from which intact RNA and DNA can be extracted. Establishing a correlation between blood and brain expression levels is especially important as we try to relate expression of the TSC alleles in blood to severity of cognitive impairment in patients with TSC. The goal of our research is to determine if the levels of expression of mutant and wild-type alleles in patients with TSC, as measured in peripheral blood, correlates with phenotypic severity. To this end, we plan to next study familial cases of TSC, where multiple affected individuals have the same identical gene mutation, but are discordant in terms of disease severity. We shall determine if, in these multiplex families, disease severity is correlated with skewing of allele specific expression. Ultimately, we hope to use the combination of mutation detection and measures of AEI in blood samples to predict disease severity (at least in relation to cognitive impairment). Early identification of patients who are at risk for developing severe disease may allow for aggressive preventive interventions, and may protect the patient from additional damaging effects of the disease.", "We have concluded from our research that variation in TSC1 and TSC2 gene allele expression is common in normal individuals, as it was easily detected in a relatively small sample population. It is likely that this variation in allele expression will also be seen in some patients carrying TSC gene mutations and may therefore help to explain the intra-familial variation in disease severity frequently observed in TSC. These ideas can be tested in multiplex families that include patients with TSC that are discordant in disease severity (particularly cognitive symptoms). After such validation, we might be able to develop a simple blood test (ratio of wild-type to mutant TSC mRNA levels) that predicts disease severity in simplex cases of TSC.", "The authors declare that they have no competing interests.", "VN conceived of the study and VN, GJ and SR participated in its design and coordination. SR performed the DNA/RNA isolation and the AEI analysis. SO performed DNA sequencing and genotyping and provided theoretical advice. GJ and VN drafted the manuscript and SR and SO edited the manuscript. DS provided key technical advice and critical review of the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2350/12/29/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adolescent", "Bifidobacterium", "Child", "Child, Preschool", "Constipation", "Defecation", "Fecal Incontinence", "Gastrointestinal Diseases", "Humans", "Male", "Pilot Projects", "Probiotics" ]

[ "Background", "Subjects", "Study design", "Outcome measures", "Analysis", "Results", "Discussion", "Competing interests", "Authors' contributions" ]

[ "Functional constipation is a common and frustrating problem in childhood with an estimated prevalence of 3% in the western world [1]. This chronic condition is characterised by infrequent defecation less than three times per week, more than two episodes of faecal incontinence per week, the passage of large and painful stools which clog the toilet and retentive posturing. Upon physical examination a palpable faecal mass is often found in the abdomen and the rectum [2,3]. It causes distress to child and family and results in severe emotional disturbance and family discord [4]. The pathophysiology underlying functional constipation is undoubtedly multi-factorial, and not well understood. Withholding behaviour is probably the major cause for the development of constipation and might be caused by the previous production of a large, hard painful stool, anal fissures, a primarily behavioural mechanism or the resistance to go to another toilet then their own [4].\nTo date, patients are treated with a combination of education, toilet training and oral laxatives. Disappointingly, only 50% of all children followed for 6 to 12 months are found to recover and were successfully taken off laxatives [5]. Another study showed that despite intensive medical and behavioural therapy, 25% of patients developing constipation before the age of 5 years continued to have severe complaints of constipation beyond puberty [6]. Furthermore, in 50% of the patients using these compounds, adverse side-effects were registered such as: abdominal pain, bloating, flatulence, diarrhoea, nausea and bad taste [7]. No data exist concerning possible long-term adverse effects such as electrolyte disturbances, mucosal damage and habituation.\nProbiotics are defined as live micro-organisms which when administered in adequate amounts, 106-109 colony forming units, confer a health benefit on the host [8]. The use of probiotics has entered mainstream medicine and has proven being an effective therapy in many different gastrointestinal disorders, including functional gastrointestinal disorders [9,10]. However, large trials investigating the efficacy and safety of probiotics in pediatric patients are lacking [11,12]. Studies in constipated adults with a Bifidus yoghurt, containing Bifidobacterium breve, Bifidobacterium bifidum and Lactobacillus acidophilus, showed a significant increase in defecation frequency without any side effects [13,14]. Furthermore, a randomized controlled trial using Bifidobacterium breve in preterm infants showed less gas accumulation in the stomach, less vomiting and improved weight gain without any side effects, suggesting a positive effect on gastrointestinal motility [15] The exact working mechanism of probiotics are not well understood. There are some hypotheses, however, why probiotics might have therapeutic potential for the treatment of constipation. Firstly, a dysbiosis in the gut flora in constipated patients has been suggested which might improve after the ingestion of probiotics [9,10]. It remains important however to understand if dysbiosis is a secondary manifestation of constipation, or if it is a factor contributing to constipation. Furthermore, probiotics can lower pH of the colon by producing lactic, acetic and other short chain fatty acids. A lower pH enhances colonic peristalsis and subsequently decreases colonic transit time [9,10].\nBased on the positive data in constipated adults, we therefore performed a pilot study to determine if Bifidobacterium breve is effective in the treatment of children with constipation.", "Children, 3 to 16 years of age, referred to the outpatient clinic of a tertiary pediatric gastroenterology department, with constipation were eligible for this study. Patients were included if they had been suffering from functional constipation according to the Rome III criteria for the last 2 months [2,3]. All children included had a defecation frequency of <3 times/week and one or more of the following criteria: faecal incontinence >1 episode/week, a large amount of stools that clog the toilet, painful defecation, withholding behaviour, or abdominal or rectal faecal impaction upon physical examination. In order to obtain a homogeneous group of patients, we included children with a defecation frequency <3 times/week in combination with at least one other ROME III criterion. Patients were not enrolled in this study if they had been treated for constipation less than 2 weeks before the start of the study. Other exclusion criteria were: a diagnosis of either IBS or functional non-retentive faecal incontinence according to the Rome III criteria; a diagnosis of mental retardation or metabolic disease (hypothyroidism), Hirschsprung's disease, spinal anomalies, anorectal pathology, previous gastrointestinal surgery. All children older than 12 years and/or parents gave informed consent. This pilot was approved by the medical ethical committee,", "Seven days prior to baseline assessment and during the treatment period, all children recorded frequency of bowel movements, stool consistency according to the Bristol stool scale, the number of faecal incontinence episodes, pain during defecation, abdominal pain, as well as adverse effects such as vomiting and diarrhoea in a standardized bowel diary. At baseline a medical history and information on the current defection pattern was collected and also a physical examination including a rectal digital exam was performed. Information and education about functional constipation was given to all patients and their care takers. Before start of the probiotic treatment, all children received once daily for 3 days a rectal enema in order to accomplish rectal disimpaction. After rectal disimpaction, all children received daily one sachet of powder containing 108-10 CFU Bifidobacterium breve Yakult, for 4 weeks. The patients were allowed to mix the powder with all liquids on condition that the liquid was not hot. During the study, all children were instructed to try to defecate on the toilet for 5-10 minutes after each meal (3 times a day). Patients were not allowed to consume any other fermented dairy product during this study or any other laxatives, except for the rescue medication Biscaodyl. Names of the forbidden products were pointed out in the diary. During the product consumption period, patients were instructed to take bisacodyl 5 mg if they did not defecate for 3 consecutive days. Clinical evaluation, frequency of adverse effects and compliance to the study protocol were carried out at enrolment and at 2, 4 and 6 weeks using standardized bowel diaries.", "The primary outcome measure was change in defecation frequency in week 4 compared to baseline. Secondary outcome measures were stool consistency, frequency of episodes of faecal incontinence, pain during defecation, frequency of abdominal pain, frequency of adverse effects (nausea, diarrhoea and bad taste), and frequency of intake of bisacodyl.", "Descriptive statistics were performed for baseline characteristics, adverse effects and Bisacodyl use. Change of frequency of bowel movements, fecal incontinence and change of stool consistency, was assessed using the non-parametric paired Wilcoxon test. For the comparison of abdominal pain between baseline and the evaluation time points, the Wilcoxon rank test was used. A p- value < 0.05 was considered to be significant. All analyses were done with SPSS (version 16.0).", "Between July 2009 and January 2010, 22 children were included into this pilot study. Two children were lost to follow-up without having any outcome data during follow-up. Both patients were therefore excluded from the final analysis. All outcomes of the remaining 20 children were collected and analyzed. The baseline characteristics are described in Table 1 (mean (range))\nBaseline patient characteristics: mean (range)\nThe defecation frequency per week significantly increased from 0.9 (0-2) at baseline to 5.3 (0-21) in week 2 (p < 0.01) and 4.9 (0-21) in week 4 (p < 0.01) (figure 1: Defecation frequency over 4 weeks.). The mean stool consistency score significantly increased from 2.6 (2-4) at baseline to 3.6 (1-6) in week 2 (p = 0.07) and 3.5 (1-6) in week 4 (p = 0.03). The episodes of faecal incontinence per week significantly decreased from 9.0 (0-35) at baseline to 2.6 (0-7) in week 2 (p = 0.01) and 1.5 (0-7) in week 4 (p < 0.01). Pain during defecation decreased from 71% (12/17) at baseline to 40% (8/20) in week 2 (p = 0.10) and 33% (6/18) in week 4 (p = 0.08). Abdominal pain episodes per week significantly decreased from 4.2 (0-7) at baseline to 2.2 (0-7) in week 2 (p = 0.02) and 1.9 (0-7) in week 4 (p = 0.01). Bisacodyl was used by 45% of patients during week 1, 25% of patients during week 2, 35% of patients during week 3 and by 20% of patients during week 4. No side effects were reported such as nausea, diarrhoea and bad taste or increased flatulence during the study period. Table 2 gives an overview of the outcome measures with p-values.\nPrimary outcome: change in defecation frequency (n = 20). X-axis: week number. Y-axis: defecation frequency per week\nMain outcome measures (mean) with p-values (n = 20)", "This pilot study showed that intake of Bifidobacterium breve for 4 weeks, significantly increased the defecation frequency. Furthermore, stool consistency, the frequency of episodes of faecal incontinence and the frequency of abdominal pain significantly changed after the use of this specific probiotic strain.\nA recent systematic review on the effects of laxative treatment and dietary measures in the management of childhood constipation found only 2 randomized controlled trials evaluating the effects of probiotics [16]. In the first small study, 45 children younger than 10 years with chronic constipation were randomly assigned to receive magnesium oxide (50 mg/kg/day (n = 18), or 8 × 108 cfu/day of the probiotic Lactobacillus casei rhamnosus (n = 18), or placebo (n = 9) twice daily for 4 weeks [17]. No statistically significant difference in the defecation frequency per day was found between the probiotic group and the magnesium oxide group. However, patients receiving either the probiotic strain or the oral laxative had a significantly higher defecation frequency compared to the placebo group (defecation frequency [times/day 0.57 ± 0.17 and 0.55 ± 0.13, respectively, compared to 0.37 ± 0.10, p = 0.03). The second trial was conducted to determine if Lactobacillus rhamnosus GG (LGG) is an effective adjunct to lactulose for treating constipation in children. A total of 48 children with constipation received 1 ml/kg/day of 70% lactulose plus 109 cfu of LGG or 1 ml/kg/day of 70% lactulose plus placebo, twice daily for 12 weeks [12]. There were no significant differences in rates of product success (defined as ≥ 3 spontaneous stools per week with no faecal incontinence) at 12 and 24 weeks between the LGG group (rates: 72% and 64%, respectively) and the placebo group (rates: 68% and 65%, respectively). `\nIn a recent trial, 44 children, at least 6 months old, with chronic constipation were randomly assigned to receive supplementation with the probiotic Lactobacillus reuteri (DSM 17938) (n = 22) or placebo (n = 22) [18]. Infants receiving Lactobacillus reuteri had a significantly higher frequency of bowel movements than infants receiving a placebo at week 8 of supplementation (2.82 per week at week 0, compared with 4.77 at week 8 in the probiotic group, absolute numbers not given for placebo group, P = 0.027). There was no significant difference between Lactobacillus reuteri and placebo groups in the stool consistency at all weeks nor in the presence of inconsolable crying episodes. All three trials did not report any adverse events in the probiotic group.\nA mixture of probiotics (containing Bifidobacteria (B.) bifidum, B. infantis, B. longum, Lactobacillus (L.) casei, L. plantarum and L. rhamnosus), increased the number of bowel movements, decreased the number of faecal incontinence episodes and improved the consistency of stools [19]. Although the results of this small pilot study are positive, a large randomized controlled trial is necessary to confirm these data.\nIt is unknown why the outcome of the above mentioned trials in constipated children is so different. We speculate that the use of different inclusion criteria for paediatric constipation with consequently different study populations and the use of different probiotic strains, given in different dosages with variable duration of the treatment period, influence study outcomes.\nIn contrast to the few paediatric studies, data suggest that adults with constipation might benefit from ingestion of B. lactis DN-173 010, L. casei Shirota, and E. coli Nissle 1917. All studies showed an increased defecation frequency and improved stool consistency [20]. These findings, however are not directly applicable to the paediatric population because constipation in children differs considerably from that in constipated adults with regard to its prevalence, onset, aetiology, symptoms, treatment, and prognosis [21].\nBesides the increase in defecation frequency and decrease in episodes of faecal incontinence, this study also showed a significant effect in softening of stools and in decreasing abdominal pain. Both effects could be a direct consequence an increase in defecation but theoretically, it could also be caused by the working mechanism of the probiotics. It has been assumed that probiotics soften the stools by stimulating water and electrolyte secretion [22,23]. Furthermore, one paediatric study and several studies in adults with irritable bowel syndrome (IBS), have demonstrated that abdominal pain decreased when using probiotics [17,24,25]. Whorwell et al. conducted a randomized trial in 360 women with IBS receiving Bifidobacterium infantis and found a significant improvement of abdominal pain which occurred irrespective of any effect on stool frequency. The authors hypothesized that the probiotics were able to diminish visceral hypersensitivity by its anti-inflammatory effect on the enteric mucosa [26].\nAccording to the available data, it is assumed that the risk of infection with the probiotic lactobacilli or bifidobacteria is similar to risks with commensal strains [10]. However, there is concern that the use of probiotics may result in harmful events in at-risk populations like immunocompromized subjects or in patients with other life-threatening illnesses, who were admitted in the intensive care unit. Based on our results and their safety profile, adding probiotics to the standard treatment of functional constipation in otherwise healthy children is promising. The major limitation of our study is that this study is a non randomized non placebo controlled small pilot study. However, since this study shows promising results it is worthwhile to perform a large RCT to unravel the efficacy of Bifidobacterium breve in constipated children.\nIn conclusion, this small pilot study suggests that Bifidobacterium breve is effective in the treatment of childhood constipation. A large randomized placebo controlled trial is now required to confirm these data.", "The authors declare that they have no competing interests.", "MMT and MAB participated in the design of the study\nAll authors collected the data.\nMMT and IDM did the statistical analysis\nMMT drafted the first manuscript\nAll authors read and approved the final manuscript" ]

[ null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Subjects", "Study design", "Outcome measures", "Analysis", "Results", "Discussion", "Competing interests", "Authors' contributions" ]

[ "Functional constipation is a common and frustrating problem in childhood with an estimated prevalence of 3% in the western world [1]. This chronic condition is characterised by infrequent defecation less than three times per week, more than two episodes of faecal incontinence per week, the passage of large and painful stools which clog the toilet and retentive posturing. Upon physical examination a palpable faecal mass is often found in the abdomen and the rectum [2,3]. It causes distress to child and family and results in severe emotional disturbance and family discord [4]. The pathophysiology underlying functional constipation is undoubtedly multi-factorial, and not well understood. Withholding behaviour is probably the major cause for the development of constipation and might be caused by the previous production of a large, hard painful stool, anal fissures, a primarily behavioural mechanism or the resistance to go to another toilet then their own [4].\nTo date, patients are treated with a combination of education, toilet training and oral laxatives. Disappointingly, only 50% of all children followed for 6 to 12 months are found to recover and were successfully taken off laxatives [5]. Another study showed that despite intensive medical and behavioural therapy, 25% of patients developing constipation before the age of 5 years continued to have severe complaints of constipation beyond puberty [6]. Furthermore, in 50% of the patients using these compounds, adverse side-effects were registered such as: abdominal pain, bloating, flatulence, diarrhoea, nausea and bad taste [7]. No data exist concerning possible long-term adverse effects such as electrolyte disturbances, mucosal damage and habituation.\nProbiotics are defined as live micro-organisms which when administered in adequate amounts, 106-109 colony forming units, confer a health benefit on the host [8]. The use of probiotics has entered mainstream medicine and has proven being an effective therapy in many different gastrointestinal disorders, including functional gastrointestinal disorders [9,10]. However, large trials investigating the efficacy and safety of probiotics in pediatric patients are lacking [11,12]. Studies in constipated adults with a Bifidus yoghurt, containing Bifidobacterium breve, Bifidobacterium bifidum and Lactobacillus acidophilus, showed a significant increase in defecation frequency without any side effects [13,14]. Furthermore, a randomized controlled trial using Bifidobacterium breve in preterm infants showed less gas accumulation in the stomach, less vomiting and improved weight gain without any side effects, suggesting a positive effect on gastrointestinal motility [15] The exact working mechanism of probiotics are not well understood. There are some hypotheses, however, why probiotics might have therapeutic potential for the treatment of constipation. Firstly, a dysbiosis in the gut flora in constipated patients has been suggested which might improve after the ingestion of probiotics [9,10]. It remains important however to understand if dysbiosis is a secondary manifestation of constipation, or if it is a factor contributing to constipation. Furthermore, probiotics can lower pH of the colon by producing lactic, acetic and other short chain fatty acids. A lower pH enhances colonic peristalsis and subsequently decreases colonic transit time [9,10].\nBased on the positive data in constipated adults, we therefore performed a pilot study to determine if Bifidobacterium breve is effective in the treatment of children with constipation.", "[SUBTITLE] Subjects [SUBSECTION] Children, 3 to 16 years of age, referred to the outpatient clinic of a tertiary pediatric gastroenterology department, with constipation were eligible for this study. Patients were included if they had been suffering from functional constipation according to the Rome III criteria for the last 2 months [2,3]. All children included had a defecation frequency of <3 times/week and one or more of the following criteria: faecal incontinence >1 episode/week, a large amount of stools that clog the toilet, painful defecation, withholding behaviour, or abdominal or rectal faecal impaction upon physical examination. In order to obtain a homogeneous group of patients, we included children with a defecation frequency <3 times/week in combination with at least one other ROME III criterion. Patients were not enrolled in this study if they had been treated for constipation less than 2 weeks before the start of the study. Other exclusion criteria were: a diagnosis of either IBS or functional non-retentive faecal incontinence according to the Rome III criteria; a diagnosis of mental retardation or metabolic disease (hypothyroidism), Hirschsprung's disease, spinal anomalies, anorectal pathology, previous gastrointestinal surgery. All children older than 12 years and/or parents gave informed consent. This pilot was approved by the medical ethical committee,\nChildren, 3 to 16 years of age, referred to the outpatient clinic of a tertiary pediatric gastroenterology department, with constipation were eligible for this study. Patients were included if they had been suffering from functional constipation according to the Rome III criteria for the last 2 months [2,3]. All children included had a defecation frequency of <3 times/week and one or more of the following criteria: faecal incontinence >1 episode/week, a large amount of stools that clog the toilet, painful defecation, withholding behaviour, or abdominal or rectal faecal impaction upon physical examination. In order to obtain a homogeneous group of patients, we included children with a defecation frequency <3 times/week in combination with at least one other ROME III criterion. Patients were not enrolled in this study if they had been treated for constipation less than 2 weeks before the start of the study. Other exclusion criteria were: a diagnosis of either IBS or functional non-retentive faecal incontinence according to the Rome III criteria; a diagnosis of mental retardation or metabolic disease (hypothyroidism), Hirschsprung's disease, spinal anomalies, anorectal pathology, previous gastrointestinal surgery. All children older than 12 years and/or parents gave informed consent. This pilot was approved by the medical ethical committee,\n[SUBTITLE] Study design [SUBSECTION] Seven days prior to baseline assessment and during the treatment period, all children recorded frequency of bowel movements, stool consistency according to the Bristol stool scale, the number of faecal incontinence episodes, pain during defecation, abdominal pain, as well as adverse effects such as vomiting and diarrhoea in a standardized bowel diary. At baseline a medical history and information on the current defection pattern was collected and also a physical examination including a rectal digital exam was performed. Information and education about functional constipation was given to all patients and their care takers. Before start of the probiotic treatment, all children received once daily for 3 days a rectal enema in order to accomplish rectal disimpaction. After rectal disimpaction, all children received daily one sachet of powder containing 108-10 CFU Bifidobacterium breve Yakult, for 4 weeks. The patients were allowed to mix the powder with all liquids on condition that the liquid was not hot. During the study, all children were instructed to try to defecate on the toilet for 5-10 minutes after each meal (3 times a day). Patients were not allowed to consume any other fermented dairy product during this study or any other laxatives, except for the rescue medication Biscaodyl. Names of the forbidden products were pointed out in the diary. During the product consumption period, patients were instructed to take bisacodyl 5 mg if they did not defecate for 3 consecutive days. Clinical evaluation, frequency of adverse effects and compliance to the study protocol were carried out at enrolment and at 2, 4 and 6 weeks using standardized bowel diaries.\nSeven days prior to baseline assessment and during the treatment period, all children recorded frequency of bowel movements, stool consistency according to the Bristol stool scale, the number of faecal incontinence episodes, pain during defecation, abdominal pain, as well as adverse effects such as vomiting and diarrhoea in a standardized bowel diary. At baseline a medical history and information on the current defection pattern was collected and also a physical examination including a rectal digital exam was performed. Information and education about functional constipation was given to all patients and their care takers. Before start of the probiotic treatment, all children received once daily for 3 days a rectal enema in order to accomplish rectal disimpaction. After rectal disimpaction, all children received daily one sachet of powder containing 108-10 CFU Bifidobacterium breve Yakult, for 4 weeks. The patients were allowed to mix the powder with all liquids on condition that the liquid was not hot. During the study, all children were instructed to try to defecate on the toilet for 5-10 minutes after each meal (3 times a day). Patients were not allowed to consume any other fermented dairy product during this study or any other laxatives, except for the rescue medication Biscaodyl. Names of the forbidden products were pointed out in the diary. During the product consumption period, patients were instructed to take bisacodyl 5 mg if they did not defecate for 3 consecutive days. Clinical evaluation, frequency of adverse effects and compliance to the study protocol were carried out at enrolment and at 2, 4 and 6 weeks using standardized bowel diaries.\n[SUBTITLE] Outcome measures [SUBSECTION] The primary outcome measure was change in defecation frequency in week 4 compared to baseline. Secondary outcome measures were stool consistency, frequency of episodes of faecal incontinence, pain during defecation, frequency of abdominal pain, frequency of adverse effects (nausea, diarrhoea and bad taste), and frequency of intake of bisacodyl.\nThe primary outcome measure was change in defecation frequency in week 4 compared to baseline. Secondary outcome measures were stool consistency, frequency of episodes of faecal incontinence, pain during defecation, frequency of abdominal pain, frequency of adverse effects (nausea, diarrhoea and bad taste), and frequency of intake of bisacodyl.\n[SUBTITLE] Analysis [SUBSECTION] Descriptive statistics were performed for baseline characteristics, adverse effects and Bisacodyl use. Change of frequency of bowel movements, fecal incontinence and change of stool consistency, was assessed using the non-parametric paired Wilcoxon test. For the comparison of abdominal pain between baseline and the evaluation time points, the Wilcoxon rank test was used. A p- value < 0.05 was considered to be significant. All analyses were done with SPSS (version 16.0).\nDescriptive statistics were performed for baseline characteristics, adverse effects and Bisacodyl use. Change of frequency of bowel movements, fecal incontinence and change of stool consistency, was assessed using the non-parametric paired Wilcoxon test. For the comparison of abdominal pain between baseline and the evaluation time points, the Wilcoxon rank test was used. A p- value < 0.05 was considered to be significant. All analyses were done with SPSS (version 16.0).", "Children, 3 to 16 years of age, referred to the outpatient clinic of a tertiary pediatric gastroenterology department, with constipation were eligible for this study. Patients were included if they had been suffering from functional constipation according to the Rome III criteria for the last 2 months [2,3]. All children included had a defecation frequency of <3 times/week and one or more of the following criteria: faecal incontinence >1 episode/week, a large amount of stools that clog the toilet, painful defecation, withholding behaviour, or abdominal or rectal faecal impaction upon physical examination. In order to obtain a homogeneous group of patients, we included children with a defecation frequency <3 times/week in combination with at least one other ROME III criterion. Patients were not enrolled in this study if they had been treated for constipation less than 2 weeks before the start of the study. Other exclusion criteria were: a diagnosis of either IBS or functional non-retentive faecal incontinence according to the Rome III criteria; a diagnosis of mental retardation or metabolic disease (hypothyroidism), Hirschsprung's disease, spinal anomalies, anorectal pathology, previous gastrointestinal surgery. All children older than 12 years and/or parents gave informed consent. This pilot was approved by the medical ethical committee,", "Seven days prior to baseline assessment and during the treatment period, all children recorded frequency of bowel movements, stool consistency according to the Bristol stool scale, the number of faecal incontinence episodes, pain during defecation, abdominal pain, as well as adverse effects such as vomiting and diarrhoea in a standardized bowel diary. At baseline a medical history and information on the current defection pattern was collected and also a physical examination including a rectal digital exam was performed. Information and education about functional constipation was given to all patients and their care takers. Before start of the probiotic treatment, all children received once daily for 3 days a rectal enema in order to accomplish rectal disimpaction. After rectal disimpaction, all children received daily one sachet of powder containing 108-10 CFU Bifidobacterium breve Yakult, for 4 weeks. The patients were allowed to mix the powder with all liquids on condition that the liquid was not hot. During the study, all children were instructed to try to defecate on the toilet for 5-10 minutes after each meal (3 times a day). Patients were not allowed to consume any other fermented dairy product during this study or any other laxatives, except for the rescue medication Biscaodyl. Names of the forbidden products were pointed out in the diary. During the product consumption period, patients were instructed to take bisacodyl 5 mg if they did not defecate for 3 consecutive days. Clinical evaluation, frequency of adverse effects and compliance to the study protocol were carried out at enrolment and at 2, 4 and 6 weeks using standardized bowel diaries.", "The primary outcome measure was change in defecation frequency in week 4 compared to baseline. Secondary outcome measures were stool consistency, frequency of episodes of faecal incontinence, pain during defecation, frequency of abdominal pain, frequency of adverse effects (nausea, diarrhoea and bad taste), and frequency of intake of bisacodyl.", "Descriptive statistics were performed for baseline characteristics, adverse effects and Bisacodyl use. Change of frequency of bowel movements, fecal incontinence and change of stool consistency, was assessed using the non-parametric paired Wilcoxon test. For the comparison of abdominal pain between baseline and the evaluation time points, the Wilcoxon rank test was used. A p- value < 0.05 was considered to be significant. All analyses were done with SPSS (version 16.0).", "Between July 2009 and January 2010, 22 children were included into this pilot study. Two children were lost to follow-up without having any outcome data during follow-up. Both patients were therefore excluded from the final analysis. All outcomes of the remaining 20 children were collected and analyzed. The baseline characteristics are described in Table 1 (mean (range))\nBaseline patient characteristics: mean (range)\nThe defecation frequency per week significantly increased from 0.9 (0-2) at baseline to 5.3 (0-21) in week 2 (p < 0.01) and 4.9 (0-21) in week 4 (p < 0.01) (figure 1: Defecation frequency over 4 weeks.). The mean stool consistency score significantly increased from 2.6 (2-4) at baseline to 3.6 (1-6) in week 2 (p = 0.07) and 3.5 (1-6) in week 4 (p = 0.03). The episodes of faecal incontinence per week significantly decreased from 9.0 (0-35) at baseline to 2.6 (0-7) in week 2 (p = 0.01) and 1.5 (0-7) in week 4 (p < 0.01). Pain during defecation decreased from 71% (12/17) at baseline to 40% (8/20) in week 2 (p = 0.10) and 33% (6/18) in week 4 (p = 0.08). Abdominal pain episodes per week significantly decreased from 4.2 (0-7) at baseline to 2.2 (0-7) in week 2 (p = 0.02) and 1.9 (0-7) in week 4 (p = 0.01). Bisacodyl was used by 45% of patients during week 1, 25% of patients during week 2, 35% of patients during week 3 and by 20% of patients during week 4. No side effects were reported such as nausea, diarrhoea and bad taste or increased flatulence during the study period. Table 2 gives an overview of the outcome measures with p-values.\nPrimary outcome: change in defecation frequency (n = 20). X-axis: week number. Y-axis: defecation frequency per week\nMain outcome measures (mean) with p-values (n = 20)", "This pilot study showed that intake of Bifidobacterium breve for 4 weeks, significantly increased the defecation frequency. Furthermore, stool consistency, the frequency of episodes of faecal incontinence and the frequency of abdominal pain significantly changed after the use of this specific probiotic strain.\nA recent systematic review on the effects of laxative treatment and dietary measures in the management of childhood constipation found only 2 randomized controlled trials evaluating the effects of probiotics [16]. In the first small study, 45 children younger than 10 years with chronic constipation were randomly assigned to receive magnesium oxide (50 mg/kg/day (n = 18), or 8 × 108 cfu/day of the probiotic Lactobacillus casei rhamnosus (n = 18), or placebo (n = 9) twice daily for 4 weeks [17]. No statistically significant difference in the defecation frequency per day was found between the probiotic group and the magnesium oxide group. However, patients receiving either the probiotic strain or the oral laxative had a significantly higher defecation frequency compared to the placebo group (defecation frequency [times/day 0.57 ± 0.17 and 0.55 ± 0.13, respectively, compared to 0.37 ± 0.10, p = 0.03). The second trial was conducted to determine if Lactobacillus rhamnosus GG (LGG) is an effective adjunct to lactulose for treating constipation in children. A total of 48 children with constipation received 1 ml/kg/day of 70% lactulose plus 109 cfu of LGG or 1 ml/kg/day of 70% lactulose plus placebo, twice daily for 12 weeks [12]. There were no significant differences in rates of product success (defined as ≥ 3 spontaneous stools per week with no faecal incontinence) at 12 and 24 weeks between the LGG group (rates: 72% and 64%, respectively) and the placebo group (rates: 68% and 65%, respectively). `\nIn a recent trial, 44 children, at least 6 months old, with chronic constipation were randomly assigned to receive supplementation with the probiotic Lactobacillus reuteri (DSM 17938) (n = 22) or placebo (n = 22) [18]. Infants receiving Lactobacillus reuteri had a significantly higher frequency of bowel movements than infants receiving a placebo at week 8 of supplementation (2.82 per week at week 0, compared with 4.77 at week 8 in the probiotic group, absolute numbers not given for placebo group, P = 0.027). There was no significant difference between Lactobacillus reuteri and placebo groups in the stool consistency at all weeks nor in the presence of inconsolable crying episodes. All three trials did not report any adverse events in the probiotic group.\nA mixture of probiotics (containing Bifidobacteria (B.) bifidum, B. infantis, B. longum, Lactobacillus (L.) casei, L. plantarum and L. rhamnosus), increased the number of bowel movements, decreased the number of faecal incontinence episodes and improved the consistency of stools [19]. Although the results of this small pilot study are positive, a large randomized controlled trial is necessary to confirm these data.\nIt is unknown why the outcome of the above mentioned trials in constipated children is so different. We speculate that the use of different inclusion criteria for paediatric constipation with consequently different study populations and the use of different probiotic strains, given in different dosages with variable duration of the treatment period, influence study outcomes.\nIn contrast to the few paediatric studies, data suggest that adults with constipation might benefit from ingestion of B. lactis DN-173 010, L. casei Shirota, and E. coli Nissle 1917. All studies showed an increased defecation frequency and improved stool consistency [20]. These findings, however are not directly applicable to the paediatric population because constipation in children differs considerably from that in constipated adults with regard to its prevalence, onset, aetiology, symptoms, treatment, and prognosis [21].\nBesides the increase in defecation frequency and decrease in episodes of faecal incontinence, this study also showed a significant effect in softening of stools and in decreasing abdominal pain. Both effects could be a direct consequence an increase in defecation but theoretically, it could also be caused by the working mechanism of the probiotics. It has been assumed that probiotics soften the stools by stimulating water and electrolyte secretion [22,23]. Furthermore, one paediatric study and several studies in adults with irritable bowel syndrome (IBS), have demonstrated that abdominal pain decreased when using probiotics [17,24,25]. Whorwell et al. conducted a randomized trial in 360 women with IBS receiving Bifidobacterium infantis and found a significant improvement of abdominal pain which occurred irrespective of any effect on stool frequency. The authors hypothesized that the probiotics were able to diminish visceral hypersensitivity by its anti-inflammatory effect on the enteric mucosa [26].\nAccording to the available data, it is assumed that the risk of infection with the probiotic lactobacilli or bifidobacteria is similar to risks with commensal strains [10]. However, there is concern that the use of probiotics may result in harmful events in at-risk populations like immunocompromized subjects or in patients with other life-threatening illnesses, who were admitted in the intensive care unit. Based on our results and their safety profile, adding probiotics to the standard treatment of functional constipation in otherwise healthy children is promising. The major limitation of our study is that this study is a non randomized non placebo controlled small pilot study. However, since this study shows promising results it is worthwhile to perform a large RCT to unravel the efficacy of Bifidobacterium breve in constipated children.\nIn conclusion, this small pilot study suggests that Bifidobacterium breve is effective in the treatment of childhood constipation. A large randomized placebo controlled trial is now required to confirm these data.", "The authors declare that they have no competing interests.", "MMT and MAB participated in the design of the study\nAll authors collected the data.\nMMT and IDM did the statistical analysis\nMMT drafted the first manuscript\nAll authors read and approved the final manuscript" ]

[ null, "methods", null, null, null, null, null, null, null, null ]

[]

[ "Aged", "Aged, 80 and over", "Algorithms", "Antirheumatic Agents", "Arthritis, Rheumatoid", "Databases as Topic", "Databases, Factual", "Female", "Humans", "Male", "Pennsylvania", "Predictive Value of Tests", "Rheumatology" ]

[ "Introduction", "Data source", "Study procedures", "Statistical analyses", "Characteristics of the study population", "Positive predictive value for various algorithms", "Abbreviations", "Competing interests", "Authors' contributions" ]

[ "Large automated databases such as health care utilization and medical record databases have been widely used as data sources for epidemiologic studies [1]. Validity and completeness of prescription drug data in health care utilization databases with the prescription drug plan have been checked several times and reported as being of high quality [2], but the accuracy of specific disease data such as diagnosis of rheumatoid arthritis (RA) in health care utilization data has been somewhat questionable.\nSeveral studies previously examined the accuracy of RA diagnoses in various data sources and reported inconsistent results [3-8]. A previous study examined the accuracy of computerized database diagnoses of RA among the Olmsted County residents in Minnesota on the basis of chart review and found a sensitivity of 89%, a specificity of 74%, and a positive predictive value (PPV) of 57% by using the American College of Rheumatology (ACR) RA criteria as the gold standard [3]. The PPV of the RA diagnosis codes alone was only 66% compared with the gold standard definition of RA diagnosis by a rheumatologist on two separate visits in a study using the Minneapolis Veterans Affairs administrative data [7]. A Danish national register-based study showed that 59% of the subjects identified by the algorithm using only discharge diagnosis codes had a clinical diagnosis of RA and that 46% of those met the ACR criteria for RA [8].\nHowever, the sensitivity and PPV were over 90% for the chart documentation of RA diagnosis in a study of Medicare diagnosis claims for RA from several rheumatology practices [4]. The PPV of the RA diagnosis codes from Medicare inpatient claims among total hip replacement recipients was 86% for the chart documentation of RA diagnosis [5]. Another administrative data-based algorithm with at least two physician visit claims for RA (with at least 30 days between the visits) had a PPV of 92% for RA based on a patient self-report questionnaire [6].\nIn this study, we developed several diagnosis code-based algorithms with and without a link to pharmacy claims for disease-modifying antirheumatic drugs (DMARDs) to define the outpatient diagnosis of RA in a health care utilization database and compared the validity of these algorithms to various gold standard definitions.", "We studied participants in the Pennsylvania Assistance Contract for the Elderly (PACE) program, established in 1984 to assist Pennsylvania residents who are 65 years or older, who are of low to moderate income, and who may suffer financial hardship in paying for their medication. The PACE program provides pharmacy benefits for all drugs, including DMARDs and biologic therapy, for qualifying residents who are 65 or older. All PACE participants receive Medicare benefits. Data use agreements were in place with Medicare and the PACE program that supplied information for the study database. This work was approved by Brigham and Women's Hospital's Institutional Review Board.", "Three different algorithms were used to identify patients with RA by using the Medicare claim data from 1994 to 2004: (a) beneficiaries with at least two claims associated with RA (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9 CM] code 714), (b) beneficiaries with at least three claims associated with RA, and (c) beneficiaries with at least two RA claims that were from a rheumatologist and that were separated by at least 7 days. All inpatient, outpatient, and procedure claims such as laboratory or radiologic tests were included. We identified rheumatologists with a Medicare provider specialty code in the database and verified them with the ACR membership directory. A subgroup of patients who filled at least one prescription for DMARDs over a period of 1 year after the RA diagnosis was then identified by using the data from both pharmacy benefit program and claim data for infusions. To compare baseline characteristics of the study subjects, we selected a group of beneficiaries who never had any claims for RA.\nAfter identifying subjects by each of the algorithms, we attempted to obtain consent to review their medical record. First, the PACE program mailed a letter to the groups of subjects identified by our algorithms to inform them that they would be contacted by our research group. A letter that provided details about the study was then sent to the subjects in each of the groups and asked whether they would consent to have the study researchers review their medical records from their physicians, including doctors who treated them for arthritis. Subjects who agreed to participate in the study signed a consent and authorization form for release of medical records. Additionally, subjects were asked to complete a physician information form to identify their primary physicians as well as specialists and their contact information. We then attempted to obtain copies of medical records.\nOnce we received the medical records, all personal identifiers were removed from the records for protection of patients' privacy. Medical records were reviewed independently by several rheumatologists at Brigham and Women's Hospital. To minimize inter-reviewer variation in data abstraction, a structured data abstraction form was developed and pilot-tested with the principal investigator (DHS). The form included items such as the seven ACR 1987 classification criteria for RA, disease onset, other rheumatologic diagnoses, medications, and laboratory data. On the basis of these data, the reviewers assessed whether a patient met the gold standard definitions of RA: (a) diagnosis of RA by a rheumatologist and (b) fulfillment of the ACR criteria for RA. Any indication in the medical record that the diagnosing rheumatologists thought that the patient had RA at that time was counted as having 'RA diagnosis per rheumatologists'. When the patients were not seen by rheumatologists, 'RA diagnosis per rheumatologists' was made by the reviewers on the basis of the data from their medical records. When the diagnosis of RA was neither documented nor clear in their medical records, the patients were considered non-RA. Areas of disagreement or uncertainty were resolved by consensus. The study period for data collection from medical records lasted from 2004 to 2008.", "We calculated PPV as the percentage of the patients who met the gold standard definitions among those identified by the algorithms. We also examined the PPVs of these algorithms combined with at least one prescription fill for a DMARD (Table 1). Ninety-five percent confidence intervals (CIs) of the PPVs were calculated by using the normal approximation of the binomial distribution. All analyses were conducted with SAS 9.1 Statistical Software (SAS Institute Inc., Cary, NC, USA).\nA list of disease-modifying antirheumatic drugs included in the study", "A total of 9,482 patients were identified with the algorithms. Only 2% of the patients consented to have medical records reviewed for our study. Subsequently, medical records were obtained in 83.1% of those who consented to the study. Demographic characteristics were similar between respondents and non-respondents. Among the non-respondents, the mean age was 80.7 years with a standard deviation (SD) of 6.8, and 85.9% were female. Table 2 describes the characteristics of study subjects identified by each algorithm. Overall, the mean age was 79.3 (SD 7.1) years, 82.9% were female, and 98.2% were Caucasians. The patients identified by the algorithm requiring at least two claims from a rheumatologist were slightly younger and had more comorbidities than the patients identified by the other algorithms.\nBaseline characteristics of study subjects\naAt least 7 days were required between the claims. DMARD, disease-modifying antirheumatic drug; RA, rheumatoid arthritis; SD, standard deviation.", "Table 3 presents the PPV of each algorithm. When 'RA diagnosis per rheumatologists' was used as the gold standard, the PPVs were 55.7% (95% CI 46.8% to 64.4%) for the algorithm of at least two claims for RA and 65.5% (95% CI 55.8% to 74.3%) for the algorithm of at least three claims for RA. When the algorithm was restricted to at least two claims that were from a rheumatologist and that were separated by at least 7 days, the PPV increased to 66.7% (95% CI 55.5% to 76.6%). The PPVs of these algorithms were generally lower, ranging from 33.6% to 40.0%, with fulfillment of four or more of the ACR RA criteria as the gold standard.\nPositive predictive values and 95% confidence intervals of the algorithms to define rheumatoid arthritis in health care utilization data\nPositive predictive values (PPVs) are presented as a percentage. aAt least 7 days were required between the claims. ACR, American College of Rheumatology; CI, confidence interval; DMARD, disease-modifying antirheumatic drug; RA, rheumatoid arthritis.\nWhen at least one DMARD prescription was required, the PPV improved to 86.2% (95% CI 74.6% to 93.9%) for the algorithm of at least two claims for RA, with 'RA diagnosis per rheumatologists' as the gold standard. The PPV was highest (88.9%, 95% CI 76.0% to 96.3%) for the algorithm of at least two claims from a rheumatologist combined with at least one DMARD prescription. When fulfillment of four or more of the ACR RA criteria was used as the gold standard, the PPVs of the algorithms combined with at least one DMARD prescription ranged from 55.6% to 60.7% (Table 3).\nLess than 20% of the patients were identified with ICD-9 714.9, which is for unspecified inflammatory polyarthropathy. In a sensitivity analysis, we excluded those patients and recalculated the PPVs of the algorithms. Overall, the PPV did not improve substantially. The PPVs were 60.7% (95% CI 51.8% to 69.5%) for the algorithm of at least two claims for RA and 70.1% (95% CI 61.0% to 79.2%) for the algorithm of at least three claims for RA using 'RA diagnosis per rheumatologists' as the gold standard. The algorithm of at least two claims from a rheumatologist had the PPV of 73.0% (95% CI 62.9% to 83.1%).", "ACR: American College of Rheumatology; CI: confidence interval; DMARD: disease-modifying antirheumatic drug; ICD-9: International Classification of Diseases-9th Revision; PACE: Pennsylvania Assistance Contract for the Elderly; PPV: positive predictive value; RA: rheumatoid arthritis; SD: standard deviation.", "DHS has received research support from Amgen (Thousand Oaks, CA, USA) and Abbott (Abbott Park, IL, USA) and support for an educational course from Bristol-Myers Squibb Company (Princeton, NJ, USA). He has non-compensation roles in two drug trials sponsored by Pfizer Inc (New York, NY, USA). The other authors declare that they have no competing interests.", "All authors participated in the study conception. AS and JMP participated in the study design and in data acquisition. JNK participated in the study design and in data analysis and interpretation. DHS participated in the study design and in data acquisition, analysis, and interpretation. SK, MEW, and HM participated in data analysis and interpretation. All authors participated in manuscript preparation and revision. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Materials and methods", "Data source", "Study procedures", "Statistical analyses", "Results", "Characteristics of the study population", "Positive predictive value for various algorithms", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions" ]

[ "Large automated databases such as health care utilization and medical record databases have been widely used as data sources for epidemiologic studies [1]. Validity and completeness of prescription drug data in health care utilization databases with the prescription drug plan have been checked several times and reported as being of high quality [2], but the accuracy of specific disease data such as diagnosis of rheumatoid arthritis (RA) in health care utilization data has been somewhat questionable.\nSeveral studies previously examined the accuracy of RA diagnoses in various data sources and reported inconsistent results [3-8]. A previous study examined the accuracy of computerized database diagnoses of RA among the Olmsted County residents in Minnesota on the basis of chart review and found a sensitivity of 89%, a specificity of 74%, and a positive predictive value (PPV) of 57% by using the American College of Rheumatology (ACR) RA criteria as the gold standard [3]. The PPV of the RA diagnosis codes alone was only 66% compared with the gold standard definition of RA diagnosis by a rheumatologist on two separate visits in a study using the Minneapolis Veterans Affairs administrative data [7]. A Danish national register-based study showed that 59% of the subjects identified by the algorithm using only discharge diagnosis codes had a clinical diagnosis of RA and that 46% of those met the ACR criteria for RA [8].\nHowever, the sensitivity and PPV were over 90% for the chart documentation of RA diagnosis in a study of Medicare diagnosis claims for RA from several rheumatology practices [4]. The PPV of the RA diagnosis codes from Medicare inpatient claims among total hip replacement recipients was 86% for the chart documentation of RA diagnosis [5]. Another administrative data-based algorithm with at least two physician visit claims for RA (with at least 30 days between the visits) had a PPV of 92% for RA based on a patient self-report questionnaire [6].\nIn this study, we developed several diagnosis code-based algorithms with and without a link to pharmacy claims for disease-modifying antirheumatic drugs (DMARDs) to define the outpatient diagnosis of RA in a health care utilization database and compared the validity of these algorithms to various gold standard definitions.", "[SUBTITLE] Data source [SUBSECTION] We studied participants in the Pennsylvania Assistance Contract for the Elderly (PACE) program, established in 1984 to assist Pennsylvania residents who are 65 years or older, who are of low to moderate income, and who may suffer financial hardship in paying for their medication. The PACE program provides pharmacy benefits for all drugs, including DMARDs and biologic therapy, for qualifying residents who are 65 or older. All PACE participants receive Medicare benefits. Data use agreements were in place with Medicare and the PACE program that supplied information for the study database. This work was approved by Brigham and Women's Hospital's Institutional Review Board.\nWe studied participants in the Pennsylvania Assistance Contract for the Elderly (PACE) program, established in 1984 to assist Pennsylvania residents who are 65 years or older, who are of low to moderate income, and who may suffer financial hardship in paying for their medication. The PACE program provides pharmacy benefits for all drugs, including DMARDs and biologic therapy, for qualifying residents who are 65 or older. All PACE participants receive Medicare benefits. Data use agreements were in place with Medicare and the PACE program that supplied information for the study database. This work was approved by Brigham and Women's Hospital's Institutional Review Board.\n[SUBTITLE] Study procedures [SUBSECTION] Three different algorithms were used to identify patients with RA by using the Medicare claim data from 1994 to 2004: (a) beneficiaries with at least two claims associated with RA (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9 CM] code 714), (b) beneficiaries with at least three claims associated with RA, and (c) beneficiaries with at least two RA claims that were from a rheumatologist and that were separated by at least 7 days. All inpatient, outpatient, and procedure claims such as laboratory or radiologic tests were included. We identified rheumatologists with a Medicare provider specialty code in the database and verified them with the ACR membership directory. A subgroup of patients who filled at least one prescription for DMARDs over a period of 1 year after the RA diagnosis was then identified by using the data from both pharmacy benefit program and claim data for infusions. To compare baseline characteristics of the study subjects, we selected a group of beneficiaries who never had any claims for RA.\nAfter identifying subjects by each of the algorithms, we attempted to obtain consent to review their medical record. First, the PACE program mailed a letter to the groups of subjects identified by our algorithms to inform them that they would be contacted by our research group. A letter that provided details about the study was then sent to the subjects in each of the groups and asked whether they would consent to have the study researchers review their medical records from their physicians, including doctors who treated them for arthritis. Subjects who agreed to participate in the study signed a consent and authorization form for release of medical records. Additionally, subjects were asked to complete a physician information form to identify their primary physicians as well as specialists and their contact information. We then attempted to obtain copies of medical records.\nOnce we received the medical records, all personal identifiers were removed from the records for protection of patients' privacy. Medical records were reviewed independently by several rheumatologists at Brigham and Women's Hospital. To minimize inter-reviewer variation in data abstraction, a structured data abstraction form was developed and pilot-tested with the principal investigator (DHS). The form included items such as the seven ACR 1987 classification criteria for RA, disease onset, other rheumatologic diagnoses, medications, and laboratory data. On the basis of these data, the reviewers assessed whether a patient met the gold standard definitions of RA: (a) diagnosis of RA by a rheumatologist and (b) fulfillment of the ACR criteria for RA. Any indication in the medical record that the diagnosing rheumatologists thought that the patient had RA at that time was counted as having 'RA diagnosis per rheumatologists'. When the patients were not seen by rheumatologists, 'RA diagnosis per rheumatologists' was made by the reviewers on the basis of the data from their medical records. When the diagnosis of RA was neither documented nor clear in their medical records, the patients were considered non-RA. Areas of disagreement or uncertainty were resolved by consensus. The study period for data collection from medical records lasted from 2004 to 2008.\nThree different algorithms were used to identify patients with RA by using the Medicare claim data from 1994 to 2004: (a) beneficiaries with at least two claims associated with RA (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9 CM] code 714), (b) beneficiaries with at least three claims associated with RA, and (c) beneficiaries with at least two RA claims that were from a rheumatologist and that were separated by at least 7 days. All inpatient, outpatient, and procedure claims such as laboratory or radiologic tests were included. We identified rheumatologists with a Medicare provider specialty code in the database and verified them with the ACR membership directory. A subgroup of patients who filled at least one prescription for DMARDs over a period of 1 year after the RA diagnosis was then identified by using the data from both pharmacy benefit program and claim data for infusions. To compare baseline characteristics of the study subjects, we selected a group of beneficiaries who never had any claims for RA.\nAfter identifying subjects by each of the algorithms, we attempted to obtain consent to review their medical record. First, the PACE program mailed a letter to the groups of subjects identified by our algorithms to inform them that they would be contacted by our research group. A letter that provided details about the study was then sent to the subjects in each of the groups and asked whether they would consent to have the study researchers review their medical records from their physicians, including doctors who treated them for arthritis. Subjects who agreed to participate in the study signed a consent and authorization form for release of medical records. Additionally, subjects were asked to complete a physician information form to identify their primary physicians as well as specialists and their contact information. We then attempted to obtain copies of medical records.\nOnce we received the medical records, all personal identifiers were removed from the records for protection of patients' privacy. Medical records were reviewed independently by several rheumatologists at Brigham and Women's Hospital. To minimize inter-reviewer variation in data abstraction, a structured data abstraction form was developed and pilot-tested with the principal investigator (DHS). The form included items such as the seven ACR 1987 classification criteria for RA, disease onset, other rheumatologic diagnoses, medications, and laboratory data. On the basis of these data, the reviewers assessed whether a patient met the gold standard definitions of RA: (a) diagnosis of RA by a rheumatologist and (b) fulfillment of the ACR criteria for RA. Any indication in the medical record that the diagnosing rheumatologists thought that the patient had RA at that time was counted as having 'RA diagnosis per rheumatologists'. When the patients were not seen by rheumatologists, 'RA diagnosis per rheumatologists' was made by the reviewers on the basis of the data from their medical records. When the diagnosis of RA was neither documented nor clear in their medical records, the patients were considered non-RA. Areas of disagreement or uncertainty were resolved by consensus. The study period for data collection from medical records lasted from 2004 to 2008.\n[SUBTITLE] Statistical analyses [SUBSECTION] We calculated PPV as the percentage of the patients who met the gold standard definitions among those identified by the algorithms. We also examined the PPVs of these algorithms combined with at least one prescription fill for a DMARD (Table 1). Ninety-five percent confidence intervals (CIs) of the PPVs were calculated by using the normal approximation of the binomial distribution. All analyses were conducted with SAS 9.1 Statistical Software (SAS Institute Inc., Cary, NC, USA).\nA list of disease-modifying antirheumatic drugs included in the study\nWe calculated PPV as the percentage of the patients who met the gold standard definitions among those identified by the algorithms. We also examined the PPVs of these algorithms combined with at least one prescription fill for a DMARD (Table 1). Ninety-five percent confidence intervals (CIs) of the PPVs were calculated by using the normal approximation of the binomial distribution. All analyses were conducted with SAS 9.1 Statistical Software (SAS Institute Inc., Cary, NC, USA).\nA list of disease-modifying antirheumatic drugs included in the study", "We studied participants in the Pennsylvania Assistance Contract for the Elderly (PACE) program, established in 1984 to assist Pennsylvania residents who are 65 years or older, who are of low to moderate income, and who may suffer financial hardship in paying for their medication. The PACE program provides pharmacy benefits for all drugs, including DMARDs and biologic therapy, for qualifying residents who are 65 or older. All PACE participants receive Medicare benefits. Data use agreements were in place with Medicare and the PACE program that supplied information for the study database. This work was approved by Brigham and Women's Hospital's Institutional Review Board.", "Three different algorithms were used to identify patients with RA by using the Medicare claim data from 1994 to 2004: (a) beneficiaries with at least two claims associated with RA (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9 CM] code 714), (b) beneficiaries with at least three claims associated with RA, and (c) beneficiaries with at least two RA claims that were from a rheumatologist and that were separated by at least 7 days. All inpatient, outpatient, and procedure claims such as laboratory or radiologic tests were included. We identified rheumatologists with a Medicare provider specialty code in the database and verified them with the ACR membership directory. A subgroup of patients who filled at least one prescription for DMARDs over a period of 1 year after the RA diagnosis was then identified by using the data from both pharmacy benefit program and claim data for infusions. To compare baseline characteristics of the study subjects, we selected a group of beneficiaries who never had any claims for RA.\nAfter identifying subjects by each of the algorithms, we attempted to obtain consent to review their medical record. First, the PACE program mailed a letter to the groups of subjects identified by our algorithms to inform them that they would be contacted by our research group. A letter that provided details about the study was then sent to the subjects in each of the groups and asked whether they would consent to have the study researchers review their medical records from their physicians, including doctors who treated them for arthritis. Subjects who agreed to participate in the study signed a consent and authorization form for release of medical records. Additionally, subjects were asked to complete a physician information form to identify their primary physicians as well as specialists and their contact information. We then attempted to obtain copies of medical records.\nOnce we received the medical records, all personal identifiers were removed from the records for protection of patients' privacy. Medical records were reviewed independently by several rheumatologists at Brigham and Women's Hospital. To minimize inter-reviewer variation in data abstraction, a structured data abstraction form was developed and pilot-tested with the principal investigator (DHS). The form included items such as the seven ACR 1987 classification criteria for RA, disease onset, other rheumatologic diagnoses, medications, and laboratory data. On the basis of these data, the reviewers assessed whether a patient met the gold standard definitions of RA: (a) diagnosis of RA by a rheumatologist and (b) fulfillment of the ACR criteria for RA. Any indication in the medical record that the diagnosing rheumatologists thought that the patient had RA at that time was counted as having 'RA diagnosis per rheumatologists'. When the patients were not seen by rheumatologists, 'RA diagnosis per rheumatologists' was made by the reviewers on the basis of the data from their medical records. When the diagnosis of RA was neither documented nor clear in their medical records, the patients were considered non-RA. Areas of disagreement or uncertainty were resolved by consensus. The study period for data collection from medical records lasted from 2004 to 2008.", "We calculated PPV as the percentage of the patients who met the gold standard definitions among those identified by the algorithms. We also examined the PPVs of these algorithms combined with at least one prescription fill for a DMARD (Table 1). Ninety-five percent confidence intervals (CIs) of the PPVs were calculated by using the normal approximation of the binomial distribution. All analyses were conducted with SAS 9.1 Statistical Software (SAS Institute Inc., Cary, NC, USA).\nA list of disease-modifying antirheumatic drugs included in the study", "[SUBTITLE] Characteristics of the study population [SUBSECTION] A total of 9,482 patients were identified with the algorithms. Only 2% of the patients consented to have medical records reviewed for our study. Subsequently, medical records were obtained in 83.1% of those who consented to the study. Demographic characteristics were similar between respondents and non-respondents. Among the non-respondents, the mean age was 80.7 years with a standard deviation (SD) of 6.8, and 85.9% were female. Table 2 describes the characteristics of study subjects identified by each algorithm. Overall, the mean age was 79.3 (SD 7.1) years, 82.9% were female, and 98.2% were Caucasians. The patients identified by the algorithm requiring at least two claims from a rheumatologist were slightly younger and had more comorbidities than the patients identified by the other algorithms.\nBaseline characteristics of study subjects\naAt least 7 days were required between the claims. DMARD, disease-modifying antirheumatic drug; RA, rheumatoid arthritis; SD, standard deviation.\nA total of 9,482 patients were identified with the algorithms. Only 2% of the patients consented to have medical records reviewed for our study. Subsequently, medical records were obtained in 83.1% of those who consented to the study. Demographic characteristics were similar between respondents and non-respondents. Among the non-respondents, the mean age was 80.7 years with a standard deviation (SD) of 6.8, and 85.9% were female. Table 2 describes the characteristics of study subjects identified by each algorithm. Overall, the mean age was 79.3 (SD 7.1) years, 82.9% were female, and 98.2% were Caucasians. The patients identified by the algorithm requiring at least two claims from a rheumatologist were slightly younger and had more comorbidities than the patients identified by the other algorithms.\nBaseline characteristics of study subjects\naAt least 7 days were required between the claims. DMARD, disease-modifying antirheumatic drug; RA, rheumatoid arthritis; SD, standard deviation.\n[SUBTITLE] Positive predictive value for various algorithms [SUBSECTION] Table 3 presents the PPV of each algorithm. When 'RA diagnosis per rheumatologists' was used as the gold standard, the PPVs were 55.7% (95% CI 46.8% to 64.4%) for the algorithm of at least two claims for RA and 65.5% (95% CI 55.8% to 74.3%) for the algorithm of at least three claims for RA. When the algorithm was restricted to at least two claims that were from a rheumatologist and that were separated by at least 7 days, the PPV increased to 66.7% (95% CI 55.5% to 76.6%). The PPVs of these algorithms were generally lower, ranging from 33.6% to 40.0%, with fulfillment of four or more of the ACR RA criteria as the gold standard.\nPositive predictive values and 95% confidence intervals of the algorithms to define rheumatoid arthritis in health care utilization data\nPositive predictive values (PPVs) are presented as a percentage. aAt least 7 days were required between the claims. ACR, American College of Rheumatology; CI, confidence interval; DMARD, disease-modifying antirheumatic drug; RA, rheumatoid arthritis.\nWhen at least one DMARD prescription was required, the PPV improved to 86.2% (95% CI 74.6% to 93.9%) for the algorithm of at least two claims for RA, with 'RA diagnosis per rheumatologists' as the gold standard. The PPV was highest (88.9%, 95% CI 76.0% to 96.3%) for the algorithm of at least two claims from a rheumatologist combined with at least one DMARD prescription. When fulfillment of four or more of the ACR RA criteria was used as the gold standard, the PPVs of the algorithms combined with at least one DMARD prescription ranged from 55.6% to 60.7% (Table 3).\nLess than 20% of the patients were identified with ICD-9 714.9, which is for unspecified inflammatory polyarthropathy. In a sensitivity analysis, we excluded those patients and recalculated the PPVs of the algorithms. Overall, the PPV did not improve substantially. The PPVs were 60.7% (95% CI 51.8% to 69.5%) for the algorithm of at least two claims for RA and 70.1% (95% CI 61.0% to 79.2%) for the algorithm of at least three claims for RA using 'RA diagnosis per rheumatologists' as the gold standard. The algorithm of at least two claims from a rheumatologist had the PPV of 73.0% (95% CI 62.9% to 83.1%).\nTable 3 presents the PPV of each algorithm. When 'RA diagnosis per rheumatologists' was used as the gold standard, the PPVs were 55.7% (95% CI 46.8% to 64.4%) for the algorithm of at least two claims for RA and 65.5% (95% CI 55.8% to 74.3%) for the algorithm of at least three claims for RA. When the algorithm was restricted to at least two claims that were from a rheumatologist and that were separated by at least 7 days, the PPV increased to 66.7% (95% CI 55.5% to 76.6%). The PPVs of these algorithms were generally lower, ranging from 33.6% to 40.0%, with fulfillment of four or more of the ACR RA criteria as the gold standard.\nPositive predictive values and 95% confidence intervals of the algorithms to define rheumatoid arthritis in health care utilization data\nPositive predictive values (PPVs) are presented as a percentage. aAt least 7 days were required between the claims. ACR, American College of Rheumatology; CI, confidence interval; DMARD, disease-modifying antirheumatic drug; RA, rheumatoid arthritis.\nWhen at least one DMARD prescription was required, the PPV improved to 86.2% (95% CI 74.6% to 93.9%) for the algorithm of at least two claims for RA, with 'RA diagnosis per rheumatologists' as the gold standard. The PPV was highest (88.9%, 95% CI 76.0% to 96.3%) for the algorithm of at least two claims from a rheumatologist combined with at least one DMARD prescription. When fulfillment of four or more of the ACR RA criteria was used as the gold standard, the PPVs of the algorithms combined with at least one DMARD prescription ranged from 55.6% to 60.7% (Table 3).\nLess than 20% of the patients were identified with ICD-9 714.9, which is for unspecified inflammatory polyarthropathy. In a sensitivity analysis, we excluded those patients and recalculated the PPVs of the algorithms. Overall, the PPV did not improve substantially. The PPVs were 60.7% (95% CI 51.8% to 69.5%) for the algorithm of at least two claims for RA and 70.1% (95% CI 61.0% to 79.2%) for the algorithm of at least three claims for RA using 'RA diagnosis per rheumatologists' as the gold standard. The algorithm of at least two claims from a rheumatologist had the PPV of 73.0% (95% CI 62.9% to 83.1%).", "A total of 9,482 patients were identified with the algorithms. Only 2% of the patients consented to have medical records reviewed for our study. Subsequently, medical records were obtained in 83.1% of those who consented to the study. Demographic characteristics were similar between respondents and non-respondents. Among the non-respondents, the mean age was 80.7 years with a standard deviation (SD) of 6.8, and 85.9% were female. Table 2 describes the characteristics of study subjects identified by each algorithm. Overall, the mean age was 79.3 (SD 7.1) years, 82.9% were female, and 98.2% were Caucasians. The patients identified by the algorithm requiring at least two claims from a rheumatologist were slightly younger and had more comorbidities than the patients identified by the other algorithms.\nBaseline characteristics of study subjects\naAt least 7 days were required between the claims. DMARD, disease-modifying antirheumatic drug; RA, rheumatoid arthritis; SD, standard deviation.", "Table 3 presents the PPV of each algorithm. When 'RA diagnosis per rheumatologists' was used as the gold standard, the PPVs were 55.7% (95% CI 46.8% to 64.4%) for the algorithm of at least two claims for RA and 65.5% (95% CI 55.8% to 74.3%) for the algorithm of at least three claims for RA. When the algorithm was restricted to at least two claims that were from a rheumatologist and that were separated by at least 7 days, the PPV increased to 66.7% (95% CI 55.5% to 76.6%). The PPVs of these algorithms were generally lower, ranging from 33.6% to 40.0%, with fulfillment of four or more of the ACR RA criteria as the gold standard.\nPositive predictive values and 95% confidence intervals of the algorithms to define rheumatoid arthritis in health care utilization data\nPositive predictive values (PPVs) are presented as a percentage. aAt least 7 days were required between the claims. ACR, American College of Rheumatology; CI, confidence interval; DMARD, disease-modifying antirheumatic drug; RA, rheumatoid arthritis.\nWhen at least one DMARD prescription was required, the PPV improved to 86.2% (95% CI 74.6% to 93.9%) for the algorithm of at least two claims for RA, with 'RA diagnosis per rheumatologists' as the gold standard. The PPV was highest (88.9%, 95% CI 76.0% to 96.3%) for the algorithm of at least two claims from a rheumatologist combined with at least one DMARD prescription. When fulfillment of four or more of the ACR RA criteria was used as the gold standard, the PPVs of the algorithms combined with at least one DMARD prescription ranged from 55.6% to 60.7% (Table 3).\nLess than 20% of the patients were identified with ICD-9 714.9, which is for unspecified inflammatory polyarthropathy. In a sensitivity analysis, we excluded those patients and recalculated the PPVs of the algorithms. Overall, the PPV did not improve substantially. The PPVs were 60.7% (95% CI 51.8% to 69.5%) for the algorithm of at least two claims for RA and 70.1% (95% CI 61.0% to 79.2%) for the algorithm of at least three claims for RA using 'RA diagnosis per rheumatologists' as the gold standard. The algorithm of at least two claims from a rheumatologist had the PPV of 73.0% (95% CI 62.9% to 83.1%).", "This study examined the PPV of various algorithms for identifying patients with RA in health care utilization data and found that the diagnosis code-based algorithms had modest PPVs, ranging from 55.7% for the least restrictive algorithm to 66.7% for the most restrictive, using the diagnosis of RA by a rheumatologist as the gold standard. However, we found that requiring a DMARD prescription improved the PPVs substantially. We also found that PPVs were lower when fulfillment of four or more of the ACR RA criteria was used as the gold standard.\nPrevious studies of Medicare claim data for the RA diagnosis showed the high PPVs over 85% compared with the chart documentation of RA diagnosis [4,5]. The better performance of the RA diagnosis codes in these studies can be explained by a difference in patient population as these studies were limited to either a hospital inpatient setting for joint replacement surgery or rheumatology specialty clinics.\nOur study has important implications. Based on our results, a diagnosis code-based algorithm alone is not sufficient to accurately identify patients with RA in the health care utilization data. Further refinement of the algorithms with a link to pharmacy claim data for a DMARD prescription can improve the PPVs of RA diagnoses in these data. Studies assessing RA-specific complications or the burden of RA solely on the basis of the ICD-9 code should be interpreted with caution.\nSeveral limitations of this study should be noted. First, generalizability can be an issue with the low response rate, although we did not find a significant difference in demographic characteristics between respondents and non-respondents. We attempted to recruit as many patients as possible and sent multiple recruitment letters over a period of 3 years, but the response rate was only 2%. One of the main reasons for this low response rate is that this study required patients in the community to provide an authorization to release their medical records to the study investigators, who were not directly or indirectly involved in their medical care. Other potential explanations for such a low response rate include older age, low socioeconomic status, admission to a nursing home, critical illness, and death. Second, our focus on the elderly can be seen as a limitation as it is possible that validity may vary by age group as our study included only those patients who were 65 or older. However, the prevalence of RA among adults who are 60 years or older in the US is approximately 2% [9]; therefore, the elderly populations contain the substantial proportion of RA patients in the population. Third, the percentage of the patients who met the ACR criteria in our review was low. It might have been underestimated as we did not have access to all the longitudinal medical records across multiple physicians. Incompleteness of information that is needed to assess the fulfillment of the individual ACR RA criteria in medical records has been previously reported [10,11]. The diagnostic performance of the ACR classification criteria for RA is also known to be problematic in a clinical setting [12].\nOur study demonstrated that the PPVs of RA diagnosis codes in the health care utilization data varied considerably across different gold standard definitions. When 'RA diagnosis per rheumatologists' was used as the gold standard, the performance of all three algorithms requiring at least one DMARD prescription was acceptable, with the PPVs of 86.2% to 88.9%. Even with fulfillment of three or more of the ACR RA criteria as the gold standard, the PPVs of our algorithms were moderate to good (72.4% to 73.3%). Given the limitations of the ACR RA classification criteria for clinical practice, it may be more appropriate to use 'RA diagnosis per rheumatologists' as the gold standard.", "Our results indicate that, to accurately identify subjects with RA in health care utilization databases, future research should consider algorithms that link ICD-9 codes to pharmacy claim data.", "ACR: American College of Rheumatology; CI: confidence interval; DMARD: disease-modifying antirheumatic drug; ICD-9: International Classification of Diseases-9th Revision; PACE: Pennsylvania Assistance Contract for the Elderly; PPV: positive predictive value; RA: rheumatoid arthritis; SD: standard deviation.", "DHS has received research support from Amgen (Thousand Oaks, CA, USA) and Abbott (Abbott Park, IL, USA) and support for an educational course from Bristol-Myers Squibb Company (Princeton, NJ, USA). He has non-compensation roles in two drug trials sponsored by Pfizer Inc (New York, NY, USA). The other authors declare that they have no competing interests.", "All authors participated in the study conception. AS and JMP participated in the study design and in data acquisition. JNK participated in the study design and in data analysis and interpretation. DHS participated in the study design and in data acquisition, analysis, and interpretation. SK, MEW, and HM participated in data analysis and interpretation. All authors participated in manuscript preparation and revision. All authors read and approved the final manuscript." ]

[ null, "materials|methods", null, null, null, "results", null, null, "discussion", "conclusions", null, null, null ]

[]

[ "Adult", "Aged", "Aged, 80 and over", "Antibodies, Monoclonal", "Antirheumatic Agents", "Arthritis, Rheumatoid", "Cell Culture Techniques", "Cytokines", "Female", "Humans", "Immunohistochemistry", "Male", "Middle Aged", "Synovial Membrane", "Toll-Like Receptor 2" ]

[ "Introduction", "Patients and RA synovial tissue", "Whole RA synovial tissue explant culture", "Immunohistochemistry", "Isolation and culture of peripheral blood (PBMC) and synovial fluid mononuclear cells (SFMC)", "Cytokine quantification", "Statistical analysis", "Patient characteristics", "Effect of Pam3CSK4 on cytokine production", "Effect of OPN301 on cytokine production in PBMC and SFMC", "OPN301 penetration and its effect on spontaneous cytokine production in RA synovial tissue explant cultures", "Abbreviations", "Competing interests", "Authors' contributions" ]

[ "Rheumatoid arthritis (RA) is a chronic inflammatory disease characterized by synovial inflammation and destruction of cartilage and bone. This process depends on cytokines and growth factors to stimulate cell survival, proliferation and extracellular matrix (ECM) degradation [1]. Activated lymphocytes play a critical role in the initiation and perpetuation of synovial inflammation. Pro-inflammatory cytokines, such as TNF-α and IL-1β, are key mediators of these processes; however, it remains unclear which mechanisms are involved in the initiation and regulation of cytokine production and other tissue-destructive mediators.\nThere is mounting evidence for the involvement of Toll-like receptors (TLRs) in RA [2,3]. Increased expression of TLR2 and TLR4 has been demonstrated in synovial cells and tissue [4-6]. TLR2 expression in RA synovial tissue has been demonstrated at sites of attachment and invasion into cartilage and bone [4], on CD16+ monocytes and synovial macrophages [5]. TLR2 mRNA is upregulated in RA synovial fibroblasts (FLS) by TNFα and IL-1β [4]. Overexpression of dominant negative forms of the essential TLR2/4 adapter molecules MyD88 and Mal/TIRAP inhibits cytokine production and matrix metalloproteinases in RA synovial cells [6]. Furthermore, several animal models use bacterial wall components and peptidoglycans (PG), known to activate TLR2, to induce experimental arthritis [7,8].\nTargeted biologic therapies, including TNF blocking drugs, have had an important effect on the therapeutic outcome of inflammatory arthritis [9]; however, a significant proportion of patients do not respond or have a sub-optimal response highlighting the need for new therapeutic targets. TLR expression on RA cells and their ability to induce pro-inflammatory cytokines, suggest TLRs may play an integral role in the pathogenesis of RA, as such TLRs represent a rational target for therapeutic intervention [3].\nIn the present study, using whole tissue synovial explant cultures ex vivo (which closely reflect the in vivo environment) and RA mononuclear cells, we demonstrate that Pam3CSK4, a TLR1/2 agonist, significantly increases release of key cytokines, an effect that is blocked by an anti-TLR2 antibody, OPN301. In RA synovial explants, we demonstrate that OPN301 penetrates the synovial tissue, localizing to the lining layer and perivascular region and significantly suppresses spontaneous release of pro-inflammatory cytokines. This effect was comparable to that of Adalimumab, a well established TNF blocking therapy. Inhibition of spontaneous pro-inflammatory cytokine production by OPN301 from RA synovial explants in the absence of a specific TLR2 agonist suggests expression of endogenous TLR ligands in RA synovial tissue. These data demonstrate that TLR2 promotes pro-inflammatory and destructive processes in RA and further support the rationale of using a TLR2 therapeutic blockade.", "Patients with RA, classified according to the American College of Rheumatology criteria [10], were recruited from rheumatology outpatient clinics at St. Vincent's University Hospital (SVUH). All patients had an actively inflamed knee joint, despite current or previous therapy. All research was carried out in accordance with the Declaration of Helsinki, following approval by the SVUH ethics committee. All patients gave written informed consent. RA synovial tissue (ST) was obtained at the time of arthroscopy under direct visualization. Blood samples and synovial fluid were collected from patients at arthroscopy or clinics.", "To investigate the effect of OPN301 (a novel mouse IgG1 monoclonal anti-TLR2 antibody, Opsona Therapeutics, (Dublin, Ireland)), on cytokine production in the arthritic joint, an ex vivo RA synovial explant model was established. This system maintains the synovial architecture and cell-cell contact and spontaneously releases pro-inflammatory mediators [11,12]. OPN301 is an Opsona Therapeutics internal designation for the anti-mouse/human TLR2 monoclonal antibody known as T2.5. OPN301 is a mouse IgG1 antibody that selectively inhibits TLR2 signaling [13]. IC50s for cytokines measured ranged from 10 to 30 ng/ml for OPN301 and 1 to 10 ng/ml for Adalimumab (Humira, Abbott Laboratories, Illinois, USA). OPN301 cross reacts with mouse TLR2 but has been shown to be ineffective at inhibiting TLR4 or IL-1 receptor signaling in HEK293 or RAW264.7 cells [14]. OPN301 is capable of inhibiting responses using ligands specific for TLR1/2 (Pam3CSK4) and TLR2/6 responses (FSL1/HKLM) in human and murine cell lines (THP1 CD14 X blue, U937, J774), and also in human, murine and monkey blood (Opsona, unpublished observations).\nFor Pam3CSK4 stimulation experiments, two biopsies were obtained from each patient (n = 11 for IL-6 and n = 8 for IL-8) and each biopsy was sectioned into four separate pieces and cultured in 96-well plates in serum-free RPMI 1640 for 24 hr and then stimulated with Pam3CSK4 (200 ng, 1 μg/ml and 10 μg/ml) for a further 24 hr. For the inhibition experiments, three biopsies were obtained from each patient (n = 13), and then each biopsy was sectioned into four separate pieces and cultured in 96-well plates in full RPMI 1640 medium (10% Foetal Calf Serum, 20 mM of 1 mmole/litre HEPES, streptomycin (100 units/ml), penicillin (100 units/ml) and Fungizone (0.25/μg/ml)) in the presence of OPN301 (1 μg/ml), Adalimumab (1 μg/ml) or IgG isotype control (Mouse IgG1 isotype control, Opsona Therapeutics; 1 μg/ml) for 72 hr at 37°C in air with 5% CO2. Following incubation, supernatants were harvested and frozen at -80°C until further use, and the wet weight of each biopsy section was obtained. Cytokine production was corrected for by wet weight of the biopsy. Biopsies were then snap frozen in optimal cutting temperature (OCT) compound and stored at -80°C for histological analysis. To assess cell viability in RA synovial explant cultures following culture, the acetomethoxy derivate of calcein (calcein AM) was used as a marker of viability. It transports through the cellular membrane of live cells, where intracellular esterases remove the acetomethoxy group resulting in a strong green fluorescence. Following culture, explants were incubated in PBS/Calcein AM (1:1,000 dilution) in the dark for 15 minutes. Biopsies were washed in PBS and examined immediately under a fluorescent microscope. Synovial cells emitted a strong green fluorescence indicative of live cells.", "Synovial biopsies obtained at arthroscopy and RA ST obtained after explant cultures were snap frozen in OCT and stored at -80°C. 7 μm OCT sections were cut with a cryostat, placed on glass slides coated with 2% 3-amino-propyl-triethoxy-silane (Sigma-Aldrich Ireland, Ltd, Dublin, Ireland) and dried overnight at room temperature. Tissue sections were allowed to reach room temperature, fixed in acetone for 10 minutes and air-dried. Non-specific binding and endogenous peroxidase activity was blocked using 10% casein and 0.3% H2O2, respectively. A routine three-stage immunoperoxidase labelling technique incorporating avidin-biotin-immunoperoxidase complex (DAKO, Glostrup, Denmark) was used. For RA biopsies obtained directly from arthroscopy, the sections were incubated with OPN301 at room temperature for one hour. Sections were also incubated with an irrelevant isotype matched mouse monoclonal antibody (mAb) as a negative control. For the RA synovial explant sections that were cultured with OPN301 or matched IgG control antibody, the primary antibody step was omitted in order to examine if tissue penetration of OPN301 mAb could be detected. All sections were incubated with mouse secondary antibody/HRP for 30 minutes, washed in PBS and colour-developed in a solution containing diaminobenzadine-tetrahydrochloride (Sigma, St Louis, MO, USA), 0.5% H2O2 in PBS buffer (pH 7.6). Slides were counterstained with haematoxylin and mounted.", "Blood and synovial fluid were obtained from patients undergoing arthroscopy or at clinics, and drawn into heparin containing tubes. PBMCs/SFMCs were isolated by Ficoll-Metrizoate density gradient centrifugation (Lymphoprep, Nycomed, Marlow Buckinghamshire, UK). Cells were seeded in 48-well plates, at a cell density of 200,000 to 400,000 cells/ml in full RPMI 1640 medium containing Pam3CSK4 (200 ng/ml, 1 μg/ml or 10 μg/ml) in the presence or absence of OPN301 or IgG isotype control for 6 hr. To assess viability of cells after stimulation with Pam3CSK4 and/or OPN301, PBMCs were stimulated with Pam3CSK4 1 μg and/or OPN301 1 μg for 24 hrs. PBMC cell suspension was diluted 1:20 in Ethidium Bromide Acridine Orange solution (EBAO). Viable and non-viable cells were counted using a dual-chamber hemocytometer and a UV-light microscope. No difference for cell viability was observed between control and Pam3CSK or OPN treated cells.", "IL-6, IL-8, IFN-γ, IL-1β, and TNF-α, levels were quantified by Multiplex Tissue Culture kit (Meso Scale Discovery (MSD), Maryland, USA) or by ELISA (R&D Systems, Oxfordshire, UK) according to the manufacturer's instructions. Absorbance was measured at 450 nm in a microtiter plate spectrophometer (Dynatech MR4000, Alexandria, VA, USA) or using MSD Sector Imager 2400.", "SPSS15 system for Windows (SPSS Inc, Chicago, Illinois, USA) was used for statistical analysis. Non-parametric Wilcoxon Signed Rank test for related samples and non-parametric Friedman analysis of variance for comparison of three or more groups were performed. P < 0.05 was determined as statistically significant. Results are expressed as mean ± SEM unless otherwise stated.", "Clinical characteristics of the study patients who underwent arthroscopy are shown in Table 1. All patients had clinically active disease as estimated by high 28-joint count Disease Activity Score (DAS28) (4.1 ± 0.9; mean ± SD) and had a swollen, inflamed knee joint. All patients except one were biologically naive. Eight patients were receiving disease modifying antirheumatic drugs (DMARDs) - either Methotrexate or Plaquenil. In addition, three were receiving oral steroids (prednisilone ≤10 mg) and four patients were receiving NonSteroidal Anti-inflammatory Drugs (NSAIDs) only. One patient had previously received Methotrexate/Adalimumab had stopped treatment due to infection. There was no relationship between response to OPN301 and treatment.\nCharacteristics of RA patients (n = 13)", "To determine the effect of the TLR2 agonist Pam3CSK4 on induction of cytokine secretion in RA cells types, the expression of pro-inflammatory cytokines IL-6 and IL-8 were assessed. Figure 1A and 1B demonstrates in RA SFMC and PBMC, significant induction of IL-6 and IL-8 secretion following Pam3CSK4 (200 ng/ml) stimulation (P < 0.05). Incubation of PBMC and SFMC with Pam3CSK4 at 1 and 10 μg/ml also significantly induced IL-6 and IL-8 production (P < 0.05) (data not shown).\nPam3CSK4 induced IL-6 and IL-8 in mononuclear cells from RA patients. SFMCs (A; n = 6) and PBMCs (B; n = 11) were stimulated with TLR2 agonist Pam3CSK4 at 200 ng/ml. Levels of IL-6 and IL-8 in the culture supernatants were determined and compared to unstimulated cells (Basal) at six hours. Values are expressed as the mean ± SEM. * P < 0.05, significantly different from control as determined using Wilcoxon Signed Rank analysis.\nIn RA synovial biopsies, TLR2 was localized to the lining layer and perivascular regions as demonstrated by staining using OPN301 as a primary antibody (Figure 2A). In paired RA synovial explant cultures, Pam3CSK4, at 200 ng/ml, 1 and 10 μg/ml significantly induced IL-6 and IL-8-cytokine production (Figure 2B). IL-6 production (mean pg/ml/mg ± SEM) increased from 14,934.9 ± 8,337.15 for an unstimulated control (Basal) to 23,783.02 ± 11,425.55 for Pam3CSK4 200 ng, P = 0.005; 66,811.71 ± 25,049.62 for Pam3CSK4 1 μg, P = 0.000 and 53,560.05 ± 13,547.85 for Pam3CSK4 10 μg, P = 0.000, respectively. IL-8 production (mean pg/ml/mg ± SEM) increased from 30,196.87 ± 8,871.838 for unstimulated control (Basal) to 81,931.94 ± 32,328.61 for Pam3CSK4 200 ng, P = 0.012; 154,736.9 ± 43,381.3 for Pam3CSK4 1 μg, P = 0.000 and 140,520.4 ± 28,061.14 for Pam3CSK4 10 μg, P = 0.004, respectively.\nTLR2 expression is localized to the perivascular region and lining layer. (A) Two representative photomicrograph (Pt 1 and Pt 2) of RA synovial tissue of 10 patients stained for TLR2 expression. Expression is localized to the perivascular region and to the lining layer, with no staining observed for matched IgG control. The bar on the lower right hand corner of each photomicrograph represents a distance of 100 μm on the top panel and 50 μm on the bottom panel. (B) RA synovial tissue explant cultures were stimulated with TLR2 angonist Pam3CSK4 at 200 ng, 1 and 10 μg/ml. Levels of IL-6 (n = 11) and IL-8 (n = 8) in the culture supernatants were determined after 24 hrs. Values are expressed as the mean ± SEM. * P < 0.05 significantly different from control as determined using Wilcoxon Signed Rank analysis.", "To assess whether OPN301 inhibits Pam3CSK induced cytokine secretion, RA PBMC and SFMC were stimulated with Pam3CSK4 (200 ng/ml) in the presence or absence of OPN301 or matched IgG isotype control. OPN301 significantly inhibited Pam3CSK4 induced IFN-γ, IL-1β, IL-6, TNF-α and IL-8 in RA SFMCs (all P < 0.05) compared to IgG isotype control (Figure 3). Similar results were obtained for RA PBMCs, where OPN301 significantly inhibited IFN-γ, IL-1β, TNF-α, IL-8, IL-6 cytokine production (all P < 0.05).\nOPN301 significantly inhibited cytokine production in SFMC compared to IgG isotype control. RA SFMCs (n = 6) were stimulated with Pam3CSK4 in the presence or absence of OPN301 (1 μg/ml) or matched IgG isotype control for six hours. Values are expressed as the mean ± SEM. * P < 0.05, significantly different from control as determined using Wilcoxon Signed Rank analysis.", "To show that OPN301 penetrates the RA ST in culture, biopsies were snap frozen following 72 h incubation with OPN301, sectioned and immunohistology performed omitting the primary antibody. Representative images of OPN301 detection in RA synovial explants are shown in Figure 4A. OPN301 was localized to the lining layer and to the perivascular regions, showing that OPN301 does penetrate the tissue in culture with localization consistent with previous studies [4-6].\nOPN301 inhibition of spontaneous cytokine release is comparable to that observed for Adalimumab. (A) One representative image of OPN301 penetration of RA synovial explant cultures of four RA patients. OPN301 was localized to the perivascular and lining layer regions with no staining observed for IgG control. The bar on the lower right hand corner of each photomicrograph represents a distance of 100 μm on the top panel and 50 μm on the bottom panel. (B) OPN301 (1 μg/ml) inhibited spontaneous cytokine release from RA synovial explant cultures (n = 13) and effect that was comparable to that observed anti-TNFα mAb Adalimumab (Hum, 1 μg/ml). Results are expressed as mean ± SEM. * P < 0.05 and ** P < 0.001 significantly different from control as determined using Wilcoxon Signed Rank analysis.\nTo determine the inhibitory effect of OPN301 on spontaneous cytokine production by ST explant cultures, RA ST explants were established and cultured in the presence of OPN301 (1 μg/ml), Adalimumab (1 μg/ml) or IgG isotype control (1 μg/ml) for 72 hr (Figure 4B). OPN301 significantly inhibited spontaneous release of IL-8 (P = 0.001), TNF-α (P = 0.003), IFN-γ (P = 0.013), and IL-1β (P = 0.039). OPN301 inhibited IL-6 production, but this did not reach significance (P = 0.056). Adalimumab significantly inhibited IL-6 (P = 0.002), IL-8 (P = 0.000), IL-1β (P = 0.020), TNF-α (P = 0.018), and IFNγ (P = 0.001). The effect of OPN301 and Adalimumab on spontaneous cytokine production was also analysed using non-parametric Friedman analysis of variance, which further confirmed significant inhibition of all cytokines in response to both antibodies. We categorized the cytokine responses in the tissue cultures as follows, low response: <20% inhibition, moderate response: 21 to 49% inhibition, and good response: >50% inhibition. For OPN301, 38% of patients had a low response, 32% had a moderate response and 30% had a good response, while biopsies treated with Adalimumab showed 37% low response, 31% moderate response and 32% good response. The rate of response reflects what is observed for TNFα in clinical practice, (approximately 30% little or no response, and 70% of patient responses varying between 20 to 100%). Therefore, the results of our ex vivo model very much reflect typical biologic response rates for RA patients in clinical practice, and thus further validates the model as a good screening method for pre-clinical 'proof of concept studies'.", "CRP: C-reactive protein; DAS28: 28-joint count Disease Activity Score; DMARDs: disease-modifying antirheumatic drugs; EBAO: ethidium bromide acridine orange solution; ECM: extracellular matrix; ESR: erthrocyte sedimentation rate; FLS: synovial fibroblasts; mAb: monoclonal antibody; NSAIDs: nonsteroidal anti-inflammatory drugs; PAMPs: pathogen associate molecular patterns; PBMC: peripheral blood mononuclear cells; PG: petidoglycan; RA: rheumatoid arthritis; SFMC: synovial fluid mononuclear cells; ST: synovial tissue; SVUH: St. Vincent's University Hospital; TLR: Toll-like receptor; VAS: visual analog scale.", "DV is in receipt of a research grant from Opsona Therapeutics Ltd. Current Opsona employees (BK, PMcG, MR, WMcC) and ex-Opsona employee JD hold shares in Opsona Therapeutics Ltd; however, the percent involved is so small that Opsona does not view this as a conflicting interest. Luke O'Neill is a founder of Opsona Therapeutics Ltd. PMcG and BK are named inventors on patent WO2009/000929 and Jerome Dellacasagrande is named inventor on patent PCT/EP2010/059667. SNAU, TS, JMcC, MC and UF have no competing interests.", "SNAU conducted most of the experiments and analysis of data. TS, JMcC, MC and UF performed some of the experiments. UF, DV, MC, PMcG, SNAU, JD, BK, WMcC, MR, TS and LO'N, participated in the study design, data analysis and manuscript preparation. UF and DV supervised the research. DV and TS recruited all patients, performed the arthroscopies and provided all clinical information. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Materials and methods", "Patients and RA synovial tissue", "Whole RA synovial tissue explant culture", "Immunohistochemistry", "Isolation and culture of peripheral blood (PBMC) and synovial fluid mononuclear cells (SFMC)", "Cytokine quantification", "Statistical analysis", "Results", "Patient characteristics", "Effect of Pam3CSK4 on cytokine production", "Effect of OPN301 on cytokine production in PBMC and SFMC", "OPN301 penetration and its effect on spontaneous cytokine production in RA synovial tissue explant cultures", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions" ]

[ "Rheumatoid arthritis (RA) is a chronic inflammatory disease characterized by synovial inflammation and destruction of cartilage and bone. This process depends on cytokines and growth factors to stimulate cell survival, proliferation and extracellular matrix (ECM) degradation [1]. Activated lymphocytes play a critical role in the initiation and perpetuation of synovial inflammation. Pro-inflammatory cytokines, such as TNF-α and IL-1β, are key mediators of these processes; however, it remains unclear which mechanisms are involved in the initiation and regulation of cytokine production and other tissue-destructive mediators.\nThere is mounting evidence for the involvement of Toll-like receptors (TLRs) in RA [2,3]. Increased expression of TLR2 and TLR4 has been demonstrated in synovial cells and tissue [4-6]. TLR2 expression in RA synovial tissue has been demonstrated at sites of attachment and invasion into cartilage and bone [4], on CD16+ monocytes and synovial macrophages [5]. TLR2 mRNA is upregulated in RA synovial fibroblasts (FLS) by TNFα and IL-1β [4]. Overexpression of dominant negative forms of the essential TLR2/4 adapter molecules MyD88 and Mal/TIRAP inhibits cytokine production and matrix metalloproteinases in RA synovial cells [6]. Furthermore, several animal models use bacterial wall components and peptidoglycans (PG), known to activate TLR2, to induce experimental arthritis [7,8].\nTargeted biologic therapies, including TNF blocking drugs, have had an important effect on the therapeutic outcome of inflammatory arthritis [9]; however, a significant proportion of patients do not respond or have a sub-optimal response highlighting the need for new therapeutic targets. TLR expression on RA cells and their ability to induce pro-inflammatory cytokines, suggest TLRs may play an integral role in the pathogenesis of RA, as such TLRs represent a rational target for therapeutic intervention [3].\nIn the present study, using whole tissue synovial explant cultures ex vivo (which closely reflect the in vivo environment) and RA mononuclear cells, we demonstrate that Pam3CSK4, a TLR1/2 agonist, significantly increases release of key cytokines, an effect that is blocked by an anti-TLR2 antibody, OPN301. In RA synovial explants, we demonstrate that OPN301 penetrates the synovial tissue, localizing to the lining layer and perivascular region and significantly suppresses spontaneous release of pro-inflammatory cytokines. This effect was comparable to that of Adalimumab, a well established TNF blocking therapy. Inhibition of spontaneous pro-inflammatory cytokine production by OPN301 from RA synovial explants in the absence of a specific TLR2 agonist suggests expression of endogenous TLR ligands in RA synovial tissue. These data demonstrate that TLR2 promotes pro-inflammatory and destructive processes in RA and further support the rationale of using a TLR2 therapeutic blockade.", "[SUBTITLE] Patients and RA synovial tissue [SUBSECTION] Patients with RA, classified according to the American College of Rheumatology criteria [10], were recruited from rheumatology outpatient clinics at St. Vincent's University Hospital (SVUH). All patients had an actively inflamed knee joint, despite current or previous therapy. All research was carried out in accordance with the Declaration of Helsinki, following approval by the SVUH ethics committee. All patients gave written informed consent. RA synovial tissue (ST) was obtained at the time of arthroscopy under direct visualization. Blood samples and synovial fluid were collected from patients at arthroscopy or clinics.\nPatients with RA, classified according to the American College of Rheumatology criteria [10], were recruited from rheumatology outpatient clinics at St. Vincent's University Hospital (SVUH). All patients had an actively inflamed knee joint, despite current or previous therapy. All research was carried out in accordance with the Declaration of Helsinki, following approval by the SVUH ethics committee. All patients gave written informed consent. RA synovial tissue (ST) was obtained at the time of arthroscopy under direct visualization. Blood samples and synovial fluid were collected from patients at arthroscopy or clinics.\n[SUBTITLE] Whole RA synovial tissue explant culture [SUBSECTION] To investigate the effect of OPN301 (a novel mouse IgG1 monoclonal anti-TLR2 antibody, Opsona Therapeutics, (Dublin, Ireland)), on cytokine production in the arthritic joint, an ex vivo RA synovial explant model was established. This system maintains the synovial architecture and cell-cell contact and spontaneously releases pro-inflammatory mediators [11,12]. OPN301 is an Opsona Therapeutics internal designation for the anti-mouse/human TLR2 monoclonal antibody known as T2.5. OPN301 is a mouse IgG1 antibody that selectively inhibits TLR2 signaling [13]. IC50s for cytokines measured ranged from 10 to 30 ng/ml for OPN301 and 1 to 10 ng/ml for Adalimumab (Humira, Abbott Laboratories, Illinois, USA). OPN301 cross reacts with mouse TLR2 but has been shown to be ineffective at inhibiting TLR4 or IL-1 receptor signaling in HEK293 or RAW264.7 cells [14]. OPN301 is capable of inhibiting responses using ligands specific for TLR1/2 (Pam3CSK4) and TLR2/6 responses (FSL1/HKLM) in human and murine cell lines (THP1 CD14 X blue, U937, J774), and also in human, murine and monkey blood (Opsona, unpublished observations).\nFor Pam3CSK4 stimulation experiments, two biopsies were obtained from each patient (n = 11 for IL-6 and n = 8 for IL-8) and each biopsy was sectioned into four separate pieces and cultured in 96-well plates in serum-free RPMI 1640 for 24 hr and then stimulated with Pam3CSK4 (200 ng, 1 μg/ml and 10 μg/ml) for a further 24 hr. For the inhibition experiments, three biopsies were obtained from each patient (n = 13), and then each biopsy was sectioned into four separate pieces and cultured in 96-well plates in full RPMI 1640 medium (10% Foetal Calf Serum, 20 mM of 1 mmole/litre HEPES, streptomycin (100 units/ml), penicillin (100 units/ml) and Fungizone (0.25/μg/ml)) in the presence of OPN301 (1 μg/ml), Adalimumab (1 μg/ml) or IgG isotype control (Mouse IgG1 isotype control, Opsona Therapeutics; 1 μg/ml) for 72 hr at 37°C in air with 5% CO2. Following incubation, supernatants were harvested and frozen at -80°C until further use, and the wet weight of each biopsy section was obtained. Cytokine production was corrected for by wet weight of the biopsy. Biopsies were then snap frozen in optimal cutting temperature (OCT) compound and stored at -80°C for histological analysis. To assess cell viability in RA synovial explant cultures following culture, the acetomethoxy derivate of calcein (calcein AM) was used as a marker of viability. It transports through the cellular membrane of live cells, where intracellular esterases remove the acetomethoxy group resulting in a strong green fluorescence. Following culture, explants were incubated in PBS/Calcein AM (1:1,000 dilution) in the dark for 15 minutes. Biopsies were washed in PBS and examined immediately under a fluorescent microscope. Synovial cells emitted a strong green fluorescence indicative of live cells.\nTo investigate the effect of OPN301 (a novel mouse IgG1 monoclonal anti-TLR2 antibody, Opsona Therapeutics, (Dublin, Ireland)), on cytokine production in the arthritic joint, an ex vivo RA synovial explant model was established. This system maintains the synovial architecture and cell-cell contact and spontaneously releases pro-inflammatory mediators [11,12]. OPN301 is an Opsona Therapeutics internal designation for the anti-mouse/human TLR2 monoclonal antibody known as T2.5. OPN301 is a mouse IgG1 antibody that selectively inhibits TLR2 signaling [13]. IC50s for cytokines measured ranged from 10 to 30 ng/ml for OPN301 and 1 to 10 ng/ml for Adalimumab (Humira, Abbott Laboratories, Illinois, USA). OPN301 cross reacts with mouse TLR2 but has been shown to be ineffective at inhibiting TLR4 or IL-1 receptor signaling in HEK293 or RAW264.7 cells [14]. OPN301 is capable of inhibiting responses using ligands specific for TLR1/2 (Pam3CSK4) and TLR2/6 responses (FSL1/HKLM) in human and murine cell lines (THP1 CD14 X blue, U937, J774), and also in human, murine and monkey blood (Opsona, unpublished observations).\nFor Pam3CSK4 stimulation experiments, two biopsies were obtained from each patient (n = 11 for IL-6 and n = 8 for IL-8) and each biopsy was sectioned into four separate pieces and cultured in 96-well plates in serum-free RPMI 1640 for 24 hr and then stimulated with Pam3CSK4 (200 ng, 1 μg/ml and 10 μg/ml) for a further 24 hr. For the inhibition experiments, three biopsies were obtained from each patient (n = 13), and then each biopsy was sectioned into four separate pieces and cultured in 96-well plates in full RPMI 1640 medium (10% Foetal Calf Serum, 20 mM of 1 mmole/litre HEPES, streptomycin (100 units/ml), penicillin (100 units/ml) and Fungizone (0.25/μg/ml)) in the presence of OPN301 (1 μg/ml), Adalimumab (1 μg/ml) or IgG isotype control (Mouse IgG1 isotype control, Opsona Therapeutics; 1 μg/ml) for 72 hr at 37°C in air with 5% CO2. Following incubation, supernatants were harvested and frozen at -80°C until further use, and the wet weight of each biopsy section was obtained. Cytokine production was corrected for by wet weight of the biopsy. Biopsies were then snap frozen in optimal cutting temperature (OCT) compound and stored at -80°C for histological analysis. To assess cell viability in RA synovial explant cultures following culture, the acetomethoxy derivate of calcein (calcein AM) was used as a marker of viability. It transports through the cellular membrane of live cells, where intracellular esterases remove the acetomethoxy group resulting in a strong green fluorescence. Following culture, explants were incubated in PBS/Calcein AM (1:1,000 dilution) in the dark for 15 minutes. Biopsies were washed in PBS and examined immediately under a fluorescent microscope. Synovial cells emitted a strong green fluorescence indicative of live cells.\n[SUBTITLE] Immunohistochemistry [SUBSECTION] Synovial biopsies obtained at arthroscopy and RA ST obtained after explant cultures were snap frozen in OCT and stored at -80°C. 7 μm OCT sections were cut with a cryostat, placed on glass slides coated with 2% 3-amino-propyl-triethoxy-silane (Sigma-Aldrich Ireland, Ltd, Dublin, Ireland) and dried overnight at room temperature. Tissue sections were allowed to reach room temperature, fixed in acetone for 10 minutes and air-dried. Non-specific binding and endogenous peroxidase activity was blocked using 10% casein and 0.3% H2O2, respectively. A routine three-stage immunoperoxidase labelling technique incorporating avidin-biotin-immunoperoxidase complex (DAKO, Glostrup, Denmark) was used. For RA biopsies obtained directly from arthroscopy, the sections were incubated with OPN301 at room temperature for one hour. Sections were also incubated with an irrelevant isotype matched mouse monoclonal antibody (mAb) as a negative control. For the RA synovial explant sections that were cultured with OPN301 or matched IgG control antibody, the primary antibody step was omitted in order to examine if tissue penetration of OPN301 mAb could be detected. All sections were incubated with mouse secondary antibody/HRP for 30 minutes, washed in PBS and colour-developed in a solution containing diaminobenzadine-tetrahydrochloride (Sigma, St Louis, MO, USA), 0.5% H2O2 in PBS buffer (pH 7.6). Slides were counterstained with haematoxylin and mounted.\nSynovial biopsies obtained at arthroscopy and RA ST obtained after explant cultures were snap frozen in OCT and stored at -80°C. 7 μm OCT sections were cut with a cryostat, placed on glass slides coated with 2% 3-amino-propyl-triethoxy-silane (Sigma-Aldrich Ireland, Ltd, Dublin, Ireland) and dried overnight at room temperature. Tissue sections were allowed to reach room temperature, fixed in acetone for 10 minutes and air-dried. Non-specific binding and endogenous peroxidase activity was blocked using 10% casein and 0.3% H2O2, respectively. A routine three-stage immunoperoxidase labelling technique incorporating avidin-biotin-immunoperoxidase complex (DAKO, Glostrup, Denmark) was used. For RA biopsies obtained directly from arthroscopy, the sections were incubated with OPN301 at room temperature for one hour. Sections were also incubated with an irrelevant isotype matched mouse monoclonal antibody (mAb) as a negative control. For the RA synovial explant sections that were cultured with OPN301 or matched IgG control antibody, the primary antibody step was omitted in order to examine if tissue penetration of OPN301 mAb could be detected. All sections were incubated with mouse secondary antibody/HRP for 30 minutes, washed in PBS and colour-developed in a solution containing diaminobenzadine-tetrahydrochloride (Sigma, St Louis, MO, USA), 0.5% H2O2 in PBS buffer (pH 7.6). Slides were counterstained with haematoxylin and mounted.\n[SUBTITLE] Isolation and culture of peripheral blood (PBMC) and synovial fluid mononuclear cells (SFMC) [SUBSECTION] Blood and synovial fluid were obtained from patients undergoing arthroscopy or at clinics, and drawn into heparin containing tubes. PBMCs/SFMCs were isolated by Ficoll-Metrizoate density gradient centrifugation (Lymphoprep, Nycomed, Marlow Buckinghamshire, UK). Cells were seeded in 48-well plates, at a cell density of 200,000 to 400,000 cells/ml in full RPMI 1640 medium containing Pam3CSK4 (200 ng/ml, 1 μg/ml or 10 μg/ml) in the presence or absence of OPN301 or IgG isotype control for 6 hr. To assess viability of cells after stimulation with Pam3CSK4 and/or OPN301, PBMCs were stimulated with Pam3CSK4 1 μg and/or OPN301 1 μg for 24 hrs. PBMC cell suspension was diluted 1:20 in Ethidium Bromide Acridine Orange solution (EBAO). Viable and non-viable cells were counted using a dual-chamber hemocytometer and a UV-light microscope. No difference for cell viability was observed between control and Pam3CSK or OPN treated cells.\nBlood and synovial fluid were obtained from patients undergoing arthroscopy or at clinics, and drawn into heparin containing tubes. PBMCs/SFMCs were isolated by Ficoll-Metrizoate density gradient centrifugation (Lymphoprep, Nycomed, Marlow Buckinghamshire, UK). Cells were seeded in 48-well plates, at a cell density of 200,000 to 400,000 cells/ml in full RPMI 1640 medium containing Pam3CSK4 (200 ng/ml, 1 μg/ml or 10 μg/ml) in the presence or absence of OPN301 or IgG isotype control for 6 hr. To assess viability of cells after stimulation with Pam3CSK4 and/or OPN301, PBMCs were stimulated with Pam3CSK4 1 μg and/or OPN301 1 μg for 24 hrs. PBMC cell suspension was diluted 1:20 in Ethidium Bromide Acridine Orange solution (EBAO). Viable and non-viable cells were counted using a dual-chamber hemocytometer and a UV-light microscope. No difference for cell viability was observed between control and Pam3CSK or OPN treated cells.\n[SUBTITLE] Cytokine quantification [SUBSECTION] IL-6, IL-8, IFN-γ, IL-1β, and TNF-α, levels were quantified by Multiplex Tissue Culture kit (Meso Scale Discovery (MSD), Maryland, USA) or by ELISA (R&D Systems, Oxfordshire, UK) according to the manufacturer's instructions. Absorbance was measured at 450 nm in a microtiter plate spectrophometer (Dynatech MR4000, Alexandria, VA, USA) or using MSD Sector Imager 2400.\nIL-6, IL-8, IFN-γ, IL-1β, and TNF-α, levels were quantified by Multiplex Tissue Culture kit (Meso Scale Discovery (MSD), Maryland, USA) or by ELISA (R&D Systems, Oxfordshire, UK) according to the manufacturer's instructions. Absorbance was measured at 450 nm in a microtiter plate spectrophometer (Dynatech MR4000, Alexandria, VA, USA) or using MSD Sector Imager 2400.\n[SUBTITLE] Statistical analysis [SUBSECTION] SPSS15 system for Windows (SPSS Inc, Chicago, Illinois, USA) was used for statistical analysis. Non-parametric Wilcoxon Signed Rank test for related samples and non-parametric Friedman analysis of variance for comparison of three or more groups were performed. P < 0.05 was determined as statistically significant. Results are expressed as mean ± SEM unless otherwise stated.\nSPSS15 system for Windows (SPSS Inc, Chicago, Illinois, USA) was used for statistical analysis. Non-parametric Wilcoxon Signed Rank test for related samples and non-parametric Friedman analysis of variance for comparison of three or more groups were performed. P < 0.05 was determined as statistically significant. Results are expressed as mean ± SEM unless otherwise stated.", "Patients with RA, classified according to the American College of Rheumatology criteria [10], were recruited from rheumatology outpatient clinics at St. Vincent's University Hospital (SVUH). All patients had an actively inflamed knee joint, despite current or previous therapy. All research was carried out in accordance with the Declaration of Helsinki, following approval by the SVUH ethics committee. All patients gave written informed consent. RA synovial tissue (ST) was obtained at the time of arthroscopy under direct visualization. Blood samples and synovial fluid were collected from patients at arthroscopy or clinics.", "To investigate the effect of OPN301 (a novel mouse IgG1 monoclonal anti-TLR2 antibody, Opsona Therapeutics, (Dublin, Ireland)), on cytokine production in the arthritic joint, an ex vivo RA synovial explant model was established. This system maintains the synovial architecture and cell-cell contact and spontaneously releases pro-inflammatory mediators [11,12]. OPN301 is an Opsona Therapeutics internal designation for the anti-mouse/human TLR2 monoclonal antibody known as T2.5. OPN301 is a mouse IgG1 antibody that selectively inhibits TLR2 signaling [13]. IC50s for cytokines measured ranged from 10 to 30 ng/ml for OPN301 and 1 to 10 ng/ml for Adalimumab (Humira, Abbott Laboratories, Illinois, USA). OPN301 cross reacts with mouse TLR2 but has been shown to be ineffective at inhibiting TLR4 or IL-1 receptor signaling in HEK293 or RAW264.7 cells [14]. OPN301 is capable of inhibiting responses using ligands specific for TLR1/2 (Pam3CSK4) and TLR2/6 responses (FSL1/HKLM) in human and murine cell lines (THP1 CD14 X blue, U937, J774), and also in human, murine and monkey blood (Opsona, unpublished observations).\nFor Pam3CSK4 stimulation experiments, two biopsies were obtained from each patient (n = 11 for IL-6 and n = 8 for IL-8) and each biopsy was sectioned into four separate pieces and cultured in 96-well plates in serum-free RPMI 1640 for 24 hr and then stimulated with Pam3CSK4 (200 ng, 1 μg/ml and 10 μg/ml) for a further 24 hr. For the inhibition experiments, three biopsies were obtained from each patient (n = 13), and then each biopsy was sectioned into four separate pieces and cultured in 96-well plates in full RPMI 1640 medium (10% Foetal Calf Serum, 20 mM of 1 mmole/litre HEPES, streptomycin (100 units/ml), penicillin (100 units/ml) and Fungizone (0.25/μg/ml)) in the presence of OPN301 (1 μg/ml), Adalimumab (1 μg/ml) or IgG isotype control (Mouse IgG1 isotype control, Opsona Therapeutics; 1 μg/ml) for 72 hr at 37°C in air with 5% CO2. Following incubation, supernatants were harvested and frozen at -80°C until further use, and the wet weight of each biopsy section was obtained. Cytokine production was corrected for by wet weight of the biopsy. Biopsies were then snap frozen in optimal cutting temperature (OCT) compound and stored at -80°C for histological analysis. To assess cell viability in RA synovial explant cultures following culture, the acetomethoxy derivate of calcein (calcein AM) was used as a marker of viability. It transports through the cellular membrane of live cells, where intracellular esterases remove the acetomethoxy group resulting in a strong green fluorescence. Following culture, explants were incubated in PBS/Calcein AM (1:1,000 dilution) in the dark for 15 minutes. Biopsies were washed in PBS and examined immediately under a fluorescent microscope. Synovial cells emitted a strong green fluorescence indicative of live cells.", "Synovial biopsies obtained at arthroscopy and RA ST obtained after explant cultures were snap frozen in OCT and stored at -80°C. 7 μm OCT sections were cut with a cryostat, placed on glass slides coated with 2% 3-amino-propyl-triethoxy-silane (Sigma-Aldrich Ireland, Ltd, Dublin, Ireland) and dried overnight at room temperature. Tissue sections were allowed to reach room temperature, fixed in acetone for 10 minutes and air-dried. Non-specific binding and endogenous peroxidase activity was blocked using 10% casein and 0.3% H2O2, respectively. A routine three-stage immunoperoxidase labelling technique incorporating avidin-biotin-immunoperoxidase complex (DAKO, Glostrup, Denmark) was used. For RA biopsies obtained directly from arthroscopy, the sections were incubated with OPN301 at room temperature for one hour. Sections were also incubated with an irrelevant isotype matched mouse monoclonal antibody (mAb) as a negative control. For the RA synovial explant sections that were cultured with OPN301 or matched IgG control antibody, the primary antibody step was omitted in order to examine if tissue penetration of OPN301 mAb could be detected. All sections were incubated with mouse secondary antibody/HRP for 30 minutes, washed in PBS and colour-developed in a solution containing diaminobenzadine-tetrahydrochloride (Sigma, St Louis, MO, USA), 0.5% H2O2 in PBS buffer (pH 7.6). Slides were counterstained with haematoxylin and mounted.", "Blood and synovial fluid were obtained from patients undergoing arthroscopy or at clinics, and drawn into heparin containing tubes. PBMCs/SFMCs were isolated by Ficoll-Metrizoate density gradient centrifugation (Lymphoprep, Nycomed, Marlow Buckinghamshire, UK). Cells were seeded in 48-well plates, at a cell density of 200,000 to 400,000 cells/ml in full RPMI 1640 medium containing Pam3CSK4 (200 ng/ml, 1 μg/ml or 10 μg/ml) in the presence or absence of OPN301 or IgG isotype control for 6 hr. To assess viability of cells after stimulation with Pam3CSK4 and/or OPN301, PBMCs were stimulated with Pam3CSK4 1 μg and/or OPN301 1 μg for 24 hrs. PBMC cell suspension was diluted 1:20 in Ethidium Bromide Acridine Orange solution (EBAO). Viable and non-viable cells were counted using a dual-chamber hemocytometer and a UV-light microscope. No difference for cell viability was observed between control and Pam3CSK or OPN treated cells.", "IL-6, IL-8, IFN-γ, IL-1β, and TNF-α, levels were quantified by Multiplex Tissue Culture kit (Meso Scale Discovery (MSD), Maryland, USA) or by ELISA (R&D Systems, Oxfordshire, UK) according to the manufacturer's instructions. Absorbance was measured at 450 nm in a microtiter plate spectrophometer (Dynatech MR4000, Alexandria, VA, USA) or using MSD Sector Imager 2400.", "SPSS15 system for Windows (SPSS Inc, Chicago, Illinois, USA) was used for statistical analysis. Non-parametric Wilcoxon Signed Rank test for related samples and non-parametric Friedman analysis of variance for comparison of three or more groups were performed. P < 0.05 was determined as statistically significant. Results are expressed as mean ± SEM unless otherwise stated.", "[SUBTITLE] Patient characteristics [SUBSECTION] Clinical characteristics of the study patients who underwent arthroscopy are shown in Table 1. All patients had clinically active disease as estimated by high 28-joint count Disease Activity Score (DAS28) (4.1 ± 0.9; mean ± SD) and had a swollen, inflamed knee joint. All patients except one were biologically naive. Eight patients were receiving disease modifying antirheumatic drugs (DMARDs) - either Methotrexate or Plaquenil. In addition, three were receiving oral steroids (prednisilone ≤10 mg) and four patients were receiving NonSteroidal Anti-inflammatory Drugs (NSAIDs) only. One patient had previously received Methotrexate/Adalimumab had stopped treatment due to infection. There was no relationship between response to OPN301 and treatment.\nCharacteristics of RA patients (n = 13)\nClinical characteristics of the study patients who underwent arthroscopy are shown in Table 1. All patients had clinically active disease as estimated by high 28-joint count Disease Activity Score (DAS28) (4.1 ± 0.9; mean ± SD) and had a swollen, inflamed knee joint. All patients except one were biologically naive. Eight patients were receiving disease modifying antirheumatic drugs (DMARDs) - either Methotrexate or Plaquenil. In addition, three were receiving oral steroids (prednisilone ≤10 mg) and four patients were receiving NonSteroidal Anti-inflammatory Drugs (NSAIDs) only. One patient had previously received Methotrexate/Adalimumab had stopped treatment due to infection. There was no relationship between response to OPN301 and treatment.\nCharacteristics of RA patients (n = 13)\n[SUBTITLE] Effect of Pam3CSK4 on cytokine production [SUBSECTION] To determine the effect of the TLR2 agonist Pam3CSK4 on induction of cytokine secretion in RA cells types, the expression of pro-inflammatory cytokines IL-6 and IL-8 were assessed. Figure 1A and 1B demonstrates in RA SFMC and PBMC, significant induction of IL-6 and IL-8 secretion following Pam3CSK4 (200 ng/ml) stimulation (P < 0.05). Incubation of PBMC and SFMC with Pam3CSK4 at 1 and 10 μg/ml also significantly induced IL-6 and IL-8 production (P < 0.05) (data not shown).\nPam3CSK4 induced IL-6 and IL-8 in mononuclear cells from RA patients. SFMCs (A; n = 6) and PBMCs (B; n = 11) were stimulated with TLR2 agonist Pam3CSK4 at 200 ng/ml. Levels of IL-6 and IL-8 in the culture supernatants were determined and compared to unstimulated cells (Basal) at six hours. Values are expressed as the mean ± SEM. * P < 0.05, significantly different from control as determined using Wilcoxon Signed Rank analysis.\nIn RA synovial biopsies, TLR2 was localized to the lining layer and perivascular regions as demonstrated by staining using OPN301 as a primary antibody (Figure 2A). In paired RA synovial explant cultures, Pam3CSK4, at 200 ng/ml, 1 and 10 μg/ml significantly induced IL-6 and IL-8-cytokine production (Figure 2B). IL-6 production (mean pg/ml/mg ± SEM) increased from 14,934.9 ± 8,337.15 for an unstimulated control (Basal) to 23,783.02 ± 11,425.55 for Pam3CSK4 200 ng, P = 0.005; 66,811.71 ± 25,049.62 for Pam3CSK4 1 μg, P = 0.000 and 53,560.05 ± 13,547.85 for Pam3CSK4 10 μg, P = 0.000, respectively. IL-8 production (mean pg/ml/mg ± SEM) increased from 30,196.87 ± 8,871.838 for unstimulated control (Basal) to 81,931.94 ± 32,328.61 for Pam3CSK4 200 ng, P = 0.012; 154,736.9 ± 43,381.3 for Pam3CSK4 1 μg, P = 0.000 and 140,520.4 ± 28,061.14 for Pam3CSK4 10 μg, P = 0.004, respectively.\nTLR2 expression is localized to the perivascular region and lining layer. (A) Two representative photomicrograph (Pt 1 and Pt 2) of RA synovial tissue of 10 patients stained for TLR2 expression. Expression is localized to the perivascular region and to the lining layer, with no staining observed for matched IgG control. The bar on the lower right hand corner of each photomicrograph represents a distance of 100 μm on the top panel and 50 μm on the bottom panel. (B) RA synovial tissue explant cultures were stimulated with TLR2 angonist Pam3CSK4 at 200 ng, 1 and 10 μg/ml. Levels of IL-6 (n = 11) and IL-8 (n = 8) in the culture supernatants were determined after 24 hrs. Values are expressed as the mean ± SEM. * P < 0.05 significantly different from control as determined using Wilcoxon Signed Rank analysis.\nTo determine the effect of the TLR2 agonist Pam3CSK4 on induction of cytokine secretion in RA cells types, the expression of pro-inflammatory cytokines IL-6 and IL-8 were assessed. Figure 1A and 1B demonstrates in RA SFMC and PBMC, significant induction of IL-6 and IL-8 secretion following Pam3CSK4 (200 ng/ml) stimulation (P < 0.05). Incubation of PBMC and SFMC with Pam3CSK4 at 1 and 10 μg/ml also significantly induced IL-6 and IL-8 production (P < 0.05) (data not shown).\nPam3CSK4 induced IL-6 and IL-8 in mononuclear cells from RA patients. SFMCs (A; n = 6) and PBMCs (B; n = 11) were stimulated with TLR2 agonist Pam3CSK4 at 200 ng/ml. Levels of IL-6 and IL-8 in the culture supernatants were determined and compared to unstimulated cells (Basal) at six hours. Values are expressed as the mean ± SEM. * P < 0.05, significantly different from control as determined using Wilcoxon Signed Rank analysis.\nIn RA synovial biopsies, TLR2 was localized to the lining layer and perivascular regions as demonstrated by staining using OPN301 as a primary antibody (Figure 2A). In paired RA synovial explant cultures, Pam3CSK4, at 200 ng/ml, 1 and 10 μg/ml significantly induced IL-6 and IL-8-cytokine production (Figure 2B). IL-6 production (mean pg/ml/mg ± SEM) increased from 14,934.9 ± 8,337.15 for an unstimulated control (Basal) to 23,783.02 ± 11,425.55 for Pam3CSK4 200 ng, P = 0.005; 66,811.71 ± 25,049.62 for Pam3CSK4 1 μg, P = 0.000 and 53,560.05 ± 13,547.85 for Pam3CSK4 10 μg, P = 0.000, respectively. IL-8 production (mean pg/ml/mg ± SEM) increased from 30,196.87 ± 8,871.838 for unstimulated control (Basal) to 81,931.94 ± 32,328.61 for Pam3CSK4 200 ng, P = 0.012; 154,736.9 ± 43,381.3 for Pam3CSK4 1 μg, P = 0.000 and 140,520.4 ± 28,061.14 for Pam3CSK4 10 μg, P = 0.004, respectively.\nTLR2 expression is localized to the perivascular region and lining layer. (A) Two representative photomicrograph (Pt 1 and Pt 2) of RA synovial tissue of 10 patients stained for TLR2 expression. Expression is localized to the perivascular region and to the lining layer, with no staining observed for matched IgG control. The bar on the lower right hand corner of each photomicrograph represents a distance of 100 μm on the top panel and 50 μm on the bottom panel. (B) RA synovial tissue explant cultures were stimulated with TLR2 angonist Pam3CSK4 at 200 ng, 1 and 10 μg/ml. Levels of IL-6 (n = 11) and IL-8 (n = 8) in the culture supernatants were determined after 24 hrs. Values are expressed as the mean ± SEM. * P < 0.05 significantly different from control as determined using Wilcoxon Signed Rank analysis.\n[SUBTITLE] Effect of OPN301 on cytokine production in PBMC and SFMC [SUBSECTION] To assess whether OPN301 inhibits Pam3CSK induced cytokine secretion, RA PBMC and SFMC were stimulated with Pam3CSK4 (200 ng/ml) in the presence or absence of OPN301 or matched IgG isotype control. OPN301 significantly inhibited Pam3CSK4 induced IFN-γ, IL-1β, IL-6, TNF-α and IL-8 in RA SFMCs (all P < 0.05) compared to IgG isotype control (Figure 3). Similar results were obtained for RA PBMCs, where OPN301 significantly inhibited IFN-γ, IL-1β, TNF-α, IL-8, IL-6 cytokine production (all P < 0.05).\nOPN301 significantly inhibited cytokine production in SFMC compared to IgG isotype control. RA SFMCs (n = 6) were stimulated with Pam3CSK4 in the presence or absence of OPN301 (1 μg/ml) or matched IgG isotype control for six hours. Values are expressed as the mean ± SEM. * P < 0.05, significantly different from control as determined using Wilcoxon Signed Rank analysis.\nTo assess whether OPN301 inhibits Pam3CSK induced cytokine secretion, RA PBMC and SFMC were stimulated with Pam3CSK4 (200 ng/ml) in the presence or absence of OPN301 or matched IgG isotype control. OPN301 significantly inhibited Pam3CSK4 induced IFN-γ, IL-1β, IL-6, TNF-α and IL-8 in RA SFMCs (all P < 0.05) compared to IgG isotype control (Figure 3). Similar results were obtained for RA PBMCs, where OPN301 significantly inhibited IFN-γ, IL-1β, TNF-α, IL-8, IL-6 cytokine production (all P < 0.05).\nOPN301 significantly inhibited cytokine production in SFMC compared to IgG isotype control. RA SFMCs (n = 6) were stimulated with Pam3CSK4 in the presence or absence of OPN301 (1 μg/ml) or matched IgG isotype control for six hours. Values are expressed as the mean ± SEM. * P < 0.05, significantly different from control as determined using Wilcoxon Signed Rank analysis.\n[SUBTITLE] OPN301 penetration and its effect on spontaneous cytokine production in RA synovial tissue explant cultures [SUBSECTION] To show that OPN301 penetrates the RA ST in culture, biopsies were snap frozen following 72 h incubation with OPN301, sectioned and immunohistology performed omitting the primary antibody. Representative images of OPN301 detection in RA synovial explants are shown in Figure 4A. OPN301 was localized to the lining layer and to the perivascular regions, showing that OPN301 does penetrate the tissue in culture with localization consistent with previous studies [4-6].\nOPN301 inhibition of spontaneous cytokine release is comparable to that observed for Adalimumab. (A) One representative image of OPN301 penetration of RA synovial explant cultures of four RA patients. OPN301 was localized to the perivascular and lining layer regions with no staining observed for IgG control. The bar on the lower right hand corner of each photomicrograph represents a distance of 100 μm on the top panel and 50 μm on the bottom panel. (B) OPN301 (1 μg/ml) inhibited spontaneous cytokine release from RA synovial explant cultures (n = 13) and effect that was comparable to that observed anti-TNFα mAb Adalimumab (Hum, 1 μg/ml). Results are expressed as mean ± SEM. * P < 0.05 and ** P < 0.001 significantly different from control as determined using Wilcoxon Signed Rank analysis.\nTo determine the inhibitory effect of OPN301 on spontaneous cytokine production by ST explant cultures, RA ST explants were established and cultured in the presence of OPN301 (1 μg/ml), Adalimumab (1 μg/ml) or IgG isotype control (1 μg/ml) for 72 hr (Figure 4B). OPN301 significantly inhibited spontaneous release of IL-8 (P = 0.001), TNF-α (P = 0.003), IFN-γ (P = 0.013), and IL-1β (P = 0.039). OPN301 inhibited IL-6 production, but this did not reach significance (P = 0.056). Adalimumab significantly inhibited IL-6 (P = 0.002), IL-8 (P = 0.000), IL-1β (P = 0.020), TNF-α (P = 0.018), and IFNγ (P = 0.001). The effect of OPN301 and Adalimumab on spontaneous cytokine production was also analysed using non-parametric Friedman analysis of variance, which further confirmed significant inhibition of all cytokines in response to both antibodies. We categorized the cytokine responses in the tissue cultures as follows, low response: <20% inhibition, moderate response: 21 to 49% inhibition, and good response: >50% inhibition. For OPN301, 38% of patients had a low response, 32% had a moderate response and 30% had a good response, while biopsies treated with Adalimumab showed 37% low response, 31% moderate response and 32% good response. The rate of response reflects what is observed for TNFα in clinical practice, (approximately 30% little or no response, and 70% of patient responses varying between 20 to 100%). Therefore, the results of our ex vivo model very much reflect typical biologic response rates for RA patients in clinical practice, and thus further validates the model as a good screening method for pre-clinical 'proof of concept studies'.\nTo show that OPN301 penetrates the RA ST in culture, biopsies were snap frozen following 72 h incubation with OPN301, sectioned and immunohistology performed omitting the primary antibody. Representative images of OPN301 detection in RA synovial explants are shown in Figure 4A. OPN301 was localized to the lining layer and to the perivascular regions, showing that OPN301 does penetrate the tissue in culture with localization consistent with previous studies [4-6].\nOPN301 inhibition of spontaneous cytokine release is comparable to that observed for Adalimumab. (A) One representative image of OPN301 penetration of RA synovial explant cultures of four RA patients. OPN301 was localized to the perivascular and lining layer regions with no staining observed for IgG control. The bar on the lower right hand corner of each photomicrograph represents a distance of 100 μm on the top panel and 50 μm on the bottom panel. (B) OPN301 (1 μg/ml) inhibited spontaneous cytokine release from RA synovial explant cultures (n = 13) and effect that was comparable to that observed anti-TNFα mAb Adalimumab (Hum, 1 μg/ml). Results are expressed as mean ± SEM. * P < 0.05 and ** P < 0.001 significantly different from control as determined using Wilcoxon Signed Rank analysis.\nTo determine the inhibitory effect of OPN301 on spontaneous cytokine production by ST explant cultures, RA ST explants were established and cultured in the presence of OPN301 (1 μg/ml), Adalimumab (1 μg/ml) or IgG isotype control (1 μg/ml) for 72 hr (Figure 4B). OPN301 significantly inhibited spontaneous release of IL-8 (P = 0.001), TNF-α (P = 0.003), IFN-γ (P = 0.013), and IL-1β (P = 0.039). OPN301 inhibited IL-6 production, but this did not reach significance (P = 0.056). Adalimumab significantly inhibited IL-6 (P = 0.002), IL-8 (P = 0.000), IL-1β (P = 0.020), TNF-α (P = 0.018), and IFNγ (P = 0.001). The effect of OPN301 and Adalimumab on spontaneous cytokine production was also analysed using non-parametric Friedman analysis of variance, which further confirmed significant inhibition of all cytokines in response to both antibodies. We categorized the cytokine responses in the tissue cultures as follows, low response: <20% inhibition, moderate response: 21 to 49% inhibition, and good response: >50% inhibition. For OPN301, 38% of patients had a low response, 32% had a moderate response and 30% had a good response, while biopsies treated with Adalimumab showed 37% low response, 31% moderate response and 32% good response. The rate of response reflects what is observed for TNFα in clinical practice, (approximately 30% little or no response, and 70% of patient responses varying between 20 to 100%). Therefore, the results of our ex vivo model very much reflect typical biologic response rates for RA patients in clinical practice, and thus further validates the model as a good screening method for pre-clinical 'proof of concept studies'.", "Clinical characteristics of the study patients who underwent arthroscopy are shown in Table 1. All patients had clinically active disease as estimated by high 28-joint count Disease Activity Score (DAS28) (4.1 ± 0.9; mean ± SD) and had a swollen, inflamed knee joint. All patients except one were biologically naive. Eight patients were receiving disease modifying antirheumatic drugs (DMARDs) - either Methotrexate or Plaquenil. In addition, three were receiving oral steroids (prednisilone ≤10 mg) and four patients were receiving NonSteroidal Anti-inflammatory Drugs (NSAIDs) only. One patient had previously received Methotrexate/Adalimumab had stopped treatment due to infection. There was no relationship between response to OPN301 and treatment.\nCharacteristics of RA patients (n = 13)", "To determine the effect of the TLR2 agonist Pam3CSK4 on induction of cytokine secretion in RA cells types, the expression of pro-inflammatory cytokines IL-6 and IL-8 were assessed. Figure 1A and 1B demonstrates in RA SFMC and PBMC, significant induction of IL-6 and IL-8 secretion following Pam3CSK4 (200 ng/ml) stimulation (P < 0.05). Incubation of PBMC and SFMC with Pam3CSK4 at 1 and 10 μg/ml also significantly induced IL-6 and IL-8 production (P < 0.05) (data not shown).\nPam3CSK4 induced IL-6 and IL-8 in mononuclear cells from RA patients. SFMCs (A; n = 6) and PBMCs (B; n = 11) were stimulated with TLR2 agonist Pam3CSK4 at 200 ng/ml. Levels of IL-6 and IL-8 in the culture supernatants were determined and compared to unstimulated cells (Basal) at six hours. Values are expressed as the mean ± SEM. * P < 0.05, significantly different from control as determined using Wilcoxon Signed Rank analysis.\nIn RA synovial biopsies, TLR2 was localized to the lining layer and perivascular regions as demonstrated by staining using OPN301 as a primary antibody (Figure 2A). In paired RA synovial explant cultures, Pam3CSK4, at 200 ng/ml, 1 and 10 μg/ml significantly induced IL-6 and IL-8-cytokine production (Figure 2B). IL-6 production (mean pg/ml/mg ± SEM) increased from 14,934.9 ± 8,337.15 for an unstimulated control (Basal) to 23,783.02 ± 11,425.55 for Pam3CSK4 200 ng, P = 0.005; 66,811.71 ± 25,049.62 for Pam3CSK4 1 μg, P = 0.000 and 53,560.05 ± 13,547.85 for Pam3CSK4 10 μg, P = 0.000, respectively. IL-8 production (mean pg/ml/mg ± SEM) increased from 30,196.87 ± 8,871.838 for unstimulated control (Basal) to 81,931.94 ± 32,328.61 for Pam3CSK4 200 ng, P = 0.012; 154,736.9 ± 43,381.3 for Pam3CSK4 1 μg, P = 0.000 and 140,520.4 ± 28,061.14 for Pam3CSK4 10 μg, P = 0.004, respectively.\nTLR2 expression is localized to the perivascular region and lining layer. (A) Two representative photomicrograph (Pt 1 and Pt 2) of RA synovial tissue of 10 patients stained for TLR2 expression. Expression is localized to the perivascular region and to the lining layer, with no staining observed for matched IgG control. The bar on the lower right hand corner of each photomicrograph represents a distance of 100 μm on the top panel and 50 μm on the bottom panel. (B) RA synovial tissue explant cultures were stimulated with TLR2 angonist Pam3CSK4 at 200 ng, 1 and 10 μg/ml. Levels of IL-6 (n = 11) and IL-8 (n = 8) in the culture supernatants were determined after 24 hrs. Values are expressed as the mean ± SEM. * P < 0.05 significantly different from control as determined using Wilcoxon Signed Rank analysis.", "To assess whether OPN301 inhibits Pam3CSK induced cytokine secretion, RA PBMC and SFMC were stimulated with Pam3CSK4 (200 ng/ml) in the presence or absence of OPN301 or matched IgG isotype control. OPN301 significantly inhibited Pam3CSK4 induced IFN-γ, IL-1β, IL-6, TNF-α and IL-8 in RA SFMCs (all P < 0.05) compared to IgG isotype control (Figure 3). Similar results were obtained for RA PBMCs, where OPN301 significantly inhibited IFN-γ, IL-1β, TNF-α, IL-8, IL-6 cytokine production (all P < 0.05).\nOPN301 significantly inhibited cytokine production in SFMC compared to IgG isotype control. RA SFMCs (n = 6) were stimulated with Pam3CSK4 in the presence or absence of OPN301 (1 μg/ml) or matched IgG isotype control for six hours. Values are expressed as the mean ± SEM. * P < 0.05, significantly different from control as determined using Wilcoxon Signed Rank analysis.", "To show that OPN301 penetrates the RA ST in culture, biopsies were snap frozen following 72 h incubation with OPN301, sectioned and immunohistology performed omitting the primary antibody. Representative images of OPN301 detection in RA synovial explants are shown in Figure 4A. OPN301 was localized to the lining layer and to the perivascular regions, showing that OPN301 does penetrate the tissue in culture with localization consistent with previous studies [4-6].\nOPN301 inhibition of spontaneous cytokine release is comparable to that observed for Adalimumab. (A) One representative image of OPN301 penetration of RA synovial explant cultures of four RA patients. OPN301 was localized to the perivascular and lining layer regions with no staining observed for IgG control. The bar on the lower right hand corner of each photomicrograph represents a distance of 100 μm on the top panel and 50 μm on the bottom panel. (B) OPN301 (1 μg/ml) inhibited spontaneous cytokine release from RA synovial explant cultures (n = 13) and effect that was comparable to that observed anti-TNFα mAb Adalimumab (Hum, 1 μg/ml). Results are expressed as mean ± SEM. * P < 0.05 and ** P < 0.001 significantly different from control as determined using Wilcoxon Signed Rank analysis.\nTo determine the inhibitory effect of OPN301 on spontaneous cytokine production by ST explant cultures, RA ST explants were established and cultured in the presence of OPN301 (1 μg/ml), Adalimumab (1 μg/ml) or IgG isotype control (1 μg/ml) for 72 hr (Figure 4B). OPN301 significantly inhibited spontaneous release of IL-8 (P = 0.001), TNF-α (P = 0.003), IFN-γ (P = 0.013), and IL-1β (P = 0.039). OPN301 inhibited IL-6 production, but this did not reach significance (P = 0.056). Adalimumab significantly inhibited IL-6 (P = 0.002), IL-8 (P = 0.000), IL-1β (P = 0.020), TNF-α (P = 0.018), and IFNγ (P = 0.001). The effect of OPN301 and Adalimumab on spontaneous cytokine production was also analysed using non-parametric Friedman analysis of variance, which further confirmed significant inhibition of all cytokines in response to both antibodies. We categorized the cytokine responses in the tissue cultures as follows, low response: <20% inhibition, moderate response: 21 to 49% inhibition, and good response: >50% inhibition. For OPN301, 38% of patients had a low response, 32% had a moderate response and 30% had a good response, while biopsies treated with Adalimumab showed 37% low response, 31% moderate response and 32% good response. The rate of response reflects what is observed for TNFα in clinical practice, (approximately 30% little or no response, and 70% of patient responses varying between 20 to 100%). Therefore, the results of our ex vivo model very much reflect typical biologic response rates for RA patients in clinical practice, and thus further validates the model as a good screening method for pre-clinical 'proof of concept studies'.", "TLRs are phylogenetically conserved receptors involved in the innate immune response to microbial pathogens through recognition of pathogen associated molecular patterns (PAMPs). Recent studies have shown that TLRs recognize endogenous ligands, found in RA serum and synovial fluid [15,16], these ligands can be released from necrotic cells during tissue damage or cell stress, leading to TLR mediated immune responses [17-19]. Expression and activation of TLR 2, 3, 4 and 9 in RA ST has been demonstrated suggesting TLRs may be involved in the pathogenesis [4,20].\nIn this study we demonstrated that the TLR2 agonist Pam3CSK4 significantly upregulated pro-inflammatory cytokine production in RA mononuclear cells, an effect that was significantly blocked by OPN301. This is consistent with previous studies using RA FLS, where PG, a bacterial derived TLR2 agonist, significantly induced angiogenic factors, pro-inflammatory chemokines and cytokines [21-23]. This effect was inhibited by approximately 40% in the presence of an anti-TLR2 mAb [23]. Furthermore, studies using dominant negative forms of the essential TLR2/4 adapter molecules MyD88 and MAL/TIRAP also showed a partial inhibitory effect, which varied considerably depending on the cytokine analyzed, suggesting intricate signaling pathways are involved in this complex inflammatory environment [6]. A key role for TLR2 in RA is further supported by evidence from animal models. TLR2 and MyD88 knockout mice are protected from SWC induced joint inflammation [24]. Intra-articular injection of the TLR2 and NOD2 ligand PG led to development of destructive arthritis in mice [8]. Functional studies have shown that stimulating TLR2 expressing RA fibroblasts with PG, leads to induction of cytokines and matrix-metalloproteinases [23]. TLR2 has also been implicated in other inflammatory diseases such as atherosclerosis and inflammatory bowel disease [25,26].\nTLR2 expression was localized to synovial tissue lining and sub-lining layers of RA patients, which is consistent with previous studies. In situ hybridization revealed TLR2 mRNA expression in the synovial lining, small vessels and in areas of infiltrating lymphocytes [4]. TLR2 and TLR4 expression has also been shown in the lining, sublining and perivascular regions of RA synovial tissue, with TLR2 expression higher than that of TLR4 [23]. TLR2 is also expressed at the sites of attachment and invasion into cartilage and bone [4]. Furthermore, in this study we show that OPN301 directly penetrated RA synovial explant cultures, localizing to the lining layer, which is comprised of synoviocyte-like fibroblasts and macrophages and to the perivascular region where angiogenesis and leukocyte extravasation occurs, critical mechanisms in the pathogenesis of RA. Consistent with this localization, TLR2 is functionally active in synovial fibroblasts and endothelial cells, where its activation results in induction of VEGF, IL-8, ICAM-1, VCAM-1 and MMPs [15,21,22,27-29]. Moreover, several studies have demonstrated that TLR2 activation of monocytes resulted in an increase in adhesive and migratory capacity of cells [29].\nIn this study we used an ex vivo RA ST explant model to investigate the role of TLR2 blockade in RA synovial inflammation. This model more closely reflects the in vivo joint environment, as it maintains tissue architecture and cell-cell contact of the complex mix of different cell types whose interaction contributes to the pro-inflammatory environment in the RA joint. Furthermore, RA synovial explants spontaneously release key pro-inflammatory cytokines and, therefore, this model is ideal for examining potential therapeutic molecules. We demonstrated that OPN301 significantly inhibited spontaneous secretion of TNFα, IL-1β, IFNγ and IL-8, suggesting that TLR2 is important in RA pathogenesis. The inhibitory properties of OPN301 and Adalimumab on spontaneous release of proinflammatory cytokines in our explant model reflect response rates observed in routine clinical practice.\nActivation of TLRs by local endogenous ligands leading to increased proinflammatory cytokine/chemokine secretion, may result in a vicious cycle of inflammation, ultimately leading to the pathological destruction of cartilage and bone seen in RA [2]. While bacterial TLR ligands have been found in RA synovial fluid and tissue, they have also been found in normal tissue [30]. Endogenous TLR ligands, which are released under inflammatory conditions and in response to tissue damage, have now been implicated in RA. Evidence for TLR4 ligands have been demonstrated, where RA synovial fluid stimulated TLR4 expressing CHO cells to regulate CD25 [31], and RNA released from necrotic synovial fluid can activate RA synovial fibroblasts in a TLR3 mediated mechansim [15]. While no ligand has been defined, the existence of a ligand is supported here and by other studies, which show that conditioned media from RA synovial explants can activate macrophages in a MyD88 and Mal dependent manner [6]. Several potential ligands have been suggested, such as Heat Shock Proteins, Fibronectin fragments, Hyaluronan oligosaccharides, HMGB1 and GP96; all of which have been identified in the RA joint [19,32-36].", "Targeting of the inflammatory cytokine TNF-α by biologic agents, such as Adalimumab, has been the most beneficial treatment strategy to date for patients with arthritis [9]. However, a significant proportion of patients fail to respond to these therapies, while others may be at risk of adverse events such as infections due to impaired immune function [9]. Our findings support further evaluation of strategies targeting TLR2 as potential therapeutic agents for the treatment of RA.", "CRP: C-reactive protein; DAS28: 28-joint count Disease Activity Score; DMARDs: disease-modifying antirheumatic drugs; EBAO: ethidium bromide acridine orange solution; ECM: extracellular matrix; ESR: erthrocyte sedimentation rate; FLS: synovial fibroblasts; mAb: monoclonal antibody; NSAIDs: nonsteroidal anti-inflammatory drugs; PAMPs: pathogen associate molecular patterns; PBMC: peripheral blood mononuclear cells; PG: petidoglycan; RA: rheumatoid arthritis; SFMC: synovial fluid mononuclear cells; ST: synovial tissue; SVUH: St. Vincent's University Hospital; TLR: Toll-like receptor; VAS: visual analog scale.", "DV is in receipt of a research grant from Opsona Therapeutics Ltd. Current Opsona employees (BK, PMcG, MR, WMcC) and ex-Opsona employee JD hold shares in Opsona Therapeutics Ltd; however, the percent involved is so small that Opsona does not view this as a conflicting interest. Luke O'Neill is a founder of Opsona Therapeutics Ltd. PMcG and BK are named inventors on patent WO2009/000929 and Jerome Dellacasagrande is named inventor on patent PCT/EP2010/059667. SNAU, TS, JMcC, MC and UF have no competing interests.", "SNAU conducted most of the experiments and analysis of data. TS, JMcC, MC and UF performed some of the experiments. UF, DV, MC, PMcG, SNAU, JD, BK, WMcC, MR, TS and LO'N, participated in the study design, data analysis and manuscript preparation. UF and DV supervised the research. DV and TS recruited all patients, performed the arthroscopies and provided all clinical information. All authors read and approved the final manuscript." ]

[ null, "materials|methods", null, null, null, null, null, null, "results", null, null, null, null, "discussion", "conclusions", null, null, null ]

[]

[ "Adolescent", "Adult", "Anisotropy", "Brain Injuries", "Brain Mapping", "Cross-Sectional Studies", "Diffusion Tensor Imaging", "Female", "Humans", "Male", "Memory Disorders", "Nerve Fibers, Myelinated", "Neural Pathways", "Neuropsychological Tests" ]

[ "Background", "Subjects", "Memory assessment", "Image acquisition and analysis", "Statistical analysis", "Results", "Comparison between TBI patients and controls", "Correlation analysis", "Correlation with clinical variables", "Correlation with declarative and working memory performance", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Diffuse axonal injury (DAI) was initially defined as widespread damage to axons throughout the white matter, evoked by intense shear and strain forces resulting from rapid acceleration and deceleration of the brain with or without impact after traumatic brain injury (TBI) [1,2]. More recently, traumatic axonal injury (TAI) has been suggested as a more appropriate term for describing axonal damage because it encompasses not only the primary axonal damage specifically caused by shear/strain injury, but also secondary alterations of white matter such as metabolic, hypoxic and microvascular damage or excitotoxicity [3,4].\nAlthough TAI has been described in neuropathological terms, magnetic resonance imaging (MRI) allows the detection of microhemorrhages and other indirect signs in regions commonly affected by this injury such as the subcortical white matter, the corpus callosum and the dorsolateral quadrant of the rostral brain-stem. It has recently been demonstrated that T2*-weighted MRI at high field strength is a useful tool for the identification of traumatic microbleeds even in the chronic stage of TBI [5]. However, diffusion tensor imaging (DTI) has been suggested as the best technique for the detection of subtle white matter changes [6] given that it can reveal significant abnormalities in white matter in patients with normal findings in conventional MRI [7,8].\nDTI is a non-invasive MRI technique that identifies the microscopic physical properties of tissues directly through the observation of translational molecular movement of water [9]. Water diffusion in cerebral white matter tends to be anisotropic, because the highly linear organization of white matter fibers restricts movement in other directions [10,11]. Fractional anisotropy (FA), one of the main DTI-derived indices, provides information of the degree of directionality of water diffusion and on microstructural white matter changes. DTI has been shown to be an efficient technique for determining white-matter integrity in several pathologies [12]. It has also been proposed as the most feasible biomarker of TAI and one of the best indicators of TBI severity [13,14]. Reductions in FA have been detected not only in moderate and severe TBI patients [15-18] but also in cases of mild TBI [19-24]. Moreover, DTI has proved to be an excellent tool for evaluating structural changes after TBI in longitudinal studies [16-18].\nThe advantages of DTI have resulted in a growing body of scientific evidence regarding the relationship between white matter damage and neuropsychological deficits in TBI. Studies conducted with pediatric samples have identified correlations between FA values and various cognitive functions, including cognitive processing speed and interference, executive functioning, IQ, verbal working memory, reading comprehension and letter naming speed [25-27]. Recently, Wu et al., [28] reported correlations between immediate recall and left cingulum bundles in adolescents after mild TBI. Some studies have related FA measurements with neuropsychological deficits in adults. Nakayama et al. [29] identified a positive correlation between the Mini-Mental State Examination (MMSE) and FA in the splenium of the corpus callosum. Salmond et al. [30] found a significant correlation between diffusivity and the impairment of learning and memory in the posterior cingulate, hippocampal formation and cortical areas. Kraus et al. [31] in a sample including all grade severities, found reduced FA in the ROIs analyzed and obtained a measure of the total regions of reduced FA that negatively correlated with the three cognitive domains evaluated. Furthermore, in a mild TBI sample, Niogi et al. [32] found a significant correlation between attentional control and FA within a ROI in the corona radiata and between memory performance and FA in the ROI placed in the uncinate both in the group of mild TBI patients and the control group. Kumar et al. [18] found correlations between the corpus callosum and neuropsychological tests involving processing speed as well as visuospatial and visuperceptive tasks. Finally, Lipton et al., [33] and Miles et al. [34] found that reductions in FA in dorsolateral prefrontal cortex correlated significantly with tests of executive functions. In summary, DTI technique, in special FA measures, has been found sensitive to reflect cognitive deficits associate with TBI.\nMemory is one of the functions that is most frequently impaired by TBI [35-37]. The concept of multiple memory systems and their different neuroanatomical substrates is currently accepted [38,39]. Declarative and working memory systems are significantly impaired after traumatic brain injury (TBI). Deficits in declarative memory - the capacity for conscious recollection of facts and events - are a common consequence of head trauma that are disproportionately suffered in comparison with other cognitive functions [35,40]. These memory difficulties improve slowly and although progress is made over the first and second year following injury, they remain apparent over time [40-42]. Neuroanatomically, declarative memory depends on the integrity of the hippocampus and its connections with the neocortex [43,44]. In neuroimaging studies with TBI patients, declarative memory has been found to correlate negatively with hippocampal [45,46] and fornix damage [47]. Working memory is defined as the ability to maintain and manipulate information temporarily [48]. Impairment of this memory is frequent in TBI patients given that implicated neural substrates, particularly the frontal cortex, are highly vulnerable in this type of injury. There is considerable evidence that working memory depends on network activity including the frontal and parietal regions and its connections. A meta-analysis of functional neuroimaging studies conducted by Owen et al. [49] provided strong evidence for the activation of frontal and parietal cortical regions by various versions of the n-back paradigm. The main fasciculus linking the parietal and frontal lobes is the superior longitudinal fasciculus (SLF) and hence it is likely that this has a role in working memory. Relations between the SLF and working memory deficits have been reported in multiple sclerosis [50] but not in TBI patients. To our knowledge there is no study investigating the impairment of white matter damage related to declarative and working memory deficits in a sample of severe and diffuse TBI.\nThe aim of this study was to investigate the role of white matter damage in declarative and working memory deficits after diffuse TBI, focusing on the main associative fasciculi [51] including those connecting the cerebral regions involved in the declarative memory and working memory networks.\nOur study had two main hypotheses: firstly, that a decreased FA in the superior longitudinal fasciculi (SLF), which is presumably involved in working memory function since it links the parietal and prefrontal regions, would correlate with working memory deficits, and, secondly, that a decreased FA in the fornix, the main fasciculus interconnecting the hippocampus with the frontal lobe, would correlate with declarative memory impairment.", "A cross-sectional study of thirty-one subjects was performed. Fifteen patients (eleven male) with severe TBI were recruited from the Head Injury Unit of the Institut de Neurorehabilitació Guttmann. Inclusion criteria were: a) age < 40 years, b) diffuse axonal injury according to clinical MRI without focal cortical lesions or larger than 1.5 cm3, c) severe TBI: defined as a minimal Glasgow Coma Scale (GCS) score ≤ 8 assessed at the first contact with the emergency services, d) emergence from posttraumatic amnesia (PTA) phase at the moment of the enrollment according to the Galveston Orientation and Attention Test (GOAT) [52], defined as two consecutive scores > 65, and f) no previous history of TBI, drug intake, neurological, or psychiatric disorders.\nThe etiology of TBI was a traffic accident in all cases. Fourteen patients were involved in car collisions, and one was a pedestrian hit by a motor vehicle. All patients had closed head injury and had not received surgery for extra- or subdural hematoma; all structural MRI scans were suggestive of TAI. The neuroradiologist (NB) took into account T1-weighted, FLAIR, and T2* GE sequences. The T2* GE sequences, which have a high level of sensitivity for detecting chronic hemosiderin, indicated evidence of TAI-related neuropathology. The method proposed by Gennarelli et al. [2] was used to classify the patients' TAI type. The grading system used was: type I, TAI only involving convexity gray-white matter junction; type II, also involving the corpus callosum in addition to the gray-white junction; and, type III, involving the rostral brainstem as well as the two previous criteria. Cases in which the midbrain was involved, but no corpus callosum lesions were apparent, were classified as type III (see Table 1).\nClinical and neuroimaging characteristics of the TBI group\nPT: Patient; GCS: Glasgow coma scale; PTA: posttraumatic amnesia; CT: computer tomography; MRI: magnetic resonance imaging; tevol: time of evolution since accident to the MRI evaluation; TAI: diffuse axonal injury; R/L: right/left; CC: corpus callosum; SHA: subarachnoidal hemorrhage.\nA control group of sixteen healthy subjects (nine male) were recruited from relatives and friends of the TBI group. This control group was matched by age, years of education and premorbid intellectual function estimated using the Vocabulary subtest of the Wechsler Adult Intelligence Scale (WAIS-III) [53], recognized as an efficient method for estimating general intelligence [54] (see Table 2). All subjects were right-handed, Caucasian-Mediterranean, and none had a previous history of neurological or psychiatric diseases.\nDemographic and clinical characteristics of TBI and control groups\nTBI = traumatic brain injury.\nThe study was approved by the Ethical and Research Committee of the Institut Universitari de Neurorehabilitacio Guttmann and all participants gave written informed consent.", "Working memory was evaluated by the Digit span and Letter-Number Sequencing (LNS) subtests of the WAIS-III [53] and a visual 2-back task [55]. Digit span was measured as the series length correctly reproduced at least once in the same order (forwards) and in reverse order (backwards). In the LNS, subjects heard lists of randomized numbers and letters (in alternating order) of increasing lengths, and were asked to reproduce the numbers and letters beginning with the lowest in each series, always with numbers first. The scores from the LNS were calculated by adding all correct items. In the 2-back task, numbers appeared on the screen for 500 ms against a black background, followed by a fixation cross for 1500 ms. The subjects were asked to decide whether the number they were looking at matched the one that they had seen two numbers earlier in the sequence. The numbers of correct responses as well as the reaction time were recorded. The d-prime index, a bias-free measure that takes both correct answers and errors into account, was also calculated to determine the accuracy of performance.\nThe Rivermead Behavioural Memory Test (RBMT) [56] was selected for its ability to explore declarative memory, and its ecological validity in assessing TBI patients. This test consists of 11 subtests including the following: remembering a name, a hidden belonging and an appointment; recognizing pictures and faces; recalling a prose passage; remembering a short route; remembering to deliver a message; and knowledge of some basic information such as the date, place and time. Those are designed as analogs of everyday tasks, reflecting the kinds of situations with which patients typically experience difficulty on a day-to-day basis. Two methods of standardizing scores across subtests allow for derivation of either a screening score, with subtest raw scores categorized on a scale of 0 ± 1 (maximum score 12 points), or a standardized profile score, with subtest raw scores categorized on a scale of 0 ± 2 (maximum score 24 points). The slightly more fine-grained standardized profile score provides a more sensitive analysis of performance [57] thus we use this score for our correlation analysis.", "MRI data sets were acquired on a 1.5 T Signa GE (General Electric, Milwaukee, WI) at the Centre de Diagnostic per la Imatge of the Hospital Clínic (CDIC), Barcelona. Diffusion weighted images were sensitized in 25 non-collinear directions with a b-value = 1000 sec/mm2, using an echo-planar (EPI) sequence (TR = 9999.996 ms, TE = 85 ms, 20 axial slices with a resolution of 0.9375 × 0.9375 mm, slice thickness = 5 mm, gap = 2 mm matrix size = 128 × 128, FOV = 100).\nData preprocessing and analysis was performed using FMRIB's software library [FSL version 4.1; Oxford Centre for Functional MRI of the Brain (FMRIB), UK; http://www.fmrib.ox.ac.uk/fsl/]. Image artefacts due to eddy current distortions were minimized by registering the diffusion images to the b0 images. The registered images were skull-stripped using the Brain Extraction Tool (BET) [58]. Fractional anisotropy maps were calculated using the FMRIB's Diffusion Toolbox v.2.0 [FDT, [59]]. After calculation of the FA map for each subject, we implemented a voxel-wise statistical analysis of the FA data using Tract-Based Spatial Statistics v1.2 (TBSS) which aims to overcome the limitations of the standard VBM-style analyses [60], particularly those regarding to its dependence on the goodness of the registration algorithm and on the choice of the spatial smoothing [61]. FA data were aligned into a common space using a non-linear registration algorithm (FNIRT) to register the images to the standard FMRIB58 FA template, which is in MNI152 standard space. Aligned FA maps were visually inspected after registration and we confirmed that the result of the previous step was correct. Next, a mean FA image was created from the images from all the subjects in this common space and narrowed to generate a mean FA skeleton that represented the center of all tracts common to the entire group. This was thresholded to FA 0.2 to include the major white matter pathways but to exclude peripheral tracts where there was significant inter-subject variability and partial volume effects with gray matter. This ensured that each subject's skeleton was in the group space while also representing the center of the subject's unique white matter bundles. The aligned FA image for each subject was then projected onto the skeleton by filling the skeleton with FA values from the nearest relevant tract centre. This is achieved for each skeleton voxel by searching perpendicular to the local skeleton structure for the maximum value in the FA image of the subject. The resulting skeletonised data was then fed into voxelwise cross-subject statistics.", "Group comparisons and correlations with neuropsychological measures were performed using Randomise v2.1 from FSL [62,63]. As seen in table one, the time of evolution since injury was very heterogeneous. In order to control possible effects of this variable in the correlation results, time of evolution was entered as a non-interest variable in the matrix. The statistical threshold was set at p < 0.05 Family Wise Error (FWE) corrected, which is a conservative procedure that allows a high control of Type I error, being the probability of one or more false positives the same as the significance level. The Threshold-Free Cluster Enhancement (TFCE) method was used to define the clusters [64]. Correlation analyses were performed with the 2-back d-prime index and the Rivermead profile score using a region of interest (ROI) approach in the following associative fasciculi: corpus callosum, superior and inferior longitudinal, inferior fronto-occipital, uncinate, and cingulate as well as the fornix and arcuate fasciculi as the major pathways that connect associative cortical regions involved in working and declarative memory. ROI masks were obtained from the Jülich histological atlas [65,66] and the JHU white-matter tractography atlas [67-69]. Areas corresponding to significant clusters were identified using the JHU white-matter tractography atlas. Mean FA values were obtained from each subject's FA skeleton map and skeletonised SLF and fornix ROIs. Mean FA values were obtained from each subject's FA skeleton map and skeletonized for all the fasciculi ROIs mentioned above.\nStatistical tests on non-imaging data were performed using SPSS (Statistical Package for the Social Sciences) v.16 (SPSS Inc., Chicago Illinois). Group differences were examined using the Student t-test, since the data were normally distributed using a significance level of p < 0.05.\nPartial correlation coefficients, controlling for the time of evolution, were used to explore the association between mean FA values and clinical variables and neuropsychological measures. Statistical significance was set at a two-tailed p ≤ 0.05.", "[SUBTITLE] Comparison between TBI patients and controls [SUBSECTION] Performance on the memory tests is described in Table 3. Statistical significance was obtained for the difference in scores in the LNS subtest (WAIS-III), d-prime index for the 2-back, and the RBMT profile. Forward and backward digits did not reach statistical significance.\nNeuropsychological performance for TBI and control groups\nTBI: traumatic brain injury; Letter-Number Sequencing (WAIS-III); Goals: number of targets correctly identified; RBMT: Rivermead Behavioral Memory Test; ns: not significant.\nGroup comparison for FA skeleton maps revealed multiple areas of significant FA reductions in TBI patients as compared to controls. All the long associative fibers were affected, including the corpus callosum, the superior and inferior longitudinal fasciculi, and the inferior fronto-occipital fasciculi. Decreased FA was also observed in shorter fibers such as the uncinate fasciculus, cingulum, fornix and anterior thalamic radiation (Figure 1, Table 4). FA was not increased in the TBI group in any cerebral region.\nResults from TBSS analysis of FA maps showing the clusters of significantly reduced FA in TBI patients compared to controls in red (TFCE, p < 0.05 FWE-corrected). Widespread white matter affectation is observed.\nDifferences between groups in mean FA from the whole skeletonised brain and the ROIs\nTBI: Traumatic Brain Injury; FA: fractional anisotropy; CC: corpus callosum; SLF: superior longitudinal fasciculi; ILF: inferior longitudinal fasciculi; IFO: inferior fronto-occipital fasciculi.\nWe obtained mean FA values of the whole skeletonized brain and all of the selected ROIs. Group comparisons for all these values reached statistical significance in all cases with p < 0.001 (Table 4).\nPerformance on the memory tests is described in Table 3. Statistical significance was obtained for the difference in scores in the LNS subtest (WAIS-III), d-prime index for the 2-back, and the RBMT profile. Forward and backward digits did not reach statistical significance.\nNeuropsychological performance for TBI and control groups\nTBI: traumatic brain injury; Letter-Number Sequencing (WAIS-III); Goals: number of targets correctly identified; RBMT: Rivermead Behavioral Memory Test; ns: not significant.\nGroup comparison for FA skeleton maps revealed multiple areas of significant FA reductions in TBI patients as compared to controls. All the long associative fibers were affected, including the corpus callosum, the superior and inferior longitudinal fasciculi, and the inferior fronto-occipital fasciculi. Decreased FA was also observed in shorter fibers such as the uncinate fasciculus, cingulum, fornix and anterior thalamic radiation (Figure 1, Table 4). FA was not increased in the TBI group in any cerebral region.\nResults from TBSS analysis of FA maps showing the clusters of significantly reduced FA in TBI patients compared to controls in red (TFCE, p < 0.05 FWE-corrected). Widespread white matter affectation is observed.\nDifferences between groups in mean FA from the whole skeletonised brain and the ROIs\nTBI: Traumatic Brain Injury; FA: fractional anisotropy; CC: corpus callosum; SLF: superior longitudinal fasciculi; ILF: inferior longitudinal fasciculi; IFO: inferior fronto-occipital fasciculi.\nWe obtained mean FA values of the whole skeletonized brain and all of the selected ROIs. Group comparisons for all these values reached statistical significance in all cases with p < 0.001 (Table 4).\n[SUBTITLE] Correlation analysis [SUBSECTION] [SUBTITLE] Correlation with clinical variables [SUBSECTION] We observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\nWe observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\n[SUBTITLE] Correlation with declarative and working memory performance [SUBSECTION] The mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.\nThe mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.\n[SUBTITLE] Correlation with clinical variables [SUBSECTION] We observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\nWe observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\n[SUBTITLE] Correlation with declarative and working memory performance [SUBSECTION] The mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.\nThe mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.", "Performance on the memory tests is described in Table 3. Statistical significance was obtained for the difference in scores in the LNS subtest (WAIS-III), d-prime index for the 2-back, and the RBMT profile. Forward and backward digits did not reach statistical significance.\nNeuropsychological performance for TBI and control groups\nTBI: traumatic brain injury; Letter-Number Sequencing (WAIS-III); Goals: number of targets correctly identified; RBMT: Rivermead Behavioral Memory Test; ns: not significant.\nGroup comparison for FA skeleton maps revealed multiple areas of significant FA reductions in TBI patients as compared to controls. All the long associative fibers were affected, including the corpus callosum, the superior and inferior longitudinal fasciculi, and the inferior fronto-occipital fasciculi. Decreased FA was also observed in shorter fibers such as the uncinate fasciculus, cingulum, fornix and anterior thalamic radiation (Figure 1, Table 4). FA was not increased in the TBI group in any cerebral region.\nResults from TBSS analysis of FA maps showing the clusters of significantly reduced FA in TBI patients compared to controls in red (TFCE, p < 0.05 FWE-corrected). Widespread white matter affectation is observed.\nDifferences between groups in mean FA from the whole skeletonised brain and the ROIs\nTBI: Traumatic Brain Injury; FA: fractional anisotropy; CC: corpus callosum; SLF: superior longitudinal fasciculi; ILF: inferior longitudinal fasciculi; IFO: inferior fronto-occipital fasciculi.\nWe obtained mean FA values of the whole skeletonized brain and all of the selected ROIs. Group comparisons for all these values reached statistical significance in all cases with p < 0.001 (Table 4).", "[SUBTITLE] Correlation with clinical variables [SUBSECTION] We observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\nWe observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\n[SUBTITLE] Correlation with declarative and working memory performance [SUBSECTION] The mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.\nThe mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.", "We observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).", "The mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.", "The present study provides evidence of the implications of TAI in declarative and working memory deficits in TBI. DTI group comparison revealed global whole brain reductions in mean FA values for patients and FA maps confirmed that almost all the major fibers were involved. Although our patients suffered global white matter integrity impairment, we found two different and restricted patterns of correlations with the FA and neuropsychological assessment. Whole brain DTI analysis showed that decreased FA throughout the brain correlated with 2-back measures but not with the Rivermead Test. Results from the ROI analyses of the main association fibers showed, as predicted, that working memory specifically correlated with the superior longitudinal fasciculi. However, it was also found to correlate with the corpus callosum, the arcuate fasciculi and with the fornix. On the other hand, declarative memory deficits only correlated with the fornix, as we had expected, and the corpus callosum. These results suggest that there are two different patterns of FA reduction related with two types of memory dysfunctions.\nWe found that superior longitudinal fasciculi damage is related with working memory but not with declarative memory deficits. These correlations were expectable since the longitudinal fasciculi connect the associative frontal and parietal regions involved in working memory functions [70-72,49]. The correlation between working memory deficit and the superior longitudinal was also described in multiple sclerosis, pathology that also involves white matter damage [50].\nIn our sample, FA reductions of corpus callosum correlated with both working and declarative memory impairments. In declarative memory the correlations were seen in the posterior region whereas in working memory the correlations involved anterior and posterior regions, thus again these results point to differential patterns of correlations for both types of memory impairment.\nAccording our results, declarative memory impairment did not depend on diffuse white matter damage since no correlations between FA maps or mean values and declarative memory values were seen. However, the ROI analysis revealed that the fornix FA impairment correlated with the Rivermead test. This result is in agreement with the role of the damage of the hippocampus and its connections in declarative memory deficits in TBI [45,46].\nOur declarative memory results partially agree with those obtained by Salmond et al. [30]. Using a voxel-based analysis with SPM tools, these authors found a significant positive correlation between declarative memory and diffusivity in the left hippocampal formation, the left posterior cingulate, and the left frontal, temporal and occipital regions. The more widespread pattern of correlations observed in their study can be explained by the use of FDR correction, which is more liberal than the FWE correction used in ours [73]. Correlations between FA values in the fornix and declarative memory impairment have been also observed in patients with multiple sclerosis [74]. Other studies investigating FA correlations with declarative memory functions in mild TBI samples have reported significant correlations with the uncinate fasciculi [30,32] and the cingulum [28]. Although we found decreased FA in these fasciculi, correlations did not reach statistical significance. These discrepancies may be explained by the varying grade of severity of the samples, the difference in the memory tests used, and DTI methodological differences.\nIn the present study, working memory deficits also correlated with the fornix in both the whole brain analysis and the ROI analyses. There is some evidence from fMRI studies that the hippocampus is involved in working memory functions in healthy subjects [75-78]. Moreover, several animal studies also suggest a role for the hippocampus in working memory [79-81]. Anatomically, prefrontal regions involved in working memory tasks receive projections from the hippocampus [82,83] and are connected directly to the ventral hippocampus and indirectly to the dorsal hippocampus via the thalamus [84-86]. This structural connectivity supports the idea that the hippocampus has a role in working memory functioning as suggested by our findings.\nFinally, significant correlations were observed between the PTA variable and white matter integrity. Whole brain map analysis showed that PTA is an excellent index predictor of the degree of impairment of the major white matter tracts and association fibers. These results suggest that the recovery of memory functions is dependent on the integrity of the complex neocortical regions. Unlike previous studies [13,14,17], no correlations were found between GCS and FA maps or mean FA values. This result was to be expected as the fact that all our patients had severe TBI meant that GCS variability would not be sufficient to reach statistical significance.\nOur study has certain limitations and our results should be regarded as preliminary. The small sample size and its specific diffuse characteristics may preclude the generalization of the results. The presence of mixed focal and diffuse pathology frequently observed in severe TBI may confound the mapping of neural and behavioral changes in these patients. As our study sample excluded significant cortical pathology, the cognitive impairment observed is more likely to be due to the diffuse pathology alone. Nevertheless, we cannot exclude the possibility that reductions in gray matter in several subcortical structures are also influencing memory deficits in TBI.", "This DTI study suggests that declarative and working memory deficits in diffuse TBI patients are related to differential patterns of FA reduction. Working memory impairment reflects the diffuse white matter damage affecting large scale networks such as the superior longitudinal fasciculi, whereas declarative memory deficits seem to be the result of more local disruption of the cerebral circuitry.", "The authors declare that they have no competing interests.", "EP, DFE and CJ made substantial contribution to conception and design, interpretation of data, drafting and writing of manuscript and further revisions of the manuscript. Neuroimaging data were analyzed by EP and DFE. Neuroimaging sequence acquisitions, neurological description and classification of data by NB. RSC and TR participated in the collection of neuropsychological data and in the collection of acute clinical data. JT and PV made a critical revision of the manuscript for important intellectual content providing additional comments and contributions. CJ supervised the study. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2377/11/24/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Subjects", "Memory assessment", "Image acquisition and analysis", "Statistical analysis", "Results", "Comparison between TBI patients and controls", "Correlation analysis", "Correlation with clinical variables", "Correlation with declarative and working memory performance", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Diffuse axonal injury (DAI) was initially defined as widespread damage to axons throughout the white matter, evoked by intense shear and strain forces resulting from rapid acceleration and deceleration of the brain with or without impact after traumatic brain injury (TBI) [1,2]. More recently, traumatic axonal injury (TAI) has been suggested as a more appropriate term for describing axonal damage because it encompasses not only the primary axonal damage specifically caused by shear/strain injury, but also secondary alterations of white matter such as metabolic, hypoxic and microvascular damage or excitotoxicity [3,4].\nAlthough TAI has been described in neuropathological terms, magnetic resonance imaging (MRI) allows the detection of microhemorrhages and other indirect signs in regions commonly affected by this injury such as the subcortical white matter, the corpus callosum and the dorsolateral quadrant of the rostral brain-stem. It has recently been demonstrated that T2*-weighted MRI at high field strength is a useful tool for the identification of traumatic microbleeds even in the chronic stage of TBI [5]. However, diffusion tensor imaging (DTI) has been suggested as the best technique for the detection of subtle white matter changes [6] given that it can reveal significant abnormalities in white matter in patients with normal findings in conventional MRI [7,8].\nDTI is a non-invasive MRI technique that identifies the microscopic physical properties of tissues directly through the observation of translational molecular movement of water [9]. Water diffusion in cerebral white matter tends to be anisotropic, because the highly linear organization of white matter fibers restricts movement in other directions [10,11]. Fractional anisotropy (FA), one of the main DTI-derived indices, provides information of the degree of directionality of water diffusion and on microstructural white matter changes. DTI has been shown to be an efficient technique for determining white-matter integrity in several pathologies [12]. It has also been proposed as the most feasible biomarker of TAI and one of the best indicators of TBI severity [13,14]. Reductions in FA have been detected not only in moderate and severe TBI patients [15-18] but also in cases of mild TBI [19-24]. Moreover, DTI has proved to be an excellent tool for evaluating structural changes after TBI in longitudinal studies [16-18].\nThe advantages of DTI have resulted in a growing body of scientific evidence regarding the relationship between white matter damage and neuropsychological deficits in TBI. Studies conducted with pediatric samples have identified correlations between FA values and various cognitive functions, including cognitive processing speed and interference, executive functioning, IQ, verbal working memory, reading comprehension and letter naming speed [25-27]. Recently, Wu et al., [28] reported correlations between immediate recall and left cingulum bundles in adolescents after mild TBI. Some studies have related FA measurements with neuropsychological deficits in adults. Nakayama et al. [29] identified a positive correlation between the Mini-Mental State Examination (MMSE) and FA in the splenium of the corpus callosum. Salmond et al. [30] found a significant correlation between diffusivity and the impairment of learning and memory in the posterior cingulate, hippocampal formation and cortical areas. Kraus et al. [31] in a sample including all grade severities, found reduced FA in the ROIs analyzed and obtained a measure of the total regions of reduced FA that negatively correlated with the three cognitive domains evaluated. Furthermore, in a mild TBI sample, Niogi et al. [32] found a significant correlation between attentional control and FA within a ROI in the corona radiata and between memory performance and FA in the ROI placed in the uncinate both in the group of mild TBI patients and the control group. Kumar et al. [18] found correlations between the corpus callosum and neuropsychological tests involving processing speed as well as visuospatial and visuperceptive tasks. Finally, Lipton et al., [33] and Miles et al. [34] found that reductions in FA in dorsolateral prefrontal cortex correlated significantly with tests of executive functions. In summary, DTI technique, in special FA measures, has been found sensitive to reflect cognitive deficits associate with TBI.\nMemory is one of the functions that is most frequently impaired by TBI [35-37]. The concept of multiple memory systems and their different neuroanatomical substrates is currently accepted [38,39]. Declarative and working memory systems are significantly impaired after traumatic brain injury (TBI). Deficits in declarative memory - the capacity for conscious recollection of facts and events - are a common consequence of head trauma that are disproportionately suffered in comparison with other cognitive functions [35,40]. These memory difficulties improve slowly and although progress is made over the first and second year following injury, they remain apparent over time [40-42]. Neuroanatomically, declarative memory depends on the integrity of the hippocampus and its connections with the neocortex [43,44]. In neuroimaging studies with TBI patients, declarative memory has been found to correlate negatively with hippocampal [45,46] and fornix damage [47]. Working memory is defined as the ability to maintain and manipulate information temporarily [48]. Impairment of this memory is frequent in TBI patients given that implicated neural substrates, particularly the frontal cortex, are highly vulnerable in this type of injury. There is considerable evidence that working memory depends on network activity including the frontal and parietal regions and its connections. A meta-analysis of functional neuroimaging studies conducted by Owen et al. [49] provided strong evidence for the activation of frontal and parietal cortical regions by various versions of the n-back paradigm. The main fasciculus linking the parietal and frontal lobes is the superior longitudinal fasciculus (SLF) and hence it is likely that this has a role in working memory. Relations between the SLF and working memory deficits have been reported in multiple sclerosis [50] but not in TBI patients. To our knowledge there is no study investigating the impairment of white matter damage related to declarative and working memory deficits in a sample of severe and diffuse TBI.\nThe aim of this study was to investigate the role of white matter damage in declarative and working memory deficits after diffuse TBI, focusing on the main associative fasciculi [51] including those connecting the cerebral regions involved in the declarative memory and working memory networks.\nOur study had two main hypotheses: firstly, that a decreased FA in the superior longitudinal fasciculi (SLF), which is presumably involved in working memory function since it links the parietal and prefrontal regions, would correlate with working memory deficits, and, secondly, that a decreased FA in the fornix, the main fasciculus interconnecting the hippocampus with the frontal lobe, would correlate with declarative memory impairment.", "[SUBTITLE] Subjects [SUBSECTION] A cross-sectional study of thirty-one subjects was performed. Fifteen patients (eleven male) with severe TBI were recruited from the Head Injury Unit of the Institut de Neurorehabilitació Guttmann. Inclusion criteria were: a) age < 40 years, b) diffuse axonal injury according to clinical MRI without focal cortical lesions or larger than 1.5 cm3, c) severe TBI: defined as a minimal Glasgow Coma Scale (GCS) score ≤ 8 assessed at the first contact with the emergency services, d) emergence from posttraumatic amnesia (PTA) phase at the moment of the enrollment according to the Galveston Orientation and Attention Test (GOAT) [52], defined as two consecutive scores > 65, and f) no previous history of TBI, drug intake, neurological, or psychiatric disorders.\nThe etiology of TBI was a traffic accident in all cases. Fourteen patients were involved in car collisions, and one was a pedestrian hit by a motor vehicle. All patients had closed head injury and had not received surgery for extra- or subdural hematoma; all structural MRI scans were suggestive of TAI. The neuroradiologist (NB) took into account T1-weighted, FLAIR, and T2* GE sequences. The T2* GE sequences, which have a high level of sensitivity for detecting chronic hemosiderin, indicated evidence of TAI-related neuropathology. The method proposed by Gennarelli et al. [2] was used to classify the patients' TAI type. The grading system used was: type I, TAI only involving convexity gray-white matter junction; type II, also involving the corpus callosum in addition to the gray-white junction; and, type III, involving the rostral brainstem as well as the two previous criteria. Cases in which the midbrain was involved, but no corpus callosum lesions were apparent, were classified as type III (see Table 1).\nClinical and neuroimaging characteristics of the TBI group\nPT: Patient; GCS: Glasgow coma scale; PTA: posttraumatic amnesia; CT: computer tomography; MRI: magnetic resonance imaging; tevol: time of evolution since accident to the MRI evaluation; TAI: diffuse axonal injury; R/L: right/left; CC: corpus callosum; SHA: subarachnoidal hemorrhage.\nA control group of sixteen healthy subjects (nine male) were recruited from relatives and friends of the TBI group. This control group was matched by age, years of education and premorbid intellectual function estimated using the Vocabulary subtest of the Wechsler Adult Intelligence Scale (WAIS-III) [53], recognized as an efficient method for estimating general intelligence [54] (see Table 2). All subjects were right-handed, Caucasian-Mediterranean, and none had a previous history of neurological or psychiatric diseases.\nDemographic and clinical characteristics of TBI and control groups\nTBI = traumatic brain injury.\nThe study was approved by the Ethical and Research Committee of the Institut Universitari de Neurorehabilitacio Guttmann and all participants gave written informed consent.\nA cross-sectional study of thirty-one subjects was performed. Fifteen patients (eleven male) with severe TBI were recruited from the Head Injury Unit of the Institut de Neurorehabilitació Guttmann. Inclusion criteria were: a) age < 40 years, b) diffuse axonal injury according to clinical MRI without focal cortical lesions or larger than 1.5 cm3, c) severe TBI: defined as a minimal Glasgow Coma Scale (GCS) score ≤ 8 assessed at the first contact with the emergency services, d) emergence from posttraumatic amnesia (PTA) phase at the moment of the enrollment according to the Galveston Orientation and Attention Test (GOAT) [52], defined as two consecutive scores > 65, and f) no previous history of TBI, drug intake, neurological, or psychiatric disorders.\nThe etiology of TBI was a traffic accident in all cases. Fourteen patients were involved in car collisions, and one was a pedestrian hit by a motor vehicle. All patients had closed head injury and had not received surgery for extra- or subdural hematoma; all structural MRI scans were suggestive of TAI. The neuroradiologist (NB) took into account T1-weighted, FLAIR, and T2* GE sequences. The T2* GE sequences, which have a high level of sensitivity for detecting chronic hemosiderin, indicated evidence of TAI-related neuropathology. The method proposed by Gennarelli et al. [2] was used to classify the patients' TAI type. The grading system used was: type I, TAI only involving convexity gray-white matter junction; type II, also involving the corpus callosum in addition to the gray-white junction; and, type III, involving the rostral brainstem as well as the two previous criteria. Cases in which the midbrain was involved, but no corpus callosum lesions were apparent, were classified as type III (see Table 1).\nClinical and neuroimaging characteristics of the TBI group\nPT: Patient; GCS: Glasgow coma scale; PTA: posttraumatic amnesia; CT: computer tomography; MRI: magnetic resonance imaging; tevol: time of evolution since accident to the MRI evaluation; TAI: diffuse axonal injury; R/L: right/left; CC: corpus callosum; SHA: subarachnoidal hemorrhage.\nA control group of sixteen healthy subjects (nine male) were recruited from relatives and friends of the TBI group. This control group was matched by age, years of education and premorbid intellectual function estimated using the Vocabulary subtest of the Wechsler Adult Intelligence Scale (WAIS-III) [53], recognized as an efficient method for estimating general intelligence [54] (see Table 2). All subjects were right-handed, Caucasian-Mediterranean, and none had a previous history of neurological or psychiatric diseases.\nDemographic and clinical characteristics of TBI and control groups\nTBI = traumatic brain injury.\nThe study was approved by the Ethical and Research Committee of the Institut Universitari de Neurorehabilitacio Guttmann and all participants gave written informed consent.\n[SUBTITLE] Memory assessment [SUBSECTION] Working memory was evaluated by the Digit span and Letter-Number Sequencing (LNS) subtests of the WAIS-III [53] and a visual 2-back task [55]. Digit span was measured as the series length correctly reproduced at least once in the same order (forwards) and in reverse order (backwards). In the LNS, subjects heard lists of randomized numbers and letters (in alternating order) of increasing lengths, and were asked to reproduce the numbers and letters beginning with the lowest in each series, always with numbers first. The scores from the LNS were calculated by adding all correct items. In the 2-back task, numbers appeared on the screen for 500 ms against a black background, followed by a fixation cross for 1500 ms. The subjects were asked to decide whether the number they were looking at matched the one that they had seen two numbers earlier in the sequence. The numbers of correct responses as well as the reaction time were recorded. The d-prime index, a bias-free measure that takes both correct answers and errors into account, was also calculated to determine the accuracy of performance.\nThe Rivermead Behavioural Memory Test (RBMT) [56] was selected for its ability to explore declarative memory, and its ecological validity in assessing TBI patients. This test consists of 11 subtests including the following: remembering a name, a hidden belonging and an appointment; recognizing pictures and faces; recalling a prose passage; remembering a short route; remembering to deliver a message; and knowledge of some basic information such as the date, place and time. Those are designed as analogs of everyday tasks, reflecting the kinds of situations with which patients typically experience difficulty on a day-to-day basis. Two methods of standardizing scores across subtests allow for derivation of either a screening score, with subtest raw scores categorized on a scale of 0 ± 1 (maximum score 12 points), or a standardized profile score, with subtest raw scores categorized on a scale of 0 ± 2 (maximum score 24 points). The slightly more fine-grained standardized profile score provides a more sensitive analysis of performance [57] thus we use this score for our correlation analysis.\nWorking memory was evaluated by the Digit span and Letter-Number Sequencing (LNS) subtests of the WAIS-III [53] and a visual 2-back task [55]. Digit span was measured as the series length correctly reproduced at least once in the same order (forwards) and in reverse order (backwards). In the LNS, subjects heard lists of randomized numbers and letters (in alternating order) of increasing lengths, and were asked to reproduce the numbers and letters beginning with the lowest in each series, always with numbers first. The scores from the LNS were calculated by adding all correct items. In the 2-back task, numbers appeared on the screen for 500 ms against a black background, followed by a fixation cross for 1500 ms. The subjects were asked to decide whether the number they were looking at matched the one that they had seen two numbers earlier in the sequence. The numbers of correct responses as well as the reaction time were recorded. The d-prime index, a bias-free measure that takes both correct answers and errors into account, was also calculated to determine the accuracy of performance.\nThe Rivermead Behavioural Memory Test (RBMT) [56] was selected for its ability to explore declarative memory, and its ecological validity in assessing TBI patients. This test consists of 11 subtests including the following: remembering a name, a hidden belonging and an appointment; recognizing pictures and faces; recalling a prose passage; remembering a short route; remembering to deliver a message; and knowledge of some basic information such as the date, place and time. Those are designed as analogs of everyday tasks, reflecting the kinds of situations with which patients typically experience difficulty on a day-to-day basis. Two methods of standardizing scores across subtests allow for derivation of either a screening score, with subtest raw scores categorized on a scale of 0 ± 1 (maximum score 12 points), or a standardized profile score, with subtest raw scores categorized on a scale of 0 ± 2 (maximum score 24 points). The slightly more fine-grained standardized profile score provides a more sensitive analysis of performance [57] thus we use this score for our correlation analysis.\n[SUBTITLE] Image acquisition and analysis [SUBSECTION] MRI data sets were acquired on a 1.5 T Signa GE (General Electric, Milwaukee, WI) at the Centre de Diagnostic per la Imatge of the Hospital Clínic (CDIC), Barcelona. Diffusion weighted images were sensitized in 25 non-collinear directions with a b-value = 1000 sec/mm2, using an echo-planar (EPI) sequence (TR = 9999.996 ms, TE = 85 ms, 20 axial slices with a resolution of 0.9375 × 0.9375 mm, slice thickness = 5 mm, gap = 2 mm matrix size = 128 × 128, FOV = 100).\nData preprocessing and analysis was performed using FMRIB's software library [FSL version 4.1; Oxford Centre for Functional MRI of the Brain (FMRIB), UK; http://www.fmrib.ox.ac.uk/fsl/]. Image artefacts due to eddy current distortions were minimized by registering the diffusion images to the b0 images. The registered images were skull-stripped using the Brain Extraction Tool (BET) [58]. Fractional anisotropy maps were calculated using the FMRIB's Diffusion Toolbox v.2.0 [FDT, [59]]. After calculation of the FA map for each subject, we implemented a voxel-wise statistical analysis of the FA data using Tract-Based Spatial Statistics v1.2 (TBSS) which aims to overcome the limitations of the standard VBM-style analyses [60], particularly those regarding to its dependence on the goodness of the registration algorithm and on the choice of the spatial smoothing [61]. FA data were aligned into a common space using a non-linear registration algorithm (FNIRT) to register the images to the standard FMRIB58 FA template, which is in MNI152 standard space. Aligned FA maps were visually inspected after registration and we confirmed that the result of the previous step was correct. Next, a mean FA image was created from the images from all the subjects in this common space and narrowed to generate a mean FA skeleton that represented the center of all tracts common to the entire group. This was thresholded to FA 0.2 to include the major white matter pathways but to exclude peripheral tracts where there was significant inter-subject variability and partial volume effects with gray matter. This ensured that each subject's skeleton was in the group space while also representing the center of the subject's unique white matter bundles. The aligned FA image for each subject was then projected onto the skeleton by filling the skeleton with FA values from the nearest relevant tract centre. This is achieved for each skeleton voxel by searching perpendicular to the local skeleton structure for the maximum value in the FA image of the subject. The resulting skeletonised data was then fed into voxelwise cross-subject statistics.\nMRI data sets were acquired on a 1.5 T Signa GE (General Electric, Milwaukee, WI) at the Centre de Diagnostic per la Imatge of the Hospital Clínic (CDIC), Barcelona. Diffusion weighted images were sensitized in 25 non-collinear directions with a b-value = 1000 sec/mm2, using an echo-planar (EPI) sequence (TR = 9999.996 ms, TE = 85 ms, 20 axial slices with a resolution of 0.9375 × 0.9375 mm, slice thickness = 5 mm, gap = 2 mm matrix size = 128 × 128, FOV = 100).\nData preprocessing and analysis was performed using FMRIB's software library [FSL version 4.1; Oxford Centre for Functional MRI of the Brain (FMRIB), UK; http://www.fmrib.ox.ac.uk/fsl/]. Image artefacts due to eddy current distortions were minimized by registering the diffusion images to the b0 images. The registered images were skull-stripped using the Brain Extraction Tool (BET) [58]. Fractional anisotropy maps were calculated using the FMRIB's Diffusion Toolbox v.2.0 [FDT, [59]]. After calculation of the FA map for each subject, we implemented a voxel-wise statistical analysis of the FA data using Tract-Based Spatial Statistics v1.2 (TBSS) which aims to overcome the limitations of the standard VBM-style analyses [60], particularly those regarding to its dependence on the goodness of the registration algorithm and on the choice of the spatial smoothing [61]. FA data were aligned into a common space using a non-linear registration algorithm (FNIRT) to register the images to the standard FMRIB58 FA template, which is in MNI152 standard space. Aligned FA maps were visually inspected after registration and we confirmed that the result of the previous step was correct. Next, a mean FA image was created from the images from all the subjects in this common space and narrowed to generate a mean FA skeleton that represented the center of all tracts common to the entire group. This was thresholded to FA 0.2 to include the major white matter pathways but to exclude peripheral tracts where there was significant inter-subject variability and partial volume effects with gray matter. This ensured that each subject's skeleton was in the group space while also representing the center of the subject's unique white matter bundles. The aligned FA image for each subject was then projected onto the skeleton by filling the skeleton with FA values from the nearest relevant tract centre. This is achieved for each skeleton voxel by searching perpendicular to the local skeleton structure for the maximum value in the FA image of the subject. The resulting skeletonised data was then fed into voxelwise cross-subject statistics.\n[SUBTITLE] Statistical analysis [SUBSECTION] Group comparisons and correlations with neuropsychological measures were performed using Randomise v2.1 from FSL [62,63]. As seen in table one, the time of evolution since injury was very heterogeneous. In order to control possible effects of this variable in the correlation results, time of evolution was entered as a non-interest variable in the matrix. The statistical threshold was set at p < 0.05 Family Wise Error (FWE) corrected, which is a conservative procedure that allows a high control of Type I error, being the probability of one or more false positives the same as the significance level. The Threshold-Free Cluster Enhancement (TFCE) method was used to define the clusters [64]. Correlation analyses were performed with the 2-back d-prime index and the Rivermead profile score using a region of interest (ROI) approach in the following associative fasciculi: corpus callosum, superior and inferior longitudinal, inferior fronto-occipital, uncinate, and cingulate as well as the fornix and arcuate fasciculi as the major pathways that connect associative cortical regions involved in working and declarative memory. ROI masks were obtained from the Jülich histological atlas [65,66] and the JHU white-matter tractography atlas [67-69]. Areas corresponding to significant clusters were identified using the JHU white-matter tractography atlas. Mean FA values were obtained from each subject's FA skeleton map and skeletonised SLF and fornix ROIs. Mean FA values were obtained from each subject's FA skeleton map and skeletonized for all the fasciculi ROIs mentioned above.\nStatistical tests on non-imaging data were performed using SPSS (Statistical Package for the Social Sciences) v.16 (SPSS Inc., Chicago Illinois). Group differences were examined using the Student t-test, since the data were normally distributed using a significance level of p < 0.05.\nPartial correlation coefficients, controlling for the time of evolution, were used to explore the association between mean FA values and clinical variables and neuropsychological measures. Statistical significance was set at a two-tailed p ≤ 0.05.\nGroup comparisons and correlations with neuropsychological measures were performed using Randomise v2.1 from FSL [62,63]. As seen in table one, the time of evolution since injury was very heterogeneous. In order to control possible effects of this variable in the correlation results, time of evolution was entered as a non-interest variable in the matrix. The statistical threshold was set at p < 0.05 Family Wise Error (FWE) corrected, which is a conservative procedure that allows a high control of Type I error, being the probability of one or more false positives the same as the significance level. The Threshold-Free Cluster Enhancement (TFCE) method was used to define the clusters [64]. Correlation analyses were performed with the 2-back d-prime index and the Rivermead profile score using a region of interest (ROI) approach in the following associative fasciculi: corpus callosum, superior and inferior longitudinal, inferior fronto-occipital, uncinate, and cingulate as well as the fornix and arcuate fasciculi as the major pathways that connect associative cortical regions involved in working and declarative memory. ROI masks were obtained from the Jülich histological atlas [65,66] and the JHU white-matter tractography atlas [67-69]. Areas corresponding to significant clusters were identified using the JHU white-matter tractography atlas. Mean FA values were obtained from each subject's FA skeleton map and skeletonised SLF and fornix ROIs. Mean FA values were obtained from each subject's FA skeleton map and skeletonized for all the fasciculi ROIs mentioned above.\nStatistical tests on non-imaging data were performed using SPSS (Statistical Package for the Social Sciences) v.16 (SPSS Inc., Chicago Illinois). Group differences were examined using the Student t-test, since the data were normally distributed using a significance level of p < 0.05.\nPartial correlation coefficients, controlling for the time of evolution, were used to explore the association between mean FA values and clinical variables and neuropsychological measures. Statistical significance was set at a two-tailed p ≤ 0.05.", "A cross-sectional study of thirty-one subjects was performed. Fifteen patients (eleven male) with severe TBI were recruited from the Head Injury Unit of the Institut de Neurorehabilitació Guttmann. Inclusion criteria were: a) age < 40 years, b) diffuse axonal injury according to clinical MRI without focal cortical lesions or larger than 1.5 cm3, c) severe TBI: defined as a minimal Glasgow Coma Scale (GCS) score ≤ 8 assessed at the first contact with the emergency services, d) emergence from posttraumatic amnesia (PTA) phase at the moment of the enrollment according to the Galveston Orientation and Attention Test (GOAT) [52], defined as two consecutive scores > 65, and f) no previous history of TBI, drug intake, neurological, or psychiatric disorders.\nThe etiology of TBI was a traffic accident in all cases. Fourteen patients were involved in car collisions, and one was a pedestrian hit by a motor vehicle. All patients had closed head injury and had not received surgery for extra- or subdural hematoma; all structural MRI scans were suggestive of TAI. The neuroradiologist (NB) took into account T1-weighted, FLAIR, and T2* GE sequences. The T2* GE sequences, which have a high level of sensitivity for detecting chronic hemosiderin, indicated evidence of TAI-related neuropathology. The method proposed by Gennarelli et al. [2] was used to classify the patients' TAI type. The grading system used was: type I, TAI only involving convexity gray-white matter junction; type II, also involving the corpus callosum in addition to the gray-white junction; and, type III, involving the rostral brainstem as well as the two previous criteria. Cases in which the midbrain was involved, but no corpus callosum lesions were apparent, were classified as type III (see Table 1).\nClinical and neuroimaging characteristics of the TBI group\nPT: Patient; GCS: Glasgow coma scale; PTA: posttraumatic amnesia; CT: computer tomography; MRI: magnetic resonance imaging; tevol: time of evolution since accident to the MRI evaluation; TAI: diffuse axonal injury; R/L: right/left; CC: corpus callosum; SHA: subarachnoidal hemorrhage.\nA control group of sixteen healthy subjects (nine male) were recruited from relatives and friends of the TBI group. This control group was matched by age, years of education and premorbid intellectual function estimated using the Vocabulary subtest of the Wechsler Adult Intelligence Scale (WAIS-III) [53], recognized as an efficient method for estimating general intelligence [54] (see Table 2). All subjects were right-handed, Caucasian-Mediterranean, and none had a previous history of neurological or psychiatric diseases.\nDemographic and clinical characteristics of TBI and control groups\nTBI = traumatic brain injury.\nThe study was approved by the Ethical and Research Committee of the Institut Universitari de Neurorehabilitacio Guttmann and all participants gave written informed consent.", "Working memory was evaluated by the Digit span and Letter-Number Sequencing (LNS) subtests of the WAIS-III [53] and a visual 2-back task [55]. Digit span was measured as the series length correctly reproduced at least once in the same order (forwards) and in reverse order (backwards). In the LNS, subjects heard lists of randomized numbers and letters (in alternating order) of increasing lengths, and were asked to reproduce the numbers and letters beginning with the lowest in each series, always with numbers first. The scores from the LNS were calculated by adding all correct items. In the 2-back task, numbers appeared on the screen for 500 ms against a black background, followed by a fixation cross for 1500 ms. The subjects were asked to decide whether the number they were looking at matched the one that they had seen two numbers earlier in the sequence. The numbers of correct responses as well as the reaction time were recorded. The d-prime index, a bias-free measure that takes both correct answers and errors into account, was also calculated to determine the accuracy of performance.\nThe Rivermead Behavioural Memory Test (RBMT) [56] was selected for its ability to explore declarative memory, and its ecological validity in assessing TBI patients. This test consists of 11 subtests including the following: remembering a name, a hidden belonging and an appointment; recognizing pictures and faces; recalling a prose passage; remembering a short route; remembering to deliver a message; and knowledge of some basic information such as the date, place and time. Those are designed as analogs of everyday tasks, reflecting the kinds of situations with which patients typically experience difficulty on a day-to-day basis. Two methods of standardizing scores across subtests allow for derivation of either a screening score, with subtest raw scores categorized on a scale of 0 ± 1 (maximum score 12 points), or a standardized profile score, with subtest raw scores categorized on a scale of 0 ± 2 (maximum score 24 points). The slightly more fine-grained standardized profile score provides a more sensitive analysis of performance [57] thus we use this score for our correlation analysis.", "MRI data sets were acquired on a 1.5 T Signa GE (General Electric, Milwaukee, WI) at the Centre de Diagnostic per la Imatge of the Hospital Clínic (CDIC), Barcelona. Diffusion weighted images were sensitized in 25 non-collinear directions with a b-value = 1000 sec/mm2, using an echo-planar (EPI) sequence (TR = 9999.996 ms, TE = 85 ms, 20 axial slices with a resolution of 0.9375 × 0.9375 mm, slice thickness = 5 mm, gap = 2 mm matrix size = 128 × 128, FOV = 100).\nData preprocessing and analysis was performed using FMRIB's software library [FSL version 4.1; Oxford Centre for Functional MRI of the Brain (FMRIB), UK; http://www.fmrib.ox.ac.uk/fsl/]. Image artefacts due to eddy current distortions were minimized by registering the diffusion images to the b0 images. The registered images were skull-stripped using the Brain Extraction Tool (BET) [58]. Fractional anisotropy maps were calculated using the FMRIB's Diffusion Toolbox v.2.0 [FDT, [59]]. After calculation of the FA map for each subject, we implemented a voxel-wise statistical analysis of the FA data using Tract-Based Spatial Statistics v1.2 (TBSS) which aims to overcome the limitations of the standard VBM-style analyses [60], particularly those regarding to its dependence on the goodness of the registration algorithm and on the choice of the spatial smoothing [61]. FA data were aligned into a common space using a non-linear registration algorithm (FNIRT) to register the images to the standard FMRIB58 FA template, which is in MNI152 standard space. Aligned FA maps were visually inspected after registration and we confirmed that the result of the previous step was correct. Next, a mean FA image was created from the images from all the subjects in this common space and narrowed to generate a mean FA skeleton that represented the center of all tracts common to the entire group. This was thresholded to FA 0.2 to include the major white matter pathways but to exclude peripheral tracts where there was significant inter-subject variability and partial volume effects with gray matter. This ensured that each subject's skeleton was in the group space while also representing the center of the subject's unique white matter bundles. The aligned FA image for each subject was then projected onto the skeleton by filling the skeleton with FA values from the nearest relevant tract centre. This is achieved for each skeleton voxel by searching perpendicular to the local skeleton structure for the maximum value in the FA image of the subject. The resulting skeletonised data was then fed into voxelwise cross-subject statistics.", "Group comparisons and correlations with neuropsychological measures were performed using Randomise v2.1 from FSL [62,63]. As seen in table one, the time of evolution since injury was very heterogeneous. In order to control possible effects of this variable in the correlation results, time of evolution was entered as a non-interest variable in the matrix. The statistical threshold was set at p < 0.05 Family Wise Error (FWE) corrected, which is a conservative procedure that allows a high control of Type I error, being the probability of one or more false positives the same as the significance level. The Threshold-Free Cluster Enhancement (TFCE) method was used to define the clusters [64]. Correlation analyses were performed with the 2-back d-prime index and the Rivermead profile score using a region of interest (ROI) approach in the following associative fasciculi: corpus callosum, superior and inferior longitudinal, inferior fronto-occipital, uncinate, and cingulate as well as the fornix and arcuate fasciculi as the major pathways that connect associative cortical regions involved in working and declarative memory. ROI masks were obtained from the Jülich histological atlas [65,66] and the JHU white-matter tractography atlas [67-69]. Areas corresponding to significant clusters were identified using the JHU white-matter tractography atlas. Mean FA values were obtained from each subject's FA skeleton map and skeletonised SLF and fornix ROIs. Mean FA values were obtained from each subject's FA skeleton map and skeletonized for all the fasciculi ROIs mentioned above.\nStatistical tests on non-imaging data were performed using SPSS (Statistical Package for the Social Sciences) v.16 (SPSS Inc., Chicago Illinois). Group differences were examined using the Student t-test, since the data were normally distributed using a significance level of p < 0.05.\nPartial correlation coefficients, controlling for the time of evolution, were used to explore the association between mean FA values and clinical variables and neuropsychological measures. Statistical significance was set at a two-tailed p ≤ 0.05.", "[SUBTITLE] Comparison between TBI patients and controls [SUBSECTION] Performance on the memory tests is described in Table 3. Statistical significance was obtained for the difference in scores in the LNS subtest (WAIS-III), d-prime index for the 2-back, and the RBMT profile. Forward and backward digits did not reach statistical significance.\nNeuropsychological performance for TBI and control groups\nTBI: traumatic brain injury; Letter-Number Sequencing (WAIS-III); Goals: number of targets correctly identified; RBMT: Rivermead Behavioral Memory Test; ns: not significant.\nGroup comparison for FA skeleton maps revealed multiple areas of significant FA reductions in TBI patients as compared to controls. All the long associative fibers were affected, including the corpus callosum, the superior and inferior longitudinal fasciculi, and the inferior fronto-occipital fasciculi. Decreased FA was also observed in shorter fibers such as the uncinate fasciculus, cingulum, fornix and anterior thalamic radiation (Figure 1, Table 4). FA was not increased in the TBI group in any cerebral region.\nResults from TBSS analysis of FA maps showing the clusters of significantly reduced FA in TBI patients compared to controls in red (TFCE, p < 0.05 FWE-corrected). Widespread white matter affectation is observed.\nDifferences between groups in mean FA from the whole skeletonised brain and the ROIs\nTBI: Traumatic Brain Injury; FA: fractional anisotropy; CC: corpus callosum; SLF: superior longitudinal fasciculi; ILF: inferior longitudinal fasciculi; IFO: inferior fronto-occipital fasciculi.\nWe obtained mean FA values of the whole skeletonized brain and all of the selected ROIs. Group comparisons for all these values reached statistical significance in all cases with p < 0.001 (Table 4).\nPerformance on the memory tests is described in Table 3. Statistical significance was obtained for the difference in scores in the LNS subtest (WAIS-III), d-prime index for the 2-back, and the RBMT profile. Forward and backward digits did not reach statistical significance.\nNeuropsychological performance for TBI and control groups\nTBI: traumatic brain injury; Letter-Number Sequencing (WAIS-III); Goals: number of targets correctly identified; RBMT: Rivermead Behavioral Memory Test; ns: not significant.\nGroup comparison for FA skeleton maps revealed multiple areas of significant FA reductions in TBI patients as compared to controls. All the long associative fibers were affected, including the corpus callosum, the superior and inferior longitudinal fasciculi, and the inferior fronto-occipital fasciculi. Decreased FA was also observed in shorter fibers such as the uncinate fasciculus, cingulum, fornix and anterior thalamic radiation (Figure 1, Table 4). FA was not increased in the TBI group in any cerebral region.\nResults from TBSS analysis of FA maps showing the clusters of significantly reduced FA in TBI patients compared to controls in red (TFCE, p < 0.05 FWE-corrected). Widespread white matter affectation is observed.\nDifferences between groups in mean FA from the whole skeletonised brain and the ROIs\nTBI: Traumatic Brain Injury; FA: fractional anisotropy; CC: corpus callosum; SLF: superior longitudinal fasciculi; ILF: inferior longitudinal fasciculi; IFO: inferior fronto-occipital fasciculi.\nWe obtained mean FA values of the whole skeletonized brain and all of the selected ROIs. Group comparisons for all these values reached statistical significance in all cases with p < 0.001 (Table 4).\n[SUBTITLE] Correlation analysis [SUBSECTION] [SUBTITLE] Correlation with clinical variables [SUBSECTION] We observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\nWe observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\n[SUBTITLE] Correlation with declarative and working memory performance [SUBSECTION] The mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.\nThe mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.\n[SUBTITLE] Correlation with clinical variables [SUBSECTION] We observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\nWe observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\n[SUBTITLE] Correlation with declarative and working memory performance [SUBSECTION] The mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.\nThe mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.", "Performance on the memory tests is described in Table 3. Statistical significance was obtained for the difference in scores in the LNS subtest (WAIS-III), d-prime index for the 2-back, and the RBMT profile. Forward and backward digits did not reach statistical significance.\nNeuropsychological performance for TBI and control groups\nTBI: traumatic brain injury; Letter-Number Sequencing (WAIS-III); Goals: number of targets correctly identified; RBMT: Rivermead Behavioral Memory Test; ns: not significant.\nGroup comparison for FA skeleton maps revealed multiple areas of significant FA reductions in TBI patients as compared to controls. All the long associative fibers were affected, including the corpus callosum, the superior and inferior longitudinal fasciculi, and the inferior fronto-occipital fasciculi. Decreased FA was also observed in shorter fibers such as the uncinate fasciculus, cingulum, fornix and anterior thalamic radiation (Figure 1, Table 4). FA was not increased in the TBI group in any cerebral region.\nResults from TBSS analysis of FA maps showing the clusters of significantly reduced FA in TBI patients compared to controls in red (TFCE, p < 0.05 FWE-corrected). Widespread white matter affectation is observed.\nDifferences between groups in mean FA from the whole skeletonised brain and the ROIs\nTBI: Traumatic Brain Injury; FA: fractional anisotropy; CC: corpus callosum; SLF: superior longitudinal fasciculi; ILF: inferior longitudinal fasciculi; IFO: inferior fronto-occipital fasciculi.\nWe obtained mean FA values of the whole skeletonized brain and all of the selected ROIs. Group comparisons for all these values reached statistical significance in all cases with p < 0.001 (Table 4).", "[SUBTITLE] Correlation with clinical variables [SUBSECTION] We observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\nWe observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).\n[SUBTITLE] Correlation with declarative and working memory performance [SUBSECTION] The mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.\nThe mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.", "We observed significant negative correlations between FA and posttraumatic amnesia (PTA) in almost all the regions that showed significant FA decreases in the group analysis. Quantitative global mean FA values also showed a high correlation with this variable (r = -0.903 p <0.001). However, no significant correlations were found in the FA maps analysis for the GCS (r= 0.206, p = 0.499).", "The mean global FA measure correlated significantly with 2-back d-prime index (r = 0.584, p = 0.028). The correlation of global FA with the RBMT profile score did not reach statistical significance.\nThe ROI procedure revealed a positive correlation between working memory performance assessed by the 2-back d- index and the FA skeletonized SLF, fornix, and corpus callosum ROIs (Figure 2, Table 5). 2-back d-prime index also correlated with the arcuate fascicle (Table 5). Declarative memory performance, assessed by RBMT, correlated with the fornix and the posterior part of the corpus callosum ROIs (Figure 3, Table 5).\nROI correlations with d-prime 2-back index in the TBI group for the SLF, fornix, and corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nTBSS results. Correlation with working and declarative memory measures in the ROIs in the TBI group\nAll regions identified at family wise error corrected p < 0.05 with a cluster size of at least 10 voxels. Coordinates are given at peak voxel coordinate in MNI-152 standard space.\nROI correlations with RBMT in the TBI group: RBMT correlated with the fornix and the corpus callosum ROIs (TFCE, p < 0.05 FWE-corrected). Correlation coefficient (r) was directly converted from t values of the TBSS output. The t and r values correspond to the most statistically significant voxel for each cluster.\nNo other correlation reached statistical significance in the TBI group. There were no significant correlations between FA and neuropsychological measures in the control group.", "The present study provides evidence of the implications of TAI in declarative and working memory deficits in TBI. DTI group comparison revealed global whole brain reductions in mean FA values for patients and FA maps confirmed that almost all the major fibers were involved. Although our patients suffered global white matter integrity impairment, we found two different and restricted patterns of correlations with the FA and neuropsychological assessment. Whole brain DTI analysis showed that decreased FA throughout the brain correlated with 2-back measures but not with the Rivermead Test. Results from the ROI analyses of the main association fibers showed, as predicted, that working memory specifically correlated with the superior longitudinal fasciculi. However, it was also found to correlate with the corpus callosum, the arcuate fasciculi and with the fornix. On the other hand, declarative memory deficits only correlated with the fornix, as we had expected, and the corpus callosum. These results suggest that there are two different patterns of FA reduction related with two types of memory dysfunctions.\nWe found that superior longitudinal fasciculi damage is related with working memory but not with declarative memory deficits. These correlations were expectable since the longitudinal fasciculi connect the associative frontal and parietal regions involved in working memory functions [70-72,49]. The correlation between working memory deficit and the superior longitudinal was also described in multiple sclerosis, pathology that also involves white matter damage [50].\nIn our sample, FA reductions of corpus callosum correlated with both working and declarative memory impairments. In declarative memory the correlations were seen in the posterior region whereas in working memory the correlations involved anterior and posterior regions, thus again these results point to differential patterns of correlations for both types of memory impairment.\nAccording our results, declarative memory impairment did not depend on diffuse white matter damage since no correlations between FA maps or mean values and declarative memory values were seen. However, the ROI analysis revealed that the fornix FA impairment correlated with the Rivermead test. This result is in agreement with the role of the damage of the hippocampus and its connections in declarative memory deficits in TBI [45,46].\nOur declarative memory results partially agree with those obtained by Salmond et al. [30]. Using a voxel-based analysis with SPM tools, these authors found a significant positive correlation between declarative memory and diffusivity in the left hippocampal formation, the left posterior cingulate, and the left frontal, temporal and occipital regions. The more widespread pattern of correlations observed in their study can be explained by the use of FDR correction, which is more liberal than the FWE correction used in ours [73]. Correlations between FA values in the fornix and declarative memory impairment have been also observed in patients with multiple sclerosis [74]. Other studies investigating FA correlations with declarative memory functions in mild TBI samples have reported significant correlations with the uncinate fasciculi [30,32] and the cingulum [28]. Although we found decreased FA in these fasciculi, correlations did not reach statistical significance. These discrepancies may be explained by the varying grade of severity of the samples, the difference in the memory tests used, and DTI methodological differences.\nIn the present study, working memory deficits also correlated with the fornix in both the whole brain analysis and the ROI analyses. There is some evidence from fMRI studies that the hippocampus is involved in working memory functions in healthy subjects [75-78]. Moreover, several animal studies also suggest a role for the hippocampus in working memory [79-81]. Anatomically, prefrontal regions involved in working memory tasks receive projections from the hippocampus [82,83] and are connected directly to the ventral hippocampus and indirectly to the dorsal hippocampus via the thalamus [84-86]. This structural connectivity supports the idea that the hippocampus has a role in working memory functioning as suggested by our findings.\nFinally, significant correlations were observed between the PTA variable and white matter integrity. Whole brain map analysis showed that PTA is an excellent index predictor of the degree of impairment of the major white matter tracts and association fibers. These results suggest that the recovery of memory functions is dependent on the integrity of the complex neocortical regions. Unlike previous studies [13,14,17], no correlations were found between GCS and FA maps or mean FA values. This result was to be expected as the fact that all our patients had severe TBI meant that GCS variability would not be sufficient to reach statistical significance.\nOur study has certain limitations and our results should be regarded as preliminary. The small sample size and its specific diffuse characteristics may preclude the generalization of the results. The presence of mixed focal and diffuse pathology frequently observed in severe TBI may confound the mapping of neural and behavioral changes in these patients. As our study sample excluded significant cortical pathology, the cognitive impairment observed is more likely to be due to the diffuse pathology alone. Nevertheless, we cannot exclude the possibility that reductions in gray matter in several subcortical structures are also influencing memory deficits in TBI.", "This DTI study suggests that declarative and working memory deficits in diffuse TBI patients are related to differential patterns of FA reduction. Working memory impairment reflects the diffuse white matter damage affecting large scale networks such as the superior longitudinal fasciculi, whereas declarative memory deficits seem to be the result of more local disruption of the cerebral circuitry.", "The authors declare that they have no competing interests.", "EP, DFE and CJ made substantial contribution to conception and design, interpretation of data, drafting and writing of manuscript and further revisions of the manuscript. Neuroimaging data were analyzed by EP and DFE. Neuroimaging sequence acquisitions, neurological description and classification of data by NB. RSC and TR participated in the collection of neuropsychological data and in the collection of acute clinical data. JT and PV made a critical revision of the manuscript for important intellectual content providing additional comments and contributions. CJ supervised the study. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2377/11/24/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Actins", "Adenocarcinoma", "Anti-Inflammatory Agents, Non-Steroidal", "Butyrates", "Carrier Proteins", "Cell Line", "Cell Movement", "Colitis, Ulcerative", "Colon", "Colonic Neoplasms", "Crohn Disease", "Humans", "Inflammatory Bowel Diseases", "Mesalamine", "Microfilament Proteins", "Regeneration" ]

[ "Background", "Ethical considerations", "Immunohistochemistry", "Statistical analysis", "Cell Lines and treatments", "Western Blot Analysis", "Cell motility assays", "Results", "Fascin expression is expressed in inflamed colonic epithelium", "Fascin is overexpressed in inflammatory bowel disease showing stronger and more widespread expression in Crohn's compared with ulcerative colitis", "Fascin is overexpressed in colonic epithelium actively undergoing restitution and regeneration", "5-aminosalicylate reduces fascin expression and cell motility whereas sodium butyrate has a converse effect", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Ulcerative colitis (UC) and Crohn's disease (CD) are forms of inflammatory bowel disease (IBD) which affect an estimated 1.4 million people in the USA and 2.2 million across Europe. UC exclusively affects the large intestine, whereas CD affects the colon in 60% of cases (known as Crohn's colitis), but can also involve other parts of the GI tract [1]. These common, chronic and debilitating conditions remain incurable, with their aetiology and pathogenesis not clearly understood. Long-term illness greatly increases the risk of colorectal cancer, which causes approximately 15% of all IBD patient deaths [2]. The onset of cancer in IBD patients is sudden, rapid, and highly aggressive, with a poor prognosis. The occurrence of dysplasia in IBD is widely accepted to be pre-malignant, but the likelihood of progression to cancer is difficult to predict [3]. Even the distinction between low grade- and high-grade dysplasia provides little indication of disease outcome [2] and there is, therefore, an urgent need for biomarkers to predict neoplasia in IBD.\nBoth UC and CD are characterised by mucosal infiltration of inflammatory cells and frequent epithelial damage. This damage can result in destruction of the mucosa and a breach in the barrier that this tissue provides against the luminal milieu. Complete remission of IBD requires both a reduction in inflammation and repair of the damaged epithelium. Inadequate or incomplete repair can result in the formation of a 'leaky barrier' which can in turn perpetuate a vicious cycle of chronic inflammation [4]. The process of 'healing' areas of the mucosa devastated by inflammation is widely accepted to be a two-stage process comprising 'restitution' and 'regeneration' [5]. Restitution is characterised by flattening and spreading of the epithelium at the margins of the ulcer, with these cells migrating across the denuded sub-mucosa to cover the damaged area. Following restitution, the regeneration programme requires widespread epithelial cell proliferation and formation of characteristic glandular structures of the intestinal crypts. Although several cytokines and growth factors have been implicated in the restoration of epithelium following injury, the cellular processes underpinning this mechanism remain poorly understood [5].\nCurrent therapy for IBD, particularly UC, centres on the long-term administration of the non-steroidal anti-inflammatory drug (NSAID) 5-amino salicylate (5-ASA). Shown to be effective in controlling intestinal inflammation in the majority of patients, 5-ASA also reduces colorectal cancer risk in patients with IBD [6,7].\nOther workers in the field have proposed sodium butyrate, a fermentation product of dietary fibre, as a potential therapy for IBD. Trials using butyrate irrigation led to symptomatic amelioration in UC patients [8,9]. Luminal levels of butyrate may be modulated through the dietary intake of fibre and are found in the millimolar range [10].\nAmong the most pressing current issues in the clinical management of IBD are a need to further our understanding of intestinal wound healing and to understand the effects of therapeutic modalities on tissue repair as this is crucial for disease remission.\nFascin (fascin-1) is a 55kDa actin-bundling protein which localises to the core actin bundles of spikes and filopodia at the leading edge of migratory cells. It is known to be overexpressed in various cancers and has been shown to increase motility in cells from several different tissues [11]. Work in this laboratory has shown fascin to be completely absent from the normal colorectal epithelium but widespread in colorectal tumours [11,12]. Fascin expression has been shown to correlate with an increased risk of malignant progression and with a poorer disease prognosis making it a potential biomarker in colorectal neoplasia [12,13].\nAs yet, there are no published reports of fascin expression in IBD, but it is our hypothesis that fascin will be involved in tissue repair in IBD. The changes in motile behaviour demonstrated by tumour cells mimic those occurring during tissue repair and require dynamic rearrangements in the actin cytoskeleton, governed by actin-binding proteins such as fascin.\nThe first aim of this study was to determine the expression of fascin in clinical samples of IBD with particular attention to areas of restitution and regeneration. Samples of resection margins from patients undergoing surgery for diverticulitis were included as these show low-grade mucosal inflammatory infiltration.\nSecondly, in order to elucidate the potential modulation of fascin by therapeutic intervention and the subsequent consequences for epithelial repair, we studied the effects of 5-ASA and sodium butyrate on fascin expression and cell motility using an in vitro cell line model.\nWe now show, for the first time, that fascin is overexpressed in IBD and expression is associated with regions of active mucosal repair. Furthermore, therapeutic modalities for IBD can affect fascin expression and colonic epithelial cell motility.", "Ethical approval was obtained from the North Somerset & South Bristol Research Ethics Committee (REC reference 04/Q2003/49). All tissue samples were obtained from the files of the Department of Histopathology, Bristol Royal Infirmary. These samples were anonymised by a third party and the investigators had no access to patient information.", "A total of 41 surgical specimens of resected colorectal tissue from IBD patients were immunohistochemically stained for fascin as described previously [11,12]. We also examined samples of normal colorectal mucosa and 11 samples of resection margins from patients undergoing surgery for diverticulitis as these display low grade mucosal inflammation.\nFascin was detected using a mouse monoclonal antibody (Dako-Cytomation, Glostrup, Denmark) as previously described [11,12]. Negative controls had no primary antibody applied.\nStained samples were scored by two independent observers in terms of the proportion of epithelial cells staining positive (1 for <20%, 2 for >20%) and also the intensity of epithelial staining relative to that observed in adjacent endothelial cells (0,1,2 for negative, weak, moderate/strong, respectively) which act as an internal positive control [11,12]. From the anonymised histology reports, disease activity was scored 0, 1, or 2 for quiescent, low/moderate, or severe activity, respectively.", "The immunohistochemistry scores were analysed for specific correlations using Kendall's tau B analysis as previously described [12]. Correlations were checked between proportion and intensity of fascin immunoreactivity, disease activity, and the presence of dysplasia. These analyses yield a value between zero and one (one being the strongest possible correlation and zero indicating no correlation at all) and a p-value to indicate the significance of the correlation. A p-value of less than 0.05 is considered statistically significant (*, p = < 0.05; **, p = < 0.01; ***, p = < 0.001).", "The HT29 cell line was originally derived from a sporadic colonic adenocarcinoma and was maintained in culture in DMEM supplemented with 10% FBS [14].\nFor treatments, 1 × 106 cells were seeded per 25 cm2 culture flask in triplicate for each dose and control, and the cells allowed to recover for 48 hours prior to treatment. 5-ASA (Sigma) was dissolved in DMEM supplemented with 2% FBS to produce a 50 mM solution and the pH adjusted to 7.4. This stock solution was diluted further in DMEM (2% FBS) for cell treatments. The 5-ASA was made up fresh immediately prior to each experiment and protected from light throughout the procedure. Sodium butyrate (Sigma) treatments were carried out as described previously [15]. Apoptosis was assessed as previously described [15]. The level of apoptosis in cultured colon cells can be assessed by measuring the proportion of cells that detach from the flask and float in the medium [15 and references therein] Apoptosis was confirmed in these floating cells by morphology (following acridine orange staining). Acridine orange staining was performed as previously described with at least 300 cells scored for each sample. The proportion of cells exhibiting apoptotic morphology was >90% in the floating cell population, and this proportion remained constant following treatment. Therefore, the proportion of cells floating in the medium can be used as a measure of the extent of apoptosis in the culture. Western blotting analysis was performed on attached cells-thereby precluding apoptotic cells from this measure of fascin expression.", "Cell lysates of 2 × 106 cells were prepared for western blotting as described previously [16]. A mouse monoclonal antibody raised against fascin was obtained from Dako (Dako-Cytomation, Carpinteria, CA). Blots were subsequently probed with anti α-tubulin (Sigma, UK) to show equal sample loading.", "Cell migration assays were carried out using a transwell filter migration assay as previously described [12]. The lower chamber was filled with Calcium Free-DMEM supplemented with 5% FBS to act as an attractant. Following a 24 hour incubation and staining with haemotoxylin, cells on the lower filter surface were considered migratory and counted in 10 fields at ×20 magnification.", "[SUBTITLE] Fascin expression is expressed in inflamed colonic epithelium [SUBSECTION] Previous studies have shown that fascin is not expressed in the epithelial cells of the normal colonic mucosa, but is broadly expressed in the endothelial cells, fibroblasts and infiltrating lymphocytes of the lamina propria and submucosa [11,12]. Our previous study of colorectal adenomas showed epithelial fascin expression focussed around the tumour stalk, provoking the hypothesis that fascin expression may be modulated by inflammatory mediators [12]. In order to test this hypothesis immunohistochemistry was first performed on 11 tissue samples of resection margins from patients undergoing surgery for diverticulitis as these samples show low level inflammation. In 10 out of the 11, epithelial fascin immunoreactivity was focally observed in epithelial cells at the very base of the colonic crypts (Figure 1A). This observation is the first time that fascin expression has been observed in non-neoplastic colorectal epithelium and led us to question the relationship between fascin and inflammatory conditions of the colon.\nFascin is overexpressed in inflammatory bowel disease and in colonic epithelium actively undergoing restitution and regeneration. Fascin immunoreactivity was detected in specimens of human colorectal mucosa by immunoperoxidase staining. A: Diverticulitis resection margin showing positive staining in the crypt base - 200 × magnification. B: Ulcerative colitis showing positive staining in the base of the elongated crypts - 100 ×. C: Crohn's colitis showing strong positive staining toward the base of the elongated crypts - 100 ×. D: Crohn's colitis showing patchy fascin immunoreactivity throughout the gland - 200 ×. E: Ulcerative colitis with low-grade dysplasia showing strong and widespread fascin expression - 100 ×. F: Ulcerative colitis with high-grade dysplasia showing strong and widespread fascin expression - 200 ×. G: Ulcerative colitis showing fascin staining in the flattened epithelial cells undertaking restitution - 200 ×. H: enlargement of an area of panel G illustrating the fascin-positive epithelial cells (arrowheads). I: fascin positive epithelial cells undertaking restitution in a sample of Crohn's colitis. J: A regenerative polyp in a sample of ulcerative colitis showing fascin positivity in the newly forming crypts - 200 ×. K & L: Fascin expression in branching crypts (arrowheads) in regenerating ulcerative colitis tissue - 200 ×.\nPrevious studies have shown that fascin is not expressed in the epithelial cells of the normal colonic mucosa, but is broadly expressed in the endothelial cells, fibroblasts and infiltrating lymphocytes of the lamina propria and submucosa [11,12]. Our previous study of colorectal adenomas showed epithelial fascin expression focussed around the tumour stalk, provoking the hypothesis that fascin expression may be modulated by inflammatory mediators [12]. In order to test this hypothesis immunohistochemistry was first performed on 11 tissue samples of resection margins from patients undergoing surgery for diverticulitis as these samples show low level inflammation. In 10 out of the 11, epithelial fascin immunoreactivity was focally observed in epithelial cells at the very base of the colonic crypts (Figure 1A). This observation is the first time that fascin expression has been observed in non-neoplastic colorectal epithelium and led us to question the relationship between fascin and inflammatory conditions of the colon.\nFascin is overexpressed in inflammatory bowel disease and in colonic epithelium actively undergoing restitution and regeneration. Fascin immunoreactivity was detected in specimens of human colorectal mucosa by immunoperoxidase staining. A: Diverticulitis resection margin showing positive staining in the crypt base - 200 × magnification. B: Ulcerative colitis showing positive staining in the base of the elongated crypts - 100 ×. C: Crohn's colitis showing strong positive staining toward the base of the elongated crypts - 100 ×. D: Crohn's colitis showing patchy fascin immunoreactivity throughout the gland - 200 ×. E: Ulcerative colitis with low-grade dysplasia showing strong and widespread fascin expression - 100 ×. F: Ulcerative colitis with high-grade dysplasia showing strong and widespread fascin expression - 200 ×. G: Ulcerative colitis showing fascin staining in the flattened epithelial cells undertaking restitution - 200 ×. H: enlargement of an area of panel G illustrating the fascin-positive epithelial cells (arrowheads). I: fascin positive epithelial cells undertaking restitution in a sample of Crohn's colitis. J: A regenerative polyp in a sample of ulcerative colitis showing fascin positivity in the newly forming crypts - 200 ×. K & L: Fascin expression in branching crypts (arrowheads) in regenerating ulcerative colitis tissue - 200 ×.\n[SUBTITLE] Fascin is overexpressed in inflammatory bowel disease showing stronger and more widespread expression in Crohn's compared with ulcerative colitis [SUBSECTION] Having shown that epithelial fascin expression was observed in the presence of low-level colonic inflammation, fascin immunohistochemistry was performed on 41 samples of resected colorectal mucosa from patients suffering from IBD. Fascin was widely overexpressed in the epithelium of IBD-involved tissue and representative results are shown in Figure 1. The data from the subsequent scoring of these samples is summarised in Figure 2. Strong fascin staining was frequently observed toward the crypt bases in the elongated crypts associated with IBD (Figure 1B and 1C) but also showed a more patchy distribution throughout the glands (Figure 1D).\nFascin expression is stronger and more widespread in Crohn's colitis compared with Ulcerative colitis. A graphical summary of the immunohistochemical study of fascin expression in IBD. Samples were scored high or low for the proportion of epithelium staining positive (low = <20%; high = >20%), and the intensity of the epithelial fascin immunoreactivity. The graphs illustrate a significant increase in both the proportion (Ttest p = 0.0017) and intensity (Ttest p = 0.0013) of fascin immunoreactivity in Crohn's compared with ulcerative colitis samples.\nStrikingly, following scoring of the immunostaining, it was found that both the proportion of positive epithelial cells and the intensity of fascin staining (relative to internal positive control) was significantly greater in samples of Crohn's compared with UC (Figure 2). Staining intensity also positively correlated with disease activity (as determined by the consulting pathologist) for both conditions (Tau B = 0.223, p = <0.05). The intensity of fascin stain observed in the sample was also found to correlate with the proportion of fascin-positive epithelium (Tau B = 0.621, p = < 0.001).\nPrevious study has shown that fascin is overexpressed in both the benign and malignant stages of colorectal neoplasia and is associated with malignant progression of these tumours [11,12]. Increased risk of colorectal cancer is an important clinical consideration in IBD, and malignant progression of areas of dysplasia remains highly unpredictable [2]. In this study, strong fascin expression was observed in IBD samples showing either low or high grade dysplasia (Figure 1E), or cancer. Both the intensity of epithelial fascin immunoreactivity (Tau B = 0.391, p = <0.05) and the proportion (Tau B = 0.406, p = < 0.01) of positive epithelial cells in the tissue significantly correlated with the presence of areas of dysplastic epithelium.\nHaving shown that epithelial fascin expression was observed in the presence of low-level colonic inflammation, fascin immunohistochemistry was performed on 41 samples of resected colorectal mucosa from patients suffering from IBD. Fascin was widely overexpressed in the epithelium of IBD-involved tissue and representative results are shown in Figure 1. The data from the subsequent scoring of these samples is summarised in Figure 2. Strong fascin staining was frequently observed toward the crypt bases in the elongated crypts associated with IBD (Figure 1B and 1C) but also showed a more patchy distribution throughout the glands (Figure 1D).\nFascin expression is stronger and more widespread in Crohn's colitis compared with Ulcerative colitis. A graphical summary of the immunohistochemical study of fascin expression in IBD. Samples were scored high or low for the proportion of epithelium staining positive (low = <20%; high = >20%), and the intensity of the epithelial fascin immunoreactivity. The graphs illustrate a significant increase in both the proportion (Ttest p = 0.0017) and intensity (Ttest p = 0.0013) of fascin immunoreactivity in Crohn's compared with ulcerative colitis samples.\nStrikingly, following scoring of the immunostaining, it was found that both the proportion of positive epithelial cells and the intensity of fascin staining (relative to internal positive control) was significantly greater in samples of Crohn's compared with UC (Figure 2). Staining intensity also positively correlated with disease activity (as determined by the consulting pathologist) for both conditions (Tau B = 0.223, p = <0.05). The intensity of fascin stain observed in the sample was also found to correlate with the proportion of fascin-positive epithelium (Tau B = 0.621, p = < 0.001).\nPrevious study has shown that fascin is overexpressed in both the benign and malignant stages of colorectal neoplasia and is associated with malignant progression of these tumours [11,12]. Increased risk of colorectal cancer is an important clinical consideration in IBD, and malignant progression of areas of dysplasia remains highly unpredictable [2]. In this study, strong fascin expression was observed in IBD samples showing either low or high grade dysplasia (Figure 1E), or cancer. Both the intensity of epithelial fascin immunoreactivity (Tau B = 0.391, p = <0.05) and the proportion (Tau B = 0.406, p = < 0.01) of positive epithelial cells in the tissue significantly correlated with the presence of areas of dysplastic epithelium.\n[SUBTITLE] Fascin is overexpressed in colonic epithelium actively undergoing restitution and regeneration [SUBSECTION] Inflammatory infiltration in IBD frequently results in complete destruction of the mucosal layer resulting in areas of ulceration. Repair of these gaps in the epithelial barrier occurs as a two-stage process, termed restitution and regeneration [5]. In the sample group used, all of the 11 sections showing signs of active restitution stained positive for fascin in the flattened epithelial cells moving to cover the denuded area (Figure 1G, H).\nFascin staining was also observed in the newly formed immature crypts of regenerative polyps (Figure 1J) and in epithelial glands undergoing crypt fission - a relatively common observation in IBD tissue undertaking regeneration (Figure 1K). Taken together, these data suggest that fascin could play an important role in IBD and could be vital for disease remission through modulating mucosal repair.\nInflammatory infiltration in IBD frequently results in complete destruction of the mucosal layer resulting in areas of ulceration. Repair of these gaps in the epithelial barrier occurs as a two-stage process, termed restitution and regeneration [5]. In the sample group used, all of the 11 sections showing signs of active restitution stained positive for fascin in the flattened epithelial cells moving to cover the denuded area (Figure 1G, H).\nFascin staining was also observed in the newly formed immature crypts of regenerative polyps (Figure 1J) and in epithelial glands undergoing crypt fission - a relatively common observation in IBD tissue undertaking regeneration (Figure 1K). Taken together, these data suggest that fascin could play an important role in IBD and could be vital for disease remission through modulating mucosal repair.\n[SUBTITLE] 5-aminosalicylate reduces fascin expression and cell motility whereas sodium butyrate has a converse effect [SUBSECTION] We have established above that fascin is overexpressed in IBD, and that expression is associated with areas undertaking repair. As fascin is known to promote cell motility in neoplastic colorectal epithelial cells [12], and motility is a key factor in epithelial restitution, we aimed to determine the effect of therapeutic modalities used in the treatment of IBD on fascin expression and cell motility in vitro.\n5-aminosalicylate (5-ASA) is widely used in the clinical management of IBD, whereas the short chain fatty acid sodium butyrate, a luminal fermentation product of dietary fibre, has been proposed for use in IBD treatment [6,9].\nIn order to determine the effect of 5-ASA or butyrate on fascin expression we treated HT29 colorectal epithelial cells with a range of doses. Following 48 hours of treatment, the cells were counted and samples prepared in order to assay fascin expression by western blotting (Figure 3A). Both 5-ASA and butyrate reduced attached cell yield and induced apoptosis in the HT29 cells as described previously [15]. Apoptotic cells were not included in the analysis of fascin expression.\n5-aminosalicylate reduces fascin expression and cell motility in colorectal epithelial cells, whereas sodium butyrate has a converse effect. A: HT29 cells were treated with a range of doses of either 5-ASA or butyrate as indicated. Attached cell yields and percentage of apoptotic cells were determined and these data represent the mean of three independent experiments performed in triplicate ±SEM. Also shown are the subsequent western blot analyses of fascin expression. These blots are representative of three independent experiments and were re-probed for α-tubulin to show even sample loading. SW480 lysate was included as a positive control having been shown previously to express relatively high levels of fascin [Qualtrough et al., 2009]. B: The effect of 10 mM 5-ASA or 1 mM butyrate treatment on cell motility was determined on HT29 cells as measured by Boyden chamber assay using 5%FBS as an attractant (**, p = <0.01; ***, p = <0.001). Three independent experiments were carried out in triplicate and the data are expressed as the mean ± S.E.M.\n5-ASA treatment caused a decrease in fascin expression compared with control, whereas butyrate strongly enhanced fascin expression in a dose-dependent manner (Figure 3A). Promoter reporter assays using the full length fascin promoter [17] showed that regulation occurred, in both cases, at the transcriptional level (data not shown).\nIn order to determine whether 5-ASA and butyrate affect cell motility in this in vitro model, cells were seeded on collagen-coated transwell filters as previously described [12] and simultaneously treated with either 5-ASA or butyrate at doses which were shown not to significantly induce apoptosis (10 mM and 1 mM, respectively-Figure 3A). Motile cells were counted 24 hours later (Figure 3B). 5-ASA significantly reduced cell motility in HT29 cells whereas, conversely, butyrate stimulated cell migration.\nWe have established above that fascin is overexpressed in IBD, and that expression is associated with areas undertaking repair. As fascin is known to promote cell motility in neoplastic colorectal epithelial cells [12], and motility is a key factor in epithelial restitution, we aimed to determine the effect of therapeutic modalities used in the treatment of IBD on fascin expression and cell motility in vitro.\n5-aminosalicylate (5-ASA) is widely used in the clinical management of IBD, whereas the short chain fatty acid sodium butyrate, a luminal fermentation product of dietary fibre, has been proposed for use in IBD treatment [6,9].\nIn order to determine the effect of 5-ASA or butyrate on fascin expression we treated HT29 colorectal epithelial cells with a range of doses. Following 48 hours of treatment, the cells were counted and samples prepared in order to assay fascin expression by western blotting (Figure 3A). Both 5-ASA and butyrate reduced attached cell yield and induced apoptosis in the HT29 cells as described previously [15]. Apoptotic cells were not included in the analysis of fascin expression.\n5-aminosalicylate reduces fascin expression and cell motility in colorectal epithelial cells, whereas sodium butyrate has a converse effect. A: HT29 cells were treated with a range of doses of either 5-ASA or butyrate as indicated. Attached cell yields and percentage of apoptotic cells were determined and these data represent the mean of three independent experiments performed in triplicate ±SEM. Also shown are the subsequent western blot analyses of fascin expression. These blots are representative of three independent experiments and were re-probed for α-tubulin to show even sample loading. SW480 lysate was included as a positive control having been shown previously to express relatively high levels of fascin [Qualtrough et al., 2009]. B: The effect of 10 mM 5-ASA or 1 mM butyrate treatment on cell motility was determined on HT29 cells as measured by Boyden chamber assay using 5%FBS as an attractant (**, p = <0.01; ***, p = <0.001). Three independent experiments were carried out in triplicate and the data are expressed as the mean ± S.E.M.\n5-ASA treatment caused a decrease in fascin expression compared with control, whereas butyrate strongly enhanced fascin expression in a dose-dependent manner (Figure 3A). Promoter reporter assays using the full length fascin promoter [17] showed that regulation occurred, in both cases, at the transcriptional level (data not shown).\nIn order to determine whether 5-ASA and butyrate affect cell motility in this in vitro model, cells were seeded on collagen-coated transwell filters as previously described [12] and simultaneously treated with either 5-ASA or butyrate at doses which were shown not to significantly induce apoptosis (10 mM and 1 mM, respectively-Figure 3A). Motile cells were counted 24 hours later (Figure 3B). 5-ASA significantly reduced cell motility in HT29 cells whereas, conversely, butyrate stimulated cell migration.", "Previous studies have shown that fascin is not expressed in the epithelial cells of the normal colonic mucosa, but is broadly expressed in the endothelial cells, fibroblasts and infiltrating lymphocytes of the lamina propria and submucosa [11,12]. Our previous study of colorectal adenomas showed epithelial fascin expression focussed around the tumour stalk, provoking the hypothesis that fascin expression may be modulated by inflammatory mediators [12]. In order to test this hypothesis immunohistochemistry was first performed on 11 tissue samples of resection margins from patients undergoing surgery for diverticulitis as these samples show low level inflammation. In 10 out of the 11, epithelial fascin immunoreactivity was focally observed in epithelial cells at the very base of the colonic crypts (Figure 1A). This observation is the first time that fascin expression has been observed in non-neoplastic colorectal epithelium and led us to question the relationship between fascin and inflammatory conditions of the colon.\nFascin is overexpressed in inflammatory bowel disease and in colonic epithelium actively undergoing restitution and regeneration. Fascin immunoreactivity was detected in specimens of human colorectal mucosa by immunoperoxidase staining. A: Diverticulitis resection margin showing positive staining in the crypt base - 200 × magnification. B: Ulcerative colitis showing positive staining in the base of the elongated crypts - 100 ×. C: Crohn's colitis showing strong positive staining toward the base of the elongated crypts - 100 ×. D: Crohn's colitis showing patchy fascin immunoreactivity throughout the gland - 200 ×. E: Ulcerative colitis with low-grade dysplasia showing strong and widespread fascin expression - 100 ×. F: Ulcerative colitis with high-grade dysplasia showing strong and widespread fascin expression - 200 ×. G: Ulcerative colitis showing fascin staining in the flattened epithelial cells undertaking restitution - 200 ×. H: enlargement of an area of panel G illustrating the fascin-positive epithelial cells (arrowheads). I: fascin positive epithelial cells undertaking restitution in a sample of Crohn's colitis. J: A regenerative polyp in a sample of ulcerative colitis showing fascin positivity in the newly forming crypts - 200 ×. K & L: Fascin expression in branching crypts (arrowheads) in regenerating ulcerative colitis tissue - 200 ×.", "Having shown that epithelial fascin expression was observed in the presence of low-level colonic inflammation, fascin immunohistochemistry was performed on 41 samples of resected colorectal mucosa from patients suffering from IBD. Fascin was widely overexpressed in the epithelium of IBD-involved tissue and representative results are shown in Figure 1. The data from the subsequent scoring of these samples is summarised in Figure 2. Strong fascin staining was frequently observed toward the crypt bases in the elongated crypts associated with IBD (Figure 1B and 1C) but also showed a more patchy distribution throughout the glands (Figure 1D).\nFascin expression is stronger and more widespread in Crohn's colitis compared with Ulcerative colitis. A graphical summary of the immunohistochemical study of fascin expression in IBD. Samples were scored high or low for the proportion of epithelium staining positive (low = <20%; high = >20%), and the intensity of the epithelial fascin immunoreactivity. The graphs illustrate a significant increase in both the proportion (Ttest p = 0.0017) and intensity (Ttest p = 0.0013) of fascin immunoreactivity in Crohn's compared with ulcerative colitis samples.\nStrikingly, following scoring of the immunostaining, it was found that both the proportion of positive epithelial cells and the intensity of fascin staining (relative to internal positive control) was significantly greater in samples of Crohn's compared with UC (Figure 2). Staining intensity also positively correlated with disease activity (as determined by the consulting pathologist) for both conditions (Tau B = 0.223, p = <0.05). The intensity of fascin stain observed in the sample was also found to correlate with the proportion of fascin-positive epithelium (Tau B = 0.621, p = < 0.001).\nPrevious study has shown that fascin is overexpressed in both the benign and malignant stages of colorectal neoplasia and is associated with malignant progression of these tumours [11,12]. Increased risk of colorectal cancer is an important clinical consideration in IBD, and malignant progression of areas of dysplasia remains highly unpredictable [2]. In this study, strong fascin expression was observed in IBD samples showing either low or high grade dysplasia (Figure 1E), or cancer. Both the intensity of epithelial fascin immunoreactivity (Tau B = 0.391, p = <0.05) and the proportion (Tau B = 0.406, p = < 0.01) of positive epithelial cells in the tissue significantly correlated with the presence of areas of dysplastic epithelium.", "Inflammatory infiltration in IBD frequently results in complete destruction of the mucosal layer resulting in areas of ulceration. Repair of these gaps in the epithelial barrier occurs as a two-stage process, termed restitution and regeneration [5]. In the sample group used, all of the 11 sections showing signs of active restitution stained positive for fascin in the flattened epithelial cells moving to cover the denuded area (Figure 1G, H).\nFascin staining was also observed in the newly formed immature crypts of regenerative polyps (Figure 1J) and in epithelial glands undergoing crypt fission - a relatively common observation in IBD tissue undertaking regeneration (Figure 1K). Taken together, these data suggest that fascin could play an important role in IBD and could be vital for disease remission through modulating mucosal repair.", "We have established above that fascin is overexpressed in IBD, and that expression is associated with areas undertaking repair. As fascin is known to promote cell motility in neoplastic colorectal epithelial cells [12], and motility is a key factor in epithelial restitution, we aimed to determine the effect of therapeutic modalities used in the treatment of IBD on fascin expression and cell motility in vitro.\n5-aminosalicylate (5-ASA) is widely used in the clinical management of IBD, whereas the short chain fatty acid sodium butyrate, a luminal fermentation product of dietary fibre, has been proposed for use in IBD treatment [6,9].\nIn order to determine the effect of 5-ASA or butyrate on fascin expression we treated HT29 colorectal epithelial cells with a range of doses. Following 48 hours of treatment, the cells were counted and samples prepared in order to assay fascin expression by western blotting (Figure 3A). Both 5-ASA and butyrate reduced attached cell yield and induced apoptosis in the HT29 cells as described previously [15]. Apoptotic cells were not included in the analysis of fascin expression.\n5-aminosalicylate reduces fascin expression and cell motility in colorectal epithelial cells, whereas sodium butyrate has a converse effect. A: HT29 cells were treated with a range of doses of either 5-ASA or butyrate as indicated. Attached cell yields and percentage of apoptotic cells were determined and these data represent the mean of three independent experiments performed in triplicate ±SEM. Also shown are the subsequent western blot analyses of fascin expression. These blots are representative of three independent experiments and were re-probed for α-tubulin to show even sample loading. SW480 lysate was included as a positive control having been shown previously to express relatively high levels of fascin [Qualtrough et al., 2009]. B: The effect of 10 mM 5-ASA or 1 mM butyrate treatment on cell motility was determined on HT29 cells as measured by Boyden chamber assay using 5%FBS as an attractant (**, p = <0.01; ***, p = <0.001). Three independent experiments were carried out in triplicate and the data are expressed as the mean ± S.E.M.\n5-ASA treatment caused a decrease in fascin expression compared with control, whereas butyrate strongly enhanced fascin expression in a dose-dependent manner (Figure 3A). Promoter reporter assays using the full length fascin promoter [17] showed that regulation occurred, in both cases, at the transcriptional level (data not shown).\nIn order to determine whether 5-ASA and butyrate affect cell motility in this in vitro model, cells were seeded on collagen-coated transwell filters as previously described [12] and simultaneously treated with either 5-ASA or butyrate at doses which were shown not to significantly induce apoptosis (10 mM and 1 mM, respectively-Figure 3A). Motile cells were counted 24 hours later (Figure 3B). 5-ASA significantly reduced cell motility in HT29 cells whereas, conversely, butyrate stimulated cell migration.", "We show here, for the first time, that the actin bundling protein fascin is overexpressed in IBD, and may play an important role in tissue repair following inflammatory damage. Our in vitro findings suggest that therapeutic approaches for IBD could influence fascin expression, subsequently producing unwanted side-effects on tissue repair and thereby influencing achievement of remission.\nWe have shown previously that fascin is absent from the normal colorectal epithelium but overexpressed in both benign and malignant tumours where it is associated with enhanced cell motility [11,12]. Our previous findings in colorectal adenomas, where the tumour stalk frequently showed the strongest fascin immunoreactivity, suggest fascin regulation occurs through epithelial-mesenchymal crosstalk, potentially through inflammatory mediators [12]. This notion of inflammatory induction of fascin expression in the epithelium is supported here by the widespread expression of fascin in tissue resected from patients with IBD. A positive correlation was observed between the disease activity recorded by the pathologist and the intensity of fascin immunoreactivity further supporting the idea of an inflammation-mediated regulation of fascin expression. Also in keeping with these results is the observation of epithelial fascin expression at the crypt base in tissue from patients undergoing surgery for diverticulitis, which display low-grade inflammation.\nThe increased proportion and intensity of fascin staining observed in samples of Crohn's compared with UC raises questions about the fundamental difference between these classifications of IBD. As it was only the epithelial immunoreactivity that was scored, this difference cannot be attributed to the ability of Crohn's colitis to affect deeper layers of the bowel wall than UC, but rather resides in the mechanism of fascin regulation.\nIBD carries an increased risk of colorectal cancer, resulting in the deaths of 15% of IBD patients [2]. Malignant progression is highly unpredictable and therefore the identification of prognostic biomarkers of tumourigenic potential is of key importance [2]. In this study we showed that both the proportion and intensity of epithelial fascin stain significantly correlated with the presence of dysplasia or cancer. Further study, of a larger sample group with long-term clinical follow-up, will be necessary to confirm the validity of fascin as a biomarker for tumorigenesis in IBD. The role of fascin in the malignant progression of sporadic colorectal tumours [12] does suggest that fascin merits further investigation in the neoplastic transformation of IBD.\nOne of the most intriguing observations in our study of fascin expression in IBD was the focal expression of fascin in regions demonstrating active tissue repair. Fascin expression was found in all areas of restitution observed, and also in regenerative polyps and branching crypts, suggesting a role throughout the various stages of the repair process. Our previous work has shown that fascin can promote motility in benign, as well as malignant colorectal epithelial cells [12], and it would have been reasonable to hypothesise that fascin would be expressed in the motile cells undertaking restitution. Furthermore, fascin has been shown to modulate cell adhesion and could therefore also function in the dynamic adhesive changes required for both restitution and crypt fission [11]. These data suggest that regulation of fascin could be important in achieving and maintaining remission in patients with IBD.\nOne of the most broadly used treatments for the maintenance of remission in IBD is 5-ASA and previous work has shown this drug to induce apoptosis in HT29 cells in a dose-dependent manner [18]. We show here for the first time that treatment of the colorectal epithelial cell line HT29 with this drug led to a decrease in fascin expression and a retardation of cellular motility.\nLong term use of 5-ASA in patients with IBD is known to reduce cancer risk [19,20]. We postulate based on our findings here that this chemopreventive effect may, in part, be explained by the down regulation of fascin expression-with fascin being known to be involved in malignant progression of the colorectal epithelium [12]. This repression of cellular motility represents a key anti-cancer effect of 5-ASA treatment.\nConversely, considering the potential role of fascin in tissue repair, this would suggest that 5-ASA could hinder wound repair and thus impede remission in patients with active disease. The published evidence is somewhat conflicting as there are reports to suggest that NSAIDs can both impair wound healing [21] and that 5-ASA can promote intestinal repair [22]. However, the possibility exists that 5-ASA could have some detrimental effects in IBD patients in addition to its obvious benefits. This paradox in the effect of 5-ASA, and also the role of fascin, could have important consequences for the clinical management of IBD. To this end, our findings suggest that increased luminal levels of butyrate could complement 5-ASA in promoting remission whilst maintaining potent chemopreventive effects against tumorigenesis.\nPrevious work has suggested that sodium butyrate could have beneficial effects for IBD sufferers [8,9]. This fermentation product of dietary fibre is known to be present in the colonic lumen in the millimolar range and levels can be modulated by intake of suitable dietary substrate [10]. We, and others, have previously shown that butyrate has potent chemopreventive effects against colorectal tumorigenesis through the modulation of differentiation and apoptosis in colorectal epithelial cells [15,23,24].\nWe show here that butyrate upregulates fascin expression and significantly stimulates colorectal epithelial cell motility. These data suggest that butyrate will aid colonic tissue repair and therefore speed remission in IBD. This short chain fatty acid has already been proposed to promote wound healing in the small intestine [25]. Although our data suggest that fascin may be involved in tumorigenesis, butyrate has been shown to elicit potent anti-tumour effects. We propose that the combinatorial use of butyrate, or dietary modulation of its levels, could favourably complement 5-ASA use in the clinical management of IBD.", "We have shown for the first time that fascin is important in the pathogenesis and remission of IBD and could have important implications for its clinical management as fascin can potentially be manipulated using existing therapeutic approaches and modulation of diet.", "The authors declare that they have no competing interests.", "DQ: Conceived, designed, and coordinated the study and acquired the necessary funding; supervised the laboratory projects of KS & DL; carried out additional immunohistochemistry and all subsequent analyses; carried out some of the in vitro experiments; drafted the manuscript. KS & DL: Carried out the immunohistochemistry and some of the in vitro studies. MP: Contributed to the design and coordination of the study and aided with manuscript preparation. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-230X/11/14/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Ethical considerations", "Immunohistochemistry", "Statistical analysis", "Cell Lines and treatments", "Western Blot Analysis", "Cell motility assays", "Results", "Fascin expression is expressed in inflamed colonic epithelium", "Fascin is overexpressed in inflammatory bowel disease showing stronger and more widespread expression in Crohn's compared with ulcerative colitis", "Fascin is overexpressed in colonic epithelium actively undergoing restitution and regeneration", "5-aminosalicylate reduces fascin expression and cell motility whereas sodium butyrate has a converse effect", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Ulcerative colitis (UC) and Crohn's disease (CD) are forms of inflammatory bowel disease (IBD) which affect an estimated 1.4 million people in the USA and 2.2 million across Europe. UC exclusively affects the large intestine, whereas CD affects the colon in 60% of cases (known as Crohn's colitis), but can also involve other parts of the GI tract [1]. These common, chronic and debilitating conditions remain incurable, with their aetiology and pathogenesis not clearly understood. Long-term illness greatly increases the risk of colorectal cancer, which causes approximately 15% of all IBD patient deaths [2]. The onset of cancer in IBD patients is sudden, rapid, and highly aggressive, with a poor prognosis. The occurrence of dysplasia in IBD is widely accepted to be pre-malignant, but the likelihood of progression to cancer is difficult to predict [3]. Even the distinction between low grade- and high-grade dysplasia provides little indication of disease outcome [2] and there is, therefore, an urgent need for biomarkers to predict neoplasia in IBD.\nBoth UC and CD are characterised by mucosal infiltration of inflammatory cells and frequent epithelial damage. This damage can result in destruction of the mucosa and a breach in the barrier that this tissue provides against the luminal milieu. Complete remission of IBD requires both a reduction in inflammation and repair of the damaged epithelium. Inadequate or incomplete repair can result in the formation of a 'leaky barrier' which can in turn perpetuate a vicious cycle of chronic inflammation [4]. The process of 'healing' areas of the mucosa devastated by inflammation is widely accepted to be a two-stage process comprising 'restitution' and 'regeneration' [5]. Restitution is characterised by flattening and spreading of the epithelium at the margins of the ulcer, with these cells migrating across the denuded sub-mucosa to cover the damaged area. Following restitution, the regeneration programme requires widespread epithelial cell proliferation and formation of characteristic glandular structures of the intestinal crypts. Although several cytokines and growth factors have been implicated in the restoration of epithelium following injury, the cellular processes underpinning this mechanism remain poorly understood [5].\nCurrent therapy for IBD, particularly UC, centres on the long-term administration of the non-steroidal anti-inflammatory drug (NSAID) 5-amino salicylate (5-ASA). Shown to be effective in controlling intestinal inflammation in the majority of patients, 5-ASA also reduces colorectal cancer risk in patients with IBD [6,7].\nOther workers in the field have proposed sodium butyrate, a fermentation product of dietary fibre, as a potential therapy for IBD. Trials using butyrate irrigation led to symptomatic amelioration in UC patients [8,9]. Luminal levels of butyrate may be modulated through the dietary intake of fibre and are found in the millimolar range [10].\nAmong the most pressing current issues in the clinical management of IBD are a need to further our understanding of intestinal wound healing and to understand the effects of therapeutic modalities on tissue repair as this is crucial for disease remission.\nFascin (fascin-1) is a 55kDa actin-bundling protein which localises to the core actin bundles of spikes and filopodia at the leading edge of migratory cells. It is known to be overexpressed in various cancers and has been shown to increase motility in cells from several different tissues [11]. Work in this laboratory has shown fascin to be completely absent from the normal colorectal epithelium but widespread in colorectal tumours [11,12]. Fascin expression has been shown to correlate with an increased risk of malignant progression and with a poorer disease prognosis making it a potential biomarker in colorectal neoplasia [12,13].\nAs yet, there are no published reports of fascin expression in IBD, but it is our hypothesis that fascin will be involved in tissue repair in IBD. The changes in motile behaviour demonstrated by tumour cells mimic those occurring during tissue repair and require dynamic rearrangements in the actin cytoskeleton, governed by actin-binding proteins such as fascin.\nThe first aim of this study was to determine the expression of fascin in clinical samples of IBD with particular attention to areas of restitution and regeneration. Samples of resection margins from patients undergoing surgery for diverticulitis were included as these show low-grade mucosal inflammatory infiltration.\nSecondly, in order to elucidate the potential modulation of fascin by therapeutic intervention and the subsequent consequences for epithelial repair, we studied the effects of 5-ASA and sodium butyrate on fascin expression and cell motility using an in vitro cell line model.\nWe now show, for the first time, that fascin is overexpressed in IBD and expression is associated with regions of active mucosal repair. Furthermore, therapeutic modalities for IBD can affect fascin expression and colonic epithelial cell motility.", "[SUBTITLE] Ethical considerations [SUBSECTION] Ethical approval was obtained from the North Somerset & South Bristol Research Ethics Committee (REC reference 04/Q2003/49). All tissue samples were obtained from the files of the Department of Histopathology, Bristol Royal Infirmary. These samples were anonymised by a third party and the investigators had no access to patient information.\nEthical approval was obtained from the North Somerset & South Bristol Research Ethics Committee (REC reference 04/Q2003/49). All tissue samples were obtained from the files of the Department of Histopathology, Bristol Royal Infirmary. These samples were anonymised by a third party and the investigators had no access to patient information.\n[SUBTITLE] Immunohistochemistry [SUBSECTION] A total of 41 surgical specimens of resected colorectal tissue from IBD patients were immunohistochemically stained for fascin as described previously [11,12]. We also examined samples of normal colorectal mucosa and 11 samples of resection margins from patients undergoing surgery for diverticulitis as these display low grade mucosal inflammation.\nFascin was detected using a mouse monoclonal antibody (Dako-Cytomation, Glostrup, Denmark) as previously described [11,12]. Negative controls had no primary antibody applied.\nStained samples were scored by two independent observers in terms of the proportion of epithelial cells staining positive (1 for <20%, 2 for >20%) and also the intensity of epithelial staining relative to that observed in adjacent endothelial cells (0,1,2 for negative, weak, moderate/strong, respectively) which act as an internal positive control [11,12]. From the anonymised histology reports, disease activity was scored 0, 1, or 2 for quiescent, low/moderate, or severe activity, respectively.\nA total of 41 surgical specimens of resected colorectal tissue from IBD patients were immunohistochemically stained for fascin as described previously [11,12]. We also examined samples of normal colorectal mucosa and 11 samples of resection margins from patients undergoing surgery for diverticulitis as these display low grade mucosal inflammation.\nFascin was detected using a mouse monoclonal antibody (Dako-Cytomation, Glostrup, Denmark) as previously described [11,12]. Negative controls had no primary antibody applied.\nStained samples were scored by two independent observers in terms of the proportion of epithelial cells staining positive (1 for <20%, 2 for >20%) and also the intensity of epithelial staining relative to that observed in adjacent endothelial cells (0,1,2 for negative, weak, moderate/strong, respectively) which act as an internal positive control [11,12]. From the anonymised histology reports, disease activity was scored 0, 1, or 2 for quiescent, low/moderate, or severe activity, respectively.\n[SUBTITLE] Statistical analysis [SUBSECTION] The immunohistochemistry scores were analysed for specific correlations using Kendall's tau B analysis as previously described [12]. Correlations were checked between proportion and intensity of fascin immunoreactivity, disease activity, and the presence of dysplasia. These analyses yield a value between zero and one (one being the strongest possible correlation and zero indicating no correlation at all) and a p-value to indicate the significance of the correlation. A p-value of less than 0.05 is considered statistically significant (*, p = < 0.05; **, p = < 0.01; ***, p = < 0.001).\nThe immunohistochemistry scores were analysed for specific correlations using Kendall's tau B analysis as previously described [12]. Correlations were checked between proportion and intensity of fascin immunoreactivity, disease activity, and the presence of dysplasia. These analyses yield a value between zero and one (one being the strongest possible correlation and zero indicating no correlation at all) and a p-value to indicate the significance of the correlation. A p-value of less than 0.05 is considered statistically significant (*, p = < 0.05; **, p = < 0.01; ***, p = < 0.001).\n[SUBTITLE] Cell Lines and treatments [SUBSECTION] The HT29 cell line was originally derived from a sporadic colonic adenocarcinoma and was maintained in culture in DMEM supplemented with 10% FBS [14].\nFor treatments, 1 × 106 cells were seeded per 25 cm2 culture flask in triplicate for each dose and control, and the cells allowed to recover for 48 hours prior to treatment. 5-ASA (Sigma) was dissolved in DMEM supplemented with 2% FBS to produce a 50 mM solution and the pH adjusted to 7.4. This stock solution was diluted further in DMEM (2% FBS) for cell treatments. The 5-ASA was made up fresh immediately prior to each experiment and protected from light throughout the procedure. Sodium butyrate (Sigma) treatments were carried out as described previously [15]. Apoptosis was assessed as previously described [15]. The level of apoptosis in cultured colon cells can be assessed by measuring the proportion of cells that detach from the flask and float in the medium [15 and references therein] Apoptosis was confirmed in these floating cells by morphology (following acridine orange staining). Acridine orange staining was performed as previously described with at least 300 cells scored for each sample. The proportion of cells exhibiting apoptotic morphology was >90% in the floating cell population, and this proportion remained constant following treatment. Therefore, the proportion of cells floating in the medium can be used as a measure of the extent of apoptosis in the culture. Western blotting analysis was performed on attached cells-thereby precluding apoptotic cells from this measure of fascin expression.\nThe HT29 cell line was originally derived from a sporadic colonic adenocarcinoma and was maintained in culture in DMEM supplemented with 10% FBS [14].\nFor treatments, 1 × 106 cells were seeded per 25 cm2 culture flask in triplicate for each dose and control, and the cells allowed to recover for 48 hours prior to treatment. 5-ASA (Sigma) was dissolved in DMEM supplemented with 2% FBS to produce a 50 mM solution and the pH adjusted to 7.4. This stock solution was diluted further in DMEM (2% FBS) for cell treatments. The 5-ASA was made up fresh immediately prior to each experiment and protected from light throughout the procedure. Sodium butyrate (Sigma) treatments were carried out as described previously [15]. Apoptosis was assessed as previously described [15]. The level of apoptosis in cultured colon cells can be assessed by measuring the proportion of cells that detach from the flask and float in the medium [15 and references therein] Apoptosis was confirmed in these floating cells by morphology (following acridine orange staining). Acridine orange staining was performed as previously described with at least 300 cells scored for each sample. The proportion of cells exhibiting apoptotic morphology was >90% in the floating cell population, and this proportion remained constant following treatment. Therefore, the proportion of cells floating in the medium can be used as a measure of the extent of apoptosis in the culture. Western blotting analysis was performed on attached cells-thereby precluding apoptotic cells from this measure of fascin expression.\n[SUBTITLE] Western Blot Analysis [SUBSECTION] Cell lysates of 2 × 106 cells were prepared for western blotting as described previously [16]. A mouse monoclonal antibody raised against fascin was obtained from Dako (Dako-Cytomation, Carpinteria, CA). Blots were subsequently probed with anti α-tubulin (Sigma, UK) to show equal sample loading.\nCell lysates of 2 × 106 cells were prepared for western blotting as described previously [16]. A mouse monoclonal antibody raised against fascin was obtained from Dako (Dako-Cytomation, Carpinteria, CA). Blots were subsequently probed with anti α-tubulin (Sigma, UK) to show equal sample loading.\n[SUBTITLE] Cell motility assays [SUBSECTION] Cell migration assays were carried out using a transwell filter migration assay as previously described [12]. The lower chamber was filled with Calcium Free-DMEM supplemented with 5% FBS to act as an attractant. Following a 24 hour incubation and staining with haemotoxylin, cells on the lower filter surface were considered migratory and counted in 10 fields at ×20 magnification.\nCell migration assays were carried out using a transwell filter migration assay as previously described [12]. The lower chamber was filled with Calcium Free-DMEM supplemented with 5% FBS to act as an attractant. Following a 24 hour incubation and staining with haemotoxylin, cells on the lower filter surface were considered migratory and counted in 10 fields at ×20 magnification.", "Ethical approval was obtained from the North Somerset & South Bristol Research Ethics Committee (REC reference 04/Q2003/49). All tissue samples were obtained from the files of the Department of Histopathology, Bristol Royal Infirmary. These samples were anonymised by a third party and the investigators had no access to patient information.", "A total of 41 surgical specimens of resected colorectal tissue from IBD patients were immunohistochemically stained for fascin as described previously [11,12]. We also examined samples of normal colorectal mucosa and 11 samples of resection margins from patients undergoing surgery for diverticulitis as these display low grade mucosal inflammation.\nFascin was detected using a mouse monoclonal antibody (Dako-Cytomation, Glostrup, Denmark) as previously described [11,12]. Negative controls had no primary antibody applied.\nStained samples were scored by two independent observers in terms of the proportion of epithelial cells staining positive (1 for <20%, 2 for >20%) and also the intensity of epithelial staining relative to that observed in adjacent endothelial cells (0,1,2 for negative, weak, moderate/strong, respectively) which act as an internal positive control [11,12]. From the anonymised histology reports, disease activity was scored 0, 1, or 2 for quiescent, low/moderate, or severe activity, respectively.", "The immunohistochemistry scores were analysed for specific correlations using Kendall's tau B analysis as previously described [12]. Correlations were checked between proportion and intensity of fascin immunoreactivity, disease activity, and the presence of dysplasia. These analyses yield a value between zero and one (one being the strongest possible correlation and zero indicating no correlation at all) and a p-value to indicate the significance of the correlation. A p-value of less than 0.05 is considered statistically significant (*, p = < 0.05; **, p = < 0.01; ***, p = < 0.001).", "The HT29 cell line was originally derived from a sporadic colonic adenocarcinoma and was maintained in culture in DMEM supplemented with 10% FBS [14].\nFor treatments, 1 × 106 cells were seeded per 25 cm2 culture flask in triplicate for each dose and control, and the cells allowed to recover for 48 hours prior to treatment. 5-ASA (Sigma) was dissolved in DMEM supplemented with 2% FBS to produce a 50 mM solution and the pH adjusted to 7.4. This stock solution was diluted further in DMEM (2% FBS) for cell treatments. The 5-ASA was made up fresh immediately prior to each experiment and protected from light throughout the procedure. Sodium butyrate (Sigma) treatments were carried out as described previously [15]. Apoptosis was assessed as previously described [15]. The level of apoptosis in cultured colon cells can be assessed by measuring the proportion of cells that detach from the flask and float in the medium [15 and references therein] Apoptosis was confirmed in these floating cells by morphology (following acridine orange staining). Acridine orange staining was performed as previously described with at least 300 cells scored for each sample. The proportion of cells exhibiting apoptotic morphology was >90% in the floating cell population, and this proportion remained constant following treatment. Therefore, the proportion of cells floating in the medium can be used as a measure of the extent of apoptosis in the culture. Western blotting analysis was performed on attached cells-thereby precluding apoptotic cells from this measure of fascin expression.", "Cell lysates of 2 × 106 cells were prepared for western blotting as described previously [16]. A mouse monoclonal antibody raised against fascin was obtained from Dako (Dako-Cytomation, Carpinteria, CA). Blots were subsequently probed with anti α-tubulin (Sigma, UK) to show equal sample loading.", "Cell migration assays were carried out using a transwell filter migration assay as previously described [12]. The lower chamber was filled with Calcium Free-DMEM supplemented with 5% FBS to act as an attractant. Following a 24 hour incubation and staining with haemotoxylin, cells on the lower filter surface were considered migratory and counted in 10 fields at ×20 magnification.", "[SUBTITLE] Fascin expression is expressed in inflamed colonic epithelium [SUBSECTION] Previous studies have shown that fascin is not expressed in the epithelial cells of the normal colonic mucosa, but is broadly expressed in the endothelial cells, fibroblasts and infiltrating lymphocytes of the lamina propria and submucosa [11,12]. Our previous study of colorectal adenomas showed epithelial fascin expression focussed around the tumour stalk, provoking the hypothesis that fascin expression may be modulated by inflammatory mediators [12]. In order to test this hypothesis immunohistochemistry was first performed on 11 tissue samples of resection margins from patients undergoing surgery for diverticulitis as these samples show low level inflammation. In 10 out of the 11, epithelial fascin immunoreactivity was focally observed in epithelial cells at the very base of the colonic crypts (Figure 1A). This observation is the first time that fascin expression has been observed in non-neoplastic colorectal epithelium and led us to question the relationship between fascin and inflammatory conditions of the colon.\nFascin is overexpressed in inflammatory bowel disease and in colonic epithelium actively undergoing restitution and regeneration. Fascin immunoreactivity was detected in specimens of human colorectal mucosa by immunoperoxidase staining. A: Diverticulitis resection margin showing positive staining in the crypt base - 200 × magnification. B: Ulcerative colitis showing positive staining in the base of the elongated crypts - 100 ×. C: Crohn's colitis showing strong positive staining toward the base of the elongated crypts - 100 ×. D: Crohn's colitis showing patchy fascin immunoreactivity throughout the gland - 200 ×. E: Ulcerative colitis with low-grade dysplasia showing strong and widespread fascin expression - 100 ×. F: Ulcerative colitis with high-grade dysplasia showing strong and widespread fascin expression - 200 ×. G: Ulcerative colitis showing fascin staining in the flattened epithelial cells undertaking restitution - 200 ×. H: enlargement of an area of panel G illustrating the fascin-positive epithelial cells (arrowheads). I: fascin positive epithelial cells undertaking restitution in a sample of Crohn's colitis. J: A regenerative polyp in a sample of ulcerative colitis showing fascin positivity in the newly forming crypts - 200 ×. K & L: Fascin expression in branching crypts (arrowheads) in regenerating ulcerative colitis tissue - 200 ×.\nPrevious studies have shown that fascin is not expressed in the epithelial cells of the normal colonic mucosa, but is broadly expressed in the endothelial cells, fibroblasts and infiltrating lymphocytes of the lamina propria and submucosa [11,12]. Our previous study of colorectal adenomas showed epithelial fascin expression focussed around the tumour stalk, provoking the hypothesis that fascin expression may be modulated by inflammatory mediators [12]. In order to test this hypothesis immunohistochemistry was first performed on 11 tissue samples of resection margins from patients undergoing surgery for diverticulitis as these samples show low level inflammation. In 10 out of the 11, epithelial fascin immunoreactivity was focally observed in epithelial cells at the very base of the colonic crypts (Figure 1A). This observation is the first time that fascin expression has been observed in non-neoplastic colorectal epithelium and led us to question the relationship between fascin and inflammatory conditions of the colon.\nFascin is overexpressed in inflammatory bowel disease and in colonic epithelium actively undergoing restitution and regeneration. Fascin immunoreactivity was detected in specimens of human colorectal mucosa by immunoperoxidase staining. A: Diverticulitis resection margin showing positive staining in the crypt base - 200 × magnification. B: Ulcerative colitis showing positive staining in the base of the elongated crypts - 100 ×. C: Crohn's colitis showing strong positive staining toward the base of the elongated crypts - 100 ×. D: Crohn's colitis showing patchy fascin immunoreactivity throughout the gland - 200 ×. E: Ulcerative colitis with low-grade dysplasia showing strong and widespread fascin expression - 100 ×. F: Ulcerative colitis with high-grade dysplasia showing strong and widespread fascin expression - 200 ×. G: Ulcerative colitis showing fascin staining in the flattened epithelial cells undertaking restitution - 200 ×. H: enlargement of an area of panel G illustrating the fascin-positive epithelial cells (arrowheads). I: fascin positive epithelial cells undertaking restitution in a sample of Crohn's colitis. J: A regenerative polyp in a sample of ulcerative colitis showing fascin positivity in the newly forming crypts - 200 ×. K & L: Fascin expression in branching crypts (arrowheads) in regenerating ulcerative colitis tissue - 200 ×.\n[SUBTITLE] Fascin is overexpressed in inflammatory bowel disease showing stronger and more widespread expression in Crohn's compared with ulcerative colitis [SUBSECTION] Having shown that epithelial fascin expression was observed in the presence of low-level colonic inflammation, fascin immunohistochemistry was performed on 41 samples of resected colorectal mucosa from patients suffering from IBD. Fascin was widely overexpressed in the epithelium of IBD-involved tissue and representative results are shown in Figure 1. The data from the subsequent scoring of these samples is summarised in Figure 2. Strong fascin staining was frequently observed toward the crypt bases in the elongated crypts associated with IBD (Figure 1B and 1C) but also showed a more patchy distribution throughout the glands (Figure 1D).\nFascin expression is stronger and more widespread in Crohn's colitis compared with Ulcerative colitis. A graphical summary of the immunohistochemical study of fascin expression in IBD. Samples were scored high or low for the proportion of epithelium staining positive (low = <20%; high = >20%), and the intensity of the epithelial fascin immunoreactivity. The graphs illustrate a significant increase in both the proportion (Ttest p = 0.0017) and intensity (Ttest p = 0.0013) of fascin immunoreactivity in Crohn's compared with ulcerative colitis samples.\nStrikingly, following scoring of the immunostaining, it was found that both the proportion of positive epithelial cells and the intensity of fascin staining (relative to internal positive control) was significantly greater in samples of Crohn's compared with UC (Figure 2). Staining intensity also positively correlated with disease activity (as determined by the consulting pathologist) for both conditions (Tau B = 0.223, p = <0.05). The intensity of fascin stain observed in the sample was also found to correlate with the proportion of fascin-positive epithelium (Tau B = 0.621, p = < 0.001).\nPrevious study has shown that fascin is overexpressed in both the benign and malignant stages of colorectal neoplasia and is associated with malignant progression of these tumours [11,12]. Increased risk of colorectal cancer is an important clinical consideration in IBD, and malignant progression of areas of dysplasia remains highly unpredictable [2]. In this study, strong fascin expression was observed in IBD samples showing either low or high grade dysplasia (Figure 1E), or cancer. Both the intensity of epithelial fascin immunoreactivity (Tau B = 0.391, p = <0.05) and the proportion (Tau B = 0.406, p = < 0.01) of positive epithelial cells in the tissue significantly correlated with the presence of areas of dysplastic epithelium.\nHaving shown that epithelial fascin expression was observed in the presence of low-level colonic inflammation, fascin immunohistochemistry was performed on 41 samples of resected colorectal mucosa from patients suffering from IBD. Fascin was widely overexpressed in the epithelium of IBD-involved tissue and representative results are shown in Figure 1. The data from the subsequent scoring of these samples is summarised in Figure 2. Strong fascin staining was frequently observed toward the crypt bases in the elongated crypts associated with IBD (Figure 1B and 1C) but also showed a more patchy distribution throughout the glands (Figure 1D).\nFascin expression is stronger and more widespread in Crohn's colitis compared with Ulcerative colitis. A graphical summary of the immunohistochemical study of fascin expression in IBD. Samples were scored high or low for the proportion of epithelium staining positive (low = <20%; high = >20%), and the intensity of the epithelial fascin immunoreactivity. The graphs illustrate a significant increase in both the proportion (Ttest p = 0.0017) and intensity (Ttest p = 0.0013) of fascin immunoreactivity in Crohn's compared with ulcerative colitis samples.\nStrikingly, following scoring of the immunostaining, it was found that both the proportion of positive epithelial cells and the intensity of fascin staining (relative to internal positive control) was significantly greater in samples of Crohn's compared with UC (Figure 2). Staining intensity also positively correlated with disease activity (as determined by the consulting pathologist) for both conditions (Tau B = 0.223, p = <0.05). The intensity of fascin stain observed in the sample was also found to correlate with the proportion of fascin-positive epithelium (Tau B = 0.621, p = < 0.001).\nPrevious study has shown that fascin is overexpressed in both the benign and malignant stages of colorectal neoplasia and is associated with malignant progression of these tumours [11,12]. Increased risk of colorectal cancer is an important clinical consideration in IBD, and malignant progression of areas of dysplasia remains highly unpredictable [2]. In this study, strong fascin expression was observed in IBD samples showing either low or high grade dysplasia (Figure 1E), or cancer. Both the intensity of epithelial fascin immunoreactivity (Tau B = 0.391, p = <0.05) and the proportion (Tau B = 0.406, p = < 0.01) of positive epithelial cells in the tissue significantly correlated with the presence of areas of dysplastic epithelium.\n[SUBTITLE] Fascin is overexpressed in colonic epithelium actively undergoing restitution and regeneration [SUBSECTION] Inflammatory infiltration in IBD frequently results in complete destruction of the mucosal layer resulting in areas of ulceration. Repair of these gaps in the epithelial barrier occurs as a two-stage process, termed restitution and regeneration [5]. In the sample group used, all of the 11 sections showing signs of active restitution stained positive for fascin in the flattened epithelial cells moving to cover the denuded area (Figure 1G, H).\nFascin staining was also observed in the newly formed immature crypts of regenerative polyps (Figure 1J) and in epithelial glands undergoing crypt fission - a relatively common observation in IBD tissue undertaking regeneration (Figure 1K). Taken together, these data suggest that fascin could play an important role in IBD and could be vital for disease remission through modulating mucosal repair.\nInflammatory infiltration in IBD frequently results in complete destruction of the mucosal layer resulting in areas of ulceration. Repair of these gaps in the epithelial barrier occurs as a two-stage process, termed restitution and regeneration [5]. In the sample group used, all of the 11 sections showing signs of active restitution stained positive for fascin in the flattened epithelial cells moving to cover the denuded area (Figure 1G, H).\nFascin staining was also observed in the newly formed immature crypts of regenerative polyps (Figure 1J) and in epithelial glands undergoing crypt fission - a relatively common observation in IBD tissue undertaking regeneration (Figure 1K). Taken together, these data suggest that fascin could play an important role in IBD and could be vital for disease remission through modulating mucosal repair.\n[SUBTITLE] 5-aminosalicylate reduces fascin expression and cell motility whereas sodium butyrate has a converse effect [SUBSECTION] We have established above that fascin is overexpressed in IBD, and that expression is associated with areas undertaking repair. As fascin is known to promote cell motility in neoplastic colorectal epithelial cells [12], and motility is a key factor in epithelial restitution, we aimed to determine the effect of therapeutic modalities used in the treatment of IBD on fascin expression and cell motility in vitro.\n5-aminosalicylate (5-ASA) is widely used in the clinical management of IBD, whereas the short chain fatty acid sodium butyrate, a luminal fermentation product of dietary fibre, has been proposed for use in IBD treatment [6,9].\nIn order to determine the effect of 5-ASA or butyrate on fascin expression we treated HT29 colorectal epithelial cells with a range of doses. Following 48 hours of treatment, the cells were counted and samples prepared in order to assay fascin expression by western blotting (Figure 3A). Both 5-ASA and butyrate reduced attached cell yield and induced apoptosis in the HT29 cells as described previously [15]. Apoptotic cells were not included in the analysis of fascin expression.\n5-aminosalicylate reduces fascin expression and cell motility in colorectal epithelial cells, whereas sodium butyrate has a converse effect. A: HT29 cells were treated with a range of doses of either 5-ASA or butyrate as indicated. Attached cell yields and percentage of apoptotic cells were determined and these data represent the mean of three independent experiments performed in triplicate ±SEM. Also shown are the subsequent western blot analyses of fascin expression. These blots are representative of three independent experiments and were re-probed for α-tubulin to show even sample loading. SW480 lysate was included as a positive control having been shown previously to express relatively high levels of fascin [Qualtrough et al., 2009]. B: The effect of 10 mM 5-ASA or 1 mM butyrate treatment on cell motility was determined on HT29 cells as measured by Boyden chamber assay using 5%FBS as an attractant (**, p = <0.01; ***, p = <0.001). Three independent experiments were carried out in triplicate and the data are expressed as the mean ± S.E.M.\n5-ASA treatment caused a decrease in fascin expression compared with control, whereas butyrate strongly enhanced fascin expression in a dose-dependent manner (Figure 3A). Promoter reporter assays using the full length fascin promoter [17] showed that regulation occurred, in both cases, at the transcriptional level (data not shown).\nIn order to determine whether 5-ASA and butyrate affect cell motility in this in vitro model, cells were seeded on collagen-coated transwell filters as previously described [12] and simultaneously treated with either 5-ASA or butyrate at doses which were shown not to significantly induce apoptosis (10 mM and 1 mM, respectively-Figure 3A). Motile cells were counted 24 hours later (Figure 3B). 5-ASA significantly reduced cell motility in HT29 cells whereas, conversely, butyrate stimulated cell migration.\nWe have established above that fascin is overexpressed in IBD, and that expression is associated with areas undertaking repair. As fascin is known to promote cell motility in neoplastic colorectal epithelial cells [12], and motility is a key factor in epithelial restitution, we aimed to determine the effect of therapeutic modalities used in the treatment of IBD on fascin expression and cell motility in vitro.\n5-aminosalicylate (5-ASA) is widely used in the clinical management of IBD, whereas the short chain fatty acid sodium butyrate, a luminal fermentation product of dietary fibre, has been proposed for use in IBD treatment [6,9].\nIn order to determine the effect of 5-ASA or butyrate on fascin expression we treated HT29 colorectal epithelial cells with a range of doses. Following 48 hours of treatment, the cells were counted and samples prepared in order to assay fascin expression by western blotting (Figure 3A). Both 5-ASA and butyrate reduced attached cell yield and induced apoptosis in the HT29 cells as described previously [15]. Apoptotic cells were not included in the analysis of fascin expression.\n5-aminosalicylate reduces fascin expression and cell motility in colorectal epithelial cells, whereas sodium butyrate has a converse effect. A: HT29 cells were treated with a range of doses of either 5-ASA or butyrate as indicated. Attached cell yields and percentage of apoptotic cells were determined and these data represent the mean of three independent experiments performed in triplicate ±SEM. Also shown are the subsequent western blot analyses of fascin expression. These blots are representative of three independent experiments and were re-probed for α-tubulin to show even sample loading. SW480 lysate was included as a positive control having been shown previously to express relatively high levels of fascin [Qualtrough et al., 2009]. B: The effect of 10 mM 5-ASA or 1 mM butyrate treatment on cell motility was determined on HT29 cells as measured by Boyden chamber assay using 5%FBS as an attractant (**, p = <0.01; ***, p = <0.001). Three independent experiments were carried out in triplicate and the data are expressed as the mean ± S.E.M.\n5-ASA treatment caused a decrease in fascin expression compared with control, whereas butyrate strongly enhanced fascin expression in a dose-dependent manner (Figure 3A). Promoter reporter assays using the full length fascin promoter [17] showed that regulation occurred, in both cases, at the transcriptional level (data not shown).\nIn order to determine whether 5-ASA and butyrate affect cell motility in this in vitro model, cells were seeded on collagen-coated transwell filters as previously described [12] and simultaneously treated with either 5-ASA or butyrate at doses which were shown not to significantly induce apoptosis (10 mM and 1 mM, respectively-Figure 3A). Motile cells were counted 24 hours later (Figure 3B). 5-ASA significantly reduced cell motility in HT29 cells whereas, conversely, butyrate stimulated cell migration.", "Previous studies have shown that fascin is not expressed in the epithelial cells of the normal colonic mucosa, but is broadly expressed in the endothelial cells, fibroblasts and infiltrating lymphocytes of the lamina propria and submucosa [11,12]. Our previous study of colorectal adenomas showed epithelial fascin expression focussed around the tumour stalk, provoking the hypothesis that fascin expression may be modulated by inflammatory mediators [12]. In order to test this hypothesis immunohistochemistry was first performed on 11 tissue samples of resection margins from patients undergoing surgery for diverticulitis as these samples show low level inflammation. In 10 out of the 11, epithelial fascin immunoreactivity was focally observed in epithelial cells at the very base of the colonic crypts (Figure 1A). This observation is the first time that fascin expression has been observed in non-neoplastic colorectal epithelium and led us to question the relationship between fascin and inflammatory conditions of the colon.\nFascin is overexpressed in inflammatory bowel disease and in colonic epithelium actively undergoing restitution and regeneration. Fascin immunoreactivity was detected in specimens of human colorectal mucosa by immunoperoxidase staining. A: Diverticulitis resection margin showing positive staining in the crypt base - 200 × magnification. B: Ulcerative colitis showing positive staining in the base of the elongated crypts - 100 ×. C: Crohn's colitis showing strong positive staining toward the base of the elongated crypts - 100 ×. D: Crohn's colitis showing patchy fascin immunoreactivity throughout the gland - 200 ×. E: Ulcerative colitis with low-grade dysplasia showing strong and widespread fascin expression - 100 ×. F: Ulcerative colitis with high-grade dysplasia showing strong and widespread fascin expression - 200 ×. G: Ulcerative colitis showing fascin staining in the flattened epithelial cells undertaking restitution - 200 ×. H: enlargement of an area of panel G illustrating the fascin-positive epithelial cells (arrowheads). I: fascin positive epithelial cells undertaking restitution in a sample of Crohn's colitis. J: A regenerative polyp in a sample of ulcerative colitis showing fascin positivity in the newly forming crypts - 200 ×. K & L: Fascin expression in branching crypts (arrowheads) in regenerating ulcerative colitis tissue - 200 ×.", "Having shown that epithelial fascin expression was observed in the presence of low-level colonic inflammation, fascin immunohistochemistry was performed on 41 samples of resected colorectal mucosa from patients suffering from IBD. Fascin was widely overexpressed in the epithelium of IBD-involved tissue and representative results are shown in Figure 1. The data from the subsequent scoring of these samples is summarised in Figure 2. Strong fascin staining was frequently observed toward the crypt bases in the elongated crypts associated with IBD (Figure 1B and 1C) but also showed a more patchy distribution throughout the glands (Figure 1D).\nFascin expression is stronger and more widespread in Crohn's colitis compared with Ulcerative colitis. A graphical summary of the immunohistochemical study of fascin expression in IBD. Samples were scored high or low for the proportion of epithelium staining positive (low = <20%; high = >20%), and the intensity of the epithelial fascin immunoreactivity. The graphs illustrate a significant increase in both the proportion (Ttest p = 0.0017) and intensity (Ttest p = 0.0013) of fascin immunoreactivity in Crohn's compared with ulcerative colitis samples.\nStrikingly, following scoring of the immunostaining, it was found that both the proportion of positive epithelial cells and the intensity of fascin staining (relative to internal positive control) was significantly greater in samples of Crohn's compared with UC (Figure 2). Staining intensity also positively correlated with disease activity (as determined by the consulting pathologist) for both conditions (Tau B = 0.223, p = <0.05). The intensity of fascin stain observed in the sample was also found to correlate with the proportion of fascin-positive epithelium (Tau B = 0.621, p = < 0.001).\nPrevious study has shown that fascin is overexpressed in both the benign and malignant stages of colorectal neoplasia and is associated with malignant progression of these tumours [11,12]. Increased risk of colorectal cancer is an important clinical consideration in IBD, and malignant progression of areas of dysplasia remains highly unpredictable [2]. In this study, strong fascin expression was observed in IBD samples showing either low or high grade dysplasia (Figure 1E), or cancer. Both the intensity of epithelial fascin immunoreactivity (Tau B = 0.391, p = <0.05) and the proportion (Tau B = 0.406, p = < 0.01) of positive epithelial cells in the tissue significantly correlated with the presence of areas of dysplastic epithelium.", "Inflammatory infiltration in IBD frequently results in complete destruction of the mucosal layer resulting in areas of ulceration. Repair of these gaps in the epithelial barrier occurs as a two-stage process, termed restitution and regeneration [5]. In the sample group used, all of the 11 sections showing signs of active restitution stained positive for fascin in the flattened epithelial cells moving to cover the denuded area (Figure 1G, H).\nFascin staining was also observed in the newly formed immature crypts of regenerative polyps (Figure 1J) and in epithelial glands undergoing crypt fission - a relatively common observation in IBD tissue undertaking regeneration (Figure 1K). Taken together, these data suggest that fascin could play an important role in IBD and could be vital for disease remission through modulating mucosal repair.", "We have established above that fascin is overexpressed in IBD, and that expression is associated with areas undertaking repair. As fascin is known to promote cell motility in neoplastic colorectal epithelial cells [12], and motility is a key factor in epithelial restitution, we aimed to determine the effect of therapeutic modalities used in the treatment of IBD on fascin expression and cell motility in vitro.\n5-aminosalicylate (5-ASA) is widely used in the clinical management of IBD, whereas the short chain fatty acid sodium butyrate, a luminal fermentation product of dietary fibre, has been proposed for use in IBD treatment [6,9].\nIn order to determine the effect of 5-ASA or butyrate on fascin expression we treated HT29 colorectal epithelial cells with a range of doses. Following 48 hours of treatment, the cells were counted and samples prepared in order to assay fascin expression by western blotting (Figure 3A). Both 5-ASA and butyrate reduced attached cell yield and induced apoptosis in the HT29 cells as described previously [15]. Apoptotic cells were not included in the analysis of fascin expression.\n5-aminosalicylate reduces fascin expression and cell motility in colorectal epithelial cells, whereas sodium butyrate has a converse effect. A: HT29 cells were treated with a range of doses of either 5-ASA or butyrate as indicated. Attached cell yields and percentage of apoptotic cells were determined and these data represent the mean of three independent experiments performed in triplicate ±SEM. Also shown are the subsequent western blot analyses of fascin expression. These blots are representative of three independent experiments and were re-probed for α-tubulin to show even sample loading. SW480 lysate was included as a positive control having been shown previously to express relatively high levels of fascin [Qualtrough et al., 2009]. B: The effect of 10 mM 5-ASA or 1 mM butyrate treatment on cell motility was determined on HT29 cells as measured by Boyden chamber assay using 5%FBS as an attractant (**, p = <0.01; ***, p = <0.001). Three independent experiments were carried out in triplicate and the data are expressed as the mean ± S.E.M.\n5-ASA treatment caused a decrease in fascin expression compared with control, whereas butyrate strongly enhanced fascin expression in a dose-dependent manner (Figure 3A). Promoter reporter assays using the full length fascin promoter [17] showed that regulation occurred, in both cases, at the transcriptional level (data not shown).\nIn order to determine whether 5-ASA and butyrate affect cell motility in this in vitro model, cells were seeded on collagen-coated transwell filters as previously described [12] and simultaneously treated with either 5-ASA or butyrate at doses which were shown not to significantly induce apoptosis (10 mM and 1 mM, respectively-Figure 3A). Motile cells were counted 24 hours later (Figure 3B). 5-ASA significantly reduced cell motility in HT29 cells whereas, conversely, butyrate stimulated cell migration.", "We show here, for the first time, that the actin bundling protein fascin is overexpressed in IBD, and may play an important role in tissue repair following inflammatory damage. Our in vitro findings suggest that therapeutic approaches for IBD could influence fascin expression, subsequently producing unwanted side-effects on tissue repair and thereby influencing achievement of remission.\nWe have shown previously that fascin is absent from the normal colorectal epithelium but overexpressed in both benign and malignant tumours where it is associated with enhanced cell motility [11,12]. Our previous findings in colorectal adenomas, where the tumour stalk frequently showed the strongest fascin immunoreactivity, suggest fascin regulation occurs through epithelial-mesenchymal crosstalk, potentially through inflammatory mediators [12]. This notion of inflammatory induction of fascin expression in the epithelium is supported here by the widespread expression of fascin in tissue resected from patients with IBD. A positive correlation was observed between the disease activity recorded by the pathologist and the intensity of fascin immunoreactivity further supporting the idea of an inflammation-mediated regulation of fascin expression. Also in keeping with these results is the observation of epithelial fascin expression at the crypt base in tissue from patients undergoing surgery for diverticulitis, which display low-grade inflammation.\nThe increased proportion and intensity of fascin staining observed in samples of Crohn's compared with UC raises questions about the fundamental difference between these classifications of IBD. As it was only the epithelial immunoreactivity that was scored, this difference cannot be attributed to the ability of Crohn's colitis to affect deeper layers of the bowel wall than UC, but rather resides in the mechanism of fascin regulation.\nIBD carries an increased risk of colorectal cancer, resulting in the deaths of 15% of IBD patients [2]. Malignant progression is highly unpredictable and therefore the identification of prognostic biomarkers of tumourigenic potential is of key importance [2]. In this study we showed that both the proportion and intensity of epithelial fascin stain significantly correlated with the presence of dysplasia or cancer. Further study, of a larger sample group with long-term clinical follow-up, will be necessary to confirm the validity of fascin as a biomarker for tumorigenesis in IBD. The role of fascin in the malignant progression of sporadic colorectal tumours [12] does suggest that fascin merits further investigation in the neoplastic transformation of IBD.\nOne of the most intriguing observations in our study of fascin expression in IBD was the focal expression of fascin in regions demonstrating active tissue repair. Fascin expression was found in all areas of restitution observed, and also in regenerative polyps and branching crypts, suggesting a role throughout the various stages of the repair process. Our previous work has shown that fascin can promote motility in benign, as well as malignant colorectal epithelial cells [12], and it would have been reasonable to hypothesise that fascin would be expressed in the motile cells undertaking restitution. Furthermore, fascin has been shown to modulate cell adhesion and could therefore also function in the dynamic adhesive changes required for both restitution and crypt fission [11]. These data suggest that regulation of fascin could be important in achieving and maintaining remission in patients with IBD.\nOne of the most broadly used treatments for the maintenance of remission in IBD is 5-ASA and previous work has shown this drug to induce apoptosis in HT29 cells in a dose-dependent manner [18]. We show here for the first time that treatment of the colorectal epithelial cell line HT29 with this drug led to a decrease in fascin expression and a retardation of cellular motility.\nLong term use of 5-ASA in patients with IBD is known to reduce cancer risk [19,20]. We postulate based on our findings here that this chemopreventive effect may, in part, be explained by the down regulation of fascin expression-with fascin being known to be involved in malignant progression of the colorectal epithelium [12]. This repression of cellular motility represents a key anti-cancer effect of 5-ASA treatment.\nConversely, considering the potential role of fascin in tissue repair, this would suggest that 5-ASA could hinder wound repair and thus impede remission in patients with active disease. The published evidence is somewhat conflicting as there are reports to suggest that NSAIDs can both impair wound healing [21] and that 5-ASA can promote intestinal repair [22]. However, the possibility exists that 5-ASA could have some detrimental effects in IBD patients in addition to its obvious benefits. This paradox in the effect of 5-ASA, and also the role of fascin, could have important consequences for the clinical management of IBD. To this end, our findings suggest that increased luminal levels of butyrate could complement 5-ASA in promoting remission whilst maintaining potent chemopreventive effects against tumorigenesis.\nPrevious work has suggested that sodium butyrate could have beneficial effects for IBD sufferers [8,9]. This fermentation product of dietary fibre is known to be present in the colonic lumen in the millimolar range and levels can be modulated by intake of suitable dietary substrate [10]. We, and others, have previously shown that butyrate has potent chemopreventive effects against colorectal tumorigenesis through the modulation of differentiation and apoptosis in colorectal epithelial cells [15,23,24].\nWe show here that butyrate upregulates fascin expression and significantly stimulates colorectal epithelial cell motility. These data suggest that butyrate will aid colonic tissue repair and therefore speed remission in IBD. This short chain fatty acid has already been proposed to promote wound healing in the small intestine [25]. Although our data suggest that fascin may be involved in tumorigenesis, butyrate has been shown to elicit potent anti-tumour effects. We propose that the combinatorial use of butyrate, or dietary modulation of its levels, could favourably complement 5-ASA use in the clinical management of IBD.", "We have shown for the first time that fascin is important in the pathogenesis and remission of IBD and could have important implications for its clinical management as fascin can potentially be manipulated using existing therapeutic approaches and modulation of diet.", "The authors declare that they have no competing interests.", "DQ: Conceived, designed, and coordinated the study and acquired the necessary funding; supervised the laboratory projects of KS & DL; carried out additional immunohistochemistry and all subsequent analyses; carried out some of the in vitro experiments; drafted the manuscript. KS & DL: Carried out the immunohistochemistry and some of the in vitro studies. MP: Contributed to the design and coordination of the study and aided with manuscript preparation. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-230X/11/14/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Chronic Disease", "Cross-Sectional Studies", "Diffusion of Innovation", "Factor Analysis, Statistical", "Health Care Surveys", "Humans", "Learning", "Patient Care Management", "Primary Health Care", "Texas" ]

[ "Background", "Development of a learning scale", "Administration of the learning survey", "Chronic care model implementation assessment", "Factor analysis of the learning survey", "Association between learning survey and ACIC", "Results", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Despite a well-developed evidence base regarding optimal treatments for many chronic diseases, including hypertension, type 2 diabetes, congestive heart failure, and chronic obstructive pulmonary disease, many patients seen in primary care settings do not receive these treatments [1,2]. Efforts to improve the delivery of evidence-based care have largely focused on provider knowledge [3-5] and decision support [6-11]. However, systematic reviews suggest that educational or knowledge-based interventions targeting individual providers to improve quality of care have been largely unsuccessful [12-14]. This finding suggests that we cannot depend only upon individual knowledge or decision-making capability of providers to improve care. Instead, focusing also on the larger systems in which patients receive care may lead to better results.\nThe chronic care model reflects this idea through its attention to not only patients and providers, but on the healthcare system itself [15]. Its focus on elements of a healthcare system that are important for chronic disease management, specifically self-management support, decision support, clinical information systems, and delivery system design, reflect the understanding that the healthcare system in which care is delivered influences chronic disease management [16]. The chronic care model, however, is not specific about the dynamics of health care systems or the evolving context in which care is delivered, nor is it specific about how these elements are implemented. Understanding these dynamics is critical to changing them, and to improving the care of patients with chronic disease [17,18].\nConceptualizing healthcare settings such as primary care as clinical microsystems gives us insight into the dynamics of clinical systems, and may make our efforts to improve chronic care delivery more effective. Clinical microsystems are the individual, functional units in which care is delivered, such as a primary care clinic, an inpatient unit, or an intensive care unit. A growing literature provides support for the application of complex adaptive system (CAS) theory to these clinical systems [19-24]. CASs are comprised of groups of individuals who learn, self-organize to complete tasks, and co-evolve with their external environment [19,22]. Additionally, they are defined by non-linearity, meaning that inputs and outputs may not be proportional or even necessarily predictable. In a CAS, the inter-dependencies among the agents are as important if not more important than the characteristics of the agents in understanding system outcomes.\nThese attributes of CAS suggest that the ability to learn is critically important. Learning is a social, shared process through which individuals incorporate new information in ways that lead them to change their mental models and adapt. The ability to learn can help people deal with an uncertain and changing environment more effectively. There is evidence to support the importance of learning in clinical microsystems: in operating room teams where learning occurs more effectively throughout the group, new techniques are more quickly adapted [25]; when learning occurs in nursing homes, patients receive better care [26].\nDespite this insight, the phenomenon of learning in clinical microsystems is not well understood. We sought to better understand the ways in which learning occurs in primary care settings and to relate learning to primary care clinic performance. To accomplish this, we first developed a scale designed to measure attributes of learning based on the literature related to learning in the organizational and educational psychology fields. We report the development of this learning scale and the factor analysis of the scale items. To understand the association between learning and clinic performance, we then analyzed the association between learning scale scores and degree of chronic care model implementation, as measured by the Assessment of Chronic Illness Care (ACIC) scale [27]. We chose the ACIC scale because we believe chronic disease management is a critical function of primary care clinics, and because ACIC scores have been linked to patient outcomes [28,29]. We hypothesized that provider and staff ratings of learning would be associated with their assessment of the extent to which the chronic care model had been implemented in their clinics.", "We convened a multidisciplinary team with expertise in improving provider behavior and organizational change. In 2006-2007, we conducted a targeted search focused on pulling together a diverse set of papers that discussed learning in terms of a social activity that is inherent in organizations, teams, and individuals. We focused on the organizational learning and educational psychology literatures, beginning with key papers that operationalize learning in organizations [30-33], learning in teams [34,35], and learning by individuals [36-38]. We expanded our review by working backwards and forwards, examining works referenced by those authors and references of those authors in subsequent publications. This literature was synthesized by three team members into a summary of themes associated with learning, shown in Additional file 1[30-47]. With the assistance of the fourth team member, items were developed to explore the presence of these learning themes in primary care settings. We believed that learning would be embedded in the following types of clinic member actions: asking questions beyond the presenting issue, sharing knowledge about a patient or a disease, staff and patient education, learning as things happen in the clinic, and learning from unexpected events or mistakes. We also believed that learning would occur through conversation and reflection. These understandings formed the basis of the questions about learning.\nWe created a new scale consisting of twenty-two items reflecting the learning themes identified in our literature review. The scale instructs respondents to indicate their level of agreement with each statement using a 5-point Likert scale. Responses for each item are scored from one (strongly agree) to five (strongly disagree). Scale items were pilot tested in three Veterans Affairs (VA) primary care clinics and two VA contract clinics in South Texas and administered to one hundred and one staff and providers across those five clinics, including front desk staff, medical assistants, nurses, and physicians. Cronbach's alpha for the learning questions based on this sample was 0.814, indicating good internal consistency. Based on feedback and questions from participants in the pilot, the wording of specific items was refined. This refinement consisted primarily of changing negatively worded items to positive ones, and using the word \"I\" consistently instead of \"we.\" The final list of items is shown in Additional file 2.", "The ABC study is a cluster randomized controlled trial testing the effectiveness of a practice facilitation intervention to improve the processes of care and outcomes for diabetic patients in forty primary care clinics in South Texas. As part of this study, a baseline survey that included the learning scale items was administered to all clinicians and office staff of these primary care clinics prior to the start of the intervention by the research team. Here we report on the results of the baseline cross-sectional survey.\nThe primary care clinics included in the ABC study are generally small, autonomous, physician-owned clinics with four or fewer primary care providers. Thirty of the clinics have only one physician, and of these thirty, eleven had one or more non-physician providers (either physician assistant or nurse practitioner). Ten clinics had two to four physicians and of those, five had at least one physician assistant or nurse practitioner. No clinics had other types of providers such as nutritionists or counselors.", "The extent to which each clinic provides optimal care for patients with chronic illnesses was measured with the Assessment of Chronic Illness Care scale (ACIC) [27]. The ACIC is a twenty-five item questionnaire that asks health care providers to rate the degree of support for each of the six elements of the Chronic Care Model (CCM) in their health care system: delivery system redesign, patient self-management support, decision support, information support, community linkages, and health system support. Response choices for each item range from zero to eleven, with eleven representing optimal chronic care support. In addition to a total score reflecting overall CCM implementation, the ACIC score can be split into six sub-scales that reflect each of the elements contained in the model. Version 3.5 of the ACIC was used in this study, and in addition to the 6 sub-scales, also includes items that address how well a practice integrates the CCM elements [48]. Preliminary data indicate the ACIC is responsive to changes chronic care delivery and correlates well with other measures of productivity and system improvements [27]. Prior research by members of this team also suggest that ACIC scores are associated with clinical outcomes such as A1c control and ten-year risk of a cardiovascular event. That is, patients who attend clinics with higher ACIC scores have lower A1c values and lower risk [28,29].\nWe included the ACIC in the baseline survey completed by all clinic members in the forty clinics enrolled in the ABC study.", "We performed a principal components factor analysis of the learning scale [49]. Eigenvalues over 1, scree plot inspection, and determination of simple structure across items were used to identify potential factors. Cronbach coefficient alpha scores in the range of 0.7 were used to identify those factors with the greatest degree of internal validity.", "We calculated Pearson correlation coefficients between subscales identified in the factor analysis, total ACIC scores, and ACIC sub-component scores related to each element of the CCM.", "Two-hundred and ninety-six respondents from 40 clinics completed the survey during the period from October 2007 to May 2010. Fifteen percent of these were physicians, and 6% non-physician providers. The remainder of the respondents were other clinic staff members, such as front desk staff or medical assistants. Characteristics of the clinics surveyed are shown in Table 1. Medicare is the government-sponsored heathcare program for persons over age 65 in the United States, and reflects the proportion of geriatric patients in each practice. The number of managed-care contracts is a reflection of the number of insurers with which each practice is contracted.\nCharacteristics of surveyed clinics\nPrincipal components factor analysis revealed three factors with Cronbach coefficient alpha scores of 0.82, 0.57, and 0.68. Factor loading ranged from 0.44 to 0.77. The factors with scores of 0.57 and 0.68 were eliminated based on being below our acceptability threshold, and items not being conceptually similar. We examined the eight items in the factor with a score of 0.82. Based on the conceptual content included in these items and their factor loading, we concluded that five of the eight items were capturing an idea of learning as a shared, back-and-forth process between clinic members. We called this concept \"reciprocal learning\" to reflect what we believed was the notion of reciprocal interdependency - an interdependency in which the output of a system is produced by the collaboration of all contributing entities, and in which these entities are dependent on each other to produce the optimal system-level output [50]. The specific items in the reciprocal learning factor are shown in Table 2. The Cronbach alpha for the five items was 0.79. The mean score for each item across clinics was 3.83 (SD = 0.72), with a range from 1.4 to 5. The specific scores for each individual item in the reciprocal learning factor are shown in Additional file 3.\nItems in the reciprocal learning subscale identified by factor analysis\nThe mean, median, and range in ACIC scores and component scores across clinic are shown in Table 3. These scores indicate that there was a broad range in the extent to which practices had implemented the CCM elements. Inspection of normalized residual plots and skewness statistics reveal that all the variables in our analysis conformed to normal distributions, as do the close correspondence between the mean and median of each variable in the table.\nMean, median and range in ACIC scores across clinics\nTable 4 shows the Pearson correlations between learning scale scores and ACIC total and component scores. Correlation between the reciprocal learning and the ACIC score and subscales ranged from 0.28 to 0.46. We adjusted this analysis to account for the clustering effect of consistency of responses within clinics to reduce the potential bias that could result from clustering. The intraclass correlations (ICC) of the variables ranged from .10 to .22 suggesting that respondents within clinic tend to answer in a similar manner therefore affecting standard error estimates. MLWin software [51] was used to obtain unbiased associations. Adjusted correlations are also shown in Table 4.\nAssociation between reciprocal learning sub-scale and ACIC total and component scores", "We sought to better understand learning in primary care clinics, and the relationship between learning and clinic performance as measured by the degree to which the CCM was present in primary care clinics. To accomplish this, we first developed and administered a twenty-two item learning scale that reflected 6 learning themes described in the organizational and educational psychology literature. We then performed a factor analysis to examine which items most closely clustered together in the scale responses. We used the resulting factors to better understand which aspects of learning were most relevant within primary care settings. This analysis identified a subset of five items that reflect a learning process that occurs between people where each learns from sharing with the other, and in which the learning acquired from one person becomes the foundation for further learning by others in a building, iterative process. Because of the mutual and iterative nature of this process, we believe it reflects the concept of reciprocal learning [50]. We found a wide range of responses across clinics to the items on the five-item reciprocal learning scale, indicating that responses to the items on this scale can be used to discriminate between the clinics.\nTo better understand the role of learning in primary care settings, we wanted to understand the possible association between learning and clinic performance. Because we view the care of chronic illness and the presence of the chronic care model elements to be critical aspects of primary care delivery, we used ACIC scores as a measure of primary care clinic performance and tested the association between reciprocal learning and the ACIC. Reciprocal learning was significantly associated with ACIC scores, suggesting that this type of learning may be particularly important for successful chronic care model implementation.\nThis conceptualization of learning moves beyond the idea of one person learning from another to that of people learning together, building on each other's understandings. These findings echo studies from operating room teams and nursing home caregivers that demonstrated the importance of each individual contributing to care in a shared way [25,34]. The literature related to learning in healthcare settings is limited, and our results should be considered a first step in the development of the concept of reciprocal learning in these settings. However, because learning is a social activity that is dependent on relationships and the ability of clinic members to have the opportunity to speak to each other, studies on relationships, conversation, and reflection [52-56] in healthcare settings complement our findings.\nOur design of developing a scale to understand learning has several limitations. First, we developed our scale based on descriptions of learning in non-healthcare disciplines. While physicians and researchers with knowledge of primary care settings applied the concepts in ways that would be meaningful to healthcare providers in the development of the survey items, we may have missed aspects of learning important to healthcare settings that were not part of other disciplines. Second, using a scale administered at a single point in time may not be optimal for describing a dynamic and evolutionary process such as learning. Despite this, our results do discriminate between clinics and point to what we believe is an important concept of reciprocal learning. Finally, our findings are limited in that we only included forty small primary care clinics in South Texas. These results may not translate as easily to larger primary care practices or more integrated group settings, or in other geographic areas.\nDespite these limitations, our findings are an important step forward in understanding the role of learning in primary care clinics. This understanding may be particularly important in light of efforts to implement patient-centered medical home (PCMH) care models [57,58] in United States primary care settings. The purpose of the PCMH is to provide patient-centered care in which all clinic members are engaged and responsible in the care of all patients. Reciprocal learning may be an important way to improve engagement of clinic members and their ability to learn from each other to improve patient care. Improved care of patients with chronic disease is an important part of the PCMH model, and the CCM elements are shared with those of the PCMH. Successfully implementing these models of care is not a simple or static process. It requires not only attention to multiple aspects of the system in which care is delivered, but also an emphasis on patients' support system, and their ability to manage their diseases. To accomplish this requires the active and proactive engagement of staff and providers to be alert and open to new ways of doing things, to understand the impact of the way they do things on others, and to learn not only from, but with each other and respond to the needs of their patients with chronic illnesses [55]. This may explain why reciprocal learning is associated with the degree to which the chronic care model was present in each clinic.\nInterpretation of our results underscores the idea that the kinds of learning required in clinical microsystems are more sophisticated than typically acknowledged. Learning is an interdependent process that occurs between and among all members of the clinic. Managing learning as an interdependent process will likely be difficult, but our findings suggest that it will be necessary to improving the care delivered to patients with chronic disease.", "We describe the construct of reciprocal learning in primary care clinics, an activity through which clinic members learn from each other in an iterative, building process. Reciprocal learning appears to be an important attribute of learning in primary care clinics, as its presence relates to the degree of chronic care model implementation. Interventions to improve reciprocal learning among clinic members may lead to improved care of patients with chronic disease and may be relevant to improving overall clinic performance. Reciprocal learning may also be important for clinics' ability to move to more patient-centered models of care.", "The authors declare that they have no competing interests.", "MJ carried out the literature review of learning with assistance from HL. MJ, HL, RRM led the development of the learning scale. MP conceived and carried out the ABC study in collaboration with PHN. RP performed the statistical analysis. LL drafted the manuscript. All authors were involved in review and interpretation of the reported results. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1472-6963/11/44/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Development of a learning scale", "Administration of the learning survey", "Chronic care model implementation assessment", "Factor analysis of the learning survey", "Association between learning survey and ACIC", "Results", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history", "Supplementary Material" ]

[ "Despite a well-developed evidence base regarding optimal treatments for many chronic diseases, including hypertension, type 2 diabetes, congestive heart failure, and chronic obstructive pulmonary disease, many patients seen in primary care settings do not receive these treatments [1,2]. Efforts to improve the delivery of evidence-based care have largely focused on provider knowledge [3-5] and decision support [6-11]. However, systematic reviews suggest that educational or knowledge-based interventions targeting individual providers to improve quality of care have been largely unsuccessful [12-14]. This finding suggests that we cannot depend only upon individual knowledge or decision-making capability of providers to improve care. Instead, focusing also on the larger systems in which patients receive care may lead to better results.\nThe chronic care model reflects this idea through its attention to not only patients and providers, but on the healthcare system itself [15]. Its focus on elements of a healthcare system that are important for chronic disease management, specifically self-management support, decision support, clinical information systems, and delivery system design, reflect the understanding that the healthcare system in which care is delivered influences chronic disease management [16]. The chronic care model, however, is not specific about the dynamics of health care systems or the evolving context in which care is delivered, nor is it specific about how these elements are implemented. Understanding these dynamics is critical to changing them, and to improving the care of patients with chronic disease [17,18].\nConceptualizing healthcare settings such as primary care as clinical microsystems gives us insight into the dynamics of clinical systems, and may make our efforts to improve chronic care delivery more effective. Clinical microsystems are the individual, functional units in which care is delivered, such as a primary care clinic, an inpatient unit, or an intensive care unit. A growing literature provides support for the application of complex adaptive system (CAS) theory to these clinical systems [19-24]. CASs are comprised of groups of individuals who learn, self-organize to complete tasks, and co-evolve with their external environment [19,22]. Additionally, they are defined by non-linearity, meaning that inputs and outputs may not be proportional or even necessarily predictable. In a CAS, the inter-dependencies among the agents are as important if not more important than the characteristics of the agents in understanding system outcomes.\nThese attributes of CAS suggest that the ability to learn is critically important. Learning is a social, shared process through which individuals incorporate new information in ways that lead them to change their mental models and adapt. The ability to learn can help people deal with an uncertain and changing environment more effectively. There is evidence to support the importance of learning in clinical microsystems: in operating room teams where learning occurs more effectively throughout the group, new techniques are more quickly adapted [25]; when learning occurs in nursing homes, patients receive better care [26].\nDespite this insight, the phenomenon of learning in clinical microsystems is not well understood. We sought to better understand the ways in which learning occurs in primary care settings and to relate learning to primary care clinic performance. To accomplish this, we first developed a scale designed to measure attributes of learning based on the literature related to learning in the organizational and educational psychology fields. We report the development of this learning scale and the factor analysis of the scale items. To understand the association between learning and clinic performance, we then analyzed the association between learning scale scores and degree of chronic care model implementation, as measured by the Assessment of Chronic Illness Care (ACIC) scale [27]. We chose the ACIC scale because we believe chronic disease management is a critical function of primary care clinics, and because ACIC scores have been linked to patient outcomes [28,29]. We hypothesized that provider and staff ratings of learning would be associated with their assessment of the extent to which the chronic care model had been implemented in their clinics.", "[SUBTITLE] Development of a learning scale [SUBSECTION] We convened a multidisciplinary team with expertise in improving provider behavior and organizational change. In 2006-2007, we conducted a targeted search focused on pulling together a diverse set of papers that discussed learning in terms of a social activity that is inherent in organizations, teams, and individuals. We focused on the organizational learning and educational psychology literatures, beginning with key papers that operationalize learning in organizations [30-33], learning in teams [34,35], and learning by individuals [36-38]. We expanded our review by working backwards and forwards, examining works referenced by those authors and references of those authors in subsequent publications. This literature was synthesized by three team members into a summary of themes associated with learning, shown in Additional file 1[30-47]. With the assistance of the fourth team member, items were developed to explore the presence of these learning themes in primary care settings. We believed that learning would be embedded in the following types of clinic member actions: asking questions beyond the presenting issue, sharing knowledge about a patient or a disease, staff and patient education, learning as things happen in the clinic, and learning from unexpected events or mistakes. We also believed that learning would occur through conversation and reflection. These understandings formed the basis of the questions about learning.\nWe created a new scale consisting of twenty-two items reflecting the learning themes identified in our literature review. The scale instructs respondents to indicate their level of agreement with each statement using a 5-point Likert scale. Responses for each item are scored from one (strongly agree) to five (strongly disagree). Scale items were pilot tested in three Veterans Affairs (VA) primary care clinics and two VA contract clinics in South Texas and administered to one hundred and one staff and providers across those five clinics, including front desk staff, medical assistants, nurses, and physicians. Cronbach's alpha for the learning questions based on this sample was 0.814, indicating good internal consistency. Based on feedback and questions from participants in the pilot, the wording of specific items was refined. This refinement consisted primarily of changing negatively worded items to positive ones, and using the word \"I\" consistently instead of \"we.\" The final list of items is shown in Additional file 2.\nWe convened a multidisciplinary team with expertise in improving provider behavior and organizational change. In 2006-2007, we conducted a targeted search focused on pulling together a diverse set of papers that discussed learning in terms of a social activity that is inherent in organizations, teams, and individuals. We focused on the organizational learning and educational psychology literatures, beginning with key papers that operationalize learning in organizations [30-33], learning in teams [34,35], and learning by individuals [36-38]. We expanded our review by working backwards and forwards, examining works referenced by those authors and references of those authors in subsequent publications. This literature was synthesized by three team members into a summary of themes associated with learning, shown in Additional file 1[30-47]. With the assistance of the fourth team member, items were developed to explore the presence of these learning themes in primary care settings. We believed that learning would be embedded in the following types of clinic member actions: asking questions beyond the presenting issue, sharing knowledge about a patient or a disease, staff and patient education, learning as things happen in the clinic, and learning from unexpected events or mistakes. We also believed that learning would occur through conversation and reflection. These understandings formed the basis of the questions about learning.\nWe created a new scale consisting of twenty-two items reflecting the learning themes identified in our literature review. The scale instructs respondents to indicate their level of agreement with each statement using a 5-point Likert scale. Responses for each item are scored from one (strongly agree) to five (strongly disagree). Scale items were pilot tested in three Veterans Affairs (VA) primary care clinics and two VA contract clinics in South Texas and administered to one hundred and one staff and providers across those five clinics, including front desk staff, medical assistants, nurses, and physicians. Cronbach's alpha for the learning questions based on this sample was 0.814, indicating good internal consistency. Based on feedback and questions from participants in the pilot, the wording of specific items was refined. This refinement consisted primarily of changing negatively worded items to positive ones, and using the word \"I\" consistently instead of \"we.\" The final list of items is shown in Additional file 2.\n[SUBTITLE] Administration of the learning survey [SUBSECTION] The ABC study is a cluster randomized controlled trial testing the effectiveness of a practice facilitation intervention to improve the processes of care and outcomes for diabetic patients in forty primary care clinics in South Texas. As part of this study, a baseline survey that included the learning scale items was administered to all clinicians and office staff of these primary care clinics prior to the start of the intervention by the research team. Here we report on the results of the baseline cross-sectional survey.\nThe primary care clinics included in the ABC study are generally small, autonomous, physician-owned clinics with four or fewer primary care providers. Thirty of the clinics have only one physician, and of these thirty, eleven had one or more non-physician providers (either physician assistant or nurse practitioner). Ten clinics had two to four physicians and of those, five had at least one physician assistant or nurse practitioner. No clinics had other types of providers such as nutritionists or counselors.\nThe ABC study is a cluster randomized controlled trial testing the effectiveness of a practice facilitation intervention to improve the processes of care and outcomes for diabetic patients in forty primary care clinics in South Texas. As part of this study, a baseline survey that included the learning scale items was administered to all clinicians and office staff of these primary care clinics prior to the start of the intervention by the research team. Here we report on the results of the baseline cross-sectional survey.\nThe primary care clinics included in the ABC study are generally small, autonomous, physician-owned clinics with four or fewer primary care providers. Thirty of the clinics have only one physician, and of these thirty, eleven had one or more non-physician providers (either physician assistant or nurse practitioner). Ten clinics had two to four physicians and of those, five had at least one physician assistant or nurse practitioner. No clinics had other types of providers such as nutritionists or counselors.\n[SUBTITLE] Chronic care model implementation assessment [SUBSECTION] The extent to which each clinic provides optimal care for patients with chronic illnesses was measured with the Assessment of Chronic Illness Care scale (ACIC) [27]. The ACIC is a twenty-five item questionnaire that asks health care providers to rate the degree of support for each of the six elements of the Chronic Care Model (CCM) in their health care system: delivery system redesign, patient self-management support, decision support, information support, community linkages, and health system support. Response choices for each item range from zero to eleven, with eleven representing optimal chronic care support. In addition to a total score reflecting overall CCM implementation, the ACIC score can be split into six sub-scales that reflect each of the elements contained in the model. Version 3.5 of the ACIC was used in this study, and in addition to the 6 sub-scales, also includes items that address how well a practice integrates the CCM elements [48]. Preliminary data indicate the ACIC is responsive to changes chronic care delivery and correlates well with other measures of productivity and system improvements [27]. Prior research by members of this team also suggest that ACIC scores are associated with clinical outcomes such as A1c control and ten-year risk of a cardiovascular event. That is, patients who attend clinics with higher ACIC scores have lower A1c values and lower risk [28,29].\nWe included the ACIC in the baseline survey completed by all clinic members in the forty clinics enrolled in the ABC study.\nThe extent to which each clinic provides optimal care for patients with chronic illnesses was measured with the Assessment of Chronic Illness Care scale (ACIC) [27]. The ACIC is a twenty-five item questionnaire that asks health care providers to rate the degree of support for each of the six elements of the Chronic Care Model (CCM) in their health care system: delivery system redesign, patient self-management support, decision support, information support, community linkages, and health system support. Response choices for each item range from zero to eleven, with eleven representing optimal chronic care support. In addition to a total score reflecting overall CCM implementation, the ACIC score can be split into six sub-scales that reflect each of the elements contained in the model. Version 3.5 of the ACIC was used in this study, and in addition to the 6 sub-scales, also includes items that address how well a practice integrates the CCM elements [48]. Preliminary data indicate the ACIC is responsive to changes chronic care delivery and correlates well with other measures of productivity and system improvements [27]. Prior research by members of this team also suggest that ACIC scores are associated with clinical outcomes such as A1c control and ten-year risk of a cardiovascular event. That is, patients who attend clinics with higher ACIC scores have lower A1c values and lower risk [28,29].\nWe included the ACIC in the baseline survey completed by all clinic members in the forty clinics enrolled in the ABC study.\n[SUBTITLE] Factor analysis of the learning survey [SUBSECTION] We performed a principal components factor analysis of the learning scale [49]. Eigenvalues over 1, scree plot inspection, and determination of simple structure across items were used to identify potential factors. Cronbach coefficient alpha scores in the range of 0.7 were used to identify those factors with the greatest degree of internal validity.\nWe performed a principal components factor analysis of the learning scale [49]. Eigenvalues over 1, scree plot inspection, and determination of simple structure across items were used to identify potential factors. Cronbach coefficient alpha scores in the range of 0.7 were used to identify those factors with the greatest degree of internal validity.\n[SUBTITLE] Association between learning survey and ACIC [SUBSECTION] We calculated Pearson correlation coefficients between subscales identified in the factor analysis, total ACIC scores, and ACIC sub-component scores related to each element of the CCM.\nWe calculated Pearson correlation coefficients between subscales identified in the factor analysis, total ACIC scores, and ACIC sub-component scores related to each element of the CCM.", "We convened a multidisciplinary team with expertise in improving provider behavior and organizational change. In 2006-2007, we conducted a targeted search focused on pulling together a diverse set of papers that discussed learning in terms of a social activity that is inherent in organizations, teams, and individuals. We focused on the organizational learning and educational psychology literatures, beginning with key papers that operationalize learning in organizations [30-33], learning in teams [34,35], and learning by individuals [36-38]. We expanded our review by working backwards and forwards, examining works referenced by those authors and references of those authors in subsequent publications. This literature was synthesized by three team members into a summary of themes associated with learning, shown in Additional file 1[30-47]. With the assistance of the fourth team member, items were developed to explore the presence of these learning themes in primary care settings. We believed that learning would be embedded in the following types of clinic member actions: asking questions beyond the presenting issue, sharing knowledge about a patient or a disease, staff and patient education, learning as things happen in the clinic, and learning from unexpected events or mistakes. We also believed that learning would occur through conversation and reflection. These understandings formed the basis of the questions about learning.\nWe created a new scale consisting of twenty-two items reflecting the learning themes identified in our literature review. The scale instructs respondents to indicate their level of agreement with each statement using a 5-point Likert scale. Responses for each item are scored from one (strongly agree) to five (strongly disagree). Scale items were pilot tested in three Veterans Affairs (VA) primary care clinics and two VA contract clinics in South Texas and administered to one hundred and one staff and providers across those five clinics, including front desk staff, medical assistants, nurses, and physicians. Cronbach's alpha for the learning questions based on this sample was 0.814, indicating good internal consistency. Based on feedback and questions from participants in the pilot, the wording of specific items was refined. This refinement consisted primarily of changing negatively worded items to positive ones, and using the word \"I\" consistently instead of \"we.\" The final list of items is shown in Additional file 2.", "The ABC study is a cluster randomized controlled trial testing the effectiveness of a practice facilitation intervention to improve the processes of care and outcomes for diabetic patients in forty primary care clinics in South Texas. As part of this study, a baseline survey that included the learning scale items was administered to all clinicians and office staff of these primary care clinics prior to the start of the intervention by the research team. Here we report on the results of the baseline cross-sectional survey.\nThe primary care clinics included in the ABC study are generally small, autonomous, physician-owned clinics with four or fewer primary care providers. Thirty of the clinics have only one physician, and of these thirty, eleven had one or more non-physician providers (either physician assistant or nurse practitioner). Ten clinics had two to four physicians and of those, five had at least one physician assistant or nurse practitioner. No clinics had other types of providers such as nutritionists or counselors.", "The extent to which each clinic provides optimal care for patients with chronic illnesses was measured with the Assessment of Chronic Illness Care scale (ACIC) [27]. The ACIC is a twenty-five item questionnaire that asks health care providers to rate the degree of support for each of the six elements of the Chronic Care Model (CCM) in their health care system: delivery system redesign, patient self-management support, decision support, information support, community linkages, and health system support. Response choices for each item range from zero to eleven, with eleven representing optimal chronic care support. In addition to a total score reflecting overall CCM implementation, the ACIC score can be split into six sub-scales that reflect each of the elements contained in the model. Version 3.5 of the ACIC was used in this study, and in addition to the 6 sub-scales, also includes items that address how well a practice integrates the CCM elements [48]. Preliminary data indicate the ACIC is responsive to changes chronic care delivery and correlates well with other measures of productivity and system improvements [27]. Prior research by members of this team also suggest that ACIC scores are associated with clinical outcomes such as A1c control and ten-year risk of a cardiovascular event. That is, patients who attend clinics with higher ACIC scores have lower A1c values and lower risk [28,29].\nWe included the ACIC in the baseline survey completed by all clinic members in the forty clinics enrolled in the ABC study.", "We performed a principal components factor analysis of the learning scale [49]. Eigenvalues over 1, scree plot inspection, and determination of simple structure across items were used to identify potential factors. Cronbach coefficient alpha scores in the range of 0.7 were used to identify those factors with the greatest degree of internal validity.", "We calculated Pearson correlation coefficients between subscales identified in the factor analysis, total ACIC scores, and ACIC sub-component scores related to each element of the CCM.", "Two-hundred and ninety-six respondents from 40 clinics completed the survey during the period from October 2007 to May 2010. Fifteen percent of these were physicians, and 6% non-physician providers. The remainder of the respondents were other clinic staff members, such as front desk staff or medical assistants. Characteristics of the clinics surveyed are shown in Table 1. Medicare is the government-sponsored heathcare program for persons over age 65 in the United States, and reflects the proportion of geriatric patients in each practice. The number of managed-care contracts is a reflection of the number of insurers with which each practice is contracted.\nCharacteristics of surveyed clinics\nPrincipal components factor analysis revealed three factors with Cronbach coefficient alpha scores of 0.82, 0.57, and 0.68. Factor loading ranged from 0.44 to 0.77. The factors with scores of 0.57 and 0.68 were eliminated based on being below our acceptability threshold, and items not being conceptually similar. We examined the eight items in the factor with a score of 0.82. Based on the conceptual content included in these items and their factor loading, we concluded that five of the eight items were capturing an idea of learning as a shared, back-and-forth process between clinic members. We called this concept \"reciprocal learning\" to reflect what we believed was the notion of reciprocal interdependency - an interdependency in which the output of a system is produced by the collaboration of all contributing entities, and in which these entities are dependent on each other to produce the optimal system-level output [50]. The specific items in the reciprocal learning factor are shown in Table 2. The Cronbach alpha for the five items was 0.79. The mean score for each item across clinics was 3.83 (SD = 0.72), with a range from 1.4 to 5. The specific scores for each individual item in the reciprocal learning factor are shown in Additional file 3.\nItems in the reciprocal learning subscale identified by factor analysis\nThe mean, median, and range in ACIC scores and component scores across clinic are shown in Table 3. These scores indicate that there was a broad range in the extent to which practices had implemented the CCM elements. Inspection of normalized residual plots and skewness statistics reveal that all the variables in our analysis conformed to normal distributions, as do the close correspondence between the mean and median of each variable in the table.\nMean, median and range in ACIC scores across clinics\nTable 4 shows the Pearson correlations between learning scale scores and ACIC total and component scores. Correlation between the reciprocal learning and the ACIC score and subscales ranged from 0.28 to 0.46. We adjusted this analysis to account for the clustering effect of consistency of responses within clinics to reduce the potential bias that could result from clustering. The intraclass correlations (ICC) of the variables ranged from .10 to .22 suggesting that respondents within clinic tend to answer in a similar manner therefore affecting standard error estimates. MLWin software [51] was used to obtain unbiased associations. Adjusted correlations are also shown in Table 4.\nAssociation between reciprocal learning sub-scale and ACIC total and component scores", "We sought to better understand learning in primary care clinics, and the relationship between learning and clinic performance as measured by the degree to which the CCM was present in primary care clinics. To accomplish this, we first developed and administered a twenty-two item learning scale that reflected 6 learning themes described in the organizational and educational psychology literature. We then performed a factor analysis to examine which items most closely clustered together in the scale responses. We used the resulting factors to better understand which aspects of learning were most relevant within primary care settings. This analysis identified a subset of five items that reflect a learning process that occurs between people where each learns from sharing with the other, and in which the learning acquired from one person becomes the foundation for further learning by others in a building, iterative process. Because of the mutual and iterative nature of this process, we believe it reflects the concept of reciprocal learning [50]. We found a wide range of responses across clinics to the items on the five-item reciprocal learning scale, indicating that responses to the items on this scale can be used to discriminate between the clinics.\nTo better understand the role of learning in primary care settings, we wanted to understand the possible association between learning and clinic performance. Because we view the care of chronic illness and the presence of the chronic care model elements to be critical aspects of primary care delivery, we used ACIC scores as a measure of primary care clinic performance and tested the association between reciprocal learning and the ACIC. Reciprocal learning was significantly associated with ACIC scores, suggesting that this type of learning may be particularly important for successful chronic care model implementation.\nThis conceptualization of learning moves beyond the idea of one person learning from another to that of people learning together, building on each other's understandings. These findings echo studies from operating room teams and nursing home caregivers that demonstrated the importance of each individual contributing to care in a shared way [25,34]. The literature related to learning in healthcare settings is limited, and our results should be considered a first step in the development of the concept of reciprocal learning in these settings. However, because learning is a social activity that is dependent on relationships and the ability of clinic members to have the opportunity to speak to each other, studies on relationships, conversation, and reflection [52-56] in healthcare settings complement our findings.\nOur design of developing a scale to understand learning has several limitations. First, we developed our scale based on descriptions of learning in non-healthcare disciplines. While physicians and researchers with knowledge of primary care settings applied the concepts in ways that would be meaningful to healthcare providers in the development of the survey items, we may have missed aspects of learning important to healthcare settings that were not part of other disciplines. Second, using a scale administered at a single point in time may not be optimal for describing a dynamic and evolutionary process such as learning. Despite this, our results do discriminate between clinics and point to what we believe is an important concept of reciprocal learning. Finally, our findings are limited in that we only included forty small primary care clinics in South Texas. These results may not translate as easily to larger primary care practices or more integrated group settings, or in other geographic areas.\nDespite these limitations, our findings are an important step forward in understanding the role of learning in primary care clinics. This understanding may be particularly important in light of efforts to implement patient-centered medical home (PCMH) care models [57,58] in United States primary care settings. The purpose of the PCMH is to provide patient-centered care in which all clinic members are engaged and responsible in the care of all patients. Reciprocal learning may be an important way to improve engagement of clinic members and their ability to learn from each other to improve patient care. Improved care of patients with chronic disease is an important part of the PCMH model, and the CCM elements are shared with those of the PCMH. Successfully implementing these models of care is not a simple or static process. It requires not only attention to multiple aspects of the system in which care is delivered, but also an emphasis on patients' support system, and their ability to manage their diseases. To accomplish this requires the active and proactive engagement of staff and providers to be alert and open to new ways of doing things, to understand the impact of the way they do things on others, and to learn not only from, but with each other and respond to the needs of their patients with chronic illnesses [55]. This may explain why reciprocal learning is associated with the degree to which the chronic care model was present in each clinic.\nInterpretation of our results underscores the idea that the kinds of learning required in clinical microsystems are more sophisticated than typically acknowledged. Learning is an interdependent process that occurs between and among all members of the clinic. Managing learning as an interdependent process will likely be difficult, but our findings suggest that it will be necessary to improving the care delivered to patients with chronic disease.", "We describe the construct of reciprocal learning in primary care clinics, an activity through which clinic members learn from each other in an iterative, building process. Reciprocal learning appears to be an important attribute of learning in primary care clinics, as its presence relates to the degree of chronic care model implementation. Interventions to improve reciprocal learning among clinic members may lead to improved care of patients with chronic disease and may be relevant to improving overall clinic performance. Reciprocal learning may also be important for clinics' ability to move to more patient-centered models of care.", "The authors declare that they have no competing interests.", "MJ carried out the literature review of learning with assistance from HL. MJ, HL, RRM led the development of the learning scale. MP conceived and carried out the ABC study in collaboration with PHN. RP performed the statistical analysis. LL drafted the manuscript. All authors were involved in review and interpretation of the reported results. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1472-6963/11/44/prepub\n", "Themes or activities related to learning identified by literature search. This file lists the six themes related to learning identified in the literature search and provides references and examples for each.\nClick here for file\nLearning Scale items. This file lists the twenty-two items in the final learning scale administered in this study.\nClick here for file\nDescription of scores on each item in the reciprocal learning scale. This file lists the items in reciprocal learning scale and lists the minimum, maximum, and mean scores and standard deviation for each item.\nClick here for file" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Adult", "Aged", "Biopsy", "Bronchoscopy", "Diagnosis, Differential", "Female", "Follow-Up Studies", "Humans", "Male", "Middle Aged", "Plasma Cell Granuloma, Pulmonary", "Pneumonectomy", "Retrospective Studies", "Thoracic Surgery, Video-Assisted", "Tomography, X-Ray Computed", "Treatment Outcome" ]

[ "Background", "Histology study", "Results", "Discussion", "Conclusions", "Consent", "Competing interests", "Authors' contributions" ]

[ "Pulmonary inflammatory pseudotumor (PIP) is a rare disease and it is still debated whether it represents an inflammatory lesion characterized by uncontrolled cell growth or a true neoplasm, as recently suggested [1,2]. PIP is characterized by a cellular polymorphism, but trans-bronchial and trans-thoracic biopsies are often inconclusive for diagnosis [3]. Surgery is often useful for both treatment and diagnosis [4]. Complete resection is considered essential to prevent relapses [5].\nIn this study, the authors report their experience on the management of patients presenting with PIP.", "For a definitive histology study the following procedures were used. The surgical specimens were fixed in 10% formalin solution, embedded in paraffin, cut into sections of 4 μm, stained with hematoxylin and eosin and then subjected to conventional histology.\nFor immunohistochemical techniques we used antibodies against vimentin, cytokeratins, desmin, smooth muscle actin and epithelial membrane antigens.\nUltrastructural study by electron microscopy was performed after histology sections were fixed with 2.5% glutaraldehyde solution, post-fixed with osmic tetroxide and embedded in an epoxidic resin.\nPatients were regularly seen at our out-patient clinic for postoperative follow-up with CT scan performed annually to rule out recurrences.", "Between January 2001 and December 2009, we treated 8 patients affected by PIP, 5 men and 3 women, aged between 38 and 69 years (mean of 58 years).\nThree patients (37%) were asymptomatic and lung nodules on radiological examination were occasionally detected. 5 patients were admitted to our surgical unit for symptoms characterized by chest pain, shortness of breath and persistent cough or hemoptysis, despite antibiotic and anti-inflammatory therapy.\nThree patients had concomitant diseases at the time of hospital presentation: One had arterial hypertension, one had chronic obstructive pulmonary disease and one had viral hepatitis. This last patient had also had previous heart surgery for endocarditis. A fourth patient had a previous history of surgery for ovarian cancer.\nBlood tests performed on admission were normal in 3 patients. The remaining 5 patients had neutrophilic leucocytosis without other non-specific signs of inflammation.\nComputed tomography (CT) scan demonstrated solitary nodules (maximum diameter <3 cm) in 5 patients (62%) and lung masses (maximum diameter >3 cm) in 3 patients (37%). In 6 patients the CT scan showed findings of parenchymal tumors without signs of infiltration; in the other 2 patients there were signs of pleural infiltration (Figure 1). Distant lesions were excluded in all cases.\nComputed tomography in patients with pulmonary inflammatory pseudotumor: A) Non-calcified right lower lobe lung tumor with irregular margins and pleural bridging. B) Non-calcified right lower lobe lung tumor with adjacent pleural thickening. C) Tumor with partial internal cavitation and pleural infiltration.\nThe 3 patients who underwent FDG-PET scan had a focus of activity with SUV (Standardized Uptake Value) values between 6.2 and 9.8.\nPreoperative bronchoscopy was negative in 7 patients. In 1 patient we found bleeding from the right basal pyramid and bronchial brushing cytology was falsely positive for carcinoma with a finding of atypical epithelial cells, lymphocytes and histiocytes. In 3 patients with peripheral nodules we performed US-guided percutaneous needle aspiration. A preoperative histology examination failed to reach a definitive diagnosis in all patients.\nAt surgery, we performed two lobectomies, one segmentectomy and five wedge resections, these being performed with videothoracoscopy (VATS), except for one patient where open surgery was used. Complete tumor resection was obtained in all patients. The maximum tumor diameter was between 2.5 and 5 cm, with the gross appearance of a well circumscribed mass without a fibrous capsule. Microscopic results were characterized by a collection of inflammatory mesenchymal cells (histiocytes, plasma cells, lymphocytes and spindle cells). Intrapulmonary and mediastinal lymph nodes were found in all cases free from invasion.\nAccording to the Matsubara classification, microscopic examination revealed 2 cases of organizing pneumonia (Figure 2), 5 cases of fibrous histiocytoma (Figure 3) and one case of lymphoplasmacytoma.\nHistology study at low and high magnification of lung pseudotumor: \"organizing pneumonia\" type following Matzubara classification. There are areas of necrosis with inflammatory infiltration of macrophages and lymphocytes.\nHistology study at low and high magnification of lung pseudotumor: \"fibrous histiocytoma\" type following Matzubara classification. There is a nodular area with large amount of histiocytes, lymphocytes, plasma cells and ialin fibrous connective tissue.\nThe postoperative course was uneventful in all cases. The length of hospital stay was between 4 and 7 days.\nAt follow-up CT scan performed annually (range 11 to 112 months) (mean 58 months), there were no residual lesions, neither local nor distant recurrences One patient died of unrelated disease 23 months after surgery. Two patients escaped from follow-up.", "Although inflammatory pseudotumors may develop in different organs, such as brain and liver, the lung is the preferred site [1,4]. PIP is a rare disease with a reported incidence between 0.04 to 1.2% of all lung cancers [4]. Many synonyms have been used for this disease, usually in relation to the most represented cell type: plasma cell granuloma, inflammatory myofibroblastic tumor, fibroxantoma, histiocytoma, or pseudoneoplastic pneumonia [4]. The true incidence is unclear, as well as the clinical history in some cases and the response to different therapies [4]. Still nowadays we discuss the nature of this lesion: inflammatory or neoplastic [6]. According to some authors PIP represents a non-neoplastic process characterized by the uncontrolled growth of inflammatory cells [4]. The exact etiology of this inflammatory reaction is unknown and many hypotheses have been raised. The hypothesis of an immune disorder seems to prevail: i.e., a response to viral infection such as to human herpes virus 8 or an antigen-antibody reaction [7]. According to others, PIP represents a true neoplasm, benign or of low-grade malignancy, in consideration of the slow and localized growth [8,9]. This opinion is supported by the detection of cases of PIP with local aggressiveness and infiltration of pulmonary vessels, heart, chest wall, vertebrae and diaphragm, or by detection of cases with distant metastasis or multicentre disease [4,10-12]. Recently discovered cytogenetic abnormalities on chromosome 2p23 would support a neoplastic etiology for this disease [13-15]. PIP is more common in young adults and does not show sex predilection [9].\nPatients may remain asymptomatic in 30 to 70% of cases, the disease being occasionally detected on chest radiological examination performed for other reasons [3]. When symptoms occur they are represented by cough, fever, hemoptysis, weight loss, chest pain and respiratory infections due to endobronchial growth or mediastinal invasion [3].\nThere are no specific radiological signs for PIP. Radiological examination may show the appearance of solitary nodules or masses that may present as either calcified and well demarcated, with no evidence of malignancy, or with irregular contours [16]. Computed tomography (CT) usually shows single nodules or single masses, and multiple locations in only 5% of cases [17]. Agrons et al. reported signs of hilar, mediastinal or bronchial infiltration in 16% of the examined cases [17].\nFluorodeoxyglucose (18F) Positron emission tomography (FDG-PET) scan shows an uptake similar to that of malignant tumors [18]. We believe the use of FDG-PET scan is also useful for the study of mediastinal lymph nodes.\nAs often seen in previously reported series, also in our experience we did not reach a preoperative diagnosis with certainty, despite performing a bronchoscopy in all cases and a transparietal needle aspiration in 3 cases.\nFrozen sections from transbronchial or transparietal biopsy are often difficult to interpret, giving an uncertain diagnosis [3,6]. Due to the large number of inflammatory cells and fibroblast proliferation, differential diagnosis would include conditions such as fibrohistiocystic neoplasm, plasmocytomas, Hodgkin's sclerosing lymphoma, primary lung cancer or sarcoma, or mediastinal fibrosis [3,6,19]. However, frozen sections are usually able to rule out malignancies [3,6]. For the reasons mentioned, surgery would be recommended not only as treatment but also to reach a definitive diagnosis [4,20,21].\nDifferent classifications have been proposed for PIP.\nAccording to Cerfolio, PIP is histologically classified into two types with respect to its local invasiveness. The first type, called non-invasive PIP, mostly presenting in asymptomatic patients, appears as a small lesion without invasiveness of blood vessels or adjacent structures and usually easily resectable with a wedge resection [4]. The second type of Cerfolio classification is called invasive PIP and is usually diagnosed in younger patients, with symptoms such as fever, fatigue and weight loss. In these cases, the PIP usually has a larger size and may present with chest wall or mediastinal invasion, requiring lobectomy or pneumonectomy for a complete surgical resection. Invasive PIP may macroscopically appear as lesions infiltrating tissue planes and histologically characterized by nuclear atypia and frequent mitosis [4].\nMatsubara et al, reporting on their experience, distinguish three subtypes of PIP according to clinicopathological characteristics: 1) organizing pneumonia type (44%), fibrous histiocytoma type (44%), and lymphoplasmacytic type (12%) [6]. These authors with their classification consider most likely an inflammatory genesis for PIP [6].\nColby et al instead classify PIP into a fibrohistiocytic subtype and a plasma cell granuloma subtype [22].\nA recent classification of the World Health Association (WHA) classifies the PIP into three main histologic patterns: 1) myxoid vascular, 2) compact cord cell and 3) hypocellular fibrous. The three different patterns may coexist in the same lesion [23].\nNeither the Matsubara nor the WHA classification seems to have a prognostic value [5].\nEither medical, radiation or surgical therapy has been used for PIP.\nCorticosteroid therapy has been proposed in the case of inoperable patients, for concurrent cardio-respiratory diseases, for unresectable lesions or in case of recurrences [24,25]. The reported results are extremely variable, ranging from ineffectiveness to complete disease regression [25,26].\nRadiation therapy is usually reserved for cases of aggressive PIP, or after incomplete excision, or for postoperative recurrence or for patients at high surgical risk [27,28]. The alternative roles of radiation therapy or chemotherapy versus surgery is controversial [3,4,21,28]. It is a shared opinion that whenever possible, surgery should be considered the standard treatment [2]. Even in recurrence, whenever possible, surgical resection is advocated, allowing even in these cases a longer disease-free interval [4].\nComplete surgical resection is advocated to prevent recurrence. Prognosis after surgical radical resection is usually excellent [4,9,21]. Long-term follow-up is still required due to the possibility of local or distant recurrence, even after several years [4,24].\nWhenever possible, wedge resection should be considered the first line treatment. It would allow saving of lung parenchyma and intraoperative histological examination to exclude malignancy [2,5]. If needed, lobectomy or pneumonectomy would be performed to ensure radical resection, or if diagnosis of malignancy may not excluded. En bloc resection may be needed in cases of chest wall invasion or main bronchus, pericardium or diaphragm involvement [2].\nA collection of case studies reported a survival rate at 5 and 10 years, respectively, of 91 and 77%, with values similar to those of a low-grade malignant neoplasm [29].\nWe should also mention that the possibility of a transformation of PIP to sarcoma has been described [30,31], as well as the occurrence of aggressive forms with unfavorable outcome [32,33]. It is likely that PIP includes a complex of diseases ranging from benign fibroistiocitoma to malignant forms, thus explaining the large clinical outcome variability reported in the literature [10].", "Inflammatory pseudotumor of the lung is a rare disease, histologically characterized by the presence of myofibroblasts and chronic inflammatory cells, such as plasma cells, lymphocytes and histiocytes. Whenever possible surgical resection represents the treatment of choice. Major resections are sometimes needed due to the tumor size or local invasiveness. Complete resection is advocated to prevent recurrence. Long term follow-up is needed.", "Written informed consent was obtained from patients for publication of this report and accompanying images. A copy of the written consent is available for review by the Editor in chief of this journal.", "The authors declare that they have no competing interests.", "All authors: 1. have made substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data; 2. have been involved in drafting the manuscript or revisiting it critically for important intellectual content; 3. have given final approval of the version to be published." ]

[ null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Histology study", "Results", "Discussion", "Conclusions", "Consent", "Competing interests", "Authors' contributions" ]

[ "Pulmonary inflammatory pseudotumor (PIP) is a rare disease and it is still debated whether it represents an inflammatory lesion characterized by uncontrolled cell growth or a true neoplasm, as recently suggested [1,2]. PIP is characterized by a cellular polymorphism, but trans-bronchial and trans-thoracic biopsies are often inconclusive for diagnosis [3]. Surgery is often useful for both treatment and diagnosis [4]. Complete resection is considered essential to prevent relapses [5].\nIn this study, the authors report their experience on the management of patients presenting with PIP.", "Data prospectively entered into the registry of our surgical thoracic unit were analyzed. We retrospectively analyzed 8 patients with PIP treated by surgery between 2001 and 2009.\nPreoperative symptoms, concomitant disease and abnormal blood test results for all patients in the study group were recorded.\nPreoperative thoracic computed tomography (CT) scans were performed in all cases and extended to the abdomen and skull. In the three most recently treated patients, a Fluorodeoxyglucose (18F) Positron emission tomography (FDG-PET) scan was performed as well.\nAll patients underwent preoperative bronchoscopy with washing and brushing and/or transbronchial biopsy and preoperative cytology examination. All patients underwent surgery either by thoracotomy or by video assisted thoracoscopy (VATS) after lung function study. An intraoperative frozen section histology study was performed in all cases.\n[SUBTITLE] Histology study [SUBSECTION] For a definitive histology study the following procedures were used. The surgical specimens were fixed in 10% formalin solution, embedded in paraffin, cut into sections of 4 μm, stained with hematoxylin and eosin and then subjected to conventional histology.\nFor immunohistochemical techniques we used antibodies against vimentin, cytokeratins, desmin, smooth muscle actin and epithelial membrane antigens.\nUltrastructural study by electron microscopy was performed after histology sections were fixed with 2.5% glutaraldehyde solution, post-fixed with osmic tetroxide and embedded in an epoxidic resin.\nPatients were regularly seen at our out-patient clinic for postoperative follow-up with CT scan performed annually to rule out recurrences.\nFor a definitive histology study the following procedures were used. The surgical specimens were fixed in 10% formalin solution, embedded in paraffin, cut into sections of 4 μm, stained with hematoxylin and eosin and then subjected to conventional histology.\nFor immunohistochemical techniques we used antibodies against vimentin, cytokeratins, desmin, smooth muscle actin and epithelial membrane antigens.\nUltrastructural study by electron microscopy was performed after histology sections were fixed with 2.5% glutaraldehyde solution, post-fixed with osmic tetroxide and embedded in an epoxidic resin.\nPatients were regularly seen at our out-patient clinic for postoperative follow-up with CT scan performed annually to rule out recurrences.", "For a definitive histology study the following procedures were used. The surgical specimens were fixed in 10% formalin solution, embedded in paraffin, cut into sections of 4 μm, stained with hematoxylin and eosin and then subjected to conventional histology.\nFor immunohistochemical techniques we used antibodies against vimentin, cytokeratins, desmin, smooth muscle actin and epithelial membrane antigens.\nUltrastructural study by electron microscopy was performed after histology sections were fixed with 2.5% glutaraldehyde solution, post-fixed with osmic tetroxide and embedded in an epoxidic resin.\nPatients were regularly seen at our out-patient clinic for postoperative follow-up with CT scan performed annually to rule out recurrences.", "Between January 2001 and December 2009, we treated 8 patients affected by PIP, 5 men and 3 women, aged between 38 and 69 years (mean of 58 years).\nThree patients (37%) were asymptomatic and lung nodules on radiological examination were occasionally detected. 5 patients were admitted to our surgical unit for symptoms characterized by chest pain, shortness of breath and persistent cough or hemoptysis, despite antibiotic and anti-inflammatory therapy.\nThree patients had concomitant diseases at the time of hospital presentation: One had arterial hypertension, one had chronic obstructive pulmonary disease and one had viral hepatitis. This last patient had also had previous heart surgery for endocarditis. A fourth patient had a previous history of surgery for ovarian cancer.\nBlood tests performed on admission were normal in 3 patients. The remaining 5 patients had neutrophilic leucocytosis without other non-specific signs of inflammation.\nComputed tomography (CT) scan demonstrated solitary nodules (maximum diameter <3 cm) in 5 patients (62%) and lung masses (maximum diameter >3 cm) in 3 patients (37%). In 6 patients the CT scan showed findings of parenchymal tumors without signs of infiltration; in the other 2 patients there were signs of pleural infiltration (Figure 1). Distant lesions were excluded in all cases.\nComputed tomography in patients with pulmonary inflammatory pseudotumor: A) Non-calcified right lower lobe lung tumor with irregular margins and pleural bridging. B) Non-calcified right lower lobe lung tumor with adjacent pleural thickening. C) Tumor with partial internal cavitation and pleural infiltration.\nThe 3 patients who underwent FDG-PET scan had a focus of activity with SUV (Standardized Uptake Value) values between 6.2 and 9.8.\nPreoperative bronchoscopy was negative in 7 patients. In 1 patient we found bleeding from the right basal pyramid and bronchial brushing cytology was falsely positive for carcinoma with a finding of atypical epithelial cells, lymphocytes and histiocytes. In 3 patients with peripheral nodules we performed US-guided percutaneous needle aspiration. A preoperative histology examination failed to reach a definitive diagnosis in all patients.\nAt surgery, we performed two lobectomies, one segmentectomy and five wedge resections, these being performed with videothoracoscopy (VATS), except for one patient where open surgery was used. Complete tumor resection was obtained in all patients. The maximum tumor diameter was between 2.5 and 5 cm, with the gross appearance of a well circumscribed mass without a fibrous capsule. Microscopic results were characterized by a collection of inflammatory mesenchymal cells (histiocytes, plasma cells, lymphocytes and spindle cells). Intrapulmonary and mediastinal lymph nodes were found in all cases free from invasion.\nAccording to the Matsubara classification, microscopic examination revealed 2 cases of organizing pneumonia (Figure 2), 5 cases of fibrous histiocytoma (Figure 3) and one case of lymphoplasmacytoma.\nHistology study at low and high magnification of lung pseudotumor: \"organizing pneumonia\" type following Matzubara classification. There are areas of necrosis with inflammatory infiltration of macrophages and lymphocytes.\nHistology study at low and high magnification of lung pseudotumor: \"fibrous histiocytoma\" type following Matzubara classification. There is a nodular area with large amount of histiocytes, lymphocytes, plasma cells and ialin fibrous connective tissue.\nThe postoperative course was uneventful in all cases. The length of hospital stay was between 4 and 7 days.\nAt follow-up CT scan performed annually (range 11 to 112 months) (mean 58 months), there were no residual lesions, neither local nor distant recurrences One patient died of unrelated disease 23 months after surgery. Two patients escaped from follow-up.", "Although inflammatory pseudotumors may develop in different organs, such as brain and liver, the lung is the preferred site [1,4]. PIP is a rare disease with a reported incidence between 0.04 to 1.2% of all lung cancers [4]. Many synonyms have been used for this disease, usually in relation to the most represented cell type: plasma cell granuloma, inflammatory myofibroblastic tumor, fibroxantoma, histiocytoma, or pseudoneoplastic pneumonia [4]. The true incidence is unclear, as well as the clinical history in some cases and the response to different therapies [4]. Still nowadays we discuss the nature of this lesion: inflammatory or neoplastic [6]. According to some authors PIP represents a non-neoplastic process characterized by the uncontrolled growth of inflammatory cells [4]. The exact etiology of this inflammatory reaction is unknown and many hypotheses have been raised. The hypothesis of an immune disorder seems to prevail: i.e., a response to viral infection such as to human herpes virus 8 or an antigen-antibody reaction [7]. According to others, PIP represents a true neoplasm, benign or of low-grade malignancy, in consideration of the slow and localized growth [8,9]. This opinion is supported by the detection of cases of PIP with local aggressiveness and infiltration of pulmonary vessels, heart, chest wall, vertebrae and diaphragm, or by detection of cases with distant metastasis or multicentre disease [4,10-12]. Recently discovered cytogenetic abnormalities on chromosome 2p23 would support a neoplastic etiology for this disease [13-15]. PIP is more common in young adults and does not show sex predilection [9].\nPatients may remain asymptomatic in 30 to 70% of cases, the disease being occasionally detected on chest radiological examination performed for other reasons [3]. When symptoms occur they are represented by cough, fever, hemoptysis, weight loss, chest pain and respiratory infections due to endobronchial growth or mediastinal invasion [3].\nThere are no specific radiological signs for PIP. Radiological examination may show the appearance of solitary nodules or masses that may present as either calcified and well demarcated, with no evidence of malignancy, or with irregular contours [16]. Computed tomography (CT) usually shows single nodules or single masses, and multiple locations in only 5% of cases [17]. Agrons et al. reported signs of hilar, mediastinal or bronchial infiltration in 16% of the examined cases [17].\nFluorodeoxyglucose (18F) Positron emission tomography (FDG-PET) scan shows an uptake similar to that of malignant tumors [18]. We believe the use of FDG-PET scan is also useful for the study of mediastinal lymph nodes.\nAs often seen in previously reported series, also in our experience we did not reach a preoperative diagnosis with certainty, despite performing a bronchoscopy in all cases and a transparietal needle aspiration in 3 cases.\nFrozen sections from transbronchial or transparietal biopsy are often difficult to interpret, giving an uncertain diagnosis [3,6]. Due to the large number of inflammatory cells and fibroblast proliferation, differential diagnosis would include conditions such as fibrohistiocystic neoplasm, plasmocytomas, Hodgkin's sclerosing lymphoma, primary lung cancer or sarcoma, or mediastinal fibrosis [3,6,19]. However, frozen sections are usually able to rule out malignancies [3,6]. For the reasons mentioned, surgery would be recommended not only as treatment but also to reach a definitive diagnosis [4,20,21].\nDifferent classifications have been proposed for PIP.\nAccording to Cerfolio, PIP is histologically classified into two types with respect to its local invasiveness. The first type, called non-invasive PIP, mostly presenting in asymptomatic patients, appears as a small lesion without invasiveness of blood vessels or adjacent structures and usually easily resectable with a wedge resection [4]. The second type of Cerfolio classification is called invasive PIP and is usually diagnosed in younger patients, with symptoms such as fever, fatigue and weight loss. In these cases, the PIP usually has a larger size and may present with chest wall or mediastinal invasion, requiring lobectomy or pneumonectomy for a complete surgical resection. Invasive PIP may macroscopically appear as lesions infiltrating tissue planes and histologically characterized by nuclear atypia and frequent mitosis [4].\nMatsubara et al, reporting on their experience, distinguish three subtypes of PIP according to clinicopathological characteristics: 1) organizing pneumonia type (44%), fibrous histiocytoma type (44%), and lymphoplasmacytic type (12%) [6]. These authors with their classification consider most likely an inflammatory genesis for PIP [6].\nColby et al instead classify PIP into a fibrohistiocytic subtype and a plasma cell granuloma subtype [22].\nA recent classification of the World Health Association (WHA) classifies the PIP into three main histologic patterns: 1) myxoid vascular, 2) compact cord cell and 3) hypocellular fibrous. The three different patterns may coexist in the same lesion [23].\nNeither the Matsubara nor the WHA classification seems to have a prognostic value [5].\nEither medical, radiation or surgical therapy has been used for PIP.\nCorticosteroid therapy has been proposed in the case of inoperable patients, for concurrent cardio-respiratory diseases, for unresectable lesions or in case of recurrences [24,25]. The reported results are extremely variable, ranging from ineffectiveness to complete disease regression [25,26].\nRadiation therapy is usually reserved for cases of aggressive PIP, or after incomplete excision, or for postoperative recurrence or for patients at high surgical risk [27,28]. The alternative roles of radiation therapy or chemotherapy versus surgery is controversial [3,4,21,28]. It is a shared opinion that whenever possible, surgery should be considered the standard treatment [2]. Even in recurrence, whenever possible, surgical resection is advocated, allowing even in these cases a longer disease-free interval [4].\nComplete surgical resection is advocated to prevent recurrence. Prognosis after surgical radical resection is usually excellent [4,9,21]. Long-term follow-up is still required due to the possibility of local or distant recurrence, even after several years [4,24].\nWhenever possible, wedge resection should be considered the first line treatment. It would allow saving of lung parenchyma and intraoperative histological examination to exclude malignancy [2,5]. If needed, lobectomy or pneumonectomy would be performed to ensure radical resection, or if diagnosis of malignancy may not excluded. En bloc resection may be needed in cases of chest wall invasion or main bronchus, pericardium or diaphragm involvement [2].\nA collection of case studies reported a survival rate at 5 and 10 years, respectively, of 91 and 77%, with values similar to those of a low-grade malignant neoplasm [29].\nWe should also mention that the possibility of a transformation of PIP to sarcoma has been described [30,31], as well as the occurrence of aggressive forms with unfavorable outcome [32,33]. It is likely that PIP includes a complex of diseases ranging from benign fibroistiocitoma to malignant forms, thus explaining the large clinical outcome variability reported in the literature [10].", "Inflammatory pseudotumor of the lung is a rare disease, histologically characterized by the presence of myofibroblasts and chronic inflammatory cells, such as plasma cells, lymphocytes and histiocytes. Whenever possible surgical resection represents the treatment of choice. Major resections are sometimes needed due to the tumor size or local invasiveness. Complete resection is advocated to prevent recurrence. Long term follow-up is needed.", "Written informed consent was obtained from patients for publication of this report and accompanying images. A copy of the written consent is available for review by the Editor in chief of this journal.", "The authors declare that they have no competing interests.", "All authors: 1. have made substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data; 2. have been involved in drafting the manuscript or revisiting it critically for important intellectual content; 3. have given final approval of the version to be published." ]

[ null, "methods", null, null, null, null, null, null, null ]

[]

[ "Administration, Inhalation", "Adult", "British Columbia", "Cocaine-Related Disorders", "Crack Cocaine", "Crime", "Female", "Harm Reduction", "Ill-Housed Persons", "Humans", "Indians, North American", "Male", "Middle Aged", "Patient Acceptance of Health Care", "Public Facilities", "Risk-Taking", "Substance Abuse Treatment Centers" ]

[ "Background", "Results", "Discussion", "Competing interests", "Authors' contributions", "Funding" ]

[ "The use of illicit drugs in public settings, including street, alleys and parks is both a public health and public order concern in many urban areas [1-3]. To date, the use of injection drugs in public settings has received the most attention from policy-makers and public health researchers [2,4,5]. Public injecting is known to present problems for citizens who reside in or around areas where public drug use is prevalent, and scientific studies have documented that using injection drugs in public settings can discourage safer injecting practices resulting in many public health problems, including increased risk for drug overdose events and HIV and other blood-borne infections [6-8]. As a result, some cities have implemented supervised injection facilities which aim to provide an alternative injecting environment that reduces both the health risks associated with injection drug use and the street disorder it can generate [9-13]. While supervised injection facilities have been noted to have measurable success in achieving these public health and public order objectives, the use of inhalable drugs, particularly crack cocaine smoking, has been growing in popularity in many street-based drug scenes [14-16].\nIn Vancouver, Canada the popularity of crack cocaine and ease of administration through smoking has made public crack cocaine use a common feature of the streets in the city's drug use epicentre, known as the Downtown Eastside [17]. Public crack cocaine smoking is posing a growing burden for law enforcement agencies responsible for maintaining public order [18]. In addition, the health and social harms associated with crack cocaine smoking are extensive. Compared to other drug user populations, crack users are more likely to engage in risky behaviors [19-21] and illegal activities [22,23] and to experience homelessness [14] and health problems [14,24-27], yet are less likely to access health and social services [28]. It has also been recently documented that daily crack cocaine smokers are at a four-fold greater risk of contracting HIV compared to their drug using peers who smoke crack cocaine less often or not at all [16].\nGiven the dramatic rise in crack cocaine smoking and the public order and public health concerns associated with it, the need for targeted interventions for people who smoke crack cocaine is unambiguous. One potential intervention that is receiving increasing attention from public health officials, health researchers and local community groups is supervised drug consumption facilities analogous to supervised injection sites but that accommodate crack smoking and distribute drug consumption materials specific to safer inhalation (such as sterile crack pipes, mouthpieces, and screens) [16,29-33]. The Canadian Institutes of Health Research recently approved funding to conduct a randomized control trial to evaluate the impact of a supervised inhalation facility on access to medical and social services, particularly addiction treatment, among Vancouver-based crack cocaine smokers [34].\nPrevious studies have assessed general willingness among local drug users to use a supervised inhalation facility [35,36]; however, these studies were not primarily concerned with street disorder and therefore did not consider the specific risks associated with smoking crack cocaine in public areas nor did they assess willingness to use an inhalation facility among public crack cocaine smokers exclusively. Therefore we conducted a study focused on public crack smokers to identify factors associated with this practice. We also sought to assess willingness to use an inhalation facility among individuals who smoke crack cocaine in public areas to determine the potential impact a supervised inhalation facility might have on street disorder in Vancouver, Canada.", "During the study period 623 participants were seen for study follow-up visits and reported smoking crack cocaine in the last six months. These included 249 (40%) women and 231 (37%) persons who identified as Aboriginal. The median number of times that participants reported smoking crack cocaine in an average day was 4 (interquartile range = 2-10). Among our sample of 623 crack smokers, a total of 382 (61%) reported using in public areas in the last six months. The characteristics of the study sample stratified by public drug use are presented in Table 1, and the univariate analyses of behavioral and socio-demographic variables associated with public drug use among crack cocaine smokers are presented in Table 2. The results of the multivariate logistic regression for factors associated with public drug use among crack cocaine smokers are also shown in Table 2. Factors that remained independently associated with our outcome of interest included: daily heroin injection, daily crack cocaine smoking, encounters with police and drug dealing (see Table 2).\nCharacteristics of crack cocaine smokers stratified by public drug use (n = 623)\nNote: a Public locations include: city streets, parks, public washrooms, parking lots, clubs or bars, and abandon buildings; c IQR = Inter Quartile Range; d Denotes activities or situations referring to previous 6 months; e Denotes activities or situations referring to previous month.\nUnivariate and multivariate analyses of factors associated with public drug use among crack cocaine smokersa (n = 623)\nNote: a Public areas included: city streets, parks, public washrooms, parking lots, clubs or bars, and abandon buildings; bOR = Odds Ratio, CI = Confidence Interval; AOR = Adjusted Odds Ratio; cUnless otherwise stated, values are based on Pearson's chi-square test for categorical variables and Wilcoxon rank sum test for continuous variables with 1 degree of freedom; d Denotes activities or situations referring to previous 6 months; e Denotes activities or situations referring to previous month. *p-value and 95% CI reported from Fisher's Exact Test as 25% of cells had expected counts less than 5.\nFor our second analysis, the demographic and behavioural characteristics of public crack cocaine smokers stratified by willingness to use a supervised inhalation room are presented in Table 3, and the univariate results of factors associated with willingness to use a supervised inhalation room are presented in Table 4. The results of the multivariate logistic regression for factors associated with willingness to use a supervised inhalation room are also shown in Table 4. Factors that remained independently associated with willingness included: female gender, risky pipe sharing and recent encounters with police (see Table 4.).\nCharacteristics of crack cocaine smokers who use drugs in public stratified by willingness to use a supervised inhalation room (n = 382)\nNote: c IQR = Inter Quartile Range; d Denotes activities or situations referring to previous 6 months; e Denotes activities or situations referring to previous month; f Drug scene exposure was defined as spending an average of 7 or more hours on the street each day in Vancouver's drug use epicenter in the previous six months\nUnivariate and multivariate analyses of factors associated with willingness to use a supervised inhalation room among participants that smoke crack cocaine and use drugs in public locations (n = 382)\nNote: aOR = Odds Ratio, CI = Confidence Interval; bUnless otherwise stated, values are based on Pearson's chi-square test for categorical variables and Wilcoxon rank sum test for continuous variables with 1 degree of freedom; cAOR = Adjusted Odds Ratio; dDenotes activities or situations referring to previous 6 months; e Denotes activities or situations referring to previous month. *p-value and 95% CI reported from Fisher's Exact Test as 25% of cells had expected counts less than 5.", "We found that the majority of crack cocaine smokers in our study reported having used drugs in public areas at some point in the last six months. This group was more likely to be higher-intensity drug users with respect to heroin injection and crack cocaine smoking, have encounters with the police and be involved in drug dealing. Of these public crack cocaine smokers, 71% reported being willing to use a supervised inhalation room if one was available. Individuals who reported being willing were more likely to be female, engage in risky pipe sharing and have encounters with the police.\nThe profile of public crack cocaine smokers as higher-intensity drug users who have interactions with the criminal justice system is reflective of previous findings describing public injection drug user populations [1,5]. The association between drug dealing and public crack use may reflect the increased amount of time individuals spend on the street when engaged in street-level drug dealing. It may also be a function of the accessibility of drugs and additional resources gained through drug dealing which may lead to greater drug consumption and hence a greater likelihood for consuming in public areas [23].\nOur finding that 71% of public crack cocaine smokers are willing to use an inhalation facility also supports previous willingness estimates conducted among the general population of Vancouver-based illicit drug users and suggests that an intervention of this nature will likely reach the target population [36]. The high degree of willingness that this study found among public crack cocaine smokers to use an inhalation facility suggests that, like supervised injection facilities, these interventions are likely to successfully encourage public drug users to relocate to indoor venues.\nThe increased likelihood of being willing to use an inhalation facility among female participants may reflect heightened vulnerability of women involved in street drug use and it is noteworthy that Vancouver's supervised injection facility has had success in attracting vulnerable female drug users and providing them with safer alternatives to street-based drug using venues. In previous research female IDU have described the unique role that Vancouver's supervised injection facility has played in promoting their physical security and health safety [42].\nInterestingly, one of the common features among both public crack cocaine smokers and those who are willing to use a supervised inhalation facility is their elevated likelihood of recently having encounters with law enforcement. This suggests that public crack cocaine smokers who are the subject of law enforcement attention are very willing to relocate to alternative off-street and health-focussed environments if they were made available. Indeed, our data indicate that 81% of public crack cocaine smokers who have had a recent encounter with police are willing to use a supervised inhalation facility.\nA key implication of these findings is that there is a large demand for supervised inhalation rooms among individuals that are potentially key contributors to drug-related street disorder. The association between public crack smoking and encounters with police suggests that interventions of this nature are likely to target a critical sub-population of drug users and could be a valuable tool for police in the management of street disorder. Previous studies have found that Vancouver police regularly refer public injection drug users to the local supervised injection facility [43]. Since our analysis indicates that local police are already frequently interacting with public crack smokers the establishment of a supervised inhalation facility could provide a unique opportunity for police to direct this vulnerable group to a low-threshold service where they will have opportunities to be linked with appropriate health and social services.\nIt is critical to note that although this study suggests that supervised inhalation facilities could aid in the reduction of public disorder, drug consumption facilities do not address the route causes of street disorder and are not appropriate substitutes for other essential health and social interventions such as supportive housing and addiction treatment. To be effective supervised inhalation facilities should be integrated into broader comprehensive approaches to addressing the problems associated with illicit drug addiction.\nThis study has a number of limitations. Firstly, VIDUS is a community recruited non-randomized sample and therefore our findings may not be generalizable to other settings. If supervised inhalation facilities are being considered in other settings, willingness studies should be conducted among the local target population and should not rely on the findings emerging from our setting. The generalizability of our findings is also limited by our study sample which was restricted to individuals with a history of injection drug use. Crack cocaine smokers who did not have a history of injection drug use were not eligible for our study. Given the harms associated with injection drug use we anticipate that if a selection effect were present it would likely bias our sample towards high risk drug users, suggesting that this group would be an appropriate target population for public health intervention. We should also note that among our study sample daily crack cocaine smoking was significantly more common than daily injecting, suggesting that despite the requirement of a history of injecting, our sample represents a primarily crack cocaine smoking population. Secondly, some of our measures relied on self-report and could be vulnerable to socially desirable reporting. This would have likely been of most relevant to our measure of willingness, since respondents might perceive a pressure to report being willing to engage with low-threshold services of this nature given the widespread activism among local drug users in our study setting to implement supervised drug consumption facilities [32]. While it is possible that some respondents may over-report willingness, a previous study comparing measures of willingness to use a supervised injection facility before it was established with later reports of actual attendance after an injection facility was established suggests that willingness measures are good predictors of later behaviour among this population [44]. Lastly, socially desirable reporting could have influenced reports of stigmatized behaviour, such as public drug use leading to an underestimation of public crack smoking. If social desirability was an issue in our study we suspect our finding would be a conservative indication of the prevalence of and harms associated with public drug use among crack cocaine smokers.\nIn summary, our study found that locally public crack smoking is a common practice that is also associated with recent encounters with police. We found that the majority of public crack smokers were willing to use an inhalation facility if one were available. Furthermore, public crack smokers who had recent encounters with police were even more likely to be willing to use an inhalation room, suggesting that supervised inhalation facilities may offer unique opportunities to decrease one component of drug-related street disorder and reduce the burden on local law enforcement agencies.", "JM has received grants from, served as an ad hoc advisor to, or spoke at various events sponsored by; Abbott, Argos Therapeutics, Bioject Inc, Boehringer Ingelheim, BMS, Gilead Sciences, GlaxoSmithKline, Hoffmann-La Roche, Janssen-Ortho, Merck Frosst, Pfizer, Schering, Serono Inc, TheraTechnologies, Tibotec, Trimeris.\nAuthors declare no other competing interests.", "The specific contributions of each author are as follows: KD, TK, and EW were responsible for study design; JQ conducted the statistical analyses; KD prepared the first draft of the analysis; TK, JB, JM and EW contributed to the main content and provided critical comments on the final draft. All authors approved the final manuscript.", "The study was supported by the US National Institutes of Health (R01DA011591) and (R01DA021525) and the Canadian Institutes of Health Research (MOP-79297, RAA-79918). Thomas Kerr is supported by the Michael Smith Foundation for Health Research and the Canadian Institutes of Health Research. Kora DeBeck is supported by a Michael Smith Foundation for Health Research Senior Graduate Trainee Award and a Canadian Institutes of Health Research Doctoral Research Award. Julio Montaner has received an Avant-Garde award (DP1DA026182) from the National Institute of Drug Abuse, US National Institutes of Health." ]

[ null, null, null, null, null, null ]

[ "Background", "Methods", "Results", "Discussion", "Competing interests", "Authors' contributions", "Funding" ]

[ "The use of illicit drugs in public settings, including street, alleys and parks is both a public health and public order concern in many urban areas [1-3]. To date, the use of injection drugs in public settings has received the most attention from policy-makers and public health researchers [2,4,5]. Public injecting is known to present problems for citizens who reside in or around areas where public drug use is prevalent, and scientific studies have documented that using injection drugs in public settings can discourage safer injecting practices resulting in many public health problems, including increased risk for drug overdose events and HIV and other blood-borne infections [6-8]. As a result, some cities have implemented supervised injection facilities which aim to provide an alternative injecting environment that reduces both the health risks associated with injection drug use and the street disorder it can generate [9-13]. While supervised injection facilities have been noted to have measurable success in achieving these public health and public order objectives, the use of inhalable drugs, particularly crack cocaine smoking, has been growing in popularity in many street-based drug scenes [14-16].\nIn Vancouver, Canada the popularity of crack cocaine and ease of administration through smoking has made public crack cocaine use a common feature of the streets in the city's drug use epicentre, known as the Downtown Eastside [17]. Public crack cocaine smoking is posing a growing burden for law enforcement agencies responsible for maintaining public order [18]. In addition, the health and social harms associated with crack cocaine smoking are extensive. Compared to other drug user populations, crack users are more likely to engage in risky behaviors [19-21] and illegal activities [22,23] and to experience homelessness [14] and health problems [14,24-27], yet are less likely to access health and social services [28]. It has also been recently documented that daily crack cocaine smokers are at a four-fold greater risk of contracting HIV compared to their drug using peers who smoke crack cocaine less often or not at all [16].\nGiven the dramatic rise in crack cocaine smoking and the public order and public health concerns associated with it, the need for targeted interventions for people who smoke crack cocaine is unambiguous. One potential intervention that is receiving increasing attention from public health officials, health researchers and local community groups is supervised drug consumption facilities analogous to supervised injection sites but that accommodate crack smoking and distribute drug consumption materials specific to safer inhalation (such as sterile crack pipes, mouthpieces, and screens) [16,29-33]. The Canadian Institutes of Health Research recently approved funding to conduct a randomized control trial to evaluate the impact of a supervised inhalation facility on access to medical and social services, particularly addiction treatment, among Vancouver-based crack cocaine smokers [34].\nPrevious studies have assessed general willingness among local drug users to use a supervised inhalation facility [35,36]; however, these studies were not primarily concerned with street disorder and therefore did not consider the specific risks associated with smoking crack cocaine in public areas nor did they assess willingness to use an inhalation facility among public crack cocaine smokers exclusively. Therefore we conducted a study focused on public crack smokers to identify factors associated with this practice. We also sought to assess willingness to use an inhalation facility among individuals who smoke crack cocaine in public areas to determine the potential impact a supervised inhalation facility might have on street disorder in Vancouver, Canada.", "Data for this study was obtained from the Vancouver Injection Drug Users Study, which is an open prospective cohort that began enrolling people who inject drugs (IDU) through street outreach as self-referral in May 1996. This study has been described in detail previously [37,38]. In brief, to be eligible participants at recruitment must reside in the Greater Vancouver Regional District, have injected illicit drugs in the previous month, and provide written informed consent. At enrollment and on bi-annual basis participants complete an interviewer-administered questionnaire, and after an examination with a study nurse provide a blood sample for serologic testing. At each study visit participants are provided with a stipend ($20 CDN) for their time. The study has received ethics approval from St. Paul's Hospital and the University of British Columbia's Research Ethics Board. The present analyses are restricted to those participants who reported smoking crack cocaine in the last six months, and were seen for study follow-up during the period of November 2008 and June 2009 as measures for one of our outcomes of interest are available only for this sample period.\nIn our first analysis among crack cocaine smokers, the outcome of interest was using drugs (non-injection) in public areas in the last six months. As in previous analyses, public areas included city streets, parks, public washrooms, parking lots, clubs or bars and abandoned buildings. To characterize our outcome of interest we a priori selected a range of socio-demographic and behavioural variables we hypothesized might be relevant to smoking crack cocaine in public areas. This selection was informed by the 'risk environment framework' and previous analyses among street-involved drug users highlighting connections between social, structural and environmental level factors and risky drug consumption practices [14,31,39-41]. Variables included: age (per year older); gender (female vs. male); Aboriginal Ancestry (yes vs. no); limited access to private space, defined as answering \"no\" to the question: \"Do you have a private indoor space for socializing with friends and acquaintances?\" or reporting that the number of guests they were allowed to have in their residence at one time was restricted to less than three (yes vs. no); daily cocaine injection (yes vs. no); daily heroin injection (yes vs. no); daily crack cocaine smoking (yes vs. no); non-fatal overdose, self identified by participants (yes vs. no); encounters with police in the last month, defined as being questioned, searched or stopped by police (yes vs. no); being a victim of violence defined as being physically assaulted (yes vs. no); sex trade involvement, defined as exchanging sex for money, shelter, drugs or other commodities (yes vs. no), and participation in drug dealing (yes vs. no). Unless otherwise stated, all drug use and behavioural variables refer to the previous six month period.\nIn a second analysis, we sought to assess and identify predictors of willingness to use a supervised inhalation room. Because we were particularly concerned with public drug use we restricted our sample to crack cocaine smokers that reported recently using non-injection drugs in public areas. To measure willingness we asked participants \"If there was a safe place to smoke your drugs (ventilated inhalation room), close to where you buy or use, would you use it?\"\nVariables of interest for our second analysis were also selected a priori based on factors we hypothesized might be associated with willingness to use an inhalation room. These included age (per year older); gender (female vs. male); Aboriginal Ancestry (yes vs. no); limited access to private space, as defined above (yes vs. no); drug scene exposure, defined as spending an average of seven or more hours on the street each day in Vancouver's drug use epicentre in the previous six months (yes vs. no); most drug use in public areas, defined based on reports that public locations were where they most frequently used drugs (yes vs. no); daily crack cocaine smoking (yes vs. no); risky pipe sharing, defined as reporting sharing a crack pipe or mouthpiece in the same six month period as having burns or sores on their mouth (yes vs. no); encounters with police in the last month (yes vs. no); and being a victim of violence (yes vs. no). As above, unless otherwise stated, all drug use and behavioural variables refer to the previous six month period.\nFor both of our first and second analyses, we used univariate and multivariate statistics to determine factors associated with our outcomes of interest. In univariate analysis categorical explanatory variables were analyzed using Pearson's chi-square test and continuous variables were analyzed using the Wilcoxon rank sum test. Fisher's exact test was used when one or more of the cell counts was less than or equal to five. To evaluate factors independently associated with our outcomes of interest, all variables that were p < 0.05 in univariate analyses were entered into the respective multivariate regression models. All statistical analyses were performed using SAS software version 9.1 (SAS, Cary, NC). All p-values are two sided.", "During the study period 623 participants were seen for study follow-up visits and reported smoking crack cocaine in the last six months. These included 249 (40%) women and 231 (37%) persons who identified as Aboriginal. The median number of times that participants reported smoking crack cocaine in an average day was 4 (interquartile range = 2-10). Among our sample of 623 crack smokers, a total of 382 (61%) reported using in public areas in the last six months. The characteristics of the study sample stratified by public drug use are presented in Table 1, and the univariate analyses of behavioral and socio-demographic variables associated with public drug use among crack cocaine smokers are presented in Table 2. The results of the multivariate logistic regression for factors associated with public drug use among crack cocaine smokers are also shown in Table 2. Factors that remained independently associated with our outcome of interest included: daily heroin injection, daily crack cocaine smoking, encounters with police and drug dealing (see Table 2).\nCharacteristics of crack cocaine smokers stratified by public drug use (n = 623)\nNote: a Public locations include: city streets, parks, public washrooms, parking lots, clubs or bars, and abandon buildings; c IQR = Inter Quartile Range; d Denotes activities or situations referring to previous 6 months; e Denotes activities or situations referring to previous month.\nUnivariate and multivariate analyses of factors associated with public drug use among crack cocaine smokersa (n = 623)\nNote: a Public areas included: city streets, parks, public washrooms, parking lots, clubs or bars, and abandon buildings; bOR = Odds Ratio, CI = Confidence Interval; AOR = Adjusted Odds Ratio; cUnless otherwise stated, values are based on Pearson's chi-square test for categorical variables and Wilcoxon rank sum test for continuous variables with 1 degree of freedom; d Denotes activities or situations referring to previous 6 months; e Denotes activities or situations referring to previous month. *p-value and 95% CI reported from Fisher's Exact Test as 25% of cells had expected counts less than 5.\nFor our second analysis, the demographic and behavioural characteristics of public crack cocaine smokers stratified by willingness to use a supervised inhalation room are presented in Table 3, and the univariate results of factors associated with willingness to use a supervised inhalation room are presented in Table 4. The results of the multivariate logistic regression for factors associated with willingness to use a supervised inhalation room are also shown in Table 4. Factors that remained independently associated with willingness included: female gender, risky pipe sharing and recent encounters with police (see Table 4.).\nCharacteristics of crack cocaine smokers who use drugs in public stratified by willingness to use a supervised inhalation room (n = 382)\nNote: c IQR = Inter Quartile Range; d Denotes activities or situations referring to previous 6 months; e Denotes activities or situations referring to previous month; f Drug scene exposure was defined as spending an average of 7 or more hours on the street each day in Vancouver's drug use epicenter in the previous six months\nUnivariate and multivariate analyses of factors associated with willingness to use a supervised inhalation room among participants that smoke crack cocaine and use drugs in public locations (n = 382)\nNote: aOR = Odds Ratio, CI = Confidence Interval; bUnless otherwise stated, values are based on Pearson's chi-square test for categorical variables and Wilcoxon rank sum test for continuous variables with 1 degree of freedom; cAOR = Adjusted Odds Ratio; dDenotes activities or situations referring to previous 6 months; e Denotes activities or situations referring to previous month. *p-value and 95% CI reported from Fisher's Exact Test as 25% of cells had expected counts less than 5.", "We found that the majority of crack cocaine smokers in our study reported having used drugs in public areas at some point in the last six months. This group was more likely to be higher-intensity drug users with respect to heroin injection and crack cocaine smoking, have encounters with the police and be involved in drug dealing. Of these public crack cocaine smokers, 71% reported being willing to use a supervised inhalation room if one was available. Individuals who reported being willing were more likely to be female, engage in risky pipe sharing and have encounters with the police.\nThe profile of public crack cocaine smokers as higher-intensity drug users who have interactions with the criminal justice system is reflective of previous findings describing public injection drug user populations [1,5]. The association between drug dealing and public crack use may reflect the increased amount of time individuals spend on the street when engaged in street-level drug dealing. It may also be a function of the accessibility of drugs and additional resources gained through drug dealing which may lead to greater drug consumption and hence a greater likelihood for consuming in public areas [23].\nOur finding that 71% of public crack cocaine smokers are willing to use an inhalation facility also supports previous willingness estimates conducted among the general population of Vancouver-based illicit drug users and suggests that an intervention of this nature will likely reach the target population [36]. The high degree of willingness that this study found among public crack cocaine smokers to use an inhalation facility suggests that, like supervised injection facilities, these interventions are likely to successfully encourage public drug users to relocate to indoor venues.\nThe increased likelihood of being willing to use an inhalation facility among female participants may reflect heightened vulnerability of women involved in street drug use and it is noteworthy that Vancouver's supervised injection facility has had success in attracting vulnerable female drug users and providing them with safer alternatives to street-based drug using venues. In previous research female IDU have described the unique role that Vancouver's supervised injection facility has played in promoting their physical security and health safety [42].\nInterestingly, one of the common features among both public crack cocaine smokers and those who are willing to use a supervised inhalation facility is their elevated likelihood of recently having encounters with law enforcement. This suggests that public crack cocaine smokers who are the subject of law enforcement attention are very willing to relocate to alternative off-street and health-focussed environments if they were made available. Indeed, our data indicate that 81% of public crack cocaine smokers who have had a recent encounter with police are willing to use a supervised inhalation facility.\nA key implication of these findings is that there is a large demand for supervised inhalation rooms among individuals that are potentially key contributors to drug-related street disorder. The association between public crack smoking and encounters with police suggests that interventions of this nature are likely to target a critical sub-population of drug users and could be a valuable tool for police in the management of street disorder. Previous studies have found that Vancouver police regularly refer public injection drug users to the local supervised injection facility [43]. Since our analysis indicates that local police are already frequently interacting with public crack smokers the establishment of a supervised inhalation facility could provide a unique opportunity for police to direct this vulnerable group to a low-threshold service where they will have opportunities to be linked with appropriate health and social services.\nIt is critical to note that although this study suggests that supervised inhalation facilities could aid in the reduction of public disorder, drug consumption facilities do not address the route causes of street disorder and are not appropriate substitutes for other essential health and social interventions such as supportive housing and addiction treatment. To be effective supervised inhalation facilities should be integrated into broader comprehensive approaches to addressing the problems associated with illicit drug addiction.\nThis study has a number of limitations. Firstly, VIDUS is a community recruited non-randomized sample and therefore our findings may not be generalizable to other settings. If supervised inhalation facilities are being considered in other settings, willingness studies should be conducted among the local target population and should not rely on the findings emerging from our setting. The generalizability of our findings is also limited by our study sample which was restricted to individuals with a history of injection drug use. Crack cocaine smokers who did not have a history of injection drug use were not eligible for our study. Given the harms associated with injection drug use we anticipate that if a selection effect were present it would likely bias our sample towards high risk drug users, suggesting that this group would be an appropriate target population for public health intervention. We should also note that among our study sample daily crack cocaine smoking was significantly more common than daily injecting, suggesting that despite the requirement of a history of injecting, our sample represents a primarily crack cocaine smoking population. Secondly, some of our measures relied on self-report and could be vulnerable to socially desirable reporting. This would have likely been of most relevant to our measure of willingness, since respondents might perceive a pressure to report being willing to engage with low-threshold services of this nature given the widespread activism among local drug users in our study setting to implement supervised drug consumption facilities [32]. While it is possible that some respondents may over-report willingness, a previous study comparing measures of willingness to use a supervised injection facility before it was established with later reports of actual attendance after an injection facility was established suggests that willingness measures are good predictors of later behaviour among this population [44]. Lastly, socially desirable reporting could have influenced reports of stigmatized behaviour, such as public drug use leading to an underestimation of public crack smoking. If social desirability was an issue in our study we suspect our finding would be a conservative indication of the prevalence of and harms associated with public drug use among crack cocaine smokers.\nIn summary, our study found that locally public crack smoking is a common practice that is also associated with recent encounters with police. We found that the majority of public crack smokers were willing to use an inhalation facility if one were available. Furthermore, public crack smokers who had recent encounters with police were even more likely to be willing to use an inhalation room, suggesting that supervised inhalation facilities may offer unique opportunities to decrease one component of drug-related street disorder and reduce the burden on local law enforcement agencies.", "JM has received grants from, served as an ad hoc advisor to, or spoke at various events sponsored by; Abbott, Argos Therapeutics, Bioject Inc, Boehringer Ingelheim, BMS, Gilead Sciences, GlaxoSmithKline, Hoffmann-La Roche, Janssen-Ortho, Merck Frosst, Pfizer, Schering, Serono Inc, TheraTechnologies, Tibotec, Trimeris.\nAuthors declare no other competing interests.", "The specific contributions of each author are as follows: KD, TK, and EW were responsible for study design; JQ conducted the statistical analyses; KD prepared the first draft of the analysis; TK, JB, JM and EW contributed to the main content and provided critical comments on the final draft. All authors approved the final manuscript.", "The study was supported by the US National Institutes of Health (R01DA011591) and (R01DA021525) and the Canadian Institutes of Health Research (MOP-79297, RAA-79918). Thomas Kerr is supported by the Michael Smith Foundation for Health Research and the Canadian Institutes of Health Research. Kora DeBeck is supported by a Michael Smith Foundation for Health Research Senior Graduate Trainee Award and a Canadian Institutes of Health Research Doctoral Research Award. Julio Montaner has received an Avant-Garde award (DP1DA026182) from the National Institute of Drug Abuse, US National Institutes of Health." ]

[ null, "methods", null, null, null, null, null ]

[]

[ "Adult", "Anti-HIV Agents", "Antiretroviral Therapy, Highly Active", "Australia", "Female", "HIV Infections", "Humans", "Male", "Middle Aged", "Viral Load" ]

[ "Background", "Statistical methods", "Predicted rates of detectable viral load", "Competing interests", "Authors' contributions" ]

[ "There has been much interest recently in the role that combination antiretroviral treatment (cART) might have in decreasing HIV transmission at a population level. A reduced HIV viral load as a consequence of cART appears to reduce the risk of heterosexual HIV transmission [1-3]. At a community level, lower rates of HIV diagnosis in San Francisco and British Columbia have accompanied lower viral loads in HIV-infected people undergoing viral load tests [4,5], and in Taiwan, rapid expansion of cART was associated with a 50% reduction in new HIV diagnoses [6].\nDespite biological plausibility and the observational results, mathematical modelling studies have had inconsistent conclusions. Some studies have suggested that early HIV diagnosis and widespread cART could reduce HIV transmission at a population level [7,8], while others have suggested that relatively small changes in sexual risk behaviour could overwhelm any benefits of cART [9-11]. A key parameter in these mathematical modelling studies is the effect of cART on HIV viral load levels, with parameter estimates usually derived from cohort studies. Such parameter estimates from cohort studies are, however, often confounded with problems with missing data and patient loss to follow up.\nThe objective of this paper is to estimate the proportions of patients with detectable HIV viral load by calendar year in patients receiving cART in the Australian HIV Observational Database (AHOD), allowing for patient covariates and differential follow-up patterns.", "Repeated measures logistic regression, with generalized estimating equations methodology, was used to account for within and between patient variability. An exchangeable variance structure was assumed, but robust variances calculated, which are robust to incorrect assumed variance structure. Maximum likelihood random effects models were also fitted, and found similar covariates to be significant.\nInitially, all covariates were included in the models. A backward stepwise approach was then used to reduce to a parsimonious set of statistically significant (2p < 0.05) covariates. Covariates were also excluded if there appeared to be collinearity problems (for example, associations appearing the wrong way in multivariate models).", "The statistical models were used to make three sets of predictions for each six-month calendar period, and predictions compared with observed rates of detectable viral load. The probability of detectable viral load was predicted for the following three scenarios:\n1. All patients included in the predictions, including all patients who were lost to follow up, who had missing values, or who died. This estimates the proportions of patients with detectable viral load if they had all survived and remained on cART to the appropriate time point\n2. All patients included in the predictions, but excluding patients who died from the time of death\n3. Limiting predictions to patients who had a viral load test result, so predictions fitted to the analyzed data.\nScenarios 1 and 3 can be thought of as likely upper and lower limits on estimates of the proportions of detectable viral load. Scenario 1, which includes all patients who are lost to follow up, who cease cART or who die, would be an upper limit. Scenario 3, which predicts based only on the analyses data, would be a lower limit as these are patients who remain in follow up and so would generally have a better outcome. Scenario 2 was expected to lie within these two limits.", "MGL has received research grants, consultancy and/or travel grants from: Boehringer Ingelheim; Bristol-Myers Squibb; Gilead; GlaxoSmithKline; Janssen-Cilag; Johnson & Johnson; Merck Sharp & Dohme; Pfizer; Roche; and CSL Ltd. IW has received research grants, consultancy payments, clinical support funds or honoraria from: Bristol-Myers Squibb: Gilead; Merck; and Abbott. All other authors have no competing interests to declare.", "All authors contributed to the development of the hypothesis and analysis plan. MGL and HM performed the statistical analysis. MGL wrote the first draft of the manuscript. All authors commented on drafts and approved the final version of the manuscript." ]

[ null, null, null, null, null ]

[ "Background", "Methods", "Statistical methods", "Predicted rates of detectable viral load", "Results", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "There has been much interest recently in the role that combination antiretroviral treatment (cART) might have in decreasing HIV transmission at a population level. A reduced HIV viral load as a consequence of cART appears to reduce the risk of heterosexual HIV transmission [1-3]. At a community level, lower rates of HIV diagnosis in San Francisco and British Columbia have accompanied lower viral loads in HIV-infected people undergoing viral load tests [4,5], and in Taiwan, rapid expansion of cART was associated with a 50% reduction in new HIV diagnoses [6].\nDespite biological plausibility and the observational results, mathematical modelling studies have had inconsistent conclusions. Some studies have suggested that early HIV diagnosis and widespread cART could reduce HIV transmission at a population level [7,8], while others have suggested that relatively small changes in sexual risk behaviour could overwhelm any benefits of cART [9-11]. A key parameter in these mathematical modelling studies is the effect of cART on HIV viral load levels, with parameter estimates usually derived from cohort studies. Such parameter estimates from cohort studies are, however, often confounded with problems with missing data and patient loss to follow up.\nThe objective of this paper is to estimate the proportions of patients with detectable HIV viral load by calendar year in patients receiving cART in the Australian HIV Observational Database (AHOD), allowing for patient covariates and differential follow-up patterns.", "Analyses were based on patients recruited to AHOD. Detailed methods have been described previously [12], but briefly, AHOD is an observational cohort study of HIV-infected patients seen at 27 clinical sites around Australia. Data are transferred electronically to the National Centre in HIV Epidemiology and Clinical Research at the University of New South Wales, Sydney, every six months for aggregation, quality control and analysis. Core data variables include: sex; date of birth; date of most recent visit; HIV exposure; hepatitis B virus (HBV) surface antigen status; hepatitis C virus (HCV) antibody status; CD4 and CD8 counts; HIV viral load; antiretroviral treatment data; AIDS-defining illnesses; and date and cause of death.\nEthics approval was obtained from the University of New South Wales Human Research Ethics Committee and all other relevant institutional review boards, and written informed consent was obtained from all patients.\nPatients were included in this analysis if they had started cART (defined as three or more antiretrovirals), and had at least one viral load assessment after 1 January 1997. Using an intention-to-treat approach, patients were considered to remain on cART if they reverted to mono or double therapy. No account was taken of changes to the antiretrovirals received. Complete treatment interruptions of more than 14 days were excluded from analyses. Any viral load tests prior to cART were also excluded. A second sensitivity analysis was limited to patient prospective follow up.\nThe endpoint analyzed was detectable viral load (defined as >400 copies/ml) in the first and second six months of each calendar year while receiving cART. Detectable viral load was defined as >400 copies/ml as follow up included periods when more sensitive viral load assays were not available. If a patient had multiple viral loads in a six-month period, then the viral load closest to the middle of the period was selected.\nThe following covariates were considered: age at baseline (<30, 30-39, 40-49, 50+ years); sex; HIV exposure (men who have sex with men, MSM + injecting drug user, IDU, heterosexual, other/unknown); AIDS prior to first cART; mono or duo antiretroviral treatment prior to first cART; HCV antibody (no/not tested, ever positive); HBV surface antigen (no/not tested, ever positive); viral load prior to first cART (0 to 365 days prior - <400, >400 copies/ml, missing); CD4 count prior to first cART (0 to 365 days prior - <100, 100-199, 200-349, 350-499, 500+ cells/mm3, missing); viral load in previous six-month period, including viral loads while not receiving antiretrovirals - if a viral load was missing, then the previous viral load was carried forward (<400, 400-10,000, 10,000+, missing); current CD4 count, including CD4 counts while not receiving ARVs - if a CD4 was missing, then the previous CD4 was carried forward (<100, 100-199, 200-349, 350-499, 500+); year first received cART (1993-96, 1997-99, 2000-2002, 2003+; this categorization was based on a preliminary analysis that looked at each year separately, with years of similar risk grouped together); year first HIV diagnosis (< = 1989, 1990-94, 1995-99, 2000+, not known); and time since first cART (0-9 months, 9-18 months, 18+ months).\nThe time since first cART covariate was not modelled in more detail beyond the early period because this would fit to patients who survive and had extended follow up. This could introduce a serious bias into the predicted rates of detectable viral load.\n[SUBTITLE] Statistical methods [SUBSECTION] Repeated measures logistic regression, with generalized estimating equations methodology, was used to account for within and between patient variability. An exchangeable variance structure was assumed, but robust variances calculated, which are robust to incorrect assumed variance structure. Maximum likelihood random effects models were also fitted, and found similar covariates to be significant.\nInitially, all covariates were included in the models. A backward stepwise approach was then used to reduce to a parsimonious set of statistically significant (2p < 0.05) covariates. Covariates were also excluded if there appeared to be collinearity problems (for example, associations appearing the wrong way in multivariate models).\nRepeated measures logistic regression, with generalized estimating equations methodology, was used to account for within and between patient variability. An exchangeable variance structure was assumed, but robust variances calculated, which are robust to incorrect assumed variance structure. Maximum likelihood random effects models were also fitted, and found similar covariates to be significant.\nInitially, all covariates were included in the models. A backward stepwise approach was then used to reduce to a parsimonious set of statistically significant (2p < 0.05) covariates. Covariates were also excluded if there appeared to be collinearity problems (for example, associations appearing the wrong way in multivariate models).\n[SUBTITLE] Predicted rates of detectable viral load [SUBSECTION] The statistical models were used to make three sets of predictions for each six-month calendar period, and predictions compared with observed rates of detectable viral load. The probability of detectable viral load was predicted for the following three scenarios:\n1. All patients included in the predictions, including all patients who were lost to follow up, who had missing values, or who died. This estimates the proportions of patients with detectable viral load if they had all survived and remained on cART to the appropriate time point\n2. All patients included in the predictions, but excluding patients who died from the time of death\n3. Limiting predictions to patients who had a viral load test result, so predictions fitted to the analyzed data.\nScenarios 1 and 3 can be thought of as likely upper and lower limits on estimates of the proportions of detectable viral load. Scenario 1, which includes all patients who are lost to follow up, who cease cART or who die, would be an upper limit. Scenario 3, which predicts based only on the analyses data, would be a lower limit as these are patients who remain in follow up and so would generally have a better outcome. Scenario 2 was expected to lie within these two limits.\nThe statistical models were used to make three sets of predictions for each six-month calendar period, and predictions compared with observed rates of detectable viral load. The probability of detectable viral load was predicted for the following three scenarios:\n1. All patients included in the predictions, including all patients who were lost to follow up, who had missing values, or who died. This estimates the proportions of patients with detectable viral load if they had all survived and remained on cART to the appropriate time point\n2. All patients included in the predictions, but excluding patients who died from the time of death\n3. Limiting predictions to patients who had a viral load test result, so predictions fitted to the analyzed data.\nScenarios 1 and 3 can be thought of as likely upper and lower limits on estimates of the proportions of detectable viral load. Scenario 1, which includes all patients who are lost to follow up, who cease cART or who die, would be an upper limit. Scenario 3, which predicts based only on the analyses data, would be a lower limit as these are patients who remain in follow up and so would generally have a better outcome. Scenario 2 was expected to lie within these two limits.", "Repeated measures logistic regression, with generalized estimating equations methodology, was used to account for within and between patient variability. An exchangeable variance structure was assumed, but robust variances calculated, which are robust to incorrect assumed variance structure. Maximum likelihood random effects models were also fitted, and found similar covariates to be significant.\nInitially, all covariates were included in the models. A backward stepwise approach was then used to reduce to a parsimonious set of statistically significant (2p < 0.05) covariates. Covariates were also excluded if there appeared to be collinearity problems (for example, associations appearing the wrong way in multivariate models).", "The statistical models were used to make three sets of predictions for each six-month calendar period, and predictions compared with observed rates of detectable viral load. The probability of detectable viral load was predicted for the following three scenarios:\n1. All patients included in the predictions, including all patients who were lost to follow up, who had missing values, or who died. This estimates the proportions of patients with detectable viral load if they had all survived and remained on cART to the appropriate time point\n2. All patients included in the predictions, but excluding patients who died from the time of death\n3. Limiting predictions to patients who had a viral load test result, so predictions fitted to the analyzed data.\nScenarios 1 and 3 can be thought of as likely upper and lower limits on estimates of the proportions of detectable viral load. Scenario 1, which includes all patients who are lost to follow up, who cease cART or who die, would be an upper limit. Scenario 3, which predicts based only on the analyses data, would be a lower limit as these are patients who remain in follow up and so would generally have a better outcome. Scenario 2 was expected to lie within these two limits.", "A total of 2439 patients were eligible for inclusion in the analysis. The median number of viral loads analyzed for each patient was 13 (interquartile range 7 to 19). A total of 654 patients (4.7 per 100 person years) were lost to follow up (defined as more than 12 months without a clinic visit) and 194 patients died (1.4 per 100 person years). Patient characteristics at first cART are summarized by year of first cART in Table 1. Patients who first received cART in the 1990s were more likely to have been diagnosed earlier with HIV, were slightly younger, and were slightly more likely to have been infected with HIV through male-to-male sex. Patients who first received cART in 1993-96 were much more likely to have previously received mono or duo ART than those who initiated cART in later time periods, and also initiated cART at lower CD4 counts and with more prior AIDS illnesses. Patients who initiated cART in 2000 or later were more likely to report heterosexual contact as their route of HIV infection. HCV and HBV coinfection appeared less common in patients who first received cART in 2003 or later.\nPatient characteristics at first cART by year of first cART\nThe final fitted multivariate model is summarized in Table 2. Factors associated with a greater risk of detectable viral load were found to be younger age, prior mono or duo ART, a detectable previous viral load, a lower current CD4 count, and the 18-month period immediately after starting cART. First cART in more recent calendar times, and more recent reported HIV diagnosis, were found to be associated with a decreased risk of detectable viral load.\nPredictors of detectable viral load (>400 copies/ml) - all patients 1997-2009\nCovariates omitted from the model:CD4 at first cART, sex, viral load at first cART, prior AIDS, HBV, HCV, HIV exposure\nObserved proportions of detectable viral load in patients receiving cART, by six-month calendar year periods, together with model-fitted predicted proportions, are shown for all patients combined in Figure 1. This shows a strong continuing decrease in the observed proportion of patients receiving cART with a detectable viral load, from more than 50% in 1997 and 1998 to around 7.7% in 2007, 6.3% in 2008, and 5.3% in the first half of 2009. However, the model-predicted proportions of detectable viral load are much higher. Under scenario 1, predicting for all patients including those who were lost to follow up or died, the predicted proportion in 2009 was 16.0%. The predicted proportions for scenarios 2 and 3 were 13.8% and 10.1%, respectively.\nAHOD detectable viral load 1997-2009\nObserved and predicted proportions of detectable viral load by period of first cART are shown in Figure 2. Across all periods of first cART, there is the same strong decreasing proportion of detectable viral load down to around 5-6% in 2009. Perhaps not surprisingly, the predicted rates are much higher for patients who first received cART in earlier periods. The predicted proportions of detectable viral load under scenario 2 in 2009 were 19.4%, 14.9%, 9.8% and 5.7% for the four periods, respectively.\nAHOD detectable viral load 1997-2009 by year of first cART\nSensitivity analyses were also performed based on patient prospective data only. These analyses found the same covariates to be included in multivariate models, and gave similar trends in observed and predicted proportions of detectable viral loads (data not presented).", "The proportion of patients in AHOD with detectable viral load while receiving cART has been observed to be decreasing, to around 6% in 2009. These analyses, which adjust for patient covariates and differential follow up, suggest that the true proportions of patients in AHOD receiving cART with detectable viral load in more recent calendar time periods are higher than the simple observed proportions. The higher estimated proportion of patients with detectable viral load in adjusted analyses is mostly due to the inclusion of patients who were lost to follow up, and observed proportions should be used with caution because of this bias.\nUnder scenario 2 (which includes in predictions patients with unmeasured viral load or who have become lost to follow up), but censors patients who have died, the predicted proportion of patients with detectable viral load in 2009 was 13.8% compared with an observed proportion of 6.3%. Although predicted proportions of detectable viral load were higher than observed proportions, a consistent finding of our analyses was that there was no evidence of increasing proportions of patients with detectable viral load, both overall and by time of first cART. This is reassuring as it suggests that there is as yet no evidence of cohorts of HIV-infected patients running out of effective treatment options.\nOur analyses specifically looked at detectable viral load by calendar time. We performed this analysis, as opposed to looking at detectable viral load from time of first cART, because of the recent interest in levels of community viral load in HIV-infected patients receiving viral load tests by calendar time, and how this might impact on HIV transmission at a population level [1-6]. In Australia, as many other countries, population-level data on rates of detectable viral load in patients receiving cART are unavailable. AHOD, a large observational cohort study that includes 15-20% of all patients in Australia receiving cART [13], is the best available source of data on this issue on which to base assumptions for mathematical models [9-11,14]. As such, analyses of this type, assessing the effect of differential follow up on observed viral load levels in AHOD, are important for developing the most accurate assumptions possible.\nCombination ART is publicly funded and freely available to all HIV-infected patients in Australia. The HIV epidemic remains very largely (85%) transmitted through male homosexual sex [15], a well-educated and informed population. In uninfected homosexual men, HIV testing was reported to take place at least annually in around 60% of men in 2006, and this proportion increased between 1998 and 2006 [16]. The absolute number of HIV-infected people in Australia receiving cART has been estimated to have increased between 2000 and 2006, though the proportion of all HIV-infected people receiving cART was estimated to have increased only slightly or remained flat [17].\nFinally, the analyses presented here suggest that in HIV-infected men receiving cART, HIV viral load has continued to decrease through the 2000s, albeit at slower rates than observed data suggest. This set of circumstances in Australia would appear to offer the best hope for cART to have an effect on reducing HIV transmission at a population level. However, over this period, total HIV diagnoses have increased in Australia, from a low of 718 new diagnoses in 1999 to around 1000 new diagnoses annually in 2006-2008 [15].\nMathematical models and back-projections analyses have both suggested that this reflects a real increase in HIV incidence in homosexual men [11,18]. If the decreasing trends in detectable viral load in AHOD patients receiving cART are representative of all HIV-infected patients receiving cART in Australia, then this suggests that in Australia, the likely reduction in HIV transmission risk in patients receiving cART through reduced HIV viral load is being counterbalanced by increasing infection risk due to behavioural changes. This underscores the importance of continued vigilance with existing HIV prevention strategies, including symptom awareness, early risk assessment, diagnosis and referral for care and treatment.\nMathematical modelling has been used to investigate trends in HIV incidence in Australia. Early models did suggest a decrease in HIV incidence among homosexual men during 1996 to 1998 due to the introduction of widespread cART, but that this was followed in 1998 to 2001 by a slow increase in incidence due to increasing rates of unprotected anal intercourse with casual partners while use of cART remained fairly stable [10]. More recent modelling suggested that the observed increase in HIV incidence in homosexual men in some Australian states might be explained by increasing rates of other sexually transmissible infections [11]. These models also estimated that 19% of incident HIV infections were transmitted from the estimated 3% of HIV-infected homosexual men in primary HIV infection, and that 31% of incident HIV infections were transmitted from the estimated 9% of HIV-infected homosexual men with undiagnosed infection [14].\nA key limitation of our analyses is the extent to which trends in AHOD are representative of all HIV-infected people in Australia. AHOD is an observational cohort study of HIV-infected people attending clinics for their care, and recruited more patients in the late 1990s and early 2000s than in recent years. Hence trends in undetectable viral load may not reflect all HIV-infected patients receiving cART. We did stratify trends by different periods of first cART to try to assess this. AHOD represents 15-20% of HIV-infected patients receiving cART, and in terms of key epidemiological characteristics, seems reasonably representative of the wider HIV epidemic in Australia [13]. However, the estimates of trends in detectable viral load on cART in AHOD presented here are different to the true estimates of community viral load that are available in other studies [4,5], but unavailable in Australia. In particular, our analyses take no account of trends in viral load in HIV-infected people who are not receiving cART. Generalization of our results to inferences about levels of community viral load in Australia should be made with caution.\nA further limitation is that AHOD, as with all observational cohorts, has missing data and some patients were lost to follow up. While we predicted trends in detectable viral load adjusted for important covariates using statistical models that allow for patients lost to follow up, there may be unmeasured and unmeasurable confounders that would affect our results. In particular, it may be that the apparent continuing decline in detectable viral load in patients receiving cART, albeit at higher levels than observed declines, is better interpreted as a plateau over the period from the mid-2000s.", "Our analyses suggest that in AHOD, true calendar trends in detectable viral in HIV-infected patients receiving cART are higher than observed trends when adjusted for confounding covariates and patients lost to follow up. Whether these predictions reflect true continuing decreases, or actually something more of a plateau, we feel is open to interpretation. It is reassuring that under all models, there was no suggestion of increasing detectable viral load, either observed or predicted. The fact that these decreasing trends in detectable viral load in patients receiving cART in AHOD have been accompanied by increases in HIV diagnoses and estimated HIV incidence suggests that, at least in Australia, the likely decrease in the risk of transmission from people receiving cART as a result of reduced HIV viral load is being counterbalanced by increasing risk of transmission due to behaviour changes.", "MGL has received research grants, consultancy and/or travel grants from: Boehringer Ingelheim; Bristol-Myers Squibb; Gilead; GlaxoSmithKline; Janssen-Cilag; Johnson & Johnson; Merck Sharp & Dohme; Pfizer; Roche; and CSL Ltd. IW has received research grants, consultancy payments, clinical support funds or honoraria from: Bristol-Myers Squibb: Gilead; Merck; and Abbott. All other authors have no competing interests to declare.", "All authors contributed to the development of the hypothesis and analysis plan. MGL and HM performed the statistical analysis. MGL wrote the first draft of the manuscript. All authors commented on drafts and approved the final version of the manuscript." ]

[ null, "methods", null, null, "results", "discussion", "conclusions", null, null ]

[]

[ "Child, Preschool", "China", "Cluster Analysis", "DNA, Viral", "Diarrhea", "Evolution, Molecular", "Human bocavirus", "Humans", "Infant", "Molecular Sequence Data", "Parvoviridae Infections", "Phylogeny", "Polymerase Chain Reaction", "Recombination, Genetic", "Sequence Analysis, DNA", "Viral Proteins" ]

[ "Background", "Ethics statement", "Patients and methods", "Detection of HBoV, HBoV2, HBoV3, and HBoV4", "Sequence and phylogenetic analysis", "Recombination Analyses", "Results", "Detection of human bocaviruses", "Phylogenetic analyses of HBoV2", "Recombination analyses among HBoV, HBoV2, HBoV3, and HBoV4", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Human bocavirus (HBoV), HBoV2, HBoV3, and HBoV4 have been discovered recently [1-4]. These viruses belong to the genus Bocavirus in the subfamily Parvovirinae of the family Parvoridae, among which the human parvovirus B19 (B19V) is the only known human pathogen [5]. HBoV was detected in respiratory tract samples in 2005[1]. In 2009, Kapoor et al. [2] reported a new bocavirus species, HBoV2, isolated from stool samples in children with nonpolio acute flaccid paralysis, and suggested that recombination between HBoV2 strains may occurs. Almost simultaneously, Arthur et al. [3] reported that HBoV3 was detected in stool samples from children with acute gastroenteritis (AGE). In addition, they proposed that HBoV3 is a hybrid of HBoV and HBoV2. Subsequently these two new viruses were detected in nasopharyngeal aspirates or stool samples in other regions [6-11]. Recently Kapoor et al reported the discovery of HBoV4 and the detection of recombination signals between and within bocavirus species [4]. Among these human bocaviruses, HBoV2 had higher prevalence and genetic diversity in stool samples than the others [3,8], but the precise phylogenetic relationships among these viruses were not clear yet. Furthermore, the prevalence and genetic feature of HBoV2 were not addressed at children with and without AGE.\nIn the present study we collected stool samples from children with and without AGE in Lanzhou, China. Samples were assayed for the presence of HBoV, HBoV2, HBoV3 and HBoV4. Partial nucleotide sequences of the NS, NP1, VP1/2 genes, and two nearly full-length genome sequences of HBoV2 were obtained.", "The study was approved for human subject protection by the Research Ethics Committee of the Lanzhou University and the Institutional Review Board at China CDC. Following informed consent was written by parent/guardian.", "From July 2006 to June 2008 in Lanzhou, China, we collected stool samples from 632 hospitalized children with diarrhea and 162 asymptomatic children. 3-5 ml stool was collected from every participant. All subjects were 5 or less years of age. Medical histories were provided by parents/guardians. The case group included subjects hospitalized for gastroenteritis in the Department of Pediatrics in our institution. Diarrhea was defined as three or more loose stools in the previous 24 h. Patients were excluded from the study if stool sample volume was insufficient for a complete evaluation of viral agents, stools had blood streaks or pus, or due to the presence of a co-morbidity. Subjects in the control group had presented to the First Hospital of Lanzhou University Pediatric Primary Care Center for a routine examination and did not have fever, diarrhea, vomiting, or respiratory illness in the previous 3-week period [12]. Control subjects received follow-up by telephone and those in whom any of the aforementioned exclusion criteria were present during the week after the initial examination were excluded. A 10% suspension of stool sample was made by mixing 0.5 g stool with 1.0 mL PBS (pH7.2). Viral RNA and DNA were extracted from stool suspensions clarified by centrifugation (1500 × g, 20 min) using a QIAamp® Viral RNA Mini kit (Qiagen, Hilden, Germany), according to the manufacturer's protocol. RNA and DNA were resuspended in 50 μL water and stored at -70°C.", "HBoV was detected using a method described by us previously [13]. HBoV2, HBoV3, and HBoV4 were detected by nested PCR as described by Kapoor et al., in which primers HBoV2-sf1, HBoV2-sr1, HBoV2-sf2, and HBoV2-sr2 were used to amplify a 495-nt region within the ORF of NS1 [2]. PCR-positive samples were confirmed by sequencing.", "To analyze genetic variation in HBoV2, two sets of nested primers were designed and used to amplify part of the NP1 and VP1 genes in positive samples (Table 1), using the NP1-F1 and NP1-R1, and VP1-F1 and VP1-R1 primers in the first round of PCR and the NP1-F2 and NP1-R2, and VP1-F2 and VP1-R2 in the second. The reaction mix contained 10 pmol each primer and 2.5 units ExTaq DNA polymerase (Takara Bio). After 5 min at 94°C, 35 cycles of amplification (94°C for 45 s, 50°C for 1 min, and 72°C for 1 min) were performed, followed by a 7 min extension at 72°C. Complete HBoV2 sequences were amplified by using specific PCR and Genome Walking Kit (TaKaRa code: D316). PCR products were cloned and the plasmid inserts were sequenced. Nucleotide and deduced amino acid sequences were compared to entries in the GenBank database. Phylogenetic analysis was conducted with Molecular Evolutionary Genetics Analysis (MEGA) version 4.1. All NS1, NP1, VP1 partial gene sequences and nearly full-length genome sequences of HBoV2 except the termini were submitted to GenBank (accession numbers GU301644-GU301683, HQ153797-HQ153804).\nPrimers used in this study.\n* The target fragment of NP1 partial gene was 586 nt, in 2285-2870 nt according to HBoV2 FJ170279 strain.\n# The target fragment of VP1 partial gene was 620 nt, in 3134-3753 nt according to HBoV2 FJ170279 strain.\n‡The target fragment of NP1 Real-time PCR 145 nt, in 2489-2633 nt according to HBoV2 FJ170279 strain.\nA total of 82 nearly full-length genome sequences of HBoV, HBoV2, HBoV3, and HBoV4 were obtained from GenBank. These were aligned and manually adjusted using ClustalW and BioEdit. Phylogenetic trees were determined by the neighbor-joining (NJ) method using the MEGA 4.1 software package. Various nucleotide substitution models were examined and yielded phylogenetic trees of similar topology; only the Kimura 2-parameter model trees were described in this report. A bootstrap resampling (1000 replications) was used to assess the reliability of individual nodes in each phylogenetic tree.", "We detected recombination using the RDP3 package. Sequences were selected based on their similarity and then aligned and manually adjusted using ClustalW and BioEdit. The alignments were scanned by various algorithms implemented in the RDP3 package, followed by manual refinement. In addition, we performed similarity plot and bootscanning analyses for potential recombination events using SimPlot 3.5. Genetic Algorithm Recombination Detection (GARD) methods were also used to detect recombination and estimate breakpoint locations. Estimated breakpoints were verified by both the Shimodaira-Hasegawa test and manually checking phylogenetic trees for nonrecombinant segments.", "[SUBTITLE] Detection of human bocaviruses [SUBSECTION] In subjects with gastroenteritis, the positive rates of HBoV, HBoV2 and HBoV3 were 4.3%, 20.4% and 0.9%, respectively, whilst HBoV4 was not detected. In the control group, HBoV and HBoV2 were detected in 2.5% and 12.3% of subjects; HBoV3 and HBoV4 were not detected.\nIn subjects with gastroenteritis, the positive rates of HBoV, HBoV2 and HBoV3 were 4.3%, 20.4% and 0.9%, respectively, whilst HBoV4 was not detected. In the control group, HBoV and HBoV2 were detected in 2.5% and 12.3% of subjects; HBoV3 and HBoV4 were not detected.\n[SUBTITLE] Phylogenetic analyses of HBoV2 [SUBSECTION] Phylogenetic analysis indicated that all of the partial NS1 gene sequences of case group were closer with the PK-2255 strain (from Pakistani children, GenBank: FJ170279) than other strains, with 97.4-99% identity. Consistently, those of control group had 98-99% identity with the PK-2255 strain and were also closer with it than others. The sequences in this study, including case and control group, had a high identity of 98.5-100% with each other. The Phylogenetic tree also indicated that there was no difference in the topological characteristics of HBoV2 between case and control groups (Figure 1).\nPhylogenetic analysis of human bocavirus1-4 and other bocavirus members partial NS1 gene sequences. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, black triangles designate sequences from control group and the others were sequences generated from the gastroenteritis children in the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nFurther study indicated that partial NP1 gene sequences were also similar to those of strain PK-2255 (98-99%), except that four had a high identity to strain FJ973558 (Figure 2), with >99% sequence identity. Interestingly, the ten HBoV2 partial VP1 gene sequences in this study were more variable than those of NS1 and NP1, being only 92.6-97.3% similar to those of strain PK-2255. Six partial VP1 gene sequences (Figure 3) were in the same cluster as strain PK-2255 (97.3-98.1% identity) and the other four sequences clustered with strain FJ973558 (96.9-97.4% identity). The nearly full-length genome sequence of LZ53819 generated in this study (GenBank number: GU301644) was more similar to strain FJ973558 than to PK-2255 (similarity 97.5%), but the NS1 gene sequence showed greater similarity to that of strain PK-2255. Above results suggested that recombination among HBoV2 genotypes may occur. While we compared the trees of non-recombinant segments associated with the break points estimated by GARD, Shimodaira-Hasegawa test showed the topologies of these trees were not significantly different.\nPhylogenetic analysis of the partial NP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nPhylogenetic analysis of the partial VP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nPhylogenetic analysis indicated that all of the partial NS1 gene sequences of case group were closer with the PK-2255 strain (from Pakistani children, GenBank: FJ170279) than other strains, with 97.4-99% identity. Consistently, those of control group had 98-99% identity with the PK-2255 strain and were also closer with it than others. The sequences in this study, including case and control group, had a high identity of 98.5-100% with each other. The Phylogenetic tree also indicated that there was no difference in the topological characteristics of HBoV2 between case and control groups (Figure 1).\nPhylogenetic analysis of human bocavirus1-4 and other bocavirus members partial NS1 gene sequences. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, black triangles designate sequences from control group and the others were sequences generated from the gastroenteritis children in the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nFurther study indicated that partial NP1 gene sequences were also similar to those of strain PK-2255 (98-99%), except that four had a high identity to strain FJ973558 (Figure 2), with >99% sequence identity. Interestingly, the ten HBoV2 partial VP1 gene sequences in this study were more variable than those of NS1 and NP1, being only 92.6-97.3% similar to those of strain PK-2255. Six partial VP1 gene sequences (Figure 3) were in the same cluster as strain PK-2255 (97.3-98.1% identity) and the other four sequences clustered with strain FJ973558 (96.9-97.4% identity). The nearly full-length genome sequence of LZ53819 generated in this study (GenBank number: GU301644) was more similar to strain FJ973558 than to PK-2255 (similarity 97.5%), but the NS1 gene sequence showed greater similarity to that of strain PK-2255. Above results suggested that recombination among HBoV2 genotypes may occur. While we compared the trees of non-recombinant segments associated with the break points estimated by GARD, Shimodaira-Hasegawa test showed the topologies of these trees were not significantly different.\nPhylogenetic analysis of the partial NP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nPhylogenetic analysis of the partial VP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\n[SUBTITLE] Recombination analyses among HBoV, HBoV2, HBoV3, and HBoV4 [SUBSECTION] The phylogenetic relationships of a total of 82 nearly full-length genome sequences obtained both from this study and GenBank were inferred. Based on sequence identity, we selected alignments of five nearly full-length genome sequences. Recombinant signals were detected in all of the HBoV, HBoV2, HBoV3 and HBoV4 genome sequences, but the signals in HBoV3 was far more significant than those in other bocaviruses. Using RDP3 and SimPlot3.5, it was found that HBoV3 (NC_012564) was a potential recombinant of HBoV (FJ858259) and HBoV4 (NC_012729) (Figure 4A, B). The breakpoint was located near the VP1 start codon. GARD analyses suggested six possible breakpoints with model average support over 0.90, including one located 18 bp downstream of the VP1 start codon (Figure 4C). This breakpoint was further examined by constructing phylogenetic trees of the two nonrecombinant segments (Figure 4C). While NC_012564 grouped with FJ858259 in the 5' segment (including NS1 and NP1), it grouped with NC_012729 in the 3' segment (consisting of VP1 and VP2), suggesting that HBoV3 is a hybrid of HBoV and HBoV4. However, the VP1 and VP2 region of HBoV3 was as similar to HBoV2 (identity 87.0% with GQ200737 and 88.6% with FJ973560) as it was to HBoV4 (88.3% to NC_012729).\nHBoV3 is a potential hybrid of HBoV and HBoV4 by recombination analysis. (A) recombination analysis was conducted by Similarity plot, (B) recombination of HBoV3 was conducted by bootscanning analysis, (C) recombination of HBoV3 was conducted by GARD analysis.\nThe phylogenetic relationships of a total of 82 nearly full-length genome sequences obtained both from this study and GenBank were inferred. Based on sequence identity, we selected alignments of five nearly full-length genome sequences. Recombinant signals were detected in all of the HBoV, HBoV2, HBoV3 and HBoV4 genome sequences, but the signals in HBoV3 was far more significant than those in other bocaviruses. Using RDP3 and SimPlot3.5, it was found that HBoV3 (NC_012564) was a potential recombinant of HBoV (FJ858259) and HBoV4 (NC_012729) (Figure 4A, B). The breakpoint was located near the VP1 start codon. GARD analyses suggested six possible breakpoints with model average support over 0.90, including one located 18 bp downstream of the VP1 start codon (Figure 4C). This breakpoint was further examined by constructing phylogenetic trees of the two nonrecombinant segments (Figure 4C). While NC_012564 grouped with FJ858259 in the 5' segment (including NS1 and NP1), it grouped with NC_012729 in the 3' segment (consisting of VP1 and VP2), suggesting that HBoV3 is a hybrid of HBoV and HBoV4. However, the VP1 and VP2 region of HBoV3 was as similar to HBoV2 (identity 87.0% with GQ200737 and 88.6% with FJ973560) as it was to HBoV4 (88.3% to NC_012729).\nHBoV3 is a potential hybrid of HBoV and HBoV4 by recombination analysis. (A) recombination analysis was conducted by Similarity plot, (B) recombination of HBoV3 was conducted by bootscanning analysis, (C) recombination of HBoV3 was conducted by GARD analysis.", "In subjects with gastroenteritis, the positive rates of HBoV, HBoV2 and HBoV3 were 4.3%, 20.4% and 0.9%, respectively, whilst HBoV4 was not detected. In the control group, HBoV and HBoV2 were detected in 2.5% and 12.3% of subjects; HBoV3 and HBoV4 were not detected.", "Phylogenetic analysis indicated that all of the partial NS1 gene sequences of case group were closer with the PK-2255 strain (from Pakistani children, GenBank: FJ170279) than other strains, with 97.4-99% identity. Consistently, those of control group had 98-99% identity with the PK-2255 strain and were also closer with it than others. The sequences in this study, including case and control group, had a high identity of 98.5-100% with each other. The Phylogenetic tree also indicated that there was no difference in the topological characteristics of HBoV2 between case and control groups (Figure 1).\nPhylogenetic analysis of human bocavirus1-4 and other bocavirus members partial NS1 gene sequences. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, black triangles designate sequences from control group and the others were sequences generated from the gastroenteritis children in the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nFurther study indicated that partial NP1 gene sequences were also similar to those of strain PK-2255 (98-99%), except that four had a high identity to strain FJ973558 (Figure 2), with >99% sequence identity. Interestingly, the ten HBoV2 partial VP1 gene sequences in this study were more variable than those of NS1 and NP1, being only 92.6-97.3% similar to those of strain PK-2255. Six partial VP1 gene sequences (Figure 3) were in the same cluster as strain PK-2255 (97.3-98.1% identity) and the other four sequences clustered with strain FJ973558 (96.9-97.4% identity). The nearly full-length genome sequence of LZ53819 generated in this study (GenBank number: GU301644) was more similar to strain FJ973558 than to PK-2255 (similarity 97.5%), but the NS1 gene sequence showed greater similarity to that of strain PK-2255. Above results suggested that recombination among HBoV2 genotypes may occur. While we compared the trees of non-recombinant segments associated with the break points estimated by GARD, Shimodaira-Hasegawa test showed the topologies of these trees were not significantly different.\nPhylogenetic analysis of the partial NP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nPhylogenetic analysis of the partial VP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.", "The phylogenetic relationships of a total of 82 nearly full-length genome sequences obtained both from this study and GenBank were inferred. Based on sequence identity, we selected alignments of five nearly full-length genome sequences. Recombinant signals were detected in all of the HBoV, HBoV2, HBoV3 and HBoV4 genome sequences, but the signals in HBoV3 was far more significant than those in other bocaviruses. Using RDP3 and SimPlot3.5, it was found that HBoV3 (NC_012564) was a potential recombinant of HBoV (FJ858259) and HBoV4 (NC_012729) (Figure 4A, B). The breakpoint was located near the VP1 start codon. GARD analyses suggested six possible breakpoints with model average support over 0.90, including one located 18 bp downstream of the VP1 start codon (Figure 4C). This breakpoint was further examined by constructing phylogenetic trees of the two nonrecombinant segments (Figure 4C). While NC_012564 grouped with FJ858259 in the 5' segment (including NS1 and NP1), it grouped with NC_012729 in the 3' segment (consisting of VP1 and VP2), suggesting that HBoV3 is a hybrid of HBoV and HBoV4. However, the VP1 and VP2 region of HBoV3 was as similar to HBoV2 (identity 87.0% with GQ200737 and 88.6% with FJ973560) as it was to HBoV4 (88.3% to NC_012729).\nHBoV3 is a potential hybrid of HBoV and HBoV4 by recombination analysis. (A) recombination analysis was conducted by Similarity plot, (B) recombination of HBoV3 was conducted by bootscanning analysis, (C) recombination of HBoV3 was conducted by GARD analysis.", "Although the primers used in this study were capable of detecting HBoV4, no HBoV4 was detected. HBoV2 had higher prevalence than HBoV and HBoV3 in both case and control groups, indicating that HBoV2 should be given more attention than other human bocaviruses. Given the fact that children in the control group that were free of symptoms were positive for HBoV and HBoV2, perhaps bocaviruses are only \"passers-by\" in intestinal tract than pathogens of gastroenteritis, or the control participants had symptomless infection of bocaviruses. This issue is not clear yet and needs more studies to resolve [7,8,13,14]. All of the partial NS1 gene sequences in this study were closer with the PK-2255 strain (from Pakistani children) than other strains. There was a high sequence identity of 98.5-100% between HBoV2 from case and control groups and the phylogenetic tree also confirmed that there was no difference in the phylogenetic characteristics of them, suggesting that a single genetic lineage of HBoV2 is circulating in both the gastroenteritis and healthy children in Lanzhou, China. And it also indicated HBoV2 can cause asymptomatic infection in Children.\nPhylogenetic analyses demonstrated that partial NS1, NP1, and VP1 gene sequences of HBoV2 were markedly similar to those of a number of reference strains. This variability was also reported by Kapoor [2], who postulated that it was due to recombination between HBoV2 strains, as has been reported for animal parvoviruses [15]. However, when phylogenetic trees of non-recombinant segments associated with breakpoints estimated by GARD were compared, Shimodaira-Hasegawa test showed the topologies were not significantly different. Considering that these sequences exhibit very high identity (more than 95%), this recombination phenomenon may be mediated by other processes, for example, variation in spatial rate and/or heterotachy. More works are required to elucidate fully the nature and extent of any recombination that occurs between HBoV2 strains.\nArthur et al. identified two recombination sites upstream of the NS1 and VP1/2 genes, using HBoV, HBoV2, and HBoV3 sequence analyses, and hypothesized that HBoV3 may be a hybrid of HBoV and HBoV2 [3]. Our phylogenetic analyses of HBoV, HBoV2, HBoV3 and HBoV4 strains confirmed that HBoV3 may be a hybrid of HBoV and HBoV4. The estimated breakpoint was located at upstream of the VP1 gene. Although bootscanning analyses and phylogenetic trees suggested that HBoV3 was a hybrid of HBoV and HBoV4, the parent strain of the VP1/VP2 region was still hard to determine. The VP1/VP2 region of HBoV3 was as similar to that of HBoV2 as it was to that of HBoV4, suggesting that HBoV3 may be a hybrid of HBoV and the common ancestor of HBoV2 and HBoV4.\nAccording to the ICTVb criteria http://www.ictvdb.org/Ictv/fs_parvo.htm, isolates with non-structural gene homologies of less than 95% are defined as a new species in the bocavirus genus. NS1 sequence variation amongst HBoV2 clusters was as high as 8%. However, NS1 sequence variation between one HBoV2 cluster and HBoV4 was only 6%, suggesting that either more than one HBoV2 cluster exists or that some of these clusters should in fact be regarded as a separate species. The high sequence identity, limited sequence data, and unclear taxonomy rendered the precise phylogenetic relationships among these human bocaviruses difficult to determine.\nThe data presented here suggested the potential phylogenetic relationships among the known human bocaviruses. However, many issues remain, including the nature and extent of recombination between HBoV2 strains, and the precise evolutionary relationships of the various human bocaviruses. More studies including more samples from different areas and years are needed to address them.", "In summary, our data suggested that HBoV2 had higher prevalence than HBoV and HBoV3 in both case and control groups. A single genetic lineage of HBoV2 is circulating in children with and without gastroenteritis in Lanzhou, China. Recombination between HBoV2 strains may occur and HBoV3 may be a hybrid virus, originating from HBoV and the common ancestor of HBoV2 and HBoV4.", "The authors declare that they have no competing interests.", "Conceived and designed the experiments: ZD, YJ. Performed the experiments: WC, ZX, JY. Analyzed the data: JC, WC, CH, MZ, ZD. Contributed reagents/materials/analysis tools: MJ, HL. Wrote the paper: WC, JC, ZD. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2334/11/50/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Ethics statement", "Patients and methods", "Detection of HBoV, HBoV2, HBoV3, and HBoV4", "Sequence and phylogenetic analysis", "Recombination Analyses", "Results", "Detection of human bocaviruses", "Phylogenetic analyses of HBoV2", "Recombination analyses among HBoV, HBoV2, HBoV3, and HBoV4", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Human bocavirus (HBoV), HBoV2, HBoV3, and HBoV4 have been discovered recently [1-4]. These viruses belong to the genus Bocavirus in the subfamily Parvovirinae of the family Parvoridae, among which the human parvovirus B19 (B19V) is the only known human pathogen [5]. HBoV was detected in respiratory tract samples in 2005[1]. In 2009, Kapoor et al. [2] reported a new bocavirus species, HBoV2, isolated from stool samples in children with nonpolio acute flaccid paralysis, and suggested that recombination between HBoV2 strains may occurs. Almost simultaneously, Arthur et al. [3] reported that HBoV3 was detected in stool samples from children with acute gastroenteritis (AGE). In addition, they proposed that HBoV3 is a hybrid of HBoV and HBoV2. Subsequently these two new viruses were detected in nasopharyngeal aspirates or stool samples in other regions [6-11]. Recently Kapoor et al reported the discovery of HBoV4 and the detection of recombination signals between and within bocavirus species [4]. Among these human bocaviruses, HBoV2 had higher prevalence and genetic diversity in stool samples than the others [3,8], but the precise phylogenetic relationships among these viruses were not clear yet. Furthermore, the prevalence and genetic feature of HBoV2 were not addressed at children with and without AGE.\nIn the present study we collected stool samples from children with and without AGE in Lanzhou, China. Samples were assayed for the presence of HBoV, HBoV2, HBoV3 and HBoV4. Partial nucleotide sequences of the NS, NP1, VP1/2 genes, and two nearly full-length genome sequences of HBoV2 were obtained.", "[SUBTITLE] Ethics statement [SUBSECTION] The study was approved for human subject protection by the Research Ethics Committee of the Lanzhou University and the Institutional Review Board at China CDC. Following informed consent was written by parent/guardian.\nThe study was approved for human subject protection by the Research Ethics Committee of the Lanzhou University and the Institutional Review Board at China CDC. Following informed consent was written by parent/guardian.\n[SUBTITLE] Patients and methods [SUBSECTION] From July 2006 to June 2008 in Lanzhou, China, we collected stool samples from 632 hospitalized children with diarrhea and 162 asymptomatic children. 3-5 ml stool was collected from every participant. All subjects were 5 or less years of age. Medical histories were provided by parents/guardians. The case group included subjects hospitalized for gastroenteritis in the Department of Pediatrics in our institution. Diarrhea was defined as three or more loose stools in the previous 24 h. Patients were excluded from the study if stool sample volume was insufficient for a complete evaluation of viral agents, stools had blood streaks or pus, or due to the presence of a co-morbidity. Subjects in the control group had presented to the First Hospital of Lanzhou University Pediatric Primary Care Center for a routine examination and did not have fever, diarrhea, vomiting, or respiratory illness in the previous 3-week period [12]. Control subjects received follow-up by telephone and those in whom any of the aforementioned exclusion criteria were present during the week after the initial examination were excluded. A 10% suspension of stool sample was made by mixing 0.5 g stool with 1.0 mL PBS (pH7.2). Viral RNA and DNA were extracted from stool suspensions clarified by centrifugation (1500 × g, 20 min) using a QIAamp® Viral RNA Mini kit (Qiagen, Hilden, Germany), according to the manufacturer's protocol. RNA and DNA were resuspended in 50 μL water and stored at -70°C.\nFrom July 2006 to June 2008 in Lanzhou, China, we collected stool samples from 632 hospitalized children with diarrhea and 162 asymptomatic children. 3-5 ml stool was collected from every participant. All subjects were 5 or less years of age. Medical histories were provided by parents/guardians. The case group included subjects hospitalized for gastroenteritis in the Department of Pediatrics in our institution. Diarrhea was defined as three or more loose stools in the previous 24 h. Patients were excluded from the study if stool sample volume was insufficient for a complete evaluation of viral agents, stools had blood streaks or pus, or due to the presence of a co-morbidity. Subjects in the control group had presented to the First Hospital of Lanzhou University Pediatric Primary Care Center for a routine examination and did not have fever, diarrhea, vomiting, or respiratory illness in the previous 3-week period [12]. Control subjects received follow-up by telephone and those in whom any of the aforementioned exclusion criteria were present during the week after the initial examination were excluded. A 10% suspension of stool sample was made by mixing 0.5 g stool with 1.0 mL PBS (pH7.2). Viral RNA and DNA were extracted from stool suspensions clarified by centrifugation (1500 × g, 20 min) using a QIAamp® Viral RNA Mini kit (Qiagen, Hilden, Germany), according to the manufacturer's protocol. RNA and DNA were resuspended in 50 μL water and stored at -70°C.\n[SUBTITLE] Detection of HBoV, HBoV2, HBoV3, and HBoV4 [SUBSECTION] HBoV was detected using a method described by us previously [13]. HBoV2, HBoV3, and HBoV4 were detected by nested PCR as described by Kapoor et al., in which primers HBoV2-sf1, HBoV2-sr1, HBoV2-sf2, and HBoV2-sr2 were used to amplify a 495-nt region within the ORF of NS1 [2]. PCR-positive samples were confirmed by sequencing.\nHBoV was detected using a method described by us previously [13]. HBoV2, HBoV3, and HBoV4 were detected by nested PCR as described by Kapoor et al., in which primers HBoV2-sf1, HBoV2-sr1, HBoV2-sf2, and HBoV2-sr2 were used to amplify a 495-nt region within the ORF of NS1 [2]. PCR-positive samples were confirmed by sequencing.\n[SUBTITLE] Sequence and phylogenetic analysis [SUBSECTION] To analyze genetic variation in HBoV2, two sets of nested primers were designed and used to amplify part of the NP1 and VP1 genes in positive samples (Table 1), using the NP1-F1 and NP1-R1, and VP1-F1 and VP1-R1 primers in the first round of PCR and the NP1-F2 and NP1-R2, and VP1-F2 and VP1-R2 in the second. The reaction mix contained 10 pmol each primer and 2.5 units ExTaq DNA polymerase (Takara Bio). After 5 min at 94°C, 35 cycles of amplification (94°C for 45 s, 50°C for 1 min, and 72°C for 1 min) were performed, followed by a 7 min extension at 72°C. Complete HBoV2 sequences were amplified by using specific PCR and Genome Walking Kit (TaKaRa code: D316). PCR products were cloned and the plasmid inserts were sequenced. Nucleotide and deduced amino acid sequences were compared to entries in the GenBank database. Phylogenetic analysis was conducted with Molecular Evolutionary Genetics Analysis (MEGA) version 4.1. All NS1, NP1, VP1 partial gene sequences and nearly full-length genome sequences of HBoV2 except the termini were submitted to GenBank (accession numbers GU301644-GU301683, HQ153797-HQ153804).\nPrimers used in this study.\n* The target fragment of NP1 partial gene was 586 nt, in 2285-2870 nt according to HBoV2 FJ170279 strain.\n# The target fragment of VP1 partial gene was 620 nt, in 3134-3753 nt according to HBoV2 FJ170279 strain.\n‡The target fragment of NP1 Real-time PCR 145 nt, in 2489-2633 nt according to HBoV2 FJ170279 strain.\nA total of 82 nearly full-length genome sequences of HBoV, HBoV2, HBoV3, and HBoV4 were obtained from GenBank. These were aligned and manually adjusted using ClustalW and BioEdit. Phylogenetic trees were determined by the neighbor-joining (NJ) method using the MEGA 4.1 software package. Various nucleotide substitution models were examined and yielded phylogenetic trees of similar topology; only the Kimura 2-parameter model trees were described in this report. A bootstrap resampling (1000 replications) was used to assess the reliability of individual nodes in each phylogenetic tree.\nTo analyze genetic variation in HBoV2, two sets of nested primers were designed and used to amplify part of the NP1 and VP1 genes in positive samples (Table 1), using the NP1-F1 and NP1-R1, and VP1-F1 and VP1-R1 primers in the first round of PCR and the NP1-F2 and NP1-R2, and VP1-F2 and VP1-R2 in the second. The reaction mix contained 10 pmol each primer and 2.5 units ExTaq DNA polymerase (Takara Bio). After 5 min at 94°C, 35 cycles of amplification (94°C for 45 s, 50°C for 1 min, and 72°C for 1 min) were performed, followed by a 7 min extension at 72°C. Complete HBoV2 sequences were amplified by using specific PCR and Genome Walking Kit (TaKaRa code: D316). PCR products were cloned and the plasmid inserts were sequenced. Nucleotide and deduced amino acid sequences were compared to entries in the GenBank database. Phylogenetic analysis was conducted with Molecular Evolutionary Genetics Analysis (MEGA) version 4.1. All NS1, NP1, VP1 partial gene sequences and nearly full-length genome sequences of HBoV2 except the termini were submitted to GenBank (accession numbers GU301644-GU301683, HQ153797-HQ153804).\nPrimers used in this study.\n* The target fragment of NP1 partial gene was 586 nt, in 2285-2870 nt according to HBoV2 FJ170279 strain.\n# The target fragment of VP1 partial gene was 620 nt, in 3134-3753 nt according to HBoV2 FJ170279 strain.\n‡The target fragment of NP1 Real-time PCR 145 nt, in 2489-2633 nt according to HBoV2 FJ170279 strain.\nA total of 82 nearly full-length genome sequences of HBoV, HBoV2, HBoV3, and HBoV4 were obtained from GenBank. These were aligned and manually adjusted using ClustalW and BioEdit. Phylogenetic trees were determined by the neighbor-joining (NJ) method using the MEGA 4.1 software package. Various nucleotide substitution models were examined and yielded phylogenetic trees of similar topology; only the Kimura 2-parameter model trees were described in this report. A bootstrap resampling (1000 replications) was used to assess the reliability of individual nodes in each phylogenetic tree.\n[SUBTITLE] Recombination Analyses [SUBSECTION] We detected recombination using the RDP3 package. Sequences were selected based on their similarity and then aligned and manually adjusted using ClustalW and BioEdit. The alignments were scanned by various algorithms implemented in the RDP3 package, followed by manual refinement. In addition, we performed similarity plot and bootscanning analyses for potential recombination events using SimPlot 3.5. Genetic Algorithm Recombination Detection (GARD) methods were also used to detect recombination and estimate breakpoint locations. Estimated breakpoints were verified by both the Shimodaira-Hasegawa test and manually checking phylogenetic trees for nonrecombinant segments.\nWe detected recombination using the RDP3 package. Sequences were selected based on their similarity and then aligned and manually adjusted using ClustalW and BioEdit. The alignments were scanned by various algorithms implemented in the RDP3 package, followed by manual refinement. In addition, we performed similarity plot and bootscanning analyses for potential recombination events using SimPlot 3.5. Genetic Algorithm Recombination Detection (GARD) methods were also used to detect recombination and estimate breakpoint locations. Estimated breakpoints were verified by both the Shimodaira-Hasegawa test and manually checking phylogenetic trees for nonrecombinant segments.", "The study was approved for human subject protection by the Research Ethics Committee of the Lanzhou University and the Institutional Review Board at China CDC. Following informed consent was written by parent/guardian.", "From July 2006 to June 2008 in Lanzhou, China, we collected stool samples from 632 hospitalized children with diarrhea and 162 asymptomatic children. 3-5 ml stool was collected from every participant. All subjects were 5 or less years of age. Medical histories were provided by parents/guardians. The case group included subjects hospitalized for gastroenteritis in the Department of Pediatrics in our institution. Diarrhea was defined as three or more loose stools in the previous 24 h. Patients were excluded from the study if stool sample volume was insufficient for a complete evaluation of viral agents, stools had blood streaks or pus, or due to the presence of a co-morbidity. Subjects in the control group had presented to the First Hospital of Lanzhou University Pediatric Primary Care Center for a routine examination and did not have fever, diarrhea, vomiting, or respiratory illness in the previous 3-week period [12]. Control subjects received follow-up by telephone and those in whom any of the aforementioned exclusion criteria were present during the week after the initial examination were excluded. A 10% suspension of stool sample was made by mixing 0.5 g stool with 1.0 mL PBS (pH7.2). Viral RNA and DNA were extracted from stool suspensions clarified by centrifugation (1500 × g, 20 min) using a QIAamp® Viral RNA Mini kit (Qiagen, Hilden, Germany), according to the manufacturer's protocol. RNA and DNA were resuspended in 50 μL water and stored at -70°C.", "HBoV was detected using a method described by us previously [13]. HBoV2, HBoV3, and HBoV4 were detected by nested PCR as described by Kapoor et al., in which primers HBoV2-sf1, HBoV2-sr1, HBoV2-sf2, and HBoV2-sr2 were used to amplify a 495-nt region within the ORF of NS1 [2]. PCR-positive samples were confirmed by sequencing.", "To analyze genetic variation in HBoV2, two sets of nested primers were designed and used to amplify part of the NP1 and VP1 genes in positive samples (Table 1), using the NP1-F1 and NP1-R1, and VP1-F1 and VP1-R1 primers in the first round of PCR and the NP1-F2 and NP1-R2, and VP1-F2 and VP1-R2 in the second. The reaction mix contained 10 pmol each primer and 2.5 units ExTaq DNA polymerase (Takara Bio). After 5 min at 94°C, 35 cycles of amplification (94°C for 45 s, 50°C for 1 min, and 72°C for 1 min) were performed, followed by a 7 min extension at 72°C. Complete HBoV2 sequences were amplified by using specific PCR and Genome Walking Kit (TaKaRa code: D316). PCR products were cloned and the plasmid inserts were sequenced. Nucleotide and deduced amino acid sequences were compared to entries in the GenBank database. Phylogenetic analysis was conducted with Molecular Evolutionary Genetics Analysis (MEGA) version 4.1. All NS1, NP1, VP1 partial gene sequences and nearly full-length genome sequences of HBoV2 except the termini were submitted to GenBank (accession numbers GU301644-GU301683, HQ153797-HQ153804).\nPrimers used in this study.\n* The target fragment of NP1 partial gene was 586 nt, in 2285-2870 nt according to HBoV2 FJ170279 strain.\n# The target fragment of VP1 partial gene was 620 nt, in 3134-3753 nt according to HBoV2 FJ170279 strain.\n‡The target fragment of NP1 Real-time PCR 145 nt, in 2489-2633 nt according to HBoV2 FJ170279 strain.\nA total of 82 nearly full-length genome sequences of HBoV, HBoV2, HBoV3, and HBoV4 were obtained from GenBank. These were aligned and manually adjusted using ClustalW and BioEdit. Phylogenetic trees were determined by the neighbor-joining (NJ) method using the MEGA 4.1 software package. Various nucleotide substitution models were examined and yielded phylogenetic trees of similar topology; only the Kimura 2-parameter model trees were described in this report. A bootstrap resampling (1000 replications) was used to assess the reliability of individual nodes in each phylogenetic tree.", "We detected recombination using the RDP3 package. Sequences were selected based on their similarity and then aligned and manually adjusted using ClustalW and BioEdit. The alignments were scanned by various algorithms implemented in the RDP3 package, followed by manual refinement. In addition, we performed similarity plot and bootscanning analyses for potential recombination events using SimPlot 3.5. Genetic Algorithm Recombination Detection (GARD) methods were also used to detect recombination and estimate breakpoint locations. Estimated breakpoints were verified by both the Shimodaira-Hasegawa test and manually checking phylogenetic trees for nonrecombinant segments.", "[SUBTITLE] Detection of human bocaviruses [SUBSECTION] In subjects with gastroenteritis, the positive rates of HBoV, HBoV2 and HBoV3 were 4.3%, 20.4% and 0.9%, respectively, whilst HBoV4 was not detected. In the control group, HBoV and HBoV2 were detected in 2.5% and 12.3% of subjects; HBoV3 and HBoV4 were not detected.\nIn subjects with gastroenteritis, the positive rates of HBoV, HBoV2 and HBoV3 were 4.3%, 20.4% and 0.9%, respectively, whilst HBoV4 was not detected. In the control group, HBoV and HBoV2 were detected in 2.5% and 12.3% of subjects; HBoV3 and HBoV4 were not detected.\n[SUBTITLE] Phylogenetic analyses of HBoV2 [SUBSECTION] Phylogenetic analysis indicated that all of the partial NS1 gene sequences of case group were closer with the PK-2255 strain (from Pakistani children, GenBank: FJ170279) than other strains, with 97.4-99% identity. Consistently, those of control group had 98-99% identity with the PK-2255 strain and were also closer with it than others. The sequences in this study, including case and control group, had a high identity of 98.5-100% with each other. The Phylogenetic tree also indicated that there was no difference in the topological characteristics of HBoV2 between case and control groups (Figure 1).\nPhylogenetic analysis of human bocavirus1-4 and other bocavirus members partial NS1 gene sequences. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, black triangles designate sequences from control group and the others were sequences generated from the gastroenteritis children in the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nFurther study indicated that partial NP1 gene sequences were also similar to those of strain PK-2255 (98-99%), except that four had a high identity to strain FJ973558 (Figure 2), with >99% sequence identity. Interestingly, the ten HBoV2 partial VP1 gene sequences in this study were more variable than those of NS1 and NP1, being only 92.6-97.3% similar to those of strain PK-2255. Six partial VP1 gene sequences (Figure 3) were in the same cluster as strain PK-2255 (97.3-98.1% identity) and the other four sequences clustered with strain FJ973558 (96.9-97.4% identity). The nearly full-length genome sequence of LZ53819 generated in this study (GenBank number: GU301644) was more similar to strain FJ973558 than to PK-2255 (similarity 97.5%), but the NS1 gene sequence showed greater similarity to that of strain PK-2255. Above results suggested that recombination among HBoV2 genotypes may occur. While we compared the trees of non-recombinant segments associated with the break points estimated by GARD, Shimodaira-Hasegawa test showed the topologies of these trees were not significantly different.\nPhylogenetic analysis of the partial NP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nPhylogenetic analysis of the partial VP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nPhylogenetic analysis indicated that all of the partial NS1 gene sequences of case group were closer with the PK-2255 strain (from Pakistani children, GenBank: FJ170279) than other strains, with 97.4-99% identity. Consistently, those of control group had 98-99% identity with the PK-2255 strain and were also closer with it than others. The sequences in this study, including case and control group, had a high identity of 98.5-100% with each other. The Phylogenetic tree also indicated that there was no difference in the topological characteristics of HBoV2 between case and control groups (Figure 1).\nPhylogenetic analysis of human bocavirus1-4 and other bocavirus members partial NS1 gene sequences. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, black triangles designate sequences from control group and the others were sequences generated from the gastroenteritis children in the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nFurther study indicated that partial NP1 gene sequences were also similar to those of strain PK-2255 (98-99%), except that four had a high identity to strain FJ973558 (Figure 2), with >99% sequence identity. Interestingly, the ten HBoV2 partial VP1 gene sequences in this study were more variable than those of NS1 and NP1, being only 92.6-97.3% similar to those of strain PK-2255. Six partial VP1 gene sequences (Figure 3) were in the same cluster as strain PK-2255 (97.3-98.1% identity) and the other four sequences clustered with strain FJ973558 (96.9-97.4% identity). The nearly full-length genome sequence of LZ53819 generated in this study (GenBank number: GU301644) was more similar to strain FJ973558 than to PK-2255 (similarity 97.5%), but the NS1 gene sequence showed greater similarity to that of strain PK-2255. Above results suggested that recombination among HBoV2 genotypes may occur. While we compared the trees of non-recombinant segments associated with the break points estimated by GARD, Shimodaira-Hasegawa test showed the topologies of these trees were not significantly different.\nPhylogenetic analysis of the partial NP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nPhylogenetic analysis of the partial VP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\n[SUBTITLE] Recombination analyses among HBoV, HBoV2, HBoV3, and HBoV4 [SUBSECTION] The phylogenetic relationships of a total of 82 nearly full-length genome sequences obtained both from this study and GenBank were inferred. Based on sequence identity, we selected alignments of five nearly full-length genome sequences. Recombinant signals were detected in all of the HBoV, HBoV2, HBoV3 and HBoV4 genome sequences, but the signals in HBoV3 was far more significant than those in other bocaviruses. Using RDP3 and SimPlot3.5, it was found that HBoV3 (NC_012564) was a potential recombinant of HBoV (FJ858259) and HBoV4 (NC_012729) (Figure 4A, B). The breakpoint was located near the VP1 start codon. GARD analyses suggested six possible breakpoints with model average support over 0.90, including one located 18 bp downstream of the VP1 start codon (Figure 4C). This breakpoint was further examined by constructing phylogenetic trees of the two nonrecombinant segments (Figure 4C). While NC_012564 grouped with FJ858259 in the 5' segment (including NS1 and NP1), it grouped with NC_012729 in the 3' segment (consisting of VP1 and VP2), suggesting that HBoV3 is a hybrid of HBoV and HBoV4. However, the VP1 and VP2 region of HBoV3 was as similar to HBoV2 (identity 87.0% with GQ200737 and 88.6% with FJ973560) as it was to HBoV4 (88.3% to NC_012729).\nHBoV3 is a potential hybrid of HBoV and HBoV4 by recombination analysis. (A) recombination analysis was conducted by Similarity plot, (B) recombination of HBoV3 was conducted by bootscanning analysis, (C) recombination of HBoV3 was conducted by GARD analysis.\nThe phylogenetic relationships of a total of 82 nearly full-length genome sequences obtained both from this study and GenBank were inferred. Based on sequence identity, we selected alignments of five nearly full-length genome sequences. Recombinant signals were detected in all of the HBoV, HBoV2, HBoV3 and HBoV4 genome sequences, but the signals in HBoV3 was far more significant than those in other bocaviruses. Using RDP3 and SimPlot3.5, it was found that HBoV3 (NC_012564) was a potential recombinant of HBoV (FJ858259) and HBoV4 (NC_012729) (Figure 4A, B). The breakpoint was located near the VP1 start codon. GARD analyses suggested six possible breakpoints with model average support over 0.90, including one located 18 bp downstream of the VP1 start codon (Figure 4C). This breakpoint was further examined by constructing phylogenetic trees of the two nonrecombinant segments (Figure 4C). While NC_012564 grouped with FJ858259 in the 5' segment (including NS1 and NP1), it grouped with NC_012729 in the 3' segment (consisting of VP1 and VP2), suggesting that HBoV3 is a hybrid of HBoV and HBoV4. However, the VP1 and VP2 region of HBoV3 was as similar to HBoV2 (identity 87.0% with GQ200737 and 88.6% with FJ973560) as it was to HBoV4 (88.3% to NC_012729).\nHBoV3 is a potential hybrid of HBoV and HBoV4 by recombination analysis. (A) recombination analysis was conducted by Similarity plot, (B) recombination of HBoV3 was conducted by bootscanning analysis, (C) recombination of HBoV3 was conducted by GARD analysis.", "In subjects with gastroenteritis, the positive rates of HBoV, HBoV2 and HBoV3 were 4.3%, 20.4% and 0.9%, respectively, whilst HBoV4 was not detected. In the control group, HBoV and HBoV2 were detected in 2.5% and 12.3% of subjects; HBoV3 and HBoV4 were not detected.", "Phylogenetic analysis indicated that all of the partial NS1 gene sequences of case group were closer with the PK-2255 strain (from Pakistani children, GenBank: FJ170279) than other strains, with 97.4-99% identity. Consistently, those of control group had 98-99% identity with the PK-2255 strain and were also closer with it than others. The sequences in this study, including case and control group, had a high identity of 98.5-100% with each other. The Phylogenetic tree also indicated that there was no difference in the topological characteristics of HBoV2 between case and control groups (Figure 1).\nPhylogenetic analysis of human bocavirus1-4 and other bocavirus members partial NS1 gene sequences. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, black triangles designate sequences from control group and the others were sequences generated from the gastroenteritis children in the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nFurther study indicated that partial NP1 gene sequences were also similar to those of strain PK-2255 (98-99%), except that four had a high identity to strain FJ973558 (Figure 2), with >99% sequence identity. Interestingly, the ten HBoV2 partial VP1 gene sequences in this study were more variable than those of NS1 and NP1, being only 92.6-97.3% similar to those of strain PK-2255. Six partial VP1 gene sequences (Figure 3) were in the same cluster as strain PK-2255 (97.3-98.1% identity) and the other four sequences clustered with strain FJ973558 (96.9-97.4% identity). The nearly full-length genome sequence of LZ53819 generated in this study (GenBank number: GU301644) was more similar to strain FJ973558 than to PK-2255 (similarity 97.5%), but the NS1 gene sequence showed greater similarity to that of strain PK-2255. Above results suggested that recombination among HBoV2 genotypes may occur. While we compared the trees of non-recombinant segments associated with the break points estimated by GARD, Shimodaira-Hasegawa test showed the topologies of these trees were not significantly different.\nPhylogenetic analysis of the partial NP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.\nPhylogenetic analysis of the partial VP1 gene sequences of human bocavirus1-4 and other bocavirus members. Phylogenetic tree was constructed by the neighbor-joining method, Kimura 2-parameter model with 1,000 bootstrap replicates, by using MEGA 4.1 package. Black dots designate reference strains, the others were sequences generated from the present study. MVC: Minute virus of canines; BPV: Bovine parvovirus.", "The phylogenetic relationships of a total of 82 nearly full-length genome sequences obtained both from this study and GenBank were inferred. Based on sequence identity, we selected alignments of five nearly full-length genome sequences. Recombinant signals were detected in all of the HBoV, HBoV2, HBoV3 and HBoV4 genome sequences, but the signals in HBoV3 was far more significant than those in other bocaviruses. Using RDP3 and SimPlot3.5, it was found that HBoV3 (NC_012564) was a potential recombinant of HBoV (FJ858259) and HBoV4 (NC_012729) (Figure 4A, B). The breakpoint was located near the VP1 start codon. GARD analyses suggested six possible breakpoints with model average support over 0.90, including one located 18 bp downstream of the VP1 start codon (Figure 4C). This breakpoint was further examined by constructing phylogenetic trees of the two nonrecombinant segments (Figure 4C). While NC_012564 grouped with FJ858259 in the 5' segment (including NS1 and NP1), it grouped with NC_012729 in the 3' segment (consisting of VP1 and VP2), suggesting that HBoV3 is a hybrid of HBoV and HBoV4. However, the VP1 and VP2 region of HBoV3 was as similar to HBoV2 (identity 87.0% with GQ200737 and 88.6% with FJ973560) as it was to HBoV4 (88.3% to NC_012729).\nHBoV3 is a potential hybrid of HBoV and HBoV4 by recombination analysis. (A) recombination analysis was conducted by Similarity plot, (B) recombination of HBoV3 was conducted by bootscanning analysis, (C) recombination of HBoV3 was conducted by GARD analysis.", "Although the primers used in this study were capable of detecting HBoV4, no HBoV4 was detected. HBoV2 had higher prevalence than HBoV and HBoV3 in both case and control groups, indicating that HBoV2 should be given more attention than other human bocaviruses. Given the fact that children in the control group that were free of symptoms were positive for HBoV and HBoV2, perhaps bocaviruses are only \"passers-by\" in intestinal tract than pathogens of gastroenteritis, or the control participants had symptomless infection of bocaviruses. This issue is not clear yet and needs more studies to resolve [7,8,13,14]. All of the partial NS1 gene sequences in this study were closer with the PK-2255 strain (from Pakistani children) than other strains. There was a high sequence identity of 98.5-100% between HBoV2 from case and control groups and the phylogenetic tree also confirmed that there was no difference in the phylogenetic characteristics of them, suggesting that a single genetic lineage of HBoV2 is circulating in both the gastroenteritis and healthy children in Lanzhou, China. And it also indicated HBoV2 can cause asymptomatic infection in Children.\nPhylogenetic analyses demonstrated that partial NS1, NP1, and VP1 gene sequences of HBoV2 were markedly similar to those of a number of reference strains. This variability was also reported by Kapoor [2], who postulated that it was due to recombination between HBoV2 strains, as has been reported for animal parvoviruses [15]. However, when phylogenetic trees of non-recombinant segments associated with breakpoints estimated by GARD were compared, Shimodaira-Hasegawa test showed the topologies were not significantly different. Considering that these sequences exhibit very high identity (more than 95%), this recombination phenomenon may be mediated by other processes, for example, variation in spatial rate and/or heterotachy. More works are required to elucidate fully the nature and extent of any recombination that occurs between HBoV2 strains.\nArthur et al. identified two recombination sites upstream of the NS1 and VP1/2 genes, using HBoV, HBoV2, and HBoV3 sequence analyses, and hypothesized that HBoV3 may be a hybrid of HBoV and HBoV2 [3]. Our phylogenetic analyses of HBoV, HBoV2, HBoV3 and HBoV4 strains confirmed that HBoV3 may be a hybrid of HBoV and HBoV4. The estimated breakpoint was located at upstream of the VP1 gene. Although bootscanning analyses and phylogenetic trees suggested that HBoV3 was a hybrid of HBoV and HBoV4, the parent strain of the VP1/VP2 region was still hard to determine. The VP1/VP2 region of HBoV3 was as similar to that of HBoV2 as it was to that of HBoV4, suggesting that HBoV3 may be a hybrid of HBoV and the common ancestor of HBoV2 and HBoV4.\nAccording to the ICTVb criteria http://www.ictvdb.org/Ictv/fs_parvo.htm, isolates with non-structural gene homologies of less than 95% are defined as a new species in the bocavirus genus. NS1 sequence variation amongst HBoV2 clusters was as high as 8%. However, NS1 sequence variation between one HBoV2 cluster and HBoV4 was only 6%, suggesting that either more than one HBoV2 cluster exists or that some of these clusters should in fact be regarded as a separate species. The high sequence identity, limited sequence data, and unclear taxonomy rendered the precise phylogenetic relationships among these human bocaviruses difficult to determine.\nThe data presented here suggested the potential phylogenetic relationships among the known human bocaviruses. However, many issues remain, including the nature and extent of recombination between HBoV2 strains, and the precise evolutionary relationships of the various human bocaviruses. More studies including more samples from different areas and years are needed to address them.", "In summary, our data suggested that HBoV2 had higher prevalence than HBoV and HBoV3 in both case and control groups. A single genetic lineage of HBoV2 is circulating in children with and without gastroenteritis in Lanzhou, China. Recombination between HBoV2 strains may occur and HBoV3 may be a hybrid virus, originating from HBoV and the common ancestor of HBoV2 and HBoV4.", "The authors declare that they have no competing interests.", "Conceived and designed the experiments: ZD, YJ. Performed the experiments: WC, ZX, JY. Analyzed the data: JC, WC, CH, MZ, ZD. Contributed reagents/materials/analysis tools: MJ, HL. Wrote the paper: WC, JC, ZD. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2334/11/50/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Animals", "Arthritis, Experimental", "Autoantibodies", "Autoimmunity", "B-Cell Activating Factor", "B-Lymphocytes", "Cattle", "Cell Line", "Collagen Type II", "Disease Models, Animal", "Disease Progression", "Genetic Therapy", "Humans", "Male", "Mice", "Mice, Inbred DBA", "T-Lymphocytes", "alpha 1-Antitrypsin" ]

[ "Background", "rAAV Vector Production", "Animals", "Histological Assessment", "Human AAT Protein and rAAV8-CB-AAT Vector Administration", "ELISA for the Detection of Serum hAAT and BAFF Levels and Antibodies against hAAT, bCII and mCII", "Cell Culture", "Quantitative PCR", "Assessment of T-cell Autoreactive Response", "Statistical Analysis", "Results", "Human AAT Protein Therapy Delayed Arthritis Development in DBA/1 Mice", "Human AAT Protein Therapy Reduced the Levels of anti-bCII and anti-mCII Autoantibodies", "Human AAT (hAAT) Gene Therapy delayed Arthritis Development", "Human AAT (hAAT) Gene Therapy Reduced the Levels of Anti-CII Autoantibodies", "Human AAT Therapy Reduced B-cell Activating Factor (BAFF) in vitro and in vivo", "Discussion", "Conclusion", "Abbreviations", "Competing interests", "Authors' contributions" ]

[ "Rheumatoid arthritis (RA) is a systemic autoimmune disease, characterized by chronic joint inflammation and synovial hyperplasia leading to bone and joint destruction. The life expectancy is lowered and quality of life is decreased in RA patients. So far little is known about the actual disease initiating stimulus; however, extensive research over the last decades have shown that multiple genetic as well as environmental factors interact and trigger the onset of RA [1,2]. The autoimmune inflammation of RA is maintained by inappropriate action of macrophages, B-cells, T-cells, and other types of cells leading to dysregulated cytokine/chemokine production. The synovial inflammation is caused by infiltration and proliferation of activated immune cells including neutrophils, macrophages, fibroblasts, mast cells, NK cells, NKT cells, T-cells as well as plasma cells [3]. Progressive joint and bone destruction is mediated through the activities of osteoclasts, chondrocytes, synovial fibroblasts and cytokine induction of destructive enzymes, chiefly matrix metalloproteinases (MMP) [4]. Current therapy mainly aims to inhibit the biological function of tumor necrosis factor-alpha (TNF-α) and lymphocyte proliferation. Due to ineffectiveness of anti-TNF-α therapy in certain patients and various side effects of methotrexate which inhibits lymphocytes proliferation, there is still the need to identify new target molecules/pathways and to develop new treatment [5]. Immunoregulatory and anti-inflammatory strategies that affect B-cell activation, T-cell activation or inhibit proinflammatory cytokines have recently shown great potential for the treatment of RA [5,6].\nHuman alpha-1 antitrypsin (hAAT) is a 52 kDa serum glycoprotein, synthesized primarily in the liver. It is also expressed in other types of cells including neutrophils, monocytes, macrophages, alveolar macrophages, intestinal epithelial cells, carcinoma cells and the cornea [7-10]. The normal serum level of hAAT is 1-2mg/ml. During inflammation, hAAT level, as an acute phase reactant, can increase 3-4 folds, suggesting an important role in responding to inflammation in the human body. Increasing evidence indicates that hAAT is immunoregulatory, anti-inflammatory and may be used for the treatment of RA. It inhibits neutrophil elastase and proteinase 3 with high efficiency, as well as cathepsin G, thrombin, trypsin and chymotrypsin with lower efficiency [11]. Most of these proteases target receptor proteins, involved in proinflammatory cytokine expression and cell signaling [12]. It also has been reported that neutrophil elastase inhibitors reduce incidence as well as severity of collagen-induced arthritis (CIA) in both rats and mice [13]. Human AAT is able to completely eliminate the acute inflammatory infiltration and connective tissue breakdown in the lung in a cigarette smoke-induced emphysema mouse model [14]. It also inhibits lipopolysaccharide (LPS)-stimulated release of TNF-α and interleukin (IL) -1β, and enhances the production of anti-inflammatory cytokine IL-10 [15-17]. Human AAT significantly protects against the lethality induced by TNF-α or endotoxin in mice [18]. It can also induce expression of IL1-Ra in human peripheral blood mononuclear cells (PBMC's) [19] and reduces ischemia-induced apoptosis and inflammation [20]. We have recently shown, that combination therapy using doxycycline and hAAT gene therapy reduces arthritis development in mice, suggesting a therapeutic effect of hAAT in an arthritis mouse model [21].\nRecombinant adeno-associated virus vectors (rAAV) have been widely used for gene therapy in animal models and human clinical trials [22], because of their unique features in safety and efficiency. It has been reported that rAAV mediated long-term and high levels of transgene expression in a wide variety of tissues, including muscle [23], lung [24], liver [25], brain [26] and eye [27]. Recently developed rAAV vectors including new serotypes of AAV, mutants AAV and double stranded AAV have provided more opportunities and challenges for their application [28-31]. Previously, we have shown hAAT gene therapy using rAAV2 and rAAV1 vectors prevented type 1 diabetes. However, the immune response to the transgene product (hAAT) complicated the therapeutic effect [32,33]. We have recently discovered that rAAV8 vector fail to transduce dendritic cells and induce immune tolerance to transgene product entailing rAAV8 as a promising vector used for therapeutic intervention [34].\nIn the present study we further investigated the feasibility of hAAT with its anti-inflammatory and immunoregulatory properties for the treatment of RA using both, protein therapy and rAAV8 mediated gene therapy.", "The rAAV-CB-hAAT vector construct was produced and packaged as previously described [27]. Briefly, this vector carries hAAT cDNA driven by the cytomegalovirus (CMV) enhancer and chicken β-actin promoter and contains AAV2 inverted terminal repeats (ITRs). It was packaged into AAV serotype 8 capsid by cotransfection of vector plasmid and helper plasmid (XYZ8) into 293 cells. rAAV8-CB-hAAT vectors were purified by iodixanol gradient centrifugation followed by anion-exchange chromatography. The physical particle titers of vector preparations were assessed by dot blot analysis.", "Six week-old male DBA/1 mice were purchased from Harlan Sprague Dawley, Inc. (Indianapolis, IN), housed in a specific pathogen-free room as approved by the University of Florida Institutional Animal Care and Use Committee. For induction of arthritis, bCII (Chondrex LLC, Redmond, WA) was dissolved in 0.05N acetic acid at a concentration of 2mg/ml by stirring overnight at 4°C and was emulsified with an equal volume of Complete Freund's Adjuvant (CFA) (Chondrex LLC, Redmond, WA). At the age of eight weeks, DBA/1 mice were immunized intradermally at the base of the tail with 0.1ml of emulsion containing 100 μg of type II collagen. Three weeks after priming (day 21), the mice were boosted with 0.1 ml of bCII (100 μg) emulsified in equal volume of incomplete Freund's Adjuvant (IFA) (Difco, Detroit, MI). For assessment of arthritis, all mice were monitored three times a week by the same person blinded to the treatment group and evaluated the incidence of arthritis and clinical score. An arthritis score system ranging from stage 0 - 4 was used: 0: no swelling or redness; 1: detectable arthritis with erythema; 2: significant swelling and redness; 3: severe swelling and redness from joint to digit; 4: joint stiffness or deformity with ankylosis [35]. The clinical score was expressed as the average cumulative value of all four paws with a maximum score per animal of 16. Severe arthritis was defined as arthritis score > 3 for the purpose of comparing data between groups.", "For the analysis of arthritis, mice were anesthetized and sacrificed by cervical dislocation on day 28 after immunization. The two hind limbs of mice in treatment and control groups were removed. Specimens were fixed in formalin and decalcified in RDO solution (Apex, Aurora, IL) for 10-20 min depending on tissue size and then checked manually for pliability. Sections 4 μm thick were cut and stained with hematoxylin and eosin according to standard methods.\nHistological evaluation was performed by two independent and blinded pathologists. Infiltration of immune cells, hyperplasia, pannus formation and bone deformation was determined for each paw using an evaluation scale ranging from 0-3 according to severity of pathohistological changes. (0: normal, 1: mild, 2: moderate, 3: severe).", "For hAAT protein therapy studies, DBA/1 mice were intraperitoneally (IP) injected with 0.5 mg (in 100 μl saline) of hAAT (Prolastin®, Bayer Corp., Elkhard, IN). The control group received saline injection. The injections were performed twice per week, starting at 6 days before the first bCII immunization until the end of study (EOS) at day 70 after the first immunization. For hAAT gene therapy studies, DBA/1 mice were IP injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse) two weeks before the first CII immunization. The control group received saline injection.", "Detection of hAAT and anti-hAAT antibodies in mouse serum was performed as previously described [32]. Purified hAAT (Athens Research & Technology, Athens, GA) was used as a standard. Anti-type II collagen antibodies in mouse serum were detected by a standard ELISA. Briefly, microtiter plates (Immulon 4, Dynex Technologies, Chantilly, VA) were coated with bCII or mCII (0.5 μg/well, Chondrex LLC, Redmond, WA) in Voller's buffer overnight at 4°C. After blocking with 3% bovine serum albumin, wells were incubated with samples at room temperature for 2 h. HRP-conjugated goat anti-mouse IgG antibodies (1:1,000 dilution, Sigma, St. Louis, MO), goat anti-mouse IgG1 antibodies (1:1,500 dilution, Roche, Indianapolis, IN) and goat anti-mouse IgG2a antibodies (1:1,500 dilution, Roche, Indianapolis, IN) were incubated for 1 h at RT. The plates were washed with PBS-Tween 20 between reactions. After adding the substrate (o-phenylenediamine, Sigma, St Louis, MO), plates were read at 490 nm on an MRX microplate reader (Dynex Technologies, Chantilly, VA). Optical densities were converted into units based on a standard curve generated with high titer sera from DBA/1 mice immunized with bCII. Detection of BAFF in serum was performed according to manufactures instructions (R&D systems, Inc. Minneapolis, MN).", "The murine macrophage cell line RAW 264.7 was cultured in serum free DMEM at 37°C in a 5% CO2 incubator. For measuring BAFF release into medium, cells were seeded at 1 × 105/ml in 12 well plates. Cells were incubated in quadruplicates with hAAT (0.5mg/ml; Prolastin®, Bayer Corp., Elkhard, IN) for 16 hours and BAFF secretion into the culture medium was determined by ELISA according to manufactures instructions (R&D systems, Inc. Minneapolis, MN).", "Total RNA from cell culture described above, was isolated using RNeasy Mini Kit (Quiagen, Valencia, CA). Samples were processed according to the manufacture's protocol. For reverse transcription, cDNA was synthesized with oligo dT16 primers and Moloney Murine Leukemia Virus Reverse Transcriptase (MMLV-RT) according to manufacture's manual (Taqman Reverse Transcription Reagents, Applied Biosystems, Foster City, CA).\ncDNA was analyzed by quantitative PCR using gene-specific primers with SYBR Green 2X PCR mix (Applied Biosystems). The sequence of the primers were as follows: BAFF (205bp), sense: 5'-TGC CTT GGA GGA GAA AGA GA-3' and antisense: 5'-GGA ATT GTT GGG CAG TGT TT-3'; GAPDH (122bp), sense: 5'-CCT GGA GAA ACC TGC CAA GTA T-3' and antisense: 5'-TGC TGT TGA AGT CGC AGG A-3'. Reactions were set up in triplicate and performed on the ABI Prism 7700 Sequence Detector (Applied Biosystems). The cycling parameters were 2 min at 95°C for denaturation, 40 cycles of 15s at 95°C and 30 s at 60°C for amplification. The threshold cycle (CT) of each target product was determined, set to the log linear range of the amplification curve and kept constant for all data analysis. Data were analyzed with Sequence Detector Software (SDS). BAFF expression was normalized to the corresponding GAPDH values for the respective treatment. Values of BAFF expression following saline treatment are designated as 1. The experiment was repeated twice.", "To test the effect of AAV8-hAAT gene therapy on splenocyte proliferation, spleens were harvested at 30 days after the first bCII immunization. Splenocytes were isolated and cultured in serum free X-VIVO medium (Cambrex, Walkersville, MD) in the presence or absence of bCII (100 μg/ml, Chondrex LLC, Redmond, WA). After 3 days culture, 1 μCi/well of [3H] TdR was added. Cells were cultured for additional 18h and [3H] TdR uptake was measured using a β- scintillation counter.\nTo measure cytokine release into the cell culture supernatant, a Beadlyte Mouse Multi-Cytokine Detection System 1 kit (Upstate, Temecula, CA, Cat # 48-005) was used according to the manufacture's instruction and in conjunction with the Luminex 100 system for cytokine determination.", "Data Analysis was performed using GraphPad Prism 4.0 (GraphPad Software) and SAS (SAS Institute). Student's t-test was used to compare differences in BAFF levels in culture medium as well as differences in mRNA expression levels. Mann-Whitney U-test was applied to analyze differences in stimulation indices, cytokine levels, pathohistological changes, serum levels of BAFF and antibodies. For comparison of arthritis score, area under the curve analysis was used and differences in arthritis incidence were determined using Kaplan-Meier survival curve and log-rank test. A p-value of p ≤ 0.05 was considered statistically significant.", "[SUBTITLE] Human AAT Protein Therapy Delayed Arthritis Development in DBA/1 Mice [SUBSECTION] In order to investigate the effect of hAAT on development of arthritis, we first examined the feasibility of hAAT protein therapy in CIA mouse model. Administration of hAAT (0.5 mg/mouse twice per week, starting at 6 days before the induction of arthritis) resulted in sustained high levels of hAAT in mouse serum (Figure 1A). Although anti-hAAT-antibodies were detected (Figure 1B), serum levels of hAAT did not decrease over time.\nAntiarthritic effect of human alpha 1 antitrypsin (hAAT) in collagen induced arthritis (CIA) model. Human AAT (Prolastin®) was intraperitoneally injected in DBA/1 mice (n = 9), twice per week starting 6 days before until day 70 after CII immunization. Control group received saline injections (n = 7) (A) Serum hAAT protein levels in DBA/1 mice were measured by ELISA (mean+SD). ↓ indicates the day of first hAAT injection. (B) Serum anti-hAAT antibody levels (anti-hAAT-IgG) in DBA/1 mice were measured by ELISA. Each dot represents antibody levels (day 49 after bCII immunization, arbitrary units) of an individual mouse. (C) Arthritis score. For each paw, 0 is normal and 4 is the most severe arthritis. The maximum score for each animal is 16. Each line represents the scores from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, *p = 0.029 by AUC analysis) (D) Incidence of severe arthritis is defined by arthritic score/mouse > 3 (**p = 0.0025 by logrank test). Dotted line, saline injected control group; Solid line, hAAT treated group.\nA few days after the second immunization with bCII (day 21), mice in control group developed arthritis in multiple joints, which was manifested by redness, severe joint swelling and joint stiffness as well as ankylosis as the disease progressed. The severity of arthritis as measured by the arthritic score rapidly increased in control group (n = 7) whereas the disease development in hAAT treatment group (n = 9) was suppressed (Figure 1C). At day 49 (7 weeks) after the immunization, area under the curve (AUC) in the hAAT group was 50.83 ± 21.64 (mean ± SEM), while in control group it was 121.5 ± 17.67 (p = 0.029, mean ± SEM, AUC analysis until day 49). Human AAT protein therapy also reduced incidence of severe arthritis (p = 0.0025, logrank test, Figure 1D). Moreover, mice in hAAT treated group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis (arthritis score > 3) started on day 47.3 ± 8.7 (mean ± SD) in hAAT treated group compared to day 36.0 ± 5.8 (mean ± SD) in control group (p = 0.01 by students t-test). Although hAAT treated mice also developed arthritis at the end (70 days after the immunization) of the experiment, these results showed that treatment of hAAT protein (Prolastin®) led to a delayed arthritis onset and amelioration of disease progression in CIA mouse model.\nIn order to investigate the effect of hAAT on development of arthritis, we first examined the feasibility of hAAT protein therapy in CIA mouse model. Administration of hAAT (0.5 mg/mouse twice per week, starting at 6 days before the induction of arthritis) resulted in sustained high levels of hAAT in mouse serum (Figure 1A). Although anti-hAAT-antibodies were detected (Figure 1B), serum levels of hAAT did not decrease over time.\nAntiarthritic effect of human alpha 1 antitrypsin (hAAT) in collagen induced arthritis (CIA) model. Human AAT (Prolastin®) was intraperitoneally injected in DBA/1 mice (n = 9), twice per week starting 6 days before until day 70 after CII immunization. Control group received saline injections (n = 7) (A) Serum hAAT protein levels in DBA/1 mice were measured by ELISA (mean+SD). ↓ indicates the day of first hAAT injection. (B) Serum anti-hAAT antibody levels (anti-hAAT-IgG) in DBA/1 mice were measured by ELISA. Each dot represents antibody levels (day 49 after bCII immunization, arbitrary units) of an individual mouse. (C) Arthritis score. For each paw, 0 is normal and 4 is the most severe arthritis. The maximum score for each animal is 16. Each line represents the scores from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, *p = 0.029 by AUC analysis) (D) Incidence of severe arthritis is defined by arthritic score/mouse > 3 (**p = 0.0025 by logrank test). Dotted line, saline injected control group; Solid line, hAAT treated group.\nA few days after the second immunization with bCII (day 21), mice in control group developed arthritis in multiple joints, which was manifested by redness, severe joint swelling and joint stiffness as well as ankylosis as the disease progressed. The severity of arthritis as measured by the arthritic score rapidly increased in control group (n = 7) whereas the disease development in hAAT treatment group (n = 9) was suppressed (Figure 1C). At day 49 (7 weeks) after the immunization, area under the curve (AUC) in the hAAT group was 50.83 ± 21.64 (mean ± SEM), while in control group it was 121.5 ± 17.67 (p = 0.029, mean ± SEM, AUC analysis until day 49). Human AAT protein therapy also reduced incidence of severe arthritis (p = 0.0025, logrank test, Figure 1D). Moreover, mice in hAAT treated group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis (arthritis score > 3) started on day 47.3 ± 8.7 (mean ± SD) in hAAT treated group compared to day 36.0 ± 5.8 (mean ± SD) in control group (p = 0.01 by students t-test). Although hAAT treated mice also developed arthritis at the end (70 days after the immunization) of the experiment, these results showed that treatment of hAAT protein (Prolastin®) led to a delayed arthritis onset and amelioration of disease progression in CIA mouse model.\n[SUBTITLE] Human AAT Protein Therapy Reduced the Levels of anti-bCII and anti-mCII Autoantibodies [SUBSECTION] It has been shown that high levels of serum anti-collagen II autoantibodies are pathognomonic and associated with the development of arthritis [36,37]. To test the effect of hAAT on autoantibody production, we evaluated the levels of anti-CII autoantibodies in total Ig, and IgG1 and IgG2a subclass at early (day 35) and late (day 49) stages of the disease. As shown in Figure 2A, hAAT treatment did not result in a significant change of total autoantibody levels against bCII (total anti-bCII-Ig). However, hAAT treatment significantly reduced the pathognomonic IgG2a (anti-bCII-IgG2a) levels at day 35 (Figure 2B), and increased IgG1 (anti-bCII-IgG1) levels at day 49 (Figure 2C). Interestingly, levels of total Ig autoantibodies against endogenous mouse collagen II (total anti-mCII-Ig) were significantly lower in hAAT protein treated group than those in control group (P < 0.05) (Figure 2D).\nAnti-collagen II (CII) antibody levels after hAAT treatment. Anti-CII antibodies at day 35 and day 49 were tested by ELISA. Closed bars represent the average levels (n = 9, relative units, mean+SD) of antibodies in hAAT protein therapy treated group. Open bars represent the average levels (n = 7, relative units, mean+SD) of antibodies in saline injected group. (A) Levels of total Ig antibodies to bCII (total anti-bCII-Ig). (B) Levels of IgG2a anti-bCII (anti-bCII-IgG2a). (C) Levels of IgG1 anti-bCII (anti-bCII-IgG1). (D) Levels of total Ig antibodies to mCII (total anti-mCII-Ig). * p < 0.05 by Mann-Whitney U- test.\nIt has been shown that high levels of serum anti-collagen II autoantibodies are pathognomonic and associated with the development of arthritis [36,37]. To test the effect of hAAT on autoantibody production, we evaluated the levels of anti-CII autoantibodies in total Ig, and IgG1 and IgG2a subclass at early (day 35) and late (day 49) stages of the disease. As shown in Figure 2A, hAAT treatment did not result in a significant change of total autoantibody levels against bCII (total anti-bCII-Ig). However, hAAT treatment significantly reduced the pathognomonic IgG2a (anti-bCII-IgG2a) levels at day 35 (Figure 2B), and increased IgG1 (anti-bCII-IgG1) levels at day 49 (Figure 2C). Interestingly, levels of total Ig autoantibodies against endogenous mouse collagen II (total anti-mCII-Ig) were significantly lower in hAAT protein treated group than those in control group (P < 0.05) (Figure 2D).\nAnti-collagen II (CII) antibody levels after hAAT treatment. Anti-CII antibodies at day 35 and day 49 were tested by ELISA. Closed bars represent the average levels (n = 9, relative units, mean+SD) of antibodies in hAAT protein therapy treated group. Open bars represent the average levels (n = 7, relative units, mean+SD) of antibodies in saline injected group. (A) Levels of total Ig antibodies to bCII (total anti-bCII-Ig). (B) Levels of IgG2a anti-bCII (anti-bCII-IgG2a). (C) Levels of IgG1 anti-bCII (anti-bCII-IgG1). (D) Levels of total Ig antibodies to mCII (total anti-mCII-Ig). * p < 0.05 by Mann-Whitney U- test.\n[SUBTITLE] Human AAT (hAAT) Gene Therapy delayed Arthritis Development [SUBSECTION] To further confirm our observation that hAAT is effective in delaying arthritis development, and to test the feasibility of hAAT gene therapy for rheumatoid arthritis, we used recombinant adeno-associated virus vector (rAAV) to deliver the hAAT gene. A single IP injection of rAAV8-CB-hAAT vector (2x1011 particles/mouse, two weeks before the first CII immunization) resulted in sustained levels of hAAT in the circulation, similar to those levels obtained following protein therapy (Figure 3A). Interestingly, following AAV8 mediated gene delivery, we did not observe the development of antibodies to hAAT which were detected during hAAT protein therapy (Figure 3B, compare vs. Figure 1B in mice with hAAT protein therapy). Similar to the results from hAAT protein therapy, however, rAAV-mediated hAAT gene therapy significantly reduced the prevalence of arthritis development at the early stage of disease (Figure 3C). Area under the curve (AUC) in the gene therapy group (n = 10) was 71.65 ± 14.04 (mean ± SEM), while in control group (n = 10) it was 123.20 ± 19.83 (mean ± SEM; p < 0.05 by AUC analysis until day 42). AAT gene therapy also reduced the incidence of severe arthritis (score > 3) at the early stage of disease (p = 0.035 by logrank test, Figure 3D). Moreover, mice in hAAT gene therapy group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis started on day 42.3 ± 7.5 (mean ± SD) in hAAT gene therapy group compared to day 33.4 ± 7.3 in control group (mean ± SD; p < 0.02 by student's t-test). These results indicate that similar to hAAT protein therapy, AAV8 mediated hAAT gene delivery also delayed arthritis onset and ameliorated early stage disease progression in CIA mouse model.\nHuman AAT gene therapy delays disease progression in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 10) or saline (n = 10) two weeks before immunization with CII. Control group received saline. Mice were sacrificed on day 56 (EOS). (A) Serum levels of hAAT. hAAT protein serum levels in vector injected group were measured by ELISA (mean+SD). ↓ indicates the days of injection. (B) Anti-hAAT antibody levels. Serum anti-hAAT antibodies (anti-hAAT) were measured by ELISA using samples obtained at 56 days after immunization. Anti-hAAT antibodies were undetectable in the vector injected group. Each dot represents antibody level (arbitrary units) of an individual mouse. (C) Arthritis score. Each line represents the average score from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, * p < 0.05 as determined by AUC analysis.) (D) Incidence of severe arthritis. Severe arthritis was defined by arthritic score > 3, (* p = 0.035 by logrank test.; 10 mice/group).\nIn an additional experiment using AAV8 mediated hAAT gene therapy, tissue protective properties of hAAT were evaluated. Similar to the previous experiment, mice in treatment group (n = 6) showed significantly reduced arthritis development at the early disease stage compared to control (n = 4) (Figure 4A, p < 0.05 by Mann-Whitney U-test). As shown in Figure 4B-F, AAV8 mediated hAAT gene therapy resulted in less infiltration of immune cells into the joint cavity accompanied with reduced synovial cell hyperplasia and pannus formation (p < 0.05 Mann-Whitney U-test).\nTissue protective effect of hAAT gene therapy in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 6) or saline (n = 4) two weeks before immunization with CII. Control group received saline. (A) Arthritis development was evaluated based on arthritis score (mean + SD). Open circle represent rAAV8-CB-hAAT vector injected group, open triangle represent control group. Mice were sacrificed on day 28 after CII immunization, hind limbs were harvested and processed for histological assessment. *p < 0.05 by Mann-Whitney U-test. (B) Histopathological evaluation of arthritis development. Mice in gene therapy group (black bars) or control group (empty bars) were evaluated according to histopathological changes by two blinded pathologists. Each hind paw was evaluated based on a scale ranging from 0-3. (mean+SD). *p < 0.05, **p < 0.01 by Mann-Whitney U-test. (INF: Infiltration of Immune Cells, HYP: Hyperplasia, P.F.: Pannus Formation, B.D.: Bone Destruction) (C,D) Representative joint section from mice receiving hAAT gene therapy. (E,F) Representative joint section from mice in control group (saline injection). Magnification: C,E: 100x; D,F: 200x.\nTo further confirm our observation that hAAT is effective in delaying arthritis development, and to test the feasibility of hAAT gene therapy for rheumatoid arthritis, we used recombinant adeno-associated virus vector (rAAV) to deliver the hAAT gene. A single IP injection of rAAV8-CB-hAAT vector (2x1011 particles/mouse, two weeks before the first CII immunization) resulted in sustained levels of hAAT in the circulation, similar to those levels obtained following protein therapy (Figure 3A). Interestingly, following AAV8 mediated gene delivery, we did not observe the development of antibodies to hAAT which were detected during hAAT protein therapy (Figure 3B, compare vs. Figure 1B in mice with hAAT protein therapy). Similar to the results from hAAT protein therapy, however, rAAV-mediated hAAT gene therapy significantly reduced the prevalence of arthritis development at the early stage of disease (Figure 3C). Area under the curve (AUC) in the gene therapy group (n = 10) was 71.65 ± 14.04 (mean ± SEM), while in control group (n = 10) it was 123.20 ± 19.83 (mean ± SEM; p < 0.05 by AUC analysis until day 42). AAT gene therapy also reduced the incidence of severe arthritis (score > 3) at the early stage of disease (p = 0.035 by logrank test, Figure 3D). Moreover, mice in hAAT gene therapy group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis started on day 42.3 ± 7.5 (mean ± SD) in hAAT gene therapy group compared to day 33.4 ± 7.3 in control group (mean ± SD; p < 0.02 by student's t-test). These results indicate that similar to hAAT protein therapy, AAV8 mediated hAAT gene delivery also delayed arthritis onset and ameliorated early stage disease progression in CIA mouse model.\nHuman AAT gene therapy delays disease progression in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 10) or saline (n = 10) two weeks before immunization with CII. Control group received saline. Mice were sacrificed on day 56 (EOS). (A) Serum levels of hAAT. hAAT protein serum levels in vector injected group were measured by ELISA (mean+SD). ↓ indicates the days of injection. (B) Anti-hAAT antibody levels. Serum anti-hAAT antibodies (anti-hAAT) were measured by ELISA using samples obtained at 56 days after immunization. Anti-hAAT antibodies were undetectable in the vector injected group. Each dot represents antibody level (arbitrary units) of an individual mouse. (C) Arthritis score. Each line represents the average score from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, * p < 0.05 as determined by AUC analysis.) (D) Incidence of severe arthritis. Severe arthritis was defined by arthritic score > 3, (* p = 0.035 by logrank test.; 10 mice/group).\nIn an additional experiment using AAV8 mediated hAAT gene therapy, tissue protective properties of hAAT were evaluated. Similar to the previous experiment, mice in treatment group (n = 6) showed significantly reduced arthritis development at the early disease stage compared to control (n = 4) (Figure 4A, p < 0.05 by Mann-Whitney U-test). As shown in Figure 4B-F, AAV8 mediated hAAT gene therapy resulted in less infiltration of immune cells into the joint cavity accompanied with reduced synovial cell hyperplasia and pannus formation (p < 0.05 Mann-Whitney U-test).\nTissue protective effect of hAAT gene therapy in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 6) or saline (n = 4) two weeks before immunization with CII. Control group received saline. (A) Arthritis development was evaluated based on arthritis score (mean + SD). Open circle represent rAAV8-CB-hAAT vector injected group, open triangle represent control group. Mice were sacrificed on day 28 after CII immunization, hind limbs were harvested and processed for histological assessment. *p < 0.05 by Mann-Whitney U-test. (B) Histopathological evaluation of arthritis development. Mice in gene therapy group (black bars) or control group (empty bars) were evaluated according to histopathological changes by two blinded pathologists. Each hind paw was evaluated based on a scale ranging from 0-3. (mean+SD). *p < 0.05, **p < 0.01 by Mann-Whitney U-test. (INF: Infiltration of Immune Cells, HYP: Hyperplasia, P.F.: Pannus Formation, B.D.: Bone Destruction) (C,D) Representative joint section from mice receiving hAAT gene therapy. (E,F) Representative joint section from mice in control group (saline injection). Magnification: C,E: 100x; D,F: 200x.\n[SUBTITLE] Human AAT (hAAT) Gene Therapy Reduced the Levels of Anti-CII Autoantibodies [SUBSECTION] As shown in Figure 5, rAAV8-mediated hAAT gene therapy resulted in a significant suppression of anti-CII autoantibody production. The levels of total Ig anti-bCII (Figure 5A, top left panel) and IgG2a anti-bCII (Figure 5A, top right panel) were significantly reduced in hAAT gene therapy group. Although IgG1 anti-bCII levels (Figure 5A, bottom left panel) were also reduced in hAAT gene therapy group, the ratio of IgG2a anti-bCII to IgG1 anti-bCII (Figure 5A, bottom right panel) significantly decreased in hAAT gene therapy group. Importantly, hAAT gene therapy also reduced levels of autoantibodies against mCII and the ratio of IgG2a anti-mCII to IgG1 anti-mCII (Figure 5B).\nEffect of hAAT gene therapy on auto-antibody production. Anti-CII antibodies at day 28, 42 and 56 were tested by ELISA. Black bars represent the average levels (n = 10, mean+SD) (relative units) of antibodies in hAAT gene therapy treated group. Open bars represent the average levels (n = 10, relative units, mean+SD) of antibodies in saline injected group. (A) Antibody levels against bovine CII (bCII). Top left panel, total Ig antibodies against bCII (total anti-bCII-Ig); Top right panel, levels of IgG2a anti-bCII (anti-bCII-IgG2a); Bottom left panel, levels of IgG1 anti-bCII (anti-bCII-IgG1); Bottom right panel, the ratio of anti-bCII-IgG2a to anti-bCII-IgG1 (anti-bCII-IgG2a/IgG1 ratio). (B) Antibody levels against mouse CII (mCII). Top left panel, total Ig antibodies against mCII (total anti-mCII-Ig); Top right panel, levels of IgG2a anti-mCII (anti-mCII-IgG2a); Bottom left panel, levels of IgG1 anti-mCII (anti-mCII-IgG1); Bottom right panel, the ratio of anti-mCII-IgG2a to anti-mCII-IgG1 (anti-mCII-IgG2a/IgG1). *p < 0.05, **p < 0.01, ***p < 0.001 by Mann-Whitney U- test.\nAs shown in Figure 5, rAAV8-mediated hAAT gene therapy resulted in a significant suppression of anti-CII autoantibody production. The levels of total Ig anti-bCII (Figure 5A, top left panel) and IgG2a anti-bCII (Figure 5A, top right panel) were significantly reduced in hAAT gene therapy group. Although IgG1 anti-bCII levels (Figure 5A, bottom left panel) were also reduced in hAAT gene therapy group, the ratio of IgG2a anti-bCII to IgG1 anti-bCII (Figure 5A, bottom right panel) significantly decreased in hAAT gene therapy group. Importantly, hAAT gene therapy also reduced levels of autoantibodies against mCII and the ratio of IgG2a anti-mCII to IgG1 anti-mCII (Figure 5B).\nEffect of hAAT gene therapy on auto-antibody production. Anti-CII antibodies at day 28, 42 and 56 were tested by ELISA. Black bars represent the average levels (n = 10, mean+SD) (relative units) of antibodies in hAAT gene therapy treated group. Open bars represent the average levels (n = 10, relative units, mean+SD) of antibodies in saline injected group. (A) Antibody levels against bovine CII (bCII). Top left panel, total Ig antibodies against bCII (total anti-bCII-Ig); Top right panel, levels of IgG2a anti-bCII (anti-bCII-IgG2a); Bottom left panel, levels of IgG1 anti-bCII (anti-bCII-IgG1); Bottom right panel, the ratio of anti-bCII-IgG2a to anti-bCII-IgG1 (anti-bCII-IgG2a/IgG1 ratio). (B) Antibody levels against mouse CII (mCII). Top left panel, total Ig antibodies against mCII (total anti-mCII-Ig); Top right panel, levels of IgG2a anti-mCII (anti-mCII-IgG2a); Bottom left panel, levels of IgG1 anti-mCII (anti-mCII-IgG1); Bottom right panel, the ratio of anti-mCII-IgG2a to anti-mCII-IgG1 (anti-mCII-IgG2a/IgG1). *p < 0.05, **p < 0.01, ***p < 0.001 by Mann-Whitney U- test.\n[SUBTITLE] Human AAT Therapy Reduced B-cell Activating Factor (BAFF) in vitro and in vivo [SUBSECTION] In order to further elucidate the underlying mechanism of the anti-arthritic effect of hAAT, we performed additional studies focusing on the effect of AAT on T-cell and B-cell activity. Since CIA is a T-cell-mediated autoimmune disease, the effect of hAAT on T-cell function was examined in a T-cell proliferation assay. As shown in Figure 6A, treatment of rAAV8-hAAT did not change the antigen specific T-cell response after isolated splenocytes were restimulated ex vivo with bCII. Similarly, bCII induced cytokine release (IFN-γ, IL-4, IL-10, TNF-α, IL-2) from isolated splenocytes did not show any significant differences between treatment and control group (Figure 6B). The effect of hAAT therapy on B-cell activity was examined by determination of serum levels of B-cell activating factor of the TNF-α family (BAFF), which has emerged as a crucial factor for B-cell expansion and function. Interestingly, both hAAT protein as well as AAV8 mediated hAAT gene therapy resulted in significantly decreased serum levels of BAFF compared to control group (Figure 6C, 6D). Since BAFF is mainly secreted from monocytes and macrophages, we tested the effect of hAAT on BAFF production in vitro. Murine macrophages (RAW264.7) were treated with hAAT. Culture medium served as control. Protein secretion into the culture medium was determined by ELISA and mRNA expression was quantified by real-time PCR. As shown in Figure 6E, BAFF levels in culture medium were significantly lower in the AAT treated group than those in the control group. Similarly, mRNA expression levels of BAFF were also significantly decreased in AAT treated group (Figure 6F). Together these results suggest that the anti-arthritic effect of AAT is in part through the inhibition of B-cell activation.\nEffects of hAAT therapy on T-cells and B-cells. (A) Proliferative response of splenocytes after stimulation with bovine type II collagen (bCII, 10 μg/ml). Splenocytes (4 × 105 cells/well, in 96-well plate) were isolated on day 28 after rAAV8-hAAT injection. Black bar, AAT gene therapy group (n = 6); open bar, control group (n = 4). Data are expressed as the stimulation index, determined by calculating the ratio of cell proliferation with antigen (measured in counts per minute, cpm) relative to that with medium alone (mean+SD). (B) Cytokine production from bCII-stimulated (100 μg/ml) splenocytes. Values are the mean+SD of each group (n = 6 for rAAV8-hAAT group, black bars; n = 4 for saline group, open bars). (C) Serum level of BAFF in hAAT treated mice (black bar, n = 9, day 35) and control mice (open bar, n = 7). Data is expressed as mean+SD. (D) BAFF serum level in rAAV8-hAAT treated mice (black bar, n = 10, day 28) and control (open bar, n = 10). In vitro effect of hAAT on (E) BAFF secretion into culture medium measured by ELISA and (F) BAFF gene expression determined by real-time PCR. Murine macrophages (RAW 264.7) were treated with hAAT (0.5mg/ml, black bar). Culture medium served as control (open bar). Both experiments were performed in quadruplicates and repeated twice. Data is expressed as mean+SD. *p < 0.05, **p < 0.01.\nIn order to further elucidate the underlying mechanism of the anti-arthritic effect of hAAT, we performed additional studies focusing on the effect of AAT on T-cell and B-cell activity. Since CIA is a T-cell-mediated autoimmune disease, the effect of hAAT on T-cell function was examined in a T-cell proliferation assay. As shown in Figure 6A, treatment of rAAV8-hAAT did not change the antigen specific T-cell response after isolated splenocytes were restimulated ex vivo with bCII. Similarly, bCII induced cytokine release (IFN-γ, IL-4, IL-10, TNF-α, IL-2) from isolated splenocytes did not show any significant differences between treatment and control group (Figure 6B). The effect of hAAT therapy on B-cell activity was examined by determination of serum levels of B-cell activating factor of the TNF-α family (BAFF), which has emerged as a crucial factor for B-cell expansion and function. Interestingly, both hAAT protein as well as AAV8 mediated hAAT gene therapy resulted in significantly decreased serum levels of BAFF compared to control group (Figure 6C, 6D). Since BAFF is mainly secreted from monocytes and macrophages, we tested the effect of hAAT on BAFF production in vitro. Murine macrophages (RAW264.7) were treated with hAAT. Culture medium served as control. Protein secretion into the culture medium was determined by ELISA and mRNA expression was quantified by real-time PCR. As shown in Figure 6E, BAFF levels in culture medium were significantly lower in the AAT treated group than those in the control group. Similarly, mRNA expression levels of BAFF were also significantly decreased in AAT treated group (Figure 6F). Together these results suggest that the anti-arthritic effect of AAT is in part through the inhibition of B-cell activation.\nEffects of hAAT therapy on T-cells and B-cells. (A) Proliferative response of splenocytes after stimulation with bovine type II collagen (bCII, 10 μg/ml). Splenocytes (4 × 105 cells/well, in 96-well plate) were isolated on day 28 after rAAV8-hAAT injection. Black bar, AAT gene therapy group (n = 6); open bar, control group (n = 4). Data are expressed as the stimulation index, determined by calculating the ratio of cell proliferation with antigen (measured in counts per minute, cpm) relative to that with medium alone (mean+SD). (B) Cytokine production from bCII-stimulated (100 μg/ml) splenocytes. Values are the mean+SD of each group (n = 6 for rAAV8-hAAT group, black bars; n = 4 for saline group, open bars). (C) Serum level of BAFF in hAAT treated mice (black bar, n = 9, day 35) and control mice (open bar, n = 7). Data is expressed as mean+SD. (D) BAFF serum level in rAAV8-hAAT treated mice (black bar, n = 10, day 28) and control (open bar, n = 10). In vitro effect of hAAT on (E) BAFF secretion into culture medium measured by ELISA and (F) BAFF gene expression determined by real-time PCR. Murine macrophages (RAW 264.7) were treated with hAAT (0.5mg/ml, black bar). Culture medium served as control (open bar). Both experiments were performed in quadruplicates and repeated twice. Data is expressed as mean+SD. *p < 0.05, **p < 0.01.", "In order to investigate the effect of hAAT on development of arthritis, we first examined the feasibility of hAAT protein therapy in CIA mouse model. Administration of hAAT (0.5 mg/mouse twice per week, starting at 6 days before the induction of arthritis) resulted in sustained high levels of hAAT in mouse serum (Figure 1A). Although anti-hAAT-antibodies were detected (Figure 1B), serum levels of hAAT did not decrease over time.\nAntiarthritic effect of human alpha 1 antitrypsin (hAAT) in collagen induced arthritis (CIA) model. Human AAT (Prolastin®) was intraperitoneally injected in DBA/1 mice (n = 9), twice per week starting 6 days before until day 70 after CII immunization. Control group received saline injections (n = 7) (A) Serum hAAT protein levels in DBA/1 mice were measured by ELISA (mean+SD). ↓ indicates the day of first hAAT injection. (B) Serum anti-hAAT antibody levels (anti-hAAT-IgG) in DBA/1 mice were measured by ELISA. Each dot represents antibody levels (day 49 after bCII immunization, arbitrary units) of an individual mouse. (C) Arthritis score. For each paw, 0 is normal and 4 is the most severe arthritis. The maximum score for each animal is 16. Each line represents the scores from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, *p = 0.029 by AUC analysis) (D) Incidence of severe arthritis is defined by arthritic score/mouse > 3 (**p = 0.0025 by logrank test). Dotted line, saline injected control group; Solid line, hAAT treated group.\nA few days after the second immunization with bCII (day 21), mice in control group developed arthritis in multiple joints, which was manifested by redness, severe joint swelling and joint stiffness as well as ankylosis as the disease progressed. The severity of arthritis as measured by the arthritic score rapidly increased in control group (n = 7) whereas the disease development in hAAT treatment group (n = 9) was suppressed (Figure 1C). At day 49 (7 weeks) after the immunization, area under the curve (AUC) in the hAAT group was 50.83 ± 21.64 (mean ± SEM), while in control group it was 121.5 ± 17.67 (p = 0.029, mean ± SEM, AUC analysis until day 49). Human AAT protein therapy also reduced incidence of severe arthritis (p = 0.0025, logrank test, Figure 1D). Moreover, mice in hAAT treated group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis (arthritis score > 3) started on day 47.3 ± 8.7 (mean ± SD) in hAAT treated group compared to day 36.0 ± 5.8 (mean ± SD) in control group (p = 0.01 by students t-test). Although hAAT treated mice also developed arthritis at the end (70 days after the immunization) of the experiment, these results showed that treatment of hAAT protein (Prolastin®) led to a delayed arthritis onset and amelioration of disease progression in CIA mouse model.", "It has been shown that high levels of serum anti-collagen II autoantibodies are pathognomonic and associated with the development of arthritis [36,37]. To test the effect of hAAT on autoantibody production, we evaluated the levels of anti-CII autoantibodies in total Ig, and IgG1 and IgG2a subclass at early (day 35) and late (day 49) stages of the disease. As shown in Figure 2A, hAAT treatment did not result in a significant change of total autoantibody levels against bCII (total anti-bCII-Ig). However, hAAT treatment significantly reduced the pathognomonic IgG2a (anti-bCII-IgG2a) levels at day 35 (Figure 2B), and increased IgG1 (anti-bCII-IgG1) levels at day 49 (Figure 2C). Interestingly, levels of total Ig autoantibodies against endogenous mouse collagen II (total anti-mCII-Ig) were significantly lower in hAAT protein treated group than those in control group (P < 0.05) (Figure 2D).\nAnti-collagen II (CII) antibody levels after hAAT treatment. Anti-CII antibodies at day 35 and day 49 were tested by ELISA. Closed bars represent the average levels (n = 9, relative units, mean+SD) of antibodies in hAAT protein therapy treated group. Open bars represent the average levels (n = 7, relative units, mean+SD) of antibodies in saline injected group. (A) Levels of total Ig antibodies to bCII (total anti-bCII-Ig). (B) Levels of IgG2a anti-bCII (anti-bCII-IgG2a). (C) Levels of IgG1 anti-bCII (anti-bCII-IgG1). (D) Levels of total Ig antibodies to mCII (total anti-mCII-Ig). * p < 0.05 by Mann-Whitney U- test.", "To further confirm our observation that hAAT is effective in delaying arthritis development, and to test the feasibility of hAAT gene therapy for rheumatoid arthritis, we used recombinant adeno-associated virus vector (rAAV) to deliver the hAAT gene. A single IP injection of rAAV8-CB-hAAT vector (2x1011 particles/mouse, two weeks before the first CII immunization) resulted in sustained levels of hAAT in the circulation, similar to those levels obtained following protein therapy (Figure 3A). Interestingly, following AAV8 mediated gene delivery, we did not observe the development of antibodies to hAAT which were detected during hAAT protein therapy (Figure 3B, compare vs. Figure 1B in mice with hAAT protein therapy). Similar to the results from hAAT protein therapy, however, rAAV-mediated hAAT gene therapy significantly reduced the prevalence of arthritis development at the early stage of disease (Figure 3C). Area under the curve (AUC) in the gene therapy group (n = 10) was 71.65 ± 14.04 (mean ± SEM), while in control group (n = 10) it was 123.20 ± 19.83 (mean ± SEM; p < 0.05 by AUC analysis until day 42). AAT gene therapy also reduced the incidence of severe arthritis (score > 3) at the early stage of disease (p = 0.035 by logrank test, Figure 3D). Moreover, mice in hAAT gene therapy group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis started on day 42.3 ± 7.5 (mean ± SD) in hAAT gene therapy group compared to day 33.4 ± 7.3 in control group (mean ± SD; p < 0.02 by student's t-test). These results indicate that similar to hAAT protein therapy, AAV8 mediated hAAT gene delivery also delayed arthritis onset and ameliorated early stage disease progression in CIA mouse model.\nHuman AAT gene therapy delays disease progression in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 10) or saline (n = 10) two weeks before immunization with CII. Control group received saline. Mice were sacrificed on day 56 (EOS). (A) Serum levels of hAAT. hAAT protein serum levels in vector injected group were measured by ELISA (mean+SD). ↓ indicates the days of injection. (B) Anti-hAAT antibody levels. Serum anti-hAAT antibodies (anti-hAAT) were measured by ELISA using samples obtained at 56 days after immunization. Anti-hAAT antibodies were undetectable in the vector injected group. Each dot represents antibody level (arbitrary units) of an individual mouse. (C) Arthritis score. Each line represents the average score from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, * p < 0.05 as determined by AUC analysis.) (D) Incidence of severe arthritis. Severe arthritis was defined by arthritic score > 3, (* p = 0.035 by logrank test.; 10 mice/group).\nIn an additional experiment using AAV8 mediated hAAT gene therapy, tissue protective properties of hAAT were evaluated. Similar to the previous experiment, mice in treatment group (n = 6) showed significantly reduced arthritis development at the early disease stage compared to control (n = 4) (Figure 4A, p < 0.05 by Mann-Whitney U-test). As shown in Figure 4B-F, AAV8 mediated hAAT gene therapy resulted in less infiltration of immune cells into the joint cavity accompanied with reduced synovial cell hyperplasia and pannus formation (p < 0.05 Mann-Whitney U-test).\nTissue protective effect of hAAT gene therapy in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 6) or saline (n = 4) two weeks before immunization with CII. Control group received saline. (A) Arthritis development was evaluated based on arthritis score (mean + SD). Open circle represent rAAV8-CB-hAAT vector injected group, open triangle represent control group. Mice were sacrificed on day 28 after CII immunization, hind limbs were harvested and processed for histological assessment. *p < 0.05 by Mann-Whitney U-test. (B) Histopathological evaluation of arthritis development. Mice in gene therapy group (black bars) or control group (empty bars) were evaluated according to histopathological changes by two blinded pathologists. Each hind paw was evaluated based on a scale ranging from 0-3. (mean+SD). *p < 0.05, **p < 0.01 by Mann-Whitney U-test. (INF: Infiltration of Immune Cells, HYP: Hyperplasia, P.F.: Pannus Formation, B.D.: Bone Destruction) (C,D) Representative joint section from mice receiving hAAT gene therapy. (E,F) Representative joint section from mice in control group (saline injection). Magnification: C,E: 100x; D,F: 200x.", "As shown in Figure 5, rAAV8-mediated hAAT gene therapy resulted in a significant suppression of anti-CII autoantibody production. The levels of total Ig anti-bCII (Figure 5A, top left panel) and IgG2a anti-bCII (Figure 5A, top right panel) were significantly reduced in hAAT gene therapy group. Although IgG1 anti-bCII levels (Figure 5A, bottom left panel) were also reduced in hAAT gene therapy group, the ratio of IgG2a anti-bCII to IgG1 anti-bCII (Figure 5A, bottom right panel) significantly decreased in hAAT gene therapy group. Importantly, hAAT gene therapy also reduced levels of autoantibodies against mCII and the ratio of IgG2a anti-mCII to IgG1 anti-mCII (Figure 5B).\nEffect of hAAT gene therapy on auto-antibody production. Anti-CII antibodies at day 28, 42 and 56 were tested by ELISA. Black bars represent the average levels (n = 10, mean+SD) (relative units) of antibodies in hAAT gene therapy treated group. Open bars represent the average levels (n = 10, relative units, mean+SD) of antibodies in saline injected group. (A) Antibody levels against bovine CII (bCII). Top left panel, total Ig antibodies against bCII (total anti-bCII-Ig); Top right panel, levels of IgG2a anti-bCII (anti-bCII-IgG2a); Bottom left panel, levels of IgG1 anti-bCII (anti-bCII-IgG1); Bottom right panel, the ratio of anti-bCII-IgG2a to anti-bCII-IgG1 (anti-bCII-IgG2a/IgG1 ratio). (B) Antibody levels against mouse CII (mCII). Top left panel, total Ig antibodies against mCII (total anti-mCII-Ig); Top right panel, levels of IgG2a anti-mCII (anti-mCII-IgG2a); Bottom left panel, levels of IgG1 anti-mCII (anti-mCII-IgG1); Bottom right panel, the ratio of anti-mCII-IgG2a to anti-mCII-IgG1 (anti-mCII-IgG2a/IgG1). *p < 0.05, **p < 0.01, ***p < 0.001 by Mann-Whitney U- test.", "In order to further elucidate the underlying mechanism of the anti-arthritic effect of hAAT, we performed additional studies focusing on the effect of AAT on T-cell and B-cell activity. Since CIA is a T-cell-mediated autoimmune disease, the effect of hAAT on T-cell function was examined in a T-cell proliferation assay. As shown in Figure 6A, treatment of rAAV8-hAAT did not change the antigen specific T-cell response after isolated splenocytes were restimulated ex vivo with bCII. Similarly, bCII induced cytokine release (IFN-γ, IL-4, IL-10, TNF-α, IL-2) from isolated splenocytes did not show any significant differences between treatment and control group (Figure 6B). The effect of hAAT therapy on B-cell activity was examined by determination of serum levels of B-cell activating factor of the TNF-α family (BAFF), which has emerged as a crucial factor for B-cell expansion and function. Interestingly, both hAAT protein as well as AAV8 mediated hAAT gene therapy resulted in significantly decreased serum levels of BAFF compared to control group (Figure 6C, 6D). Since BAFF is mainly secreted from monocytes and macrophages, we tested the effect of hAAT on BAFF production in vitro. Murine macrophages (RAW264.7) were treated with hAAT. Culture medium served as control. Protein secretion into the culture medium was determined by ELISA and mRNA expression was quantified by real-time PCR. As shown in Figure 6E, BAFF levels in culture medium were significantly lower in the AAT treated group than those in the control group. Similarly, mRNA expression levels of BAFF were also significantly decreased in AAT treated group (Figure 6F). Together these results suggest that the anti-arthritic effect of AAT is in part through the inhibition of B-cell activation.\nEffects of hAAT therapy on T-cells and B-cells. (A) Proliferative response of splenocytes after stimulation with bovine type II collagen (bCII, 10 μg/ml). Splenocytes (4 × 105 cells/well, in 96-well plate) were isolated on day 28 after rAAV8-hAAT injection. Black bar, AAT gene therapy group (n = 6); open bar, control group (n = 4). Data are expressed as the stimulation index, determined by calculating the ratio of cell proliferation with antigen (measured in counts per minute, cpm) relative to that with medium alone (mean+SD). (B) Cytokine production from bCII-stimulated (100 μg/ml) splenocytes. Values are the mean+SD of each group (n = 6 for rAAV8-hAAT group, black bars; n = 4 for saline group, open bars). (C) Serum level of BAFF in hAAT treated mice (black bar, n = 9, day 35) and control mice (open bar, n = 7). Data is expressed as mean+SD. (D) BAFF serum level in rAAV8-hAAT treated mice (black bar, n = 10, day 28) and control (open bar, n = 10). In vitro effect of hAAT on (E) BAFF secretion into culture medium measured by ELISA and (F) BAFF gene expression determined by real-time PCR. Murine macrophages (RAW 264.7) were treated with hAAT (0.5mg/ml, black bar). Culture medium served as control (open bar). Both experiments were performed in quadruplicates and repeated twice. Data is expressed as mean+SD. *p < 0.05, **p < 0.01.", "RA is a complex systemic autoimmune disease of unknown etiology. Although recently developed biologics that target TNF-alpha have provided dramatic improvement in controlling disease activity in many patients, continued searches for more efficient and safer treatments are still needed. In the present study we showed that hAAT, administered as protein or through rAAV8 mediated gene therapy, reduced levels of serum anti-CII auto-antibodies and B-cell activating factor (BAFF) and significantly delayed arthritis development in a mouse model.\nAlthough the exact mechanisms underlying the therapeutic effect remain to be further investigated, several mechanisms may be involved. One is through the inhibition of proinflammatory cytokine production. It is well known that various proinflammatory cytokines, including TNF-α and IL1-β, play major roles in the pathogenesis of RA [3]. Strategies targeting these cytokines have proven to be effective in treatment of RA [38]. Previous work done by Janciauskiene and her colleagues clearly demonstrated that hAAT inhibited LPS-induced TNF-α, IL-6 and IL-1β production by human monocytes [15,16]. In addition, hAAT completely suppressed macrophage inflammatory protein-2 (MIP-2)/monocyte chemotactic protein-1 (MCP-1) gene expression in lung [39]. Human AAT also enhanced anti-inflammatory cytokine IL-10 production from monocytes [15]. As a consequence of interfering with the cytokine/chemokine network, hAAT may also inhibit polymorphonuclear leukocyte (PMN) invasion into the joint. Churg et al. demonstrated that hAAT inhibited silica-induced PMN influx into the lung and partially suppressed nuclear transcription factor B (NF-κB) translocation and increased inhibitor of NF-κB (I-࿠κB) levels in a mouse model of acute PMN mediated inflammation [39]. Thus, it is possible that the effects of hAAT on pro-inflammatory cytokine production contribute to suppression of autoimmune-mediated inflammation.\nIn previous studies we showed that hAAT reduced anti-insulin auto-antibodies (IAA) and attenuated cell-mediated autoimmunity [32,33]. Consistent with these results, the present study showed that hAAT reduced the levels of anti-CII auto-antibodies and the IgG2a/IgG1 ratios of anti-CII auto-antibodies (mCII and bCII). We have observed that the effect of hAAT to suppress arthritis development is more profound in early stage of arthritis development. This is supported by the effect of hAAT on pathognomonic IgG2a antibody development at early time points (Fig.2) as well as the observation that mice eventually develop arthritis overtime. Therefore, hAAT maybe especially suitable for combination therapies. We did not observe significant effect of AAT on T-cell proliferation and cytokine production in vitro (Figure 6A and 6B) indicating that AAT may have limited direct effect on T-cells. These data also suggest that AAT may more directly affect B-cell activity. Indeed, we have shown that AAT therapies significantly reduced B-cell activating factor of the TNF-α family (BAFF) in vitro and in vivo. BAFF is an important factor that modulates B-cell tolerance and homeostasis. It has been shown that soluble BAFF is elevated in serum and target organs of CIA model [40] and BAFF antagonists suppressed arthritis development in murine models of rheumatoid arthritis [41]. In addition, increased BAFF levels were found in serum of RA patients which correlated with serum levels of rheumatoid factor [42]. The exact mechanism that AAT suppresses BAFF production remains to be elucidated.\nAnother possible mechanism of hAAT suppressing arthritis development is through inhibition of proteinases to prevent tissue injury and joint destruction. Human AAT is well known as a serine proteinase inhibitor (serpin). It inhibits proteinase 3, neutrophil elastase, and cathepsin G. These serine proteases are released by joint invading neutrophils following inflammatory stimuli and have shown to be involved in arthritis development [12,13,43,44]. Human AAT can also reduce ischemia-induced apoptosis, inflammation, and acute phase response in the kidney [20]. We have recently shown that hAAT directly inhibits caspase 3 activity and protects islet cells from cytokine and chemically-induced apoptosis [45].\nIn the protein therapy studies, we used Prolastin®, which is clinical grade hAAT purified from human plasma. Repeated IP injection of hAAT induced strong humoral immune response against hAAT in DBA/1 mice (Figure 1B), similar to what has been observed in previous studies [46,47]. It is possible that non-specific inflammation caused by repeated IP injection is responsible for inhibition of arthritis. In order to rule out this possibility, we performed rAAV8 mediated hAAT gene therapy. AAV serotype 8 vector is unique for this purpose because it can mediate long term and high levels of transgene expression in the liver and muscle, but is not able to transduce dendritic cells and has low immunogenicity [48,49]. Indeed, after a single injection of rAAV8-CB-hAAT vector, sustained high levels of hAAT were detected in the circulation, while no detectable levels of anti-hAAT antibodies were present (Figure 3B) in contrast to mice that received hAAT protein therapy. These results are consistent with our recent observations in NOD mice and imply new applications of rAAV8 vectors [34]. The detailed mechanism that rAAV8 vector mediates no immune response to the transgene product remains elusive. Importantly, we have observed protective effects and reductions of auto-antibodies by hAAT gene therapy. These results strongly support our hypothesis that hAAT is able to reduce inflammation in autoimmune diseases, such as RA and type 1 diabetes.", "Our results from protein and gene therapy showed that hAAT is effective in delaying arthritis development in a mouse model of CIA. They indicate that hAAT has immunoregulatory and immunomodulatory effects and has great potential as a new treatment for RA. We also have shown that rAAV8 mediated gene therapy resulted in a reduced immune response to the transgene product. Future studies will focus on improvement of the therapeutic effect by optimizing the dose and timing of hAAT or rAAV8 vector delivery, and by combination therapy with other anti-arthritic drugs.", "hAAT: human Alpha-1 Antitrypsin; CIA: Collagen Induced Arthritis; IFA: Incomplete Freund's Adjuvant; CFA: Complete Freund's Adjuvant; RA: Rheumatoid Arthritis; NOD: Non Obese Diabetic; bCII: bovine type II Collagen; mCII: mouse type II Collagen; TNF-α: Tumor Necrosis Factor-alpha; IL: Interleukin; LPS: Lipopolysaccharide; PBMC: Peripheral Blood Mononuclear Cells; BAFF: B-cell Activation Factor of the TNF-α Family; rAAV: Recombinant Adeno-Associated Virus; MMP: Matrix- Metalloproteinase; ELISA: Enzyme-Linked Immunosorbent Assay", "Christian Grimstein and Sihong Song may be entitled to future patent royalties from technology described in this paper.", "CG conceived of the study, participated in its design, carried out animal experiments, cell proliferation assay, immunoassays, performed statistical analysis and drafted the manuscript. YKC conceived of the study, participated in its design and performed animal experiments and cell proliferation assay. CW helped performing cell proliferation assay, MS participated in discussion and helped to revise the manuscript, MA, MCT and MB participated in design and discussion of the study, SS conceived of the study participated in its design and helped to revise the manuscript. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "rAAV Vector Production", "Animals", "Histological Assessment", "Human AAT Protein and rAAV8-CB-AAT Vector Administration", "ELISA for the Detection of Serum hAAT and BAFF Levels and Antibodies against hAAT, bCII and mCII", "Cell Culture", "Quantitative PCR", "Assessment of T-cell Autoreactive Response", "Statistical Analysis", "Results", "Human AAT Protein Therapy Delayed Arthritis Development in DBA/1 Mice", "Human AAT Protein Therapy Reduced the Levels of anti-bCII and anti-mCII Autoantibodies", "Human AAT (hAAT) Gene Therapy delayed Arthritis Development", "Human AAT (hAAT) Gene Therapy Reduced the Levels of Anti-CII Autoantibodies", "Human AAT Therapy Reduced B-cell Activating Factor (BAFF) in vitro and in vivo", "Discussion", "Conclusion", "Abbreviations", "Competing interests", "Authors' contributions" ]

[ "Rheumatoid arthritis (RA) is a systemic autoimmune disease, characterized by chronic joint inflammation and synovial hyperplasia leading to bone and joint destruction. The life expectancy is lowered and quality of life is decreased in RA patients. So far little is known about the actual disease initiating stimulus; however, extensive research over the last decades have shown that multiple genetic as well as environmental factors interact and trigger the onset of RA [1,2]. The autoimmune inflammation of RA is maintained by inappropriate action of macrophages, B-cells, T-cells, and other types of cells leading to dysregulated cytokine/chemokine production. The synovial inflammation is caused by infiltration and proliferation of activated immune cells including neutrophils, macrophages, fibroblasts, mast cells, NK cells, NKT cells, T-cells as well as plasma cells [3]. Progressive joint and bone destruction is mediated through the activities of osteoclasts, chondrocytes, synovial fibroblasts and cytokine induction of destructive enzymes, chiefly matrix metalloproteinases (MMP) [4]. Current therapy mainly aims to inhibit the biological function of tumor necrosis factor-alpha (TNF-α) and lymphocyte proliferation. Due to ineffectiveness of anti-TNF-α therapy in certain patients and various side effects of methotrexate which inhibits lymphocytes proliferation, there is still the need to identify new target molecules/pathways and to develop new treatment [5]. Immunoregulatory and anti-inflammatory strategies that affect B-cell activation, T-cell activation or inhibit proinflammatory cytokines have recently shown great potential for the treatment of RA [5,6].\nHuman alpha-1 antitrypsin (hAAT) is a 52 kDa serum glycoprotein, synthesized primarily in the liver. It is also expressed in other types of cells including neutrophils, monocytes, macrophages, alveolar macrophages, intestinal epithelial cells, carcinoma cells and the cornea [7-10]. The normal serum level of hAAT is 1-2mg/ml. During inflammation, hAAT level, as an acute phase reactant, can increase 3-4 folds, suggesting an important role in responding to inflammation in the human body. Increasing evidence indicates that hAAT is immunoregulatory, anti-inflammatory and may be used for the treatment of RA. It inhibits neutrophil elastase and proteinase 3 with high efficiency, as well as cathepsin G, thrombin, trypsin and chymotrypsin with lower efficiency [11]. Most of these proteases target receptor proteins, involved in proinflammatory cytokine expression and cell signaling [12]. It also has been reported that neutrophil elastase inhibitors reduce incidence as well as severity of collagen-induced arthritis (CIA) in both rats and mice [13]. Human AAT is able to completely eliminate the acute inflammatory infiltration and connective tissue breakdown in the lung in a cigarette smoke-induced emphysema mouse model [14]. It also inhibits lipopolysaccharide (LPS)-stimulated release of TNF-α and interleukin (IL) -1β, and enhances the production of anti-inflammatory cytokine IL-10 [15-17]. Human AAT significantly protects against the lethality induced by TNF-α or endotoxin in mice [18]. It can also induce expression of IL1-Ra in human peripheral blood mononuclear cells (PBMC's) [19] and reduces ischemia-induced apoptosis and inflammation [20]. We have recently shown, that combination therapy using doxycycline and hAAT gene therapy reduces arthritis development in mice, suggesting a therapeutic effect of hAAT in an arthritis mouse model [21].\nRecombinant adeno-associated virus vectors (rAAV) have been widely used for gene therapy in animal models and human clinical trials [22], because of their unique features in safety and efficiency. It has been reported that rAAV mediated long-term and high levels of transgene expression in a wide variety of tissues, including muscle [23], lung [24], liver [25], brain [26] and eye [27]. Recently developed rAAV vectors including new serotypes of AAV, mutants AAV and double stranded AAV have provided more opportunities and challenges for their application [28-31]. Previously, we have shown hAAT gene therapy using rAAV2 and rAAV1 vectors prevented type 1 diabetes. However, the immune response to the transgene product (hAAT) complicated the therapeutic effect [32,33]. We have recently discovered that rAAV8 vector fail to transduce dendritic cells and induce immune tolerance to transgene product entailing rAAV8 as a promising vector used for therapeutic intervention [34].\nIn the present study we further investigated the feasibility of hAAT with its anti-inflammatory and immunoregulatory properties for the treatment of RA using both, protein therapy and rAAV8 mediated gene therapy.", "[SUBTITLE] rAAV Vector Production [SUBSECTION] The rAAV-CB-hAAT vector construct was produced and packaged as previously described [27]. Briefly, this vector carries hAAT cDNA driven by the cytomegalovirus (CMV) enhancer and chicken β-actin promoter and contains AAV2 inverted terminal repeats (ITRs). It was packaged into AAV serotype 8 capsid by cotransfection of vector plasmid and helper plasmid (XYZ8) into 293 cells. rAAV8-CB-hAAT vectors were purified by iodixanol gradient centrifugation followed by anion-exchange chromatography. The physical particle titers of vector preparations were assessed by dot blot analysis.\nThe rAAV-CB-hAAT vector construct was produced and packaged as previously described [27]. Briefly, this vector carries hAAT cDNA driven by the cytomegalovirus (CMV) enhancer and chicken β-actin promoter and contains AAV2 inverted terminal repeats (ITRs). It was packaged into AAV serotype 8 capsid by cotransfection of vector plasmid and helper plasmid (XYZ8) into 293 cells. rAAV8-CB-hAAT vectors were purified by iodixanol gradient centrifugation followed by anion-exchange chromatography. The physical particle titers of vector preparations were assessed by dot blot analysis.\n[SUBTITLE] Animals [SUBSECTION] Six week-old male DBA/1 mice were purchased from Harlan Sprague Dawley, Inc. (Indianapolis, IN), housed in a specific pathogen-free room as approved by the University of Florida Institutional Animal Care and Use Committee. For induction of arthritis, bCII (Chondrex LLC, Redmond, WA) was dissolved in 0.05N acetic acid at a concentration of 2mg/ml by stirring overnight at 4°C and was emulsified with an equal volume of Complete Freund's Adjuvant (CFA) (Chondrex LLC, Redmond, WA). At the age of eight weeks, DBA/1 mice were immunized intradermally at the base of the tail with 0.1ml of emulsion containing 100 μg of type II collagen. Three weeks after priming (day 21), the mice were boosted with 0.1 ml of bCII (100 μg) emulsified in equal volume of incomplete Freund's Adjuvant (IFA) (Difco, Detroit, MI). For assessment of arthritis, all mice were monitored three times a week by the same person blinded to the treatment group and evaluated the incidence of arthritis and clinical score. An arthritis score system ranging from stage 0 - 4 was used: 0: no swelling or redness; 1: detectable arthritis with erythema; 2: significant swelling and redness; 3: severe swelling and redness from joint to digit; 4: joint stiffness or deformity with ankylosis [35]. The clinical score was expressed as the average cumulative value of all four paws with a maximum score per animal of 16. Severe arthritis was defined as arthritis score > 3 for the purpose of comparing data between groups.\nSix week-old male DBA/1 mice were purchased from Harlan Sprague Dawley, Inc. (Indianapolis, IN), housed in a specific pathogen-free room as approved by the University of Florida Institutional Animal Care and Use Committee. For induction of arthritis, bCII (Chondrex LLC, Redmond, WA) was dissolved in 0.05N acetic acid at a concentration of 2mg/ml by stirring overnight at 4°C and was emulsified with an equal volume of Complete Freund's Adjuvant (CFA) (Chondrex LLC, Redmond, WA). At the age of eight weeks, DBA/1 mice were immunized intradermally at the base of the tail with 0.1ml of emulsion containing 100 μg of type II collagen. Three weeks after priming (day 21), the mice were boosted with 0.1 ml of bCII (100 μg) emulsified in equal volume of incomplete Freund's Adjuvant (IFA) (Difco, Detroit, MI). For assessment of arthritis, all mice were monitored three times a week by the same person blinded to the treatment group and evaluated the incidence of arthritis and clinical score. An arthritis score system ranging from stage 0 - 4 was used: 0: no swelling or redness; 1: detectable arthritis with erythema; 2: significant swelling and redness; 3: severe swelling and redness from joint to digit; 4: joint stiffness or deformity with ankylosis [35]. The clinical score was expressed as the average cumulative value of all four paws with a maximum score per animal of 16. Severe arthritis was defined as arthritis score > 3 for the purpose of comparing data between groups.\n[SUBTITLE] Histological Assessment [SUBSECTION] For the analysis of arthritis, mice were anesthetized and sacrificed by cervical dislocation on day 28 after immunization. The two hind limbs of mice in treatment and control groups were removed. Specimens were fixed in formalin and decalcified in RDO solution (Apex, Aurora, IL) for 10-20 min depending on tissue size and then checked manually for pliability. Sections 4 μm thick were cut and stained with hematoxylin and eosin according to standard methods.\nHistological evaluation was performed by two independent and blinded pathologists. Infiltration of immune cells, hyperplasia, pannus formation and bone deformation was determined for each paw using an evaluation scale ranging from 0-3 according to severity of pathohistological changes. (0: normal, 1: mild, 2: moderate, 3: severe).\nFor the analysis of arthritis, mice were anesthetized and sacrificed by cervical dislocation on day 28 after immunization. The two hind limbs of mice in treatment and control groups were removed. Specimens were fixed in formalin and decalcified in RDO solution (Apex, Aurora, IL) for 10-20 min depending on tissue size and then checked manually for pliability. Sections 4 μm thick were cut and stained with hematoxylin and eosin according to standard methods.\nHistological evaluation was performed by two independent and blinded pathologists. Infiltration of immune cells, hyperplasia, pannus formation and bone deformation was determined for each paw using an evaluation scale ranging from 0-3 according to severity of pathohistological changes. (0: normal, 1: mild, 2: moderate, 3: severe).\n[SUBTITLE] Human AAT Protein and rAAV8-CB-AAT Vector Administration [SUBSECTION] For hAAT protein therapy studies, DBA/1 mice were intraperitoneally (IP) injected with 0.5 mg (in 100 μl saline) of hAAT (Prolastin®, Bayer Corp., Elkhard, IN). The control group received saline injection. The injections were performed twice per week, starting at 6 days before the first bCII immunization until the end of study (EOS) at day 70 after the first immunization. For hAAT gene therapy studies, DBA/1 mice were IP injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse) two weeks before the first CII immunization. The control group received saline injection.\nFor hAAT protein therapy studies, DBA/1 mice were intraperitoneally (IP) injected with 0.5 mg (in 100 μl saline) of hAAT (Prolastin®, Bayer Corp., Elkhard, IN). The control group received saline injection. The injections were performed twice per week, starting at 6 days before the first bCII immunization until the end of study (EOS) at day 70 after the first immunization. For hAAT gene therapy studies, DBA/1 mice were IP injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse) two weeks before the first CII immunization. The control group received saline injection.\n[SUBTITLE] ELISA for the Detection of Serum hAAT and BAFF Levels and Antibodies against hAAT, bCII and mCII [SUBSECTION] Detection of hAAT and anti-hAAT antibodies in mouse serum was performed as previously described [32]. Purified hAAT (Athens Research & Technology, Athens, GA) was used as a standard. Anti-type II collagen antibodies in mouse serum were detected by a standard ELISA. Briefly, microtiter plates (Immulon 4, Dynex Technologies, Chantilly, VA) were coated with bCII or mCII (0.5 μg/well, Chondrex LLC, Redmond, WA) in Voller's buffer overnight at 4°C. After blocking with 3% bovine serum albumin, wells were incubated with samples at room temperature for 2 h. HRP-conjugated goat anti-mouse IgG antibodies (1:1,000 dilution, Sigma, St. Louis, MO), goat anti-mouse IgG1 antibodies (1:1,500 dilution, Roche, Indianapolis, IN) and goat anti-mouse IgG2a antibodies (1:1,500 dilution, Roche, Indianapolis, IN) were incubated for 1 h at RT. The plates were washed with PBS-Tween 20 between reactions. After adding the substrate (o-phenylenediamine, Sigma, St Louis, MO), plates were read at 490 nm on an MRX microplate reader (Dynex Technologies, Chantilly, VA). Optical densities were converted into units based on a standard curve generated with high titer sera from DBA/1 mice immunized with bCII. Detection of BAFF in serum was performed according to manufactures instructions (R&D systems, Inc. Minneapolis, MN).\nDetection of hAAT and anti-hAAT antibodies in mouse serum was performed as previously described [32]. Purified hAAT (Athens Research & Technology, Athens, GA) was used as a standard. Anti-type II collagen antibodies in mouse serum were detected by a standard ELISA. Briefly, microtiter plates (Immulon 4, Dynex Technologies, Chantilly, VA) were coated with bCII or mCII (0.5 μg/well, Chondrex LLC, Redmond, WA) in Voller's buffer overnight at 4°C. After blocking with 3% bovine serum albumin, wells were incubated with samples at room temperature for 2 h. HRP-conjugated goat anti-mouse IgG antibodies (1:1,000 dilution, Sigma, St. Louis, MO), goat anti-mouse IgG1 antibodies (1:1,500 dilution, Roche, Indianapolis, IN) and goat anti-mouse IgG2a antibodies (1:1,500 dilution, Roche, Indianapolis, IN) were incubated for 1 h at RT. The plates were washed with PBS-Tween 20 between reactions. After adding the substrate (o-phenylenediamine, Sigma, St Louis, MO), plates were read at 490 nm on an MRX microplate reader (Dynex Technologies, Chantilly, VA). Optical densities were converted into units based on a standard curve generated with high titer sera from DBA/1 mice immunized with bCII. Detection of BAFF in serum was performed according to manufactures instructions (R&D systems, Inc. Minneapolis, MN).\n[SUBTITLE] Cell Culture [SUBSECTION] The murine macrophage cell line RAW 264.7 was cultured in serum free DMEM at 37°C in a 5% CO2 incubator. For measuring BAFF release into medium, cells were seeded at 1 × 105/ml in 12 well plates. Cells were incubated in quadruplicates with hAAT (0.5mg/ml; Prolastin®, Bayer Corp., Elkhard, IN) for 16 hours and BAFF secretion into the culture medium was determined by ELISA according to manufactures instructions (R&D systems, Inc. Minneapolis, MN).\nThe murine macrophage cell line RAW 264.7 was cultured in serum free DMEM at 37°C in a 5% CO2 incubator. For measuring BAFF release into medium, cells were seeded at 1 × 105/ml in 12 well plates. Cells were incubated in quadruplicates with hAAT (0.5mg/ml; Prolastin®, Bayer Corp., Elkhard, IN) for 16 hours and BAFF secretion into the culture medium was determined by ELISA according to manufactures instructions (R&D systems, Inc. Minneapolis, MN).\n[SUBTITLE] Quantitative PCR [SUBSECTION] Total RNA from cell culture described above, was isolated using RNeasy Mini Kit (Quiagen, Valencia, CA). Samples were processed according to the manufacture's protocol. For reverse transcription, cDNA was synthesized with oligo dT16 primers and Moloney Murine Leukemia Virus Reverse Transcriptase (MMLV-RT) according to manufacture's manual (Taqman Reverse Transcription Reagents, Applied Biosystems, Foster City, CA).\ncDNA was analyzed by quantitative PCR using gene-specific primers with SYBR Green 2X PCR mix (Applied Biosystems). The sequence of the primers were as follows: BAFF (205bp), sense: 5'-TGC CTT GGA GGA GAA AGA GA-3' and antisense: 5'-GGA ATT GTT GGG CAG TGT TT-3'; GAPDH (122bp), sense: 5'-CCT GGA GAA ACC TGC CAA GTA T-3' and antisense: 5'-TGC TGT TGA AGT CGC AGG A-3'. Reactions were set up in triplicate and performed on the ABI Prism 7700 Sequence Detector (Applied Biosystems). The cycling parameters were 2 min at 95°C for denaturation, 40 cycles of 15s at 95°C and 30 s at 60°C for amplification. The threshold cycle (CT) of each target product was determined, set to the log linear range of the amplification curve and kept constant for all data analysis. Data were analyzed with Sequence Detector Software (SDS). BAFF expression was normalized to the corresponding GAPDH values for the respective treatment. Values of BAFF expression following saline treatment are designated as 1. The experiment was repeated twice.\nTotal RNA from cell culture described above, was isolated using RNeasy Mini Kit (Quiagen, Valencia, CA). Samples were processed according to the manufacture's protocol. For reverse transcription, cDNA was synthesized with oligo dT16 primers and Moloney Murine Leukemia Virus Reverse Transcriptase (MMLV-RT) according to manufacture's manual (Taqman Reverse Transcription Reagents, Applied Biosystems, Foster City, CA).\ncDNA was analyzed by quantitative PCR using gene-specific primers with SYBR Green 2X PCR mix (Applied Biosystems). The sequence of the primers were as follows: BAFF (205bp), sense: 5'-TGC CTT GGA GGA GAA AGA GA-3' and antisense: 5'-GGA ATT GTT GGG CAG TGT TT-3'; GAPDH (122bp), sense: 5'-CCT GGA GAA ACC TGC CAA GTA T-3' and antisense: 5'-TGC TGT TGA AGT CGC AGG A-3'. Reactions were set up in triplicate and performed on the ABI Prism 7700 Sequence Detector (Applied Biosystems). The cycling parameters were 2 min at 95°C for denaturation, 40 cycles of 15s at 95°C and 30 s at 60°C for amplification. The threshold cycle (CT) of each target product was determined, set to the log linear range of the amplification curve and kept constant for all data analysis. Data were analyzed with Sequence Detector Software (SDS). BAFF expression was normalized to the corresponding GAPDH values for the respective treatment. Values of BAFF expression following saline treatment are designated as 1. The experiment was repeated twice.\n[SUBTITLE] Assessment of T-cell Autoreactive Response [SUBSECTION] To test the effect of AAV8-hAAT gene therapy on splenocyte proliferation, spleens were harvested at 30 days after the first bCII immunization. Splenocytes were isolated and cultured in serum free X-VIVO medium (Cambrex, Walkersville, MD) in the presence or absence of bCII (100 μg/ml, Chondrex LLC, Redmond, WA). After 3 days culture, 1 μCi/well of [3H] TdR was added. Cells were cultured for additional 18h and [3H] TdR uptake was measured using a β- scintillation counter.\nTo measure cytokine release into the cell culture supernatant, a Beadlyte Mouse Multi-Cytokine Detection System 1 kit (Upstate, Temecula, CA, Cat # 48-005) was used according to the manufacture's instruction and in conjunction with the Luminex 100 system for cytokine determination.\nTo test the effect of AAV8-hAAT gene therapy on splenocyte proliferation, spleens were harvested at 30 days after the first bCII immunization. Splenocytes were isolated and cultured in serum free X-VIVO medium (Cambrex, Walkersville, MD) in the presence or absence of bCII (100 μg/ml, Chondrex LLC, Redmond, WA). After 3 days culture, 1 μCi/well of [3H] TdR was added. Cells were cultured for additional 18h and [3H] TdR uptake was measured using a β- scintillation counter.\nTo measure cytokine release into the cell culture supernatant, a Beadlyte Mouse Multi-Cytokine Detection System 1 kit (Upstate, Temecula, CA, Cat # 48-005) was used according to the manufacture's instruction and in conjunction with the Luminex 100 system for cytokine determination.\n[SUBTITLE] Statistical Analysis [SUBSECTION] Data Analysis was performed using GraphPad Prism 4.0 (GraphPad Software) and SAS (SAS Institute). Student's t-test was used to compare differences in BAFF levels in culture medium as well as differences in mRNA expression levels. Mann-Whitney U-test was applied to analyze differences in stimulation indices, cytokine levels, pathohistological changes, serum levels of BAFF and antibodies. For comparison of arthritis score, area under the curve analysis was used and differences in arthritis incidence were determined using Kaplan-Meier survival curve and log-rank test. A p-value of p ≤ 0.05 was considered statistically significant.\nData Analysis was performed using GraphPad Prism 4.0 (GraphPad Software) and SAS (SAS Institute). Student's t-test was used to compare differences in BAFF levels in culture medium as well as differences in mRNA expression levels. Mann-Whitney U-test was applied to analyze differences in stimulation indices, cytokine levels, pathohistological changes, serum levels of BAFF and antibodies. For comparison of arthritis score, area under the curve analysis was used and differences in arthritis incidence were determined using Kaplan-Meier survival curve and log-rank test. A p-value of p ≤ 0.05 was considered statistically significant.", "The rAAV-CB-hAAT vector construct was produced and packaged as previously described [27]. Briefly, this vector carries hAAT cDNA driven by the cytomegalovirus (CMV) enhancer and chicken β-actin promoter and contains AAV2 inverted terminal repeats (ITRs). It was packaged into AAV serotype 8 capsid by cotransfection of vector plasmid and helper plasmid (XYZ8) into 293 cells. rAAV8-CB-hAAT vectors were purified by iodixanol gradient centrifugation followed by anion-exchange chromatography. The physical particle titers of vector preparations were assessed by dot blot analysis.", "Six week-old male DBA/1 mice were purchased from Harlan Sprague Dawley, Inc. (Indianapolis, IN), housed in a specific pathogen-free room as approved by the University of Florida Institutional Animal Care and Use Committee. For induction of arthritis, bCII (Chondrex LLC, Redmond, WA) was dissolved in 0.05N acetic acid at a concentration of 2mg/ml by stirring overnight at 4°C and was emulsified with an equal volume of Complete Freund's Adjuvant (CFA) (Chondrex LLC, Redmond, WA). At the age of eight weeks, DBA/1 mice were immunized intradermally at the base of the tail with 0.1ml of emulsion containing 100 μg of type II collagen. Three weeks after priming (day 21), the mice were boosted with 0.1 ml of bCII (100 μg) emulsified in equal volume of incomplete Freund's Adjuvant (IFA) (Difco, Detroit, MI). For assessment of arthritis, all mice were monitored three times a week by the same person blinded to the treatment group and evaluated the incidence of arthritis and clinical score. An arthritis score system ranging from stage 0 - 4 was used: 0: no swelling or redness; 1: detectable arthritis with erythema; 2: significant swelling and redness; 3: severe swelling and redness from joint to digit; 4: joint stiffness or deformity with ankylosis [35]. The clinical score was expressed as the average cumulative value of all four paws with a maximum score per animal of 16. Severe arthritis was defined as arthritis score > 3 for the purpose of comparing data between groups.", "For the analysis of arthritis, mice were anesthetized and sacrificed by cervical dislocation on day 28 after immunization. The two hind limbs of mice in treatment and control groups were removed. Specimens were fixed in formalin and decalcified in RDO solution (Apex, Aurora, IL) for 10-20 min depending on tissue size and then checked manually for pliability. Sections 4 μm thick were cut and stained with hematoxylin and eosin according to standard methods.\nHistological evaluation was performed by two independent and blinded pathologists. Infiltration of immune cells, hyperplasia, pannus formation and bone deformation was determined for each paw using an evaluation scale ranging from 0-3 according to severity of pathohistological changes. (0: normal, 1: mild, 2: moderate, 3: severe).", "For hAAT protein therapy studies, DBA/1 mice were intraperitoneally (IP) injected with 0.5 mg (in 100 μl saline) of hAAT (Prolastin®, Bayer Corp., Elkhard, IN). The control group received saline injection. The injections were performed twice per week, starting at 6 days before the first bCII immunization until the end of study (EOS) at day 70 after the first immunization. For hAAT gene therapy studies, DBA/1 mice were IP injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse) two weeks before the first CII immunization. The control group received saline injection.", "Detection of hAAT and anti-hAAT antibodies in mouse serum was performed as previously described [32]. Purified hAAT (Athens Research & Technology, Athens, GA) was used as a standard. Anti-type II collagen antibodies in mouse serum were detected by a standard ELISA. Briefly, microtiter plates (Immulon 4, Dynex Technologies, Chantilly, VA) were coated with bCII or mCII (0.5 μg/well, Chondrex LLC, Redmond, WA) in Voller's buffer overnight at 4°C. After blocking with 3% bovine serum albumin, wells were incubated with samples at room temperature for 2 h. HRP-conjugated goat anti-mouse IgG antibodies (1:1,000 dilution, Sigma, St. Louis, MO), goat anti-mouse IgG1 antibodies (1:1,500 dilution, Roche, Indianapolis, IN) and goat anti-mouse IgG2a antibodies (1:1,500 dilution, Roche, Indianapolis, IN) were incubated for 1 h at RT. The plates were washed with PBS-Tween 20 between reactions. After adding the substrate (o-phenylenediamine, Sigma, St Louis, MO), plates were read at 490 nm on an MRX microplate reader (Dynex Technologies, Chantilly, VA). Optical densities were converted into units based on a standard curve generated with high titer sera from DBA/1 mice immunized with bCII. Detection of BAFF in serum was performed according to manufactures instructions (R&D systems, Inc. Minneapolis, MN).", "The murine macrophage cell line RAW 264.7 was cultured in serum free DMEM at 37°C in a 5% CO2 incubator. For measuring BAFF release into medium, cells were seeded at 1 × 105/ml in 12 well plates. Cells were incubated in quadruplicates with hAAT (0.5mg/ml; Prolastin®, Bayer Corp., Elkhard, IN) for 16 hours and BAFF secretion into the culture medium was determined by ELISA according to manufactures instructions (R&D systems, Inc. Minneapolis, MN).", "Total RNA from cell culture described above, was isolated using RNeasy Mini Kit (Quiagen, Valencia, CA). Samples were processed according to the manufacture's protocol. For reverse transcription, cDNA was synthesized with oligo dT16 primers and Moloney Murine Leukemia Virus Reverse Transcriptase (MMLV-RT) according to manufacture's manual (Taqman Reverse Transcription Reagents, Applied Biosystems, Foster City, CA).\ncDNA was analyzed by quantitative PCR using gene-specific primers with SYBR Green 2X PCR mix (Applied Biosystems). The sequence of the primers were as follows: BAFF (205bp), sense: 5'-TGC CTT GGA GGA GAA AGA GA-3' and antisense: 5'-GGA ATT GTT GGG CAG TGT TT-3'; GAPDH (122bp), sense: 5'-CCT GGA GAA ACC TGC CAA GTA T-3' and antisense: 5'-TGC TGT TGA AGT CGC AGG A-3'. Reactions were set up in triplicate and performed on the ABI Prism 7700 Sequence Detector (Applied Biosystems). The cycling parameters were 2 min at 95°C for denaturation, 40 cycles of 15s at 95°C and 30 s at 60°C for amplification. The threshold cycle (CT) of each target product was determined, set to the log linear range of the amplification curve and kept constant for all data analysis. Data were analyzed with Sequence Detector Software (SDS). BAFF expression was normalized to the corresponding GAPDH values for the respective treatment. Values of BAFF expression following saline treatment are designated as 1. The experiment was repeated twice.", "To test the effect of AAV8-hAAT gene therapy on splenocyte proliferation, spleens were harvested at 30 days after the first bCII immunization. Splenocytes were isolated and cultured in serum free X-VIVO medium (Cambrex, Walkersville, MD) in the presence or absence of bCII (100 μg/ml, Chondrex LLC, Redmond, WA). After 3 days culture, 1 μCi/well of [3H] TdR was added. Cells were cultured for additional 18h and [3H] TdR uptake was measured using a β- scintillation counter.\nTo measure cytokine release into the cell culture supernatant, a Beadlyte Mouse Multi-Cytokine Detection System 1 kit (Upstate, Temecula, CA, Cat # 48-005) was used according to the manufacture's instruction and in conjunction with the Luminex 100 system for cytokine determination.", "Data Analysis was performed using GraphPad Prism 4.0 (GraphPad Software) and SAS (SAS Institute). Student's t-test was used to compare differences in BAFF levels in culture medium as well as differences in mRNA expression levels. Mann-Whitney U-test was applied to analyze differences in stimulation indices, cytokine levels, pathohistological changes, serum levels of BAFF and antibodies. For comparison of arthritis score, area under the curve analysis was used and differences in arthritis incidence were determined using Kaplan-Meier survival curve and log-rank test. A p-value of p ≤ 0.05 was considered statistically significant.", "[SUBTITLE] Human AAT Protein Therapy Delayed Arthritis Development in DBA/1 Mice [SUBSECTION] In order to investigate the effect of hAAT on development of arthritis, we first examined the feasibility of hAAT protein therapy in CIA mouse model. Administration of hAAT (0.5 mg/mouse twice per week, starting at 6 days before the induction of arthritis) resulted in sustained high levels of hAAT in mouse serum (Figure 1A). Although anti-hAAT-antibodies were detected (Figure 1B), serum levels of hAAT did not decrease over time.\nAntiarthritic effect of human alpha 1 antitrypsin (hAAT) in collagen induced arthritis (CIA) model. Human AAT (Prolastin®) was intraperitoneally injected in DBA/1 mice (n = 9), twice per week starting 6 days before until day 70 after CII immunization. Control group received saline injections (n = 7) (A) Serum hAAT protein levels in DBA/1 mice were measured by ELISA (mean+SD). ↓ indicates the day of first hAAT injection. (B) Serum anti-hAAT antibody levels (anti-hAAT-IgG) in DBA/1 mice were measured by ELISA. Each dot represents antibody levels (day 49 after bCII immunization, arbitrary units) of an individual mouse. (C) Arthritis score. For each paw, 0 is normal and 4 is the most severe arthritis. The maximum score for each animal is 16. Each line represents the scores from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, *p = 0.029 by AUC analysis) (D) Incidence of severe arthritis is defined by arthritic score/mouse > 3 (**p = 0.0025 by logrank test). Dotted line, saline injected control group; Solid line, hAAT treated group.\nA few days after the second immunization with bCII (day 21), mice in control group developed arthritis in multiple joints, which was manifested by redness, severe joint swelling and joint stiffness as well as ankylosis as the disease progressed. The severity of arthritis as measured by the arthritic score rapidly increased in control group (n = 7) whereas the disease development in hAAT treatment group (n = 9) was suppressed (Figure 1C). At day 49 (7 weeks) after the immunization, area under the curve (AUC) in the hAAT group was 50.83 ± 21.64 (mean ± SEM), while in control group it was 121.5 ± 17.67 (p = 0.029, mean ± SEM, AUC analysis until day 49). Human AAT protein therapy also reduced incidence of severe arthritis (p = 0.0025, logrank test, Figure 1D). Moreover, mice in hAAT treated group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis (arthritis score > 3) started on day 47.3 ± 8.7 (mean ± SD) in hAAT treated group compared to day 36.0 ± 5.8 (mean ± SD) in control group (p = 0.01 by students t-test). Although hAAT treated mice also developed arthritis at the end (70 days after the immunization) of the experiment, these results showed that treatment of hAAT protein (Prolastin®) led to a delayed arthritis onset and amelioration of disease progression in CIA mouse model.\nIn order to investigate the effect of hAAT on development of arthritis, we first examined the feasibility of hAAT protein therapy in CIA mouse model. Administration of hAAT (0.5 mg/mouse twice per week, starting at 6 days before the induction of arthritis) resulted in sustained high levels of hAAT in mouse serum (Figure 1A). Although anti-hAAT-antibodies were detected (Figure 1B), serum levels of hAAT did not decrease over time.\nAntiarthritic effect of human alpha 1 antitrypsin (hAAT) in collagen induced arthritis (CIA) model. Human AAT (Prolastin®) was intraperitoneally injected in DBA/1 mice (n = 9), twice per week starting 6 days before until day 70 after CII immunization. Control group received saline injections (n = 7) (A) Serum hAAT protein levels in DBA/1 mice were measured by ELISA (mean+SD). ↓ indicates the day of first hAAT injection. (B) Serum anti-hAAT antibody levels (anti-hAAT-IgG) in DBA/1 mice were measured by ELISA. Each dot represents antibody levels (day 49 after bCII immunization, arbitrary units) of an individual mouse. (C) Arthritis score. For each paw, 0 is normal and 4 is the most severe arthritis. The maximum score for each animal is 16. Each line represents the scores from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, *p = 0.029 by AUC analysis) (D) Incidence of severe arthritis is defined by arthritic score/mouse > 3 (**p = 0.0025 by logrank test). Dotted line, saline injected control group; Solid line, hAAT treated group.\nA few days after the second immunization with bCII (day 21), mice in control group developed arthritis in multiple joints, which was manifested by redness, severe joint swelling and joint stiffness as well as ankylosis as the disease progressed. The severity of arthritis as measured by the arthritic score rapidly increased in control group (n = 7) whereas the disease development in hAAT treatment group (n = 9) was suppressed (Figure 1C). At day 49 (7 weeks) after the immunization, area under the curve (AUC) in the hAAT group was 50.83 ± 21.64 (mean ± SEM), while in control group it was 121.5 ± 17.67 (p = 0.029, mean ± SEM, AUC analysis until day 49). Human AAT protein therapy also reduced incidence of severe arthritis (p = 0.0025, logrank test, Figure 1D). Moreover, mice in hAAT treated group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis (arthritis score > 3) started on day 47.3 ± 8.7 (mean ± SD) in hAAT treated group compared to day 36.0 ± 5.8 (mean ± SD) in control group (p = 0.01 by students t-test). Although hAAT treated mice also developed arthritis at the end (70 days after the immunization) of the experiment, these results showed that treatment of hAAT protein (Prolastin®) led to a delayed arthritis onset and amelioration of disease progression in CIA mouse model.\n[SUBTITLE] Human AAT Protein Therapy Reduced the Levels of anti-bCII and anti-mCII Autoantibodies [SUBSECTION] It has been shown that high levels of serum anti-collagen II autoantibodies are pathognomonic and associated with the development of arthritis [36,37]. To test the effect of hAAT on autoantibody production, we evaluated the levels of anti-CII autoantibodies in total Ig, and IgG1 and IgG2a subclass at early (day 35) and late (day 49) stages of the disease. As shown in Figure 2A, hAAT treatment did not result in a significant change of total autoantibody levels against bCII (total anti-bCII-Ig). However, hAAT treatment significantly reduced the pathognomonic IgG2a (anti-bCII-IgG2a) levels at day 35 (Figure 2B), and increased IgG1 (anti-bCII-IgG1) levels at day 49 (Figure 2C). Interestingly, levels of total Ig autoantibodies against endogenous mouse collagen II (total anti-mCII-Ig) were significantly lower in hAAT protein treated group than those in control group (P < 0.05) (Figure 2D).\nAnti-collagen II (CII) antibody levels after hAAT treatment. Anti-CII antibodies at day 35 and day 49 were tested by ELISA. Closed bars represent the average levels (n = 9, relative units, mean+SD) of antibodies in hAAT protein therapy treated group. Open bars represent the average levels (n = 7, relative units, mean+SD) of antibodies in saline injected group. (A) Levels of total Ig antibodies to bCII (total anti-bCII-Ig). (B) Levels of IgG2a anti-bCII (anti-bCII-IgG2a). (C) Levels of IgG1 anti-bCII (anti-bCII-IgG1). (D) Levels of total Ig antibodies to mCII (total anti-mCII-Ig). * p < 0.05 by Mann-Whitney U- test.\nIt has been shown that high levels of serum anti-collagen II autoantibodies are pathognomonic and associated with the development of arthritis [36,37]. To test the effect of hAAT on autoantibody production, we evaluated the levels of anti-CII autoantibodies in total Ig, and IgG1 and IgG2a subclass at early (day 35) and late (day 49) stages of the disease. As shown in Figure 2A, hAAT treatment did not result in a significant change of total autoantibody levels against bCII (total anti-bCII-Ig). However, hAAT treatment significantly reduced the pathognomonic IgG2a (anti-bCII-IgG2a) levels at day 35 (Figure 2B), and increased IgG1 (anti-bCII-IgG1) levels at day 49 (Figure 2C). Interestingly, levels of total Ig autoantibodies against endogenous mouse collagen II (total anti-mCII-Ig) were significantly lower in hAAT protein treated group than those in control group (P < 0.05) (Figure 2D).\nAnti-collagen II (CII) antibody levels after hAAT treatment. Anti-CII antibodies at day 35 and day 49 were tested by ELISA. Closed bars represent the average levels (n = 9, relative units, mean+SD) of antibodies in hAAT protein therapy treated group. Open bars represent the average levels (n = 7, relative units, mean+SD) of antibodies in saline injected group. (A) Levels of total Ig antibodies to bCII (total anti-bCII-Ig). (B) Levels of IgG2a anti-bCII (anti-bCII-IgG2a). (C) Levels of IgG1 anti-bCII (anti-bCII-IgG1). (D) Levels of total Ig antibodies to mCII (total anti-mCII-Ig). * p < 0.05 by Mann-Whitney U- test.\n[SUBTITLE] Human AAT (hAAT) Gene Therapy delayed Arthritis Development [SUBSECTION] To further confirm our observation that hAAT is effective in delaying arthritis development, and to test the feasibility of hAAT gene therapy for rheumatoid arthritis, we used recombinant adeno-associated virus vector (rAAV) to deliver the hAAT gene. A single IP injection of rAAV8-CB-hAAT vector (2x1011 particles/mouse, two weeks before the first CII immunization) resulted in sustained levels of hAAT in the circulation, similar to those levels obtained following protein therapy (Figure 3A). Interestingly, following AAV8 mediated gene delivery, we did not observe the development of antibodies to hAAT which were detected during hAAT protein therapy (Figure 3B, compare vs. Figure 1B in mice with hAAT protein therapy). Similar to the results from hAAT protein therapy, however, rAAV-mediated hAAT gene therapy significantly reduced the prevalence of arthritis development at the early stage of disease (Figure 3C). Area under the curve (AUC) in the gene therapy group (n = 10) was 71.65 ± 14.04 (mean ± SEM), while in control group (n = 10) it was 123.20 ± 19.83 (mean ± SEM; p < 0.05 by AUC analysis until day 42). AAT gene therapy also reduced the incidence of severe arthritis (score > 3) at the early stage of disease (p = 0.035 by logrank test, Figure 3D). Moreover, mice in hAAT gene therapy group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis started on day 42.3 ± 7.5 (mean ± SD) in hAAT gene therapy group compared to day 33.4 ± 7.3 in control group (mean ± SD; p < 0.02 by student's t-test). These results indicate that similar to hAAT protein therapy, AAV8 mediated hAAT gene delivery also delayed arthritis onset and ameliorated early stage disease progression in CIA mouse model.\nHuman AAT gene therapy delays disease progression in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 10) or saline (n = 10) two weeks before immunization with CII. Control group received saline. Mice were sacrificed on day 56 (EOS). (A) Serum levels of hAAT. hAAT protein serum levels in vector injected group were measured by ELISA (mean+SD). ↓ indicates the days of injection. (B) Anti-hAAT antibody levels. Serum anti-hAAT antibodies (anti-hAAT) were measured by ELISA using samples obtained at 56 days after immunization. Anti-hAAT antibodies were undetectable in the vector injected group. Each dot represents antibody level (arbitrary units) of an individual mouse. (C) Arthritis score. Each line represents the average score from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, * p < 0.05 as determined by AUC analysis.) (D) Incidence of severe arthritis. Severe arthritis was defined by arthritic score > 3, (* p = 0.035 by logrank test.; 10 mice/group).\nIn an additional experiment using AAV8 mediated hAAT gene therapy, tissue protective properties of hAAT were evaluated. Similar to the previous experiment, mice in treatment group (n = 6) showed significantly reduced arthritis development at the early disease stage compared to control (n = 4) (Figure 4A, p < 0.05 by Mann-Whitney U-test). As shown in Figure 4B-F, AAV8 mediated hAAT gene therapy resulted in less infiltration of immune cells into the joint cavity accompanied with reduced synovial cell hyperplasia and pannus formation (p < 0.05 Mann-Whitney U-test).\nTissue protective effect of hAAT gene therapy in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 6) or saline (n = 4) two weeks before immunization with CII. Control group received saline. (A) Arthritis development was evaluated based on arthritis score (mean + SD). Open circle represent rAAV8-CB-hAAT vector injected group, open triangle represent control group. Mice were sacrificed on day 28 after CII immunization, hind limbs were harvested and processed for histological assessment. *p < 0.05 by Mann-Whitney U-test. (B) Histopathological evaluation of arthritis development. Mice in gene therapy group (black bars) or control group (empty bars) were evaluated according to histopathological changes by two blinded pathologists. Each hind paw was evaluated based on a scale ranging from 0-3. (mean+SD). *p < 0.05, **p < 0.01 by Mann-Whitney U-test. (INF: Infiltration of Immune Cells, HYP: Hyperplasia, P.F.: Pannus Formation, B.D.: Bone Destruction) (C,D) Representative joint section from mice receiving hAAT gene therapy. (E,F) Representative joint section from mice in control group (saline injection). Magnification: C,E: 100x; D,F: 200x.\nTo further confirm our observation that hAAT is effective in delaying arthritis development, and to test the feasibility of hAAT gene therapy for rheumatoid arthritis, we used recombinant adeno-associated virus vector (rAAV) to deliver the hAAT gene. A single IP injection of rAAV8-CB-hAAT vector (2x1011 particles/mouse, two weeks before the first CII immunization) resulted in sustained levels of hAAT in the circulation, similar to those levels obtained following protein therapy (Figure 3A). Interestingly, following AAV8 mediated gene delivery, we did not observe the development of antibodies to hAAT which were detected during hAAT protein therapy (Figure 3B, compare vs. Figure 1B in mice with hAAT protein therapy). Similar to the results from hAAT protein therapy, however, rAAV-mediated hAAT gene therapy significantly reduced the prevalence of arthritis development at the early stage of disease (Figure 3C). Area under the curve (AUC) in the gene therapy group (n = 10) was 71.65 ± 14.04 (mean ± SEM), while in control group (n = 10) it was 123.20 ± 19.83 (mean ± SEM; p < 0.05 by AUC analysis until day 42). AAT gene therapy also reduced the incidence of severe arthritis (score > 3) at the early stage of disease (p = 0.035 by logrank test, Figure 3D). Moreover, mice in hAAT gene therapy group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis started on day 42.3 ± 7.5 (mean ± SD) in hAAT gene therapy group compared to day 33.4 ± 7.3 in control group (mean ± SD; p < 0.02 by student's t-test). These results indicate that similar to hAAT protein therapy, AAV8 mediated hAAT gene delivery also delayed arthritis onset and ameliorated early stage disease progression in CIA mouse model.\nHuman AAT gene therapy delays disease progression in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 10) or saline (n = 10) two weeks before immunization with CII. Control group received saline. Mice were sacrificed on day 56 (EOS). (A) Serum levels of hAAT. hAAT protein serum levels in vector injected group were measured by ELISA (mean+SD). ↓ indicates the days of injection. (B) Anti-hAAT antibody levels. Serum anti-hAAT antibodies (anti-hAAT) were measured by ELISA using samples obtained at 56 days after immunization. Anti-hAAT antibodies were undetectable in the vector injected group. Each dot represents antibody level (arbitrary units) of an individual mouse. (C) Arthritis score. Each line represents the average score from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, * p < 0.05 as determined by AUC analysis.) (D) Incidence of severe arthritis. Severe arthritis was defined by arthritic score > 3, (* p = 0.035 by logrank test.; 10 mice/group).\nIn an additional experiment using AAV8 mediated hAAT gene therapy, tissue protective properties of hAAT were evaluated. Similar to the previous experiment, mice in treatment group (n = 6) showed significantly reduced arthritis development at the early disease stage compared to control (n = 4) (Figure 4A, p < 0.05 by Mann-Whitney U-test). As shown in Figure 4B-F, AAV8 mediated hAAT gene therapy resulted in less infiltration of immune cells into the joint cavity accompanied with reduced synovial cell hyperplasia and pannus formation (p < 0.05 Mann-Whitney U-test).\nTissue protective effect of hAAT gene therapy in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 6) or saline (n = 4) two weeks before immunization with CII. Control group received saline. (A) Arthritis development was evaluated based on arthritis score (mean + SD). Open circle represent rAAV8-CB-hAAT vector injected group, open triangle represent control group. Mice were sacrificed on day 28 after CII immunization, hind limbs were harvested and processed for histological assessment. *p < 0.05 by Mann-Whitney U-test. (B) Histopathological evaluation of arthritis development. Mice in gene therapy group (black bars) or control group (empty bars) were evaluated according to histopathological changes by two blinded pathologists. Each hind paw was evaluated based on a scale ranging from 0-3. (mean+SD). *p < 0.05, **p < 0.01 by Mann-Whitney U-test. (INF: Infiltration of Immune Cells, HYP: Hyperplasia, P.F.: Pannus Formation, B.D.: Bone Destruction) (C,D) Representative joint section from mice receiving hAAT gene therapy. (E,F) Representative joint section from mice in control group (saline injection). Magnification: C,E: 100x; D,F: 200x.\n[SUBTITLE] Human AAT (hAAT) Gene Therapy Reduced the Levels of Anti-CII Autoantibodies [SUBSECTION] As shown in Figure 5, rAAV8-mediated hAAT gene therapy resulted in a significant suppression of anti-CII autoantibody production. The levels of total Ig anti-bCII (Figure 5A, top left panel) and IgG2a anti-bCII (Figure 5A, top right panel) were significantly reduced in hAAT gene therapy group. Although IgG1 anti-bCII levels (Figure 5A, bottom left panel) were also reduced in hAAT gene therapy group, the ratio of IgG2a anti-bCII to IgG1 anti-bCII (Figure 5A, bottom right panel) significantly decreased in hAAT gene therapy group. Importantly, hAAT gene therapy also reduced levels of autoantibodies against mCII and the ratio of IgG2a anti-mCII to IgG1 anti-mCII (Figure 5B).\nEffect of hAAT gene therapy on auto-antibody production. Anti-CII antibodies at day 28, 42 and 56 were tested by ELISA. Black bars represent the average levels (n = 10, mean+SD) (relative units) of antibodies in hAAT gene therapy treated group. Open bars represent the average levels (n = 10, relative units, mean+SD) of antibodies in saline injected group. (A) Antibody levels against bovine CII (bCII). Top left panel, total Ig antibodies against bCII (total anti-bCII-Ig); Top right panel, levels of IgG2a anti-bCII (anti-bCII-IgG2a); Bottom left panel, levels of IgG1 anti-bCII (anti-bCII-IgG1); Bottom right panel, the ratio of anti-bCII-IgG2a to anti-bCII-IgG1 (anti-bCII-IgG2a/IgG1 ratio). (B) Antibody levels against mouse CII (mCII). Top left panel, total Ig antibodies against mCII (total anti-mCII-Ig); Top right panel, levels of IgG2a anti-mCII (anti-mCII-IgG2a); Bottom left panel, levels of IgG1 anti-mCII (anti-mCII-IgG1); Bottom right panel, the ratio of anti-mCII-IgG2a to anti-mCII-IgG1 (anti-mCII-IgG2a/IgG1). *p < 0.05, **p < 0.01, ***p < 0.001 by Mann-Whitney U- test.\nAs shown in Figure 5, rAAV8-mediated hAAT gene therapy resulted in a significant suppression of anti-CII autoantibody production. The levels of total Ig anti-bCII (Figure 5A, top left panel) and IgG2a anti-bCII (Figure 5A, top right panel) were significantly reduced in hAAT gene therapy group. Although IgG1 anti-bCII levels (Figure 5A, bottom left panel) were also reduced in hAAT gene therapy group, the ratio of IgG2a anti-bCII to IgG1 anti-bCII (Figure 5A, bottom right panel) significantly decreased in hAAT gene therapy group. Importantly, hAAT gene therapy also reduced levels of autoantibodies against mCII and the ratio of IgG2a anti-mCII to IgG1 anti-mCII (Figure 5B).\nEffect of hAAT gene therapy on auto-antibody production. Anti-CII antibodies at day 28, 42 and 56 were tested by ELISA. Black bars represent the average levels (n = 10, mean+SD) (relative units) of antibodies in hAAT gene therapy treated group. Open bars represent the average levels (n = 10, relative units, mean+SD) of antibodies in saline injected group. (A) Antibody levels against bovine CII (bCII). Top left panel, total Ig antibodies against bCII (total anti-bCII-Ig); Top right panel, levels of IgG2a anti-bCII (anti-bCII-IgG2a); Bottom left panel, levels of IgG1 anti-bCII (anti-bCII-IgG1); Bottom right panel, the ratio of anti-bCII-IgG2a to anti-bCII-IgG1 (anti-bCII-IgG2a/IgG1 ratio). (B) Antibody levels against mouse CII (mCII). Top left panel, total Ig antibodies against mCII (total anti-mCII-Ig); Top right panel, levels of IgG2a anti-mCII (anti-mCII-IgG2a); Bottom left panel, levels of IgG1 anti-mCII (anti-mCII-IgG1); Bottom right panel, the ratio of anti-mCII-IgG2a to anti-mCII-IgG1 (anti-mCII-IgG2a/IgG1). *p < 0.05, **p < 0.01, ***p < 0.001 by Mann-Whitney U- test.\n[SUBTITLE] Human AAT Therapy Reduced B-cell Activating Factor (BAFF) in vitro and in vivo [SUBSECTION] In order to further elucidate the underlying mechanism of the anti-arthritic effect of hAAT, we performed additional studies focusing on the effect of AAT on T-cell and B-cell activity. Since CIA is a T-cell-mediated autoimmune disease, the effect of hAAT on T-cell function was examined in a T-cell proliferation assay. As shown in Figure 6A, treatment of rAAV8-hAAT did not change the antigen specific T-cell response after isolated splenocytes were restimulated ex vivo with bCII. Similarly, bCII induced cytokine release (IFN-γ, IL-4, IL-10, TNF-α, IL-2) from isolated splenocytes did not show any significant differences between treatment and control group (Figure 6B). The effect of hAAT therapy on B-cell activity was examined by determination of serum levels of B-cell activating factor of the TNF-α family (BAFF), which has emerged as a crucial factor for B-cell expansion and function. Interestingly, both hAAT protein as well as AAV8 mediated hAAT gene therapy resulted in significantly decreased serum levels of BAFF compared to control group (Figure 6C, 6D). Since BAFF is mainly secreted from monocytes and macrophages, we tested the effect of hAAT on BAFF production in vitro. Murine macrophages (RAW264.7) were treated with hAAT. Culture medium served as control. Protein secretion into the culture medium was determined by ELISA and mRNA expression was quantified by real-time PCR. As shown in Figure 6E, BAFF levels in culture medium were significantly lower in the AAT treated group than those in the control group. Similarly, mRNA expression levels of BAFF were also significantly decreased in AAT treated group (Figure 6F). Together these results suggest that the anti-arthritic effect of AAT is in part through the inhibition of B-cell activation.\nEffects of hAAT therapy on T-cells and B-cells. (A) Proliferative response of splenocytes after stimulation with bovine type II collagen (bCII, 10 μg/ml). Splenocytes (4 × 105 cells/well, in 96-well plate) were isolated on day 28 after rAAV8-hAAT injection. Black bar, AAT gene therapy group (n = 6); open bar, control group (n = 4). Data are expressed as the stimulation index, determined by calculating the ratio of cell proliferation with antigen (measured in counts per minute, cpm) relative to that with medium alone (mean+SD). (B) Cytokine production from bCII-stimulated (100 μg/ml) splenocytes. Values are the mean+SD of each group (n = 6 for rAAV8-hAAT group, black bars; n = 4 for saline group, open bars). (C) Serum level of BAFF in hAAT treated mice (black bar, n = 9, day 35) and control mice (open bar, n = 7). Data is expressed as mean+SD. (D) BAFF serum level in rAAV8-hAAT treated mice (black bar, n = 10, day 28) and control (open bar, n = 10). In vitro effect of hAAT on (E) BAFF secretion into culture medium measured by ELISA and (F) BAFF gene expression determined by real-time PCR. Murine macrophages (RAW 264.7) were treated with hAAT (0.5mg/ml, black bar). Culture medium served as control (open bar). Both experiments were performed in quadruplicates and repeated twice. Data is expressed as mean+SD. *p < 0.05, **p < 0.01.\nIn order to further elucidate the underlying mechanism of the anti-arthritic effect of hAAT, we performed additional studies focusing on the effect of AAT on T-cell and B-cell activity. Since CIA is a T-cell-mediated autoimmune disease, the effect of hAAT on T-cell function was examined in a T-cell proliferation assay. As shown in Figure 6A, treatment of rAAV8-hAAT did not change the antigen specific T-cell response after isolated splenocytes were restimulated ex vivo with bCII. Similarly, bCII induced cytokine release (IFN-γ, IL-4, IL-10, TNF-α, IL-2) from isolated splenocytes did not show any significant differences between treatment and control group (Figure 6B). The effect of hAAT therapy on B-cell activity was examined by determination of serum levels of B-cell activating factor of the TNF-α family (BAFF), which has emerged as a crucial factor for B-cell expansion and function. Interestingly, both hAAT protein as well as AAV8 mediated hAAT gene therapy resulted in significantly decreased serum levels of BAFF compared to control group (Figure 6C, 6D). Since BAFF is mainly secreted from monocytes and macrophages, we tested the effect of hAAT on BAFF production in vitro. Murine macrophages (RAW264.7) were treated with hAAT. Culture medium served as control. Protein secretion into the culture medium was determined by ELISA and mRNA expression was quantified by real-time PCR. As shown in Figure 6E, BAFF levels in culture medium were significantly lower in the AAT treated group than those in the control group. Similarly, mRNA expression levels of BAFF were also significantly decreased in AAT treated group (Figure 6F). Together these results suggest that the anti-arthritic effect of AAT is in part through the inhibition of B-cell activation.\nEffects of hAAT therapy on T-cells and B-cells. (A) Proliferative response of splenocytes after stimulation with bovine type II collagen (bCII, 10 μg/ml). Splenocytes (4 × 105 cells/well, in 96-well plate) were isolated on day 28 after rAAV8-hAAT injection. Black bar, AAT gene therapy group (n = 6); open bar, control group (n = 4). Data are expressed as the stimulation index, determined by calculating the ratio of cell proliferation with antigen (measured in counts per minute, cpm) relative to that with medium alone (mean+SD). (B) Cytokine production from bCII-stimulated (100 μg/ml) splenocytes. Values are the mean+SD of each group (n = 6 for rAAV8-hAAT group, black bars; n = 4 for saline group, open bars). (C) Serum level of BAFF in hAAT treated mice (black bar, n = 9, day 35) and control mice (open bar, n = 7). Data is expressed as mean+SD. (D) BAFF serum level in rAAV8-hAAT treated mice (black bar, n = 10, day 28) and control (open bar, n = 10). In vitro effect of hAAT on (E) BAFF secretion into culture medium measured by ELISA and (F) BAFF gene expression determined by real-time PCR. Murine macrophages (RAW 264.7) were treated with hAAT (0.5mg/ml, black bar). Culture medium served as control (open bar). Both experiments were performed in quadruplicates and repeated twice. Data is expressed as mean+SD. *p < 0.05, **p < 0.01.", "In order to investigate the effect of hAAT on development of arthritis, we first examined the feasibility of hAAT protein therapy in CIA mouse model. Administration of hAAT (0.5 mg/mouse twice per week, starting at 6 days before the induction of arthritis) resulted in sustained high levels of hAAT in mouse serum (Figure 1A). Although anti-hAAT-antibodies were detected (Figure 1B), serum levels of hAAT did not decrease over time.\nAntiarthritic effect of human alpha 1 antitrypsin (hAAT) in collagen induced arthritis (CIA) model. Human AAT (Prolastin®) was intraperitoneally injected in DBA/1 mice (n = 9), twice per week starting 6 days before until day 70 after CII immunization. Control group received saline injections (n = 7) (A) Serum hAAT protein levels in DBA/1 mice were measured by ELISA (mean+SD). ↓ indicates the day of first hAAT injection. (B) Serum anti-hAAT antibody levels (anti-hAAT-IgG) in DBA/1 mice were measured by ELISA. Each dot represents antibody levels (day 49 after bCII immunization, arbitrary units) of an individual mouse. (C) Arthritis score. For each paw, 0 is normal and 4 is the most severe arthritis. The maximum score for each animal is 16. Each line represents the scores from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, *p = 0.029 by AUC analysis) (D) Incidence of severe arthritis is defined by arthritic score/mouse > 3 (**p = 0.0025 by logrank test). Dotted line, saline injected control group; Solid line, hAAT treated group.\nA few days after the second immunization with bCII (day 21), mice in control group developed arthritis in multiple joints, which was manifested by redness, severe joint swelling and joint stiffness as well as ankylosis as the disease progressed. The severity of arthritis as measured by the arthritic score rapidly increased in control group (n = 7) whereas the disease development in hAAT treatment group (n = 9) was suppressed (Figure 1C). At day 49 (7 weeks) after the immunization, area under the curve (AUC) in the hAAT group was 50.83 ± 21.64 (mean ± SEM), while in control group it was 121.5 ± 17.67 (p = 0.029, mean ± SEM, AUC analysis until day 49). Human AAT protein therapy also reduced incidence of severe arthritis (p = 0.0025, logrank test, Figure 1D). Moreover, mice in hAAT treated group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis (arthritis score > 3) started on day 47.3 ± 8.7 (mean ± SD) in hAAT treated group compared to day 36.0 ± 5.8 (mean ± SD) in control group (p = 0.01 by students t-test). Although hAAT treated mice also developed arthritis at the end (70 days after the immunization) of the experiment, these results showed that treatment of hAAT protein (Prolastin®) led to a delayed arthritis onset and amelioration of disease progression in CIA mouse model.", "It has been shown that high levels of serum anti-collagen II autoantibodies are pathognomonic and associated with the development of arthritis [36,37]. To test the effect of hAAT on autoantibody production, we evaluated the levels of anti-CII autoantibodies in total Ig, and IgG1 and IgG2a subclass at early (day 35) and late (day 49) stages of the disease. As shown in Figure 2A, hAAT treatment did not result in a significant change of total autoantibody levels against bCII (total anti-bCII-Ig). However, hAAT treatment significantly reduced the pathognomonic IgG2a (anti-bCII-IgG2a) levels at day 35 (Figure 2B), and increased IgG1 (anti-bCII-IgG1) levels at day 49 (Figure 2C). Interestingly, levels of total Ig autoantibodies against endogenous mouse collagen II (total anti-mCII-Ig) were significantly lower in hAAT protein treated group than those in control group (P < 0.05) (Figure 2D).\nAnti-collagen II (CII) antibody levels after hAAT treatment. Anti-CII antibodies at day 35 and day 49 were tested by ELISA. Closed bars represent the average levels (n = 9, relative units, mean+SD) of antibodies in hAAT protein therapy treated group. Open bars represent the average levels (n = 7, relative units, mean+SD) of antibodies in saline injected group. (A) Levels of total Ig antibodies to bCII (total anti-bCII-Ig). (B) Levels of IgG2a anti-bCII (anti-bCII-IgG2a). (C) Levels of IgG1 anti-bCII (anti-bCII-IgG1). (D) Levels of total Ig antibodies to mCII (total anti-mCII-Ig). * p < 0.05 by Mann-Whitney U- test.", "To further confirm our observation that hAAT is effective in delaying arthritis development, and to test the feasibility of hAAT gene therapy for rheumatoid arthritis, we used recombinant adeno-associated virus vector (rAAV) to deliver the hAAT gene. A single IP injection of rAAV8-CB-hAAT vector (2x1011 particles/mouse, two weeks before the first CII immunization) resulted in sustained levels of hAAT in the circulation, similar to those levels obtained following protein therapy (Figure 3A). Interestingly, following AAV8 mediated gene delivery, we did not observe the development of antibodies to hAAT which were detected during hAAT protein therapy (Figure 3B, compare vs. Figure 1B in mice with hAAT protein therapy). Similar to the results from hAAT protein therapy, however, rAAV-mediated hAAT gene therapy significantly reduced the prevalence of arthritis development at the early stage of disease (Figure 3C). Area under the curve (AUC) in the gene therapy group (n = 10) was 71.65 ± 14.04 (mean ± SEM), while in control group (n = 10) it was 123.20 ± 19.83 (mean ± SEM; p < 0.05 by AUC analysis until day 42). AAT gene therapy also reduced the incidence of severe arthritis (score > 3) at the early stage of disease (p = 0.035 by logrank test, Figure 3D). Moreover, mice in hAAT gene therapy group had significantly delayed onset of arthritis compared with control group. On average, the clinical signs of severe arthritis started on day 42.3 ± 7.5 (mean ± SD) in hAAT gene therapy group compared to day 33.4 ± 7.3 in control group (mean ± SD; p < 0.02 by student's t-test). These results indicate that similar to hAAT protein therapy, AAV8 mediated hAAT gene delivery also delayed arthritis onset and ameliorated early stage disease progression in CIA mouse model.\nHuman AAT gene therapy delays disease progression in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 10) or saline (n = 10) two weeks before immunization with CII. Control group received saline. Mice were sacrificed on day 56 (EOS). (A) Serum levels of hAAT. hAAT protein serum levels in vector injected group were measured by ELISA (mean+SD). ↓ indicates the days of injection. (B) Anti-hAAT antibody levels. Serum anti-hAAT antibodies (anti-hAAT) were measured by ELISA using samples obtained at 56 days after immunization. Anti-hAAT antibodies were undetectable in the vector injected group. Each dot represents antibody level (arbitrary units) of an individual mouse. (C) Arthritis score. Each line represents the average score from hAAT treated group (open triangles, mean-SD) or control group (open circles, mean+SD, * p < 0.05 as determined by AUC analysis.) (D) Incidence of severe arthritis. Severe arthritis was defined by arthritic score > 3, (* p = 0.035 by logrank test.; 10 mice/group).\nIn an additional experiment using AAV8 mediated hAAT gene therapy, tissue protective properties of hAAT were evaluated. Similar to the previous experiment, mice in treatment group (n = 6) showed significantly reduced arthritis development at the early disease stage compared to control (n = 4) (Figure 4A, p < 0.05 by Mann-Whitney U-test). As shown in Figure 4B-F, AAV8 mediated hAAT gene therapy resulted in less infiltration of immune cells into the joint cavity accompanied with reduced synovial cell hyperplasia and pannus formation (p < 0.05 Mann-Whitney U-test).\nTissue protective effect of hAAT gene therapy in CIA mouse model. DBA/1 mice were intraperitoneally injected with rAAV8-CB-hAAT vector (2 × 1011 particles/mouse, n = 6) or saline (n = 4) two weeks before immunization with CII. Control group received saline. (A) Arthritis development was evaluated based on arthritis score (mean + SD). Open circle represent rAAV8-CB-hAAT vector injected group, open triangle represent control group. Mice were sacrificed on day 28 after CII immunization, hind limbs were harvested and processed for histological assessment. *p < 0.05 by Mann-Whitney U-test. (B) Histopathological evaluation of arthritis development. Mice in gene therapy group (black bars) or control group (empty bars) were evaluated according to histopathological changes by two blinded pathologists. Each hind paw was evaluated based on a scale ranging from 0-3. (mean+SD). *p < 0.05, **p < 0.01 by Mann-Whitney U-test. (INF: Infiltration of Immune Cells, HYP: Hyperplasia, P.F.: Pannus Formation, B.D.: Bone Destruction) (C,D) Representative joint section from mice receiving hAAT gene therapy. (E,F) Representative joint section from mice in control group (saline injection). Magnification: C,E: 100x; D,F: 200x.", "As shown in Figure 5, rAAV8-mediated hAAT gene therapy resulted in a significant suppression of anti-CII autoantibody production. The levels of total Ig anti-bCII (Figure 5A, top left panel) and IgG2a anti-bCII (Figure 5A, top right panel) were significantly reduced in hAAT gene therapy group. Although IgG1 anti-bCII levels (Figure 5A, bottom left panel) were also reduced in hAAT gene therapy group, the ratio of IgG2a anti-bCII to IgG1 anti-bCII (Figure 5A, bottom right panel) significantly decreased in hAAT gene therapy group. Importantly, hAAT gene therapy also reduced levels of autoantibodies against mCII and the ratio of IgG2a anti-mCII to IgG1 anti-mCII (Figure 5B).\nEffect of hAAT gene therapy on auto-antibody production. Anti-CII antibodies at day 28, 42 and 56 were tested by ELISA. Black bars represent the average levels (n = 10, mean+SD) (relative units) of antibodies in hAAT gene therapy treated group. Open bars represent the average levels (n = 10, relative units, mean+SD) of antibodies in saline injected group. (A) Antibody levels against bovine CII (bCII). Top left panel, total Ig antibodies against bCII (total anti-bCII-Ig); Top right panel, levels of IgG2a anti-bCII (anti-bCII-IgG2a); Bottom left panel, levels of IgG1 anti-bCII (anti-bCII-IgG1); Bottom right panel, the ratio of anti-bCII-IgG2a to anti-bCII-IgG1 (anti-bCII-IgG2a/IgG1 ratio). (B) Antibody levels against mouse CII (mCII). Top left panel, total Ig antibodies against mCII (total anti-mCII-Ig); Top right panel, levels of IgG2a anti-mCII (anti-mCII-IgG2a); Bottom left panel, levels of IgG1 anti-mCII (anti-mCII-IgG1); Bottom right panel, the ratio of anti-mCII-IgG2a to anti-mCII-IgG1 (anti-mCII-IgG2a/IgG1). *p < 0.05, **p < 0.01, ***p < 0.001 by Mann-Whitney U- test.", "In order to further elucidate the underlying mechanism of the anti-arthritic effect of hAAT, we performed additional studies focusing on the effect of AAT on T-cell and B-cell activity. Since CIA is a T-cell-mediated autoimmune disease, the effect of hAAT on T-cell function was examined in a T-cell proliferation assay. As shown in Figure 6A, treatment of rAAV8-hAAT did not change the antigen specific T-cell response after isolated splenocytes were restimulated ex vivo with bCII. Similarly, bCII induced cytokine release (IFN-γ, IL-4, IL-10, TNF-α, IL-2) from isolated splenocytes did not show any significant differences between treatment and control group (Figure 6B). The effect of hAAT therapy on B-cell activity was examined by determination of serum levels of B-cell activating factor of the TNF-α family (BAFF), which has emerged as a crucial factor for B-cell expansion and function. Interestingly, both hAAT protein as well as AAV8 mediated hAAT gene therapy resulted in significantly decreased serum levels of BAFF compared to control group (Figure 6C, 6D). Since BAFF is mainly secreted from monocytes and macrophages, we tested the effect of hAAT on BAFF production in vitro. Murine macrophages (RAW264.7) were treated with hAAT. Culture medium served as control. Protein secretion into the culture medium was determined by ELISA and mRNA expression was quantified by real-time PCR. As shown in Figure 6E, BAFF levels in culture medium were significantly lower in the AAT treated group than those in the control group. Similarly, mRNA expression levels of BAFF were also significantly decreased in AAT treated group (Figure 6F). Together these results suggest that the anti-arthritic effect of AAT is in part through the inhibition of B-cell activation.\nEffects of hAAT therapy on T-cells and B-cells. (A) Proliferative response of splenocytes after stimulation with bovine type II collagen (bCII, 10 μg/ml). Splenocytes (4 × 105 cells/well, in 96-well plate) were isolated on day 28 after rAAV8-hAAT injection. Black bar, AAT gene therapy group (n = 6); open bar, control group (n = 4). Data are expressed as the stimulation index, determined by calculating the ratio of cell proliferation with antigen (measured in counts per minute, cpm) relative to that with medium alone (mean+SD). (B) Cytokine production from bCII-stimulated (100 μg/ml) splenocytes. Values are the mean+SD of each group (n = 6 for rAAV8-hAAT group, black bars; n = 4 for saline group, open bars). (C) Serum level of BAFF in hAAT treated mice (black bar, n = 9, day 35) and control mice (open bar, n = 7). Data is expressed as mean+SD. (D) BAFF serum level in rAAV8-hAAT treated mice (black bar, n = 10, day 28) and control (open bar, n = 10). In vitro effect of hAAT on (E) BAFF secretion into culture medium measured by ELISA and (F) BAFF gene expression determined by real-time PCR. Murine macrophages (RAW 264.7) were treated with hAAT (0.5mg/ml, black bar). Culture medium served as control (open bar). Both experiments were performed in quadruplicates and repeated twice. Data is expressed as mean+SD. *p < 0.05, **p < 0.01.", "RA is a complex systemic autoimmune disease of unknown etiology. Although recently developed biologics that target TNF-alpha have provided dramatic improvement in controlling disease activity in many patients, continued searches for more efficient and safer treatments are still needed. In the present study we showed that hAAT, administered as protein or through rAAV8 mediated gene therapy, reduced levels of serum anti-CII auto-antibodies and B-cell activating factor (BAFF) and significantly delayed arthritis development in a mouse model.\nAlthough the exact mechanisms underlying the therapeutic effect remain to be further investigated, several mechanisms may be involved. One is through the inhibition of proinflammatory cytokine production. It is well known that various proinflammatory cytokines, including TNF-α and IL1-β, play major roles in the pathogenesis of RA [3]. Strategies targeting these cytokines have proven to be effective in treatment of RA [38]. Previous work done by Janciauskiene and her colleagues clearly demonstrated that hAAT inhibited LPS-induced TNF-α, IL-6 and IL-1β production by human monocytes [15,16]. In addition, hAAT completely suppressed macrophage inflammatory protein-2 (MIP-2)/monocyte chemotactic protein-1 (MCP-1) gene expression in lung [39]. Human AAT also enhanced anti-inflammatory cytokine IL-10 production from monocytes [15]. As a consequence of interfering with the cytokine/chemokine network, hAAT may also inhibit polymorphonuclear leukocyte (PMN) invasion into the joint. Churg et al. demonstrated that hAAT inhibited silica-induced PMN influx into the lung and partially suppressed nuclear transcription factor B (NF-κB) translocation and increased inhibitor of NF-κB (I-࿠κB) levels in a mouse model of acute PMN mediated inflammation [39]. Thus, it is possible that the effects of hAAT on pro-inflammatory cytokine production contribute to suppression of autoimmune-mediated inflammation.\nIn previous studies we showed that hAAT reduced anti-insulin auto-antibodies (IAA) and attenuated cell-mediated autoimmunity [32,33]. Consistent with these results, the present study showed that hAAT reduced the levels of anti-CII auto-antibodies and the IgG2a/IgG1 ratios of anti-CII auto-antibodies (mCII and bCII). We have observed that the effect of hAAT to suppress arthritis development is more profound in early stage of arthritis development. This is supported by the effect of hAAT on pathognomonic IgG2a antibody development at early time points (Fig.2) as well as the observation that mice eventually develop arthritis overtime. Therefore, hAAT maybe especially suitable for combination therapies. We did not observe significant effect of AAT on T-cell proliferation and cytokine production in vitro (Figure 6A and 6B) indicating that AAT may have limited direct effect on T-cells. These data also suggest that AAT may more directly affect B-cell activity. Indeed, we have shown that AAT therapies significantly reduced B-cell activating factor of the TNF-α family (BAFF) in vitro and in vivo. BAFF is an important factor that modulates B-cell tolerance and homeostasis. It has been shown that soluble BAFF is elevated in serum and target organs of CIA model [40] and BAFF antagonists suppressed arthritis development in murine models of rheumatoid arthritis [41]. In addition, increased BAFF levels were found in serum of RA patients which correlated with serum levels of rheumatoid factor [42]. The exact mechanism that AAT suppresses BAFF production remains to be elucidated.\nAnother possible mechanism of hAAT suppressing arthritis development is through inhibition of proteinases to prevent tissue injury and joint destruction. Human AAT is well known as a serine proteinase inhibitor (serpin). It inhibits proteinase 3, neutrophil elastase, and cathepsin G. These serine proteases are released by joint invading neutrophils following inflammatory stimuli and have shown to be involved in arthritis development [12,13,43,44]. Human AAT can also reduce ischemia-induced apoptosis, inflammation, and acute phase response in the kidney [20]. We have recently shown that hAAT directly inhibits caspase 3 activity and protects islet cells from cytokine and chemically-induced apoptosis [45].\nIn the protein therapy studies, we used Prolastin®, which is clinical grade hAAT purified from human plasma. Repeated IP injection of hAAT induced strong humoral immune response against hAAT in DBA/1 mice (Figure 1B), similar to what has been observed in previous studies [46,47]. It is possible that non-specific inflammation caused by repeated IP injection is responsible for inhibition of arthritis. In order to rule out this possibility, we performed rAAV8 mediated hAAT gene therapy. AAV serotype 8 vector is unique for this purpose because it can mediate long term and high levels of transgene expression in the liver and muscle, but is not able to transduce dendritic cells and has low immunogenicity [48,49]. Indeed, after a single injection of rAAV8-CB-hAAT vector, sustained high levels of hAAT were detected in the circulation, while no detectable levels of anti-hAAT antibodies were present (Figure 3B) in contrast to mice that received hAAT protein therapy. These results are consistent with our recent observations in NOD mice and imply new applications of rAAV8 vectors [34]. The detailed mechanism that rAAV8 vector mediates no immune response to the transgene product remains elusive. Importantly, we have observed protective effects and reductions of auto-antibodies by hAAT gene therapy. These results strongly support our hypothesis that hAAT is able to reduce inflammation in autoimmune diseases, such as RA and type 1 diabetes.", "Our results from protein and gene therapy showed that hAAT is effective in delaying arthritis development in a mouse model of CIA. They indicate that hAAT has immunoregulatory and immunomodulatory effects and has great potential as a new treatment for RA. We also have shown that rAAV8 mediated gene therapy resulted in a reduced immune response to the transgene product. Future studies will focus on improvement of the therapeutic effect by optimizing the dose and timing of hAAT or rAAV8 vector delivery, and by combination therapy with other anti-arthritic drugs.", "hAAT: human Alpha-1 Antitrypsin; CIA: Collagen Induced Arthritis; IFA: Incomplete Freund's Adjuvant; CFA: Complete Freund's Adjuvant; RA: Rheumatoid Arthritis; NOD: Non Obese Diabetic; bCII: bovine type II Collagen; mCII: mouse type II Collagen; TNF-α: Tumor Necrosis Factor-alpha; IL: Interleukin; LPS: Lipopolysaccharide; PBMC: Peripheral Blood Mononuclear Cells; BAFF: B-cell Activation Factor of the TNF-α Family; rAAV: Recombinant Adeno-Associated Virus; MMP: Matrix- Metalloproteinase; ELISA: Enzyme-Linked Immunosorbent Assay", "Christian Grimstein and Sihong Song may be entitled to future patent royalties from technology described in this paper.", "CG conceived of the study, participated in its design, carried out animal experiments, cell proliferation assay, immunoassays, performed statistical analysis and drafted the manuscript. YKC conceived of the study, participated in its design and performed animal experiments and cell proliferation assay. CW helped performing cell proliferation assay, MS participated in discussion and helped to revise the manuscript, MA, MCT and MB participated in design and discussion of the study, SS conceived of the study participated in its design and helped to revise the manuscript. All authors read and approved the final manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adult", "Biomechanical Phenomena", "Exercise Test", "Gait", "Humans", "Male", "Walking" ]

[ "Introduction", "Participants", "Apparatus", "Procedures", "Stride intervals and kinematic variability", "Detrended Fluctuation Analysis", "Local dynamic stability", "Statistical analysis", "Results", "Treadmill effect", "Correlations", "Discussion", "Technical issues", "Differences between treadmill and overground walking", "Correlations between variability indicators", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Walking is a repetitive movement which is characterized by a low variability [1]. This motor skill requires not only conscious neuromotor tasks but also complex automated regulation, both interacting to produce steady gait pattern. Classically, gait variability (i.a. kinematic variability) has been assessed from the differences among the strides (Standard Deviation SD, coefficient of variation CV), i.e. each stride considered as an independent event resulting from a random process. However, this approach fails to account for the presence of feedback loops in the motor control of walking: the walking pattern at a given gait cycle may have consequences on subsequent strides. As a result, correlations between consecutive gait cycles and non-linear dependencies are expected.\nDuring the last decades, various new mathematical tools have been used to better characterise the non-linear features of gait variability. With the Detrended Fluctuation Analysis (DFA [2-4]) it has been observed that the stride interval (i.e. time to complete a gait cycle) at any time was related (in a statistical sense) to intervals at relatively remote times (persistent pattern over more than 100 strides). This dependence (memory effect) decayed in a power-law fashion, similar to scale-free, fractal-like phenomena (fractal dynamics [1,3-5]), also known as 1/fβ noise [6]).\nAnother non-linear approach was proposed to characterize the dynamic variability in continuous walking. The sensitivity of a dynamical system to small perturbations can be quantified by the system maximal Lyapunov exponent, which characterizes the average rate of divergence in pseudo-periodic processes [7]. This method allows to evaluate the ability of locomotor system to maintain continuous motion by accommodating infinitesimally small perturbations that occur naturally during walking [8]. This includes external perturbations induced by small variations in the walking surface, as well as internal perturbations resulting from the natural noise present in the neuromuscular system [8].\nMany theoretical questions are still open about the validity and application of these methods. For instance, DFA results are difficult to interpret [9], and no definitive conclusion on the presence of long range correlations should be drawn relying only on it. In addition, the underlying mechanism of long range correlations in stride interval is not fully understood [3,10]. West & Latka suggested that the observed scaling in inter-stride interval data may not be due to long-term memory alone, but may, in fact, be due partly to the statistics [11]. It was also suggested that the use of multi-fractal spectrum could be a better approach than mono-fractal analysis, such as DFA [12,13]. There are also several methodological issues to compute consistent and reliable stability index [14,15].\nIn parallel with the ongoing theoretical research on non-linear analysis of physiological time series, the use of non-linear bio-markers in applied clinical research has been already fruitful. In the field of human locomotion, it has been demonstrated that gait variability could serve as a sensitive and clinically relevant tool in the evaluation of mobility and the response to therapeutic interventions. For instance, gait variability (SD and dynamics) is altered in clinically relevant syndromes, such as falling and neuro-degenerative disease [16,17]. Gait instability measurement apparently predict falls in idiopathic elderly fallers [18]. Improvements in muscle function are associated with enhanced gait stability in elderly [19].\nMotorized treadmills are widely used in biomechanical studies of human locomotion. They allow the documentation of a large number of successive strides under controlled environment, with a selectable steady-state locomotion speed. In the rehabilitation field, treadmill walking is used in locomotor therapy, for instance with partial body weight support in spinal cord injury or stroke rehabilitation [20,21]. Since the classical work of Van Ingen Schenau [22], it is admitted that overground and treadmill locomotion are similar if treadmill belt speed is constant. Nevertheless, both walking types present small differences in kinematics [23,24], kinetics [25] and energetics [26]. It was also observed that treadmill locomotion induced shorter step lengths and higher cadences than walking on the floor at the same speed [26,27]. There is still a matter of debate to interpret such subtle differences [28,29].\nIt is obvious that treadmill walking (TW) induces specific kinaesthetic and perceptual information. Previous studies confirmed that vision plays a central role in the control of locomotion [30,31]. These differences in visual afferences between TW and Overground Walking (OW) may induce a modification in motor control, and consequently in gait variability.\nIn 2000, Dingwell et al. analyzed TW local dynamic stability (maximal Lyapunov exponent) in 10 healthy subjects [8,32]. They highlighted significant differences between TW and OW by evaluating local dynamic stability of lower limbs kinematics [8]. The effect was low in upper body accelerations. Later [32], they calculated more specifically short term stability and found a strong effect of TW in trunk accelerations. On the other hand, they found a greater kinematic variability at the lower limb level in OW as compared to TW, but no significant difference in trunk kinematics.\nIn 2005, Terrier et al. [1], by using high accuracy GPS, described low stride-to-stride variability of speed, step length and step duration in free walking. They observed that the constraint of rhythmical auditory signal (\"metronome walking\") did not alter kinematic variability, but modify the fractal dynamics (DFA) of the stride interval (anti-persistent pattern).\nBased on these previous works, the working hypothesis of the present article is 1) that the constraint of TW (constant speed, narrow pathway) may induce a less persistent pattern in the stride interval, by analogy to the constraint induced by a metronome; 2) that TW may increase the local dynamic stability of walking, due to the diminution of degrees of freedom in the more constrained artificial environment [32,33], 3) that, for the same reasons, TW may slightly reduce kinematic variability [32,33] 4) that no correlation exist between the 3 variability indexes, because they are related to different aspects of the locomotion process.\nThe purpose of the present study was to analyze, by using trunk accelerometry, differences between TW and OW in terms of stride-to-stride kinematic variability (SD), fractal dynamics (by DFA) and local dynamic stability (maximal Lyapunov exponent). In addition, we assessed the strength of the relationships between these variables (canonical correlation analysis).", "Twenty healthy male subjects, with no neurological deficit or orthopaedic impairment, participated to the study. Most of them were recruited among participants of a previous \"treadmill\" study implying only males subjects [34]. Their characteristics were (mean ± SD): age 35 ± 7 yr, body mass 79 ± 10 kg, and height 1.80 ± 0.06 m. All subjects were well trained to walk on a treadmill before the beginning of the study. The experimental protocol was approved by the local ethics committee (commission d'éthique du Valais).", "The motion sensor (Physilog system, BioAGM, Switzerland [35]) was a triaxial accelerometer connected to a data logger recording body accelerations in medio-lateral (ML), vertical (V) and antero-posterior (AP) directions. The dimensions of the logger were 130 × 68 × 30 mm and the weight was 285 g. The accelerometers are piezoresistive sensors coupled with amplifiers (± 5 g, 500 mV/g) and mounted on a belt. The signals were sampled at 200 Hz with 12-bit resolution. After each experiment, the data were downloaded to a PC computer and converted in earth acceleration units (g) according to a previous calibration. Data analysis was then performed by using Matlab (Mathworks, Natick MA, USA) and Stata 11.0 (StataCorp LP, TX, USA)", "The subjects performed 10 min. treadmill walking (TW) and 10 min. overground walking (OW) in a random order. A rest period of five minutes (sitting still) was imposed between the two trials. The motor-driven treadmill was a Technogym, (Runrace, Italy). The imposed speed was 1.25 m/s (4.5 km/h) for all subjects: in the context of a previous study [34], we assessed average running and walking preferred speed on the same treadmill in 88 male subjects; an average of 1.26 ± 0.13 m/s was observed. A thirty second warm-up was performed before the beginning of the measurement. For the OW test, the subjects walked along a standardized 800 m indoor circuit along hospital corridors and halls. The circuit exhibited only 90° turns. A large part (about 400 m) of the circuit was constituted by a long corridor. Other people working in the hospital were present in the halls. Hence, the OW trials mimicked actual condition of walking. Subects were asked to walk at their Preferred Walking Speed (PWS) with a regular pace. Under both conditions, the accelerometer was attached to the low back (L4-L5 region) with an elastic belt, and the logger was worn on the side of the body. Subjects wore their own low-rise comfortable walking shoes.", "Five seconds were removed at the beginning and at the end of the 10 min. acceleration measurements in order to avoid non-stationary periods. Heel strike was detected in the raw acceleration AP signal with a peak detection method designed to minimize the risk of false step detection: first, we generated a low-pass filtered version of the signal (4 order Butterworth, 3 Hz, zero-phase filtering). The time of each local minimum was detected. By superimposing the Filtered Signal (FS) to the original, Unfiltered Signal (US), we tracked the nearest peak in US of each local minimum in FS. US peak time was then chosen as the limit between two steps (Figure 1A). The strides were defined as two consecutive steps. On average, the number of strides was 543 per trial.\nMethod: Step detection, stride intervals and Detrended Fluctuation Analysis. One subject performed 10 min of free walking. A: 2.5s sample of the antero-posterior acceleration signal; red dotted line is a low pass filtered (<3 Hz) version of the raw signal (black continuous line). Cross and black circle indicate how the algorithm specifically detect the heel strike (see method section for further explanation). The duration of two consecutive steps is defined as stride interval. B: Time series of stride intervals during the 10 min walking test. Average stride time (mean) and CV (SD/mean * 100) is also presented. C: Detredend Fluctuation Analysis (DFA). The fractal dynamics of the time series (B) is characterized by the scaling exponent α, computed by comparing the fluctuation (F(n)) at different scales (n) in a log-log plot.\nTime series of the stride intervals were used to compute a traditional variability index (Coefficient of Variation of the stride time, CV = SD/Mean*100, Figure 1B). Moreover, the variability of the acceleration pattern among strides was evaluated as follows (Figure 2): each stride was normalized to 200 sample points by using a polyphase filter implementation (Matlab command Resample); the average stride-to-stride Standard Deviation across all data points ((SD(i) ∀ i ∈ [1 ....200])) was evaluated (MeanSD = 〈SD(i)〉).\nMethod: variability, MeanSD. One subject (same as in Figure 1) performed 10 min of free walking. Each stride (see Figure 1A) was normalized to 200 samples (0% to 100% gait cycle). Top: Average acceleration pattern of the normalized strides (N = 513). Bottom: Standard Deviation (SD) of the normalized strides (N = 513). MeanSD is the average SD of the 200 samples.", "The presence of long range correlations in the time series of stride intervals (fractal dynamics) was assessed by the use of the non-linear DFA method. Strictly speaking, this non-linear method should be used in addition to other statistical tools to definitively conclude that a process is a true 1/fβ noise with power-law decrease of long range auto-correlations [6,9]. However, DFA has been successfully used as relevant biomarker in numerous studies [1,16,17,36,37]. Detrended Fluctuation Analysis is based on a classic root-mean square analysis of a random walk, but is specifically designed to be less likely affected by nonstationarities. Full details of the methodology are published elsewhere [1-4]. In short, the integrated time series of length N is divided into boxes of equal length, n. In each box of length n, a least squares line is fit to the data (representing the trend in that box). The y coordinate of the straight line segments is denoted by yn(k). Next, the integrated time series, y(k), was detrended, by subtracting the local trend, yn(k), in each box. The root-mean-square fluctuation of this integrated and detrended time series is calculated by\n\n\n(1)\n\n\nF\n(\nn\n)\n=\n\n\n\n1\nN\n\n\n\n∑\n\nk\n=\n1\n\nN\n\n\n\n\n[\ny\n(\nk\n)\n−\n\ny\nn\n\n(\nk\n)\n]\n\n2\n\n\n\n\n\n\n\n\n\nThis computation is repeated over all box sizes (from 4 to 200) to characterize the relationship between F(n), the average fluctuation, and the box size, n. The fluctuations can be characterized by the scaling exponent α, which is the slope of the line relating log F(n) to log(n) (F(n) ~ nα), Figure 1C). Long range correlations are present in the original time series when α lies between 0.5 and 1 [3,4].\nIn a finite length time series, an uncorrelated process could exhibit \"by chance\" a scaling exponent different from the theoretical 0.5 value. To statistically differentiate the stride time series from a random uncorrelated process, we applied the surrogate data method [1,3]. This method increases the confidence that the analyzed series exhibits long-range correlation. Twenty different surrogate data sets were generated by shuffling the original time series in a random order. On each data set, DFA analysis was performed to calculate α value. The standard deviation and mean of this sample was calculated and compared to α exponent of the original series. The result is considered significant if the original α is 2 standard deviation away from the mean of the surrogate data set.", "The method for quantifying the local dynamical stability of the gait by using largest Lyapunov exponent has been extensively described in literature [8]. It examines structural characteristics of a time series that is embedded in an appropriately constructed state space. A valid state space contains a sufficient number of independent coordinates to define the state of the system unequivocally [38]. According to the Takens' theorem, an appropriate state space can be reconstructed from a single time series using the original data and its time delayed copies (figure 3A) [38].\nMethod: dynamic stability, maximal Lyapunov exponent. A: Two dimensional state space of the antero-posterior acceleration signal (5s) reconstructed from the original data set and its time delayed copy (Δt = 11 samples). B: Magnification of the state space. An initial local perturbation at dj(0) diverge across i time steps as measured by dj(i). C: Short term (λS*) and long term (λL*) finite-time maximal Lyapunov exponent computed from average logarithmic divergence.\n\n\n(2)\n\n\nX\n(\nt\n)\n=\n[\nx\n(\nt\n)\n,\nx\n(\nt\n+\nT\n)\n,\nx\n(\nt\n+\n2\nT\n)\n,\n…\n,\nx\n(\nt\n+\n(\n\nd\nE\n\n−\n1\n)\nT\n]\n\n\n\n\nWhere X(t) is the dE-dimensional state vector, x(t) are the original data, T is the time delay, and dE is the embedding dimension. The time delays (T) were calculated individually for each of the 120 acceleration data set (3-axis, 2 conditions, and 20 individuals) from the first minimum of the Average Mutual Information (AMI) function [8,39]. Embedding dimensions (dE) were computed from a Global False Nearest Neighbors (GFNN) analysis [8,40]. Because the result was similar for all acceleration time series, we use a constant dimension (dE = 6) [8,32]]. The Lyapunov exponent is the mean exponential rate of divergence of initially nearby points in the reconstructed space (Figure 3B). Because the determination of the maximal Lypunov exponent requires intensive computing power, 7 min of the 10 min walking test (from 1.5 to 8.5 min.) was selected and the raw data were down-sampled to 100 Hz. The determination of the Lyapunov exponent was then achieved by using the algorithm introduced by Rosenstein and colleagues [7], which provided dedicated software to compute divergence as a function of time in finite-time series [41] (Figure 3B). The maximum finite-time Lyapunov exponents (λ*) were estimated from the slopes of linear fits in the divergence diagrams (Figure 3C). Strictly speaking, because divergence diagrams (Figure 3C) are non-linear, multiple slopes could be defined and so no true single maximum Lyapunov exponent exists. The slopes (exponents) quantify local divergence (and hence stability) of the observed dynamics at different time scale, and should not be interpreted as a classical maximal Lyapunov exponent in chaos theory.\nSince each subject exhibited a different average step frequency, the time was normalized by average stride time for each subject and each condition (Figure 3C). As suggested by Dingwell and colleagues [32], we use two different time scales for assessing short-term and long-term dynamic stability: short term exponents (λS*) was computed over the first stride (0 to 1), and long term exponents (λL*) between 4 and 10 strides (Figure 3C).", "Mean and Standard Deviation (SD) were computed to describe the data (table 1). Ninety-five percent Confidence Intervals (CI) were calculated as ± 1.96 times the Standard Error of the Mean (SEM, N = 20).\nThe effect size of TW as compared to OW was expressed in both absolute (mean difference) and standardized (mean difference divided by SD) terms. The standardized effect size was the Hedge's g, which is a modified version of the Cohen's d for inferential measure [42]. Paired t-tests between OW and TW were performed, and the p-values are shown in the last column of table 1. The precision of the effect sizes was estimated with CI (Figure 4). CI were ± 1.96 times the asymptotic estimates of the standard error (SE) of g [42]. The arbitrary limit of 0.5 was uses to delineate small effect size, as defined by Cohen [42]. The extent of the data (quartiles and median) and individual differences between conditions are shown in Figure 5 for λ*. In order to facilitate results interpretation by reducing the risk of type I statistical error, a Hotelling T2 test was used. This is a multivariate generalization of paired t-test [43]. The null hypothesis is that a vector of p differences is equal to a vector of zeros. Two multivariate sets were tested: meanSD (p = 3) and λ* (p = 6).\nCanonical correlation analyses (CCA, table 2&3) were performed in order to assess the strength of the relationships between different sets of variables [43]. This multivariate method allows one to find linear combinations (variates) in two sets of variables, which have maximum correlation (canonical correlation coefficient or canonical root) with each other. For each condition (OW and TW), two sets of p variables were analyzed: kinematic variability (set#1, p = 3) including MeanSD in ML, V and AP directions, and dynamic stability (set#2, p = 6), including short term and long term lyapunov exponent (λS*, λL*) in ML, V and AP directions. In addition, α scaling exponent was also analyzed with the same method vs. set#1 and set#2. In this case, CCA is equivalent to multiple regression analysis. Significance of the canonical correlations was assessed with the Wilks' lambda statistics.\nTo enhance the interpretation of CCA, different parameters were computed: the standardized canonical weights are the linear coefficients for each set after Z-transform of the variables; canonical loadings are the correlation coefficients between each variable and their respective linear composites; redundancy expresses the amount of variance in one set explained by a linear composite of the other set.", "[SUBTITLE] Treadmill effect [SUBSECTION] As presented in table 1, TW did not modify the stride-to-stride kinematic variability of normalized acceleration pattern, either considering multivariate T2 statistics (p = 0.87) or individual results for each direction. TW was on average performed at slightly lower cadence than Overground Walking (OW, 3% relative difference). The variability of stride interval was similar under both conditions. DFA of stride intervals revealed that TW changed the fractal dynamics of walking (-11% relative difference). Globally, multivariate analysis showed that the data are compatible with the assumption that TW modified dynamic stability of the gait (T2 (6, 20) p = 0.0002). Five from six particular λ* exponents exhibited significant differences.\nComparison between Overground and Treadmill Walking\nThe Descriptive statistics of variability indexes are expressed as mean, Standard Deviation (SD) and 95% Confidence Interval (mean ± 1.96 times the Standard Error of the Mean). The effect size is given as Absolute (Abs.) and Normalized (Norm.) values, i.e. respectively the difference between Overground (OW) and Treadmill (TW) conditions (Abs.) and the difference normalized by SD (Hedge's g). The t-test column shows the p values of paired t-tests between TW and OW conditions. T2-test is the Hotelling multivariate test by regrouping MeanSD and λ*. Significant results (p < 0.05) are printed in bold. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior, i.e. the 3 directions of the triaxial accelerometer.\nFigure 4 shows the accuracy of the effect size estimation. Non-linear estimators of gait variability (α, λ*) exhibit mostly medium effect size.\nDifferences between overground and treadmill walking. Effect size and confidence intervals. Black circles are the standardized effect size (Hedge's g), as reported in table 1. Horizontal lines are the 95% confidence intervals. The arbitrary limit of 0.5 (vertical dotted line) corresponds to a medium effect as defined by Cohen.\nFigure 5 shows the individual results of the local dynamic stability (λ*). Stability was clearly increased (lower λ*) for a majority of subjects except for long-range Antero-Posterior stability λL*.\nIndividual changes of dynamic stability (λ*). Lyapunov exponent λL* and λS* of the 20 subjects are presented for Overground Walking (OW) and Treadmill Walking (TW). Discontinuous lines join OW and TW results. Boxplots show the quartiles and the median.\nFigure 6 presents the individual results of surrogate testing of fractal dynamics. The response to TW was not homogenous among subjects. Four subjects (20%) exhibited a significant turn of long range correlations to uncorrelated pattern. For ten more subjects (50%), a reduction was observed (more than 0.05), but outside the significant limits.\nDetrended Fluctuation Analysis: surrogate data tests. The time series of stride intervals (Figure 1B) of each subject (#1 to #20) were analyzed by DFA (figure 1C) to determine the scaling exponent α indicating the presence of a long range correlation pattern in stride intervals. Black and white circles are respectively the scaling exponent for Overground Walking (OW) and Treadmill Walking (TW). Each time series was randomly shuffled twenty times to produce 20 surrogate time series. The average of these series is near 0.5 (random process with no correlation). The vertical bars show the extent of 2 times the SD of the 20 surrogate time series. Scaling exponent larger than this value can be considered significantly different from a random uncorrelated series.\nAs presented in table 1, TW did not modify the stride-to-stride kinematic variability of normalized acceleration pattern, either considering multivariate T2 statistics (p = 0.87) or individual results for each direction. TW was on average performed at slightly lower cadence than Overground Walking (OW, 3% relative difference). The variability of stride interval was similar under both conditions. DFA of stride intervals revealed that TW changed the fractal dynamics of walking (-11% relative difference). Globally, multivariate analysis showed that the data are compatible with the assumption that TW modified dynamic stability of the gait (T2 (6, 20) p = 0.0002). Five from six particular λ* exponents exhibited significant differences.\nComparison between Overground and Treadmill Walking\nThe Descriptive statistics of variability indexes are expressed as mean, Standard Deviation (SD) and 95% Confidence Interval (mean ± 1.96 times the Standard Error of the Mean). The effect size is given as Absolute (Abs.) and Normalized (Norm.) values, i.e. respectively the difference between Overground (OW) and Treadmill (TW) conditions (Abs.) and the difference normalized by SD (Hedge's g). The t-test column shows the p values of paired t-tests between TW and OW conditions. T2-test is the Hotelling multivariate test by regrouping MeanSD and λ*. Significant results (p < 0.05) are printed in bold. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior, i.e. the 3 directions of the triaxial accelerometer.\nFigure 4 shows the accuracy of the effect size estimation. Non-linear estimators of gait variability (α, λ*) exhibit mostly medium effect size.\nDifferences between overground and treadmill walking. Effect size and confidence intervals. Black circles are the standardized effect size (Hedge's g), as reported in table 1. Horizontal lines are the 95% confidence intervals. The arbitrary limit of 0.5 (vertical dotted line) corresponds to a medium effect as defined by Cohen.\nFigure 5 shows the individual results of the local dynamic stability (λ*). Stability was clearly increased (lower λ*) for a majority of subjects except for long-range Antero-Posterior stability λL*.\nIndividual changes of dynamic stability (λ*). Lyapunov exponent λL* and λS* of the 20 subjects are presented for Overground Walking (OW) and Treadmill Walking (TW). Discontinuous lines join OW and TW results. Boxplots show the quartiles and the median.\nFigure 6 presents the individual results of surrogate testing of fractal dynamics. The response to TW was not homogenous among subjects. Four subjects (20%) exhibited a significant turn of long range correlations to uncorrelated pattern. For ten more subjects (50%), a reduction was observed (more than 0.05), but outside the significant limits.\nDetrended Fluctuation Analysis: surrogate data tests. The time series of stride intervals (Figure 1B) of each subject (#1 to #20) were analyzed by DFA (figure 1C) to determine the scaling exponent α indicating the presence of a long range correlation pattern in stride intervals. Black and white circles are respectively the scaling exponent for Overground Walking (OW) and Treadmill Walking (TW). Each time series was randomly shuffled twenty times to produce 20 surrogate time series. The average of these series is near 0.5 (random process with no correlation). The vertical bars show the extent of 2 times the SD of the 20 surrogate time series. Scaling exponent larger than this value can be considered significantly different from a random uncorrelated series.\n[SUBTITLE] Correlations [SUBSECTION] Table 2 shows the correlation matrix (Perason's r) of the variables under both conditions. It can be observed that correlations exist between the same variables measured along different axes (for instance MeanSD ML vs. MeanSD V, r = 0.92), what makes difficult the global interpretation of potential correlation among the different variability indexes.\nCorrelation matrix\nPearson's r correlation coefficients between the variables. SD = Mean Standard Deviation (MeanSD). λS* = maximal Lyapunov exponent, short term dynamic stability. λL* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior. Significant correlation are bold printed (p < 0.05).\nIn table 3, the results of 6 CCA are shown in details in order to explore global correlation hypotheses. The data seem compatible with the hypothesis that a negative correlation exists between kinematic variability (MeanSD) and local dynamic stability (λ*) under TW condition. Namely, two significant canonical roots (R2 = 0.88 and 0.62) indicates that the canonical variates share an important variance. In addition, the canonical loadings show that the canonical model extract a substantial portion of the variance from the variables (70% from the set#1 and 27% from the set#2). Finally, the redundancy analysis reveals that at least 70% of the variance of the set#2 (stability) can be explained by the set#1 (kinematic variability). The five other CCA did not produce clear evidence for significant relationship between the analyzed sets of variables. Three CCA showed low and non significant canonical roots. Two CCA exhibited barely significant correlation, but the analysis of loadings showed that the canonical model did not explain a large part of the variances in the sets.\nCanonical Correlation Analysis (CCA)\nCanonical correlation analysis between 6 sets of variables. SD = Mean Standard Deviation. λS* = maximal Lyapunov exponent, short term dynamic stability. λ:* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior.\nTable 2 shows the correlation matrix (Perason's r) of the variables under both conditions. It can be observed that correlations exist between the same variables measured along different axes (for instance MeanSD ML vs. MeanSD V, r = 0.92), what makes difficult the global interpretation of potential correlation among the different variability indexes.\nCorrelation matrix\nPearson's r correlation coefficients between the variables. SD = Mean Standard Deviation (MeanSD). λS* = maximal Lyapunov exponent, short term dynamic stability. λL* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior. Significant correlation are bold printed (p < 0.05).\nIn table 3, the results of 6 CCA are shown in details in order to explore global correlation hypotheses. The data seem compatible with the hypothesis that a negative correlation exists between kinematic variability (MeanSD) and local dynamic stability (λ*) under TW condition. Namely, two significant canonical roots (R2 = 0.88 and 0.62) indicates that the canonical variates share an important variance. In addition, the canonical loadings show that the canonical model extract a substantial portion of the variance from the variables (70% from the set#1 and 27% from the set#2). Finally, the redundancy analysis reveals that at least 70% of the variance of the set#2 (stability) can be explained by the set#1 (kinematic variability). The five other CCA did not produce clear evidence for significant relationship between the analyzed sets of variables. Three CCA showed low and non significant canonical roots. Two CCA exhibited barely significant correlation, but the analysis of loadings showed that the canonical model did not explain a large part of the variances in the sets.\nCanonical Correlation Analysis (CCA)\nCanonical correlation analysis between 6 sets of variables. SD = Mean Standard Deviation. λS* = maximal Lyapunov exponent, short term dynamic stability. λ:* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior.", "As presented in table 1, TW did not modify the stride-to-stride kinematic variability of normalized acceleration pattern, either considering multivariate T2 statistics (p = 0.87) or individual results for each direction. TW was on average performed at slightly lower cadence than Overground Walking (OW, 3% relative difference). The variability of stride interval was similar under both conditions. DFA of stride intervals revealed that TW changed the fractal dynamics of walking (-11% relative difference). Globally, multivariate analysis showed that the data are compatible with the assumption that TW modified dynamic stability of the gait (T2 (6, 20) p = 0.0002). Five from six particular λ* exponents exhibited significant differences.\nComparison between Overground and Treadmill Walking\nThe Descriptive statistics of variability indexes are expressed as mean, Standard Deviation (SD) and 95% Confidence Interval (mean ± 1.96 times the Standard Error of the Mean). The effect size is given as Absolute (Abs.) and Normalized (Norm.) values, i.e. respectively the difference between Overground (OW) and Treadmill (TW) conditions (Abs.) and the difference normalized by SD (Hedge's g). The t-test column shows the p values of paired t-tests between TW and OW conditions. T2-test is the Hotelling multivariate test by regrouping MeanSD and λ*. Significant results (p < 0.05) are printed in bold. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior, i.e. the 3 directions of the triaxial accelerometer.\nFigure 4 shows the accuracy of the effect size estimation. Non-linear estimators of gait variability (α, λ*) exhibit mostly medium effect size.\nDifferences between overground and treadmill walking. Effect size and confidence intervals. Black circles are the standardized effect size (Hedge's g), as reported in table 1. Horizontal lines are the 95% confidence intervals. The arbitrary limit of 0.5 (vertical dotted line) corresponds to a medium effect as defined by Cohen.\nFigure 5 shows the individual results of the local dynamic stability (λ*). Stability was clearly increased (lower λ*) for a majority of subjects except for long-range Antero-Posterior stability λL*.\nIndividual changes of dynamic stability (λ*). Lyapunov exponent λL* and λS* of the 20 subjects are presented for Overground Walking (OW) and Treadmill Walking (TW). Discontinuous lines join OW and TW results. Boxplots show the quartiles and the median.\nFigure 6 presents the individual results of surrogate testing of fractal dynamics. The response to TW was not homogenous among subjects. Four subjects (20%) exhibited a significant turn of long range correlations to uncorrelated pattern. For ten more subjects (50%), a reduction was observed (more than 0.05), but outside the significant limits.\nDetrended Fluctuation Analysis: surrogate data tests. The time series of stride intervals (Figure 1B) of each subject (#1 to #20) were analyzed by DFA (figure 1C) to determine the scaling exponent α indicating the presence of a long range correlation pattern in stride intervals. Black and white circles are respectively the scaling exponent for Overground Walking (OW) and Treadmill Walking (TW). Each time series was randomly shuffled twenty times to produce 20 surrogate time series. The average of these series is near 0.5 (random process with no correlation). The vertical bars show the extent of 2 times the SD of the 20 surrogate time series. Scaling exponent larger than this value can be considered significantly different from a random uncorrelated series.", "Table 2 shows the correlation matrix (Perason's r) of the variables under both conditions. It can be observed that correlations exist between the same variables measured along different axes (for instance MeanSD ML vs. MeanSD V, r = 0.92), what makes difficult the global interpretation of potential correlation among the different variability indexes.\nCorrelation matrix\nPearson's r correlation coefficients between the variables. SD = Mean Standard Deviation (MeanSD). λS* = maximal Lyapunov exponent, short term dynamic stability. λL* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior. Significant correlation are bold printed (p < 0.05).\nIn table 3, the results of 6 CCA are shown in details in order to explore global correlation hypotheses. The data seem compatible with the hypothesis that a negative correlation exists between kinematic variability (MeanSD) and local dynamic stability (λ*) under TW condition. Namely, two significant canonical roots (R2 = 0.88 and 0.62) indicates that the canonical variates share an important variance. In addition, the canonical loadings show that the canonical model extract a substantial portion of the variance from the variables (70% from the set#1 and 27% from the set#2). Finally, the redundancy analysis reveals that at least 70% of the variance of the set#2 (stability) can be explained by the set#1 (kinematic variability). The five other CCA did not produce clear evidence for significant relationship between the analyzed sets of variables. Three CCA showed low and non significant canonical roots. Two CCA exhibited barely significant correlation, but the analysis of loadings showed that the canonical model did not explain a large part of the variances in the sets.\nCanonical Correlation Analysis (CCA)\nCanonical correlation analysis between 6 sets of variables. SD = Mean Standard Deviation. λS* = maximal Lyapunov exponent, short term dynamic stability. λ:* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior.", "The purpose of the present study was to analyze three gait variability indexes under two walking conditions in order to highlight modifications induced by motorized treadmill and to analyze the relationship between the indexes.\nAccording to the working hypothesis, the results are summarized as follows:\n1) As compared to Overground Walking (OW), Treadmill Walking (TW) significantly reduced the average scaling exponent (lower α), but did not reverse the correlated pattern to a random or anti-persistent pattern in a majority of subjects.\n2) TW significantly increased local dynamic stability (lower λ*).\n3) TW did not significantly modify the kinematic variability (MeanSD).\n4) No evident relationship was observed between variability indexes during OW at preferred walking speed, but in TW significant negative correlation was found between kinematic variability (MeanSD) and stability (λ*).\nOverall, Conventional variability analysis (MeanSD) failed to report differences between OW and TW, whereas non-linear approaches were able to show significant changes. The variability indexes were poorly correlated together (with one exception), which might signify that each index was related to a different aspect of motor control.\n[SUBTITLE] Technical issues [SUBSECTION] For the present study, portable trunk accelerometry was chosen because it offers the possibility to record long-term free walking. Hence, the results concern the gait stability measured from accelerations of the low-back. Comparisons with other results should take into account that that the different gait stability studies use different kinematic variables (acceleration [7,10], positions [44], angle [8]) and different body location (thorax, head, knee, and ankle) to assess λ*. We found λ* similar to those measured by others [8,32], suggesting that the results are rather independent on the measurements methods.\nIn fractal dynamics studies, the first step is the detection of the periodic pattern of the gait in order to compute time series of stride intervals. Several methodologies have been used to measure long-term time series of stride intervals, such as foot switches [3,5], goniometer [45], video analysis [46], or high accuracy GPS [1]. Because the same variable is used (i.e. time duration of the gait cycle) for DFA analyses, data from different studies are probably comparable.\nIn order to increase the likelihood to point out significant correlations among variability indexes, we designed the experiment to obtain a substantial degree of standardization: we imposed the same speed (1.25 m/s, 4.5 km/h) for all subjects on the treadmill. This speed was chosen on the basis of a previous experiment (partially published yet [34]), which showed that the preferred speed in the same experimental conditions (same room, same treadmill) was 1.26 ± 0.13 m/s (n = 88). Similar values are found in the literature: 1.25 m/s (n = 8) [47], 1.19 m/s (n = 26) [48].\nWalking speed was not standardized between TW and OW, as in other studies [32]. However, by selecting treadmill speed at the same speed of overground preferred speed, the results would be that subjects walk at higher speed than their preferred speed on the treadmill. Several studies showed a substantial difference between both conditions: Dal et al. [48] demonstrated that preferred walking speed determined on a treadmill is slower than overground (21% relative difference); Marsh et al. [49]showed that, when older adults were allowed to choose a preferred walking pace, they walked faster (+61%), used longer strides, and had a faster rate walking overground than when they walked on a treadmill. As a result, speed normalization could introduce unwanted bias. Our experimental design was therefore a compromise, which standardized speed among subjects in TW condition, but also which selected walking speed close to preferred speed, making both OW and TW conditions comparable.\nIn addition, Indirect clues seem to indicate that TW and OW conditions were quite similar: 1) stride time (which is related to walking speed) were close (3% difference, small effect size), 2) stride time variability (CV) was the same (no significant differences), 3) no correlation was observed between stride time and other parameters (results not shown),\nFor the present study, portable trunk accelerometry was chosen because it offers the possibility to record long-term free walking. Hence, the results concern the gait stability measured from accelerations of the low-back. Comparisons with other results should take into account that that the different gait stability studies use different kinematic variables (acceleration [7,10], positions [44], angle [8]) and different body location (thorax, head, knee, and ankle) to assess λ*. We found λ* similar to those measured by others [8,32], suggesting that the results are rather independent on the measurements methods.\nIn fractal dynamics studies, the first step is the detection of the periodic pattern of the gait in order to compute time series of stride intervals. Several methodologies have been used to measure long-term time series of stride intervals, such as foot switches [3,5], goniometer [45], video analysis [46], or high accuracy GPS [1]. Because the same variable is used (i.e. time duration of the gait cycle) for DFA analyses, data from different studies are probably comparable.\nIn order to increase the likelihood to point out significant correlations among variability indexes, we designed the experiment to obtain a substantial degree of standardization: we imposed the same speed (1.25 m/s, 4.5 km/h) for all subjects on the treadmill. This speed was chosen on the basis of a previous experiment (partially published yet [34]), which showed that the preferred speed in the same experimental conditions (same room, same treadmill) was 1.26 ± 0.13 m/s (n = 88). Similar values are found in the literature: 1.25 m/s (n = 8) [47], 1.19 m/s (n = 26) [48].\nWalking speed was not standardized between TW and OW, as in other studies [32]. However, by selecting treadmill speed at the same speed of overground preferred speed, the results would be that subjects walk at higher speed than their preferred speed on the treadmill. Several studies showed a substantial difference between both conditions: Dal et al. [48] demonstrated that preferred walking speed determined on a treadmill is slower than overground (21% relative difference); Marsh et al. [49]showed that, when older adults were allowed to choose a preferred walking pace, they walked faster (+61%), used longer strides, and had a faster rate walking overground than when they walked on a treadmill. As a result, speed normalization could introduce unwanted bias. Our experimental design was therefore a compromise, which standardized speed among subjects in TW condition, but also which selected walking speed close to preferred speed, making both OW and TW conditions comparable.\nIn addition, Indirect clues seem to indicate that TW and OW conditions were quite similar: 1) stride time (which is related to walking speed) were close (3% difference, small effect size), 2) stride time variability (CV) was the same (no significant differences), 3) no correlation was observed between stride time and other parameters (results not shown),\n[SUBTITLE] Differences between treadmill and overground walking [SUBSECTION] Kinematic variability, fractal dynamics (DFA) and local dynamic stability (Lyapunov exponents) quantify different aspects of locomotor control [32]. Kinematic variability describes the range in which the locomotor system operates. DFA quantify temporal dynamics of discrete events (i.e stride interval) over hundreds of consecutives strides; it assesses the presence of long-range correlations between strides, and hence analyzes the characteristics of feedbacks in locomotor control. Lyapunov exponents quantify the temporal dynamics in continuous time based on the theory of deterministic chaos; it evaluates the degree of divergence in the signal, and hence the resilience of the locomotor system to small perturbations. Therefore, it can be expected that these variability indexes did not react in the same way under various conditions.\nThese assumptions were experimentally verified in various studies that observed changes of λ* and kinematic variability between different experimental conditions or between different populations. For instance it was observed that patients with peripheral neuropathy present altered dynamic stability but normal kinematic variability [33,50]. Other investigators have shown that an exercise training intervention in elderly people could improve dynamic stability but not decrease kinematic variability [19].\nDespite differences in the method of measurement (low-back vs. thorax acceleration) and in the experimental design (speed normalization), our results are generally in accordance with the results of Dingwell et al [32]. They analyzed only 10 healthy individuals, therefore statistical significance for small effects was more difficult to reach than in the present study. They showed a significant treadmill effect in short-term stability (lower λS*). A slight but not significant effect for long-term vertical stability (lower λL*) was found. They observed that kinematic variability (MeanSD) for upper body accelerations was generally greater for OW than TW, but this trend was only significant for antero-posterior accelerations. They explained that underlying causes of differences between TW and OW were unclear: on one hand, the motorized treadmill imposed a constant nominal speed on the subjects and constrained them to walk along a much narrower and straighter path than during OW; but on the other hand, differences may have been induced by intra-stride fluctuations in treadmill belt speed, differences in mechanical compliance between the walking surfaces, and changes in visual and vestibular perceptual information. In light of the results of the present study, we hypothesize that motor control is able to maintain the same range of kinematic variability in both TW and OW conditions (same kinematic variability), probably because of compensating effects: in TW, destabilizing factors (intra-stride belt speed fluctuations, disturbing mechanical compliance, alteration of perceptual information) are balanced by stabilizing factors (constant speed, narrow and straight path). Conversely, motor control strategy adapting the gait to TW seems to specifically alter non-linear dependencies among consecutive strides: the stabilizing factors override the destabilizing ones.\nIn a subsequent study, Dingwell & Marin [51] analyzed speed effect on dynamical stability (λS* and λL*) and kinematic variability (MeanSD). Walking speed was normalized by individual PWS on a treadmill. Speed range was 0.6PW to 1.4PWS by steps of 0.2. They found significant speed effect for both λ* and MeanSD: however the effect was small for 0.8-1.2 PWS. Under our experimental conditions [34], we observed that inter-indivudual variability of PWS on the treadmill was low: 90% of individuals walked in the range of 0.87-1.13 mean PWS. As a result, the speed effect among individuals in the present study was probably low. This is also indirectly confirmed by the low inter-individual variability of stride duration (CV = 6%).\nFractal dynamics of stride intervals has been extensively studied by Hausdorff et al. [36]. Them and other [1,3,52] have observed that constrained walking (paced cadence with a metronome), deeply modified the scaling exponent. By analogy, because treadmill also constraints the gait by imposing a constant speed, a similar effect could be expected. The results of the present study showed, in a majority of subjects, a lowering of scaling exponent to a less correlated pattern. The effect was not as strong as with paced walking [1]. The explanation could be that treadmill constrained walking speed, while metronome constrained walking pace; it could be hypothesized that the adaptation of locomotor control to external cues specifically modify correlation pattern of the constrained walking parameter, as suggested by the results of Terrier et al. [1], but this remains to be investigated.\nKinematic variability, fractal dynamics (DFA) and local dynamic stability (Lyapunov exponents) quantify different aspects of locomotor control [32]. Kinematic variability describes the range in which the locomotor system operates. DFA quantify temporal dynamics of discrete events (i.e stride interval) over hundreds of consecutives strides; it assesses the presence of long-range correlations between strides, and hence analyzes the characteristics of feedbacks in locomotor control. Lyapunov exponents quantify the temporal dynamics in continuous time based on the theory of deterministic chaos; it evaluates the degree of divergence in the signal, and hence the resilience of the locomotor system to small perturbations. Therefore, it can be expected that these variability indexes did not react in the same way under various conditions.\nThese assumptions were experimentally verified in various studies that observed changes of λ* and kinematic variability between different experimental conditions or between different populations. For instance it was observed that patients with peripheral neuropathy present altered dynamic stability but normal kinematic variability [33,50]. Other investigators have shown that an exercise training intervention in elderly people could improve dynamic stability but not decrease kinematic variability [19].\nDespite differences in the method of measurement (low-back vs. thorax acceleration) and in the experimental design (speed normalization), our results are generally in accordance with the results of Dingwell et al [32]. They analyzed only 10 healthy individuals, therefore statistical significance for small effects was more difficult to reach than in the present study. They showed a significant treadmill effect in short-term stability (lower λS*). A slight but not significant effect for long-term vertical stability (lower λL*) was found. They observed that kinematic variability (MeanSD) for upper body accelerations was generally greater for OW than TW, but this trend was only significant for antero-posterior accelerations. They explained that underlying causes of differences between TW and OW were unclear: on one hand, the motorized treadmill imposed a constant nominal speed on the subjects and constrained them to walk along a much narrower and straighter path than during OW; but on the other hand, differences may have been induced by intra-stride fluctuations in treadmill belt speed, differences in mechanical compliance between the walking surfaces, and changes in visual and vestibular perceptual information. In light of the results of the present study, we hypothesize that motor control is able to maintain the same range of kinematic variability in both TW and OW conditions (same kinematic variability), probably because of compensating effects: in TW, destabilizing factors (intra-stride belt speed fluctuations, disturbing mechanical compliance, alteration of perceptual information) are balanced by stabilizing factors (constant speed, narrow and straight path). Conversely, motor control strategy adapting the gait to TW seems to specifically alter non-linear dependencies among consecutive strides: the stabilizing factors override the destabilizing ones.\nIn a subsequent study, Dingwell & Marin [51] analyzed speed effect on dynamical stability (λS* and λL*) and kinematic variability (MeanSD). Walking speed was normalized by individual PWS on a treadmill. Speed range was 0.6PW to 1.4PWS by steps of 0.2. They found significant speed effect for both λ* and MeanSD: however the effect was small for 0.8-1.2 PWS. Under our experimental conditions [34], we observed that inter-indivudual variability of PWS on the treadmill was low: 90% of individuals walked in the range of 0.87-1.13 mean PWS. As a result, the speed effect among individuals in the present study was probably low. This is also indirectly confirmed by the low inter-individual variability of stride duration (CV = 6%).\nFractal dynamics of stride intervals has been extensively studied by Hausdorff et al. [36]. Them and other [1,3,52] have observed that constrained walking (paced cadence with a metronome), deeply modified the scaling exponent. By analogy, because treadmill also constraints the gait by imposing a constant speed, a similar effect could be expected. The results of the present study showed, in a majority of subjects, a lowering of scaling exponent to a less correlated pattern. The effect was not as strong as with paced walking [1]. The explanation could be that treadmill constrained walking speed, while metronome constrained walking pace; it could be hypothesized that the adaptation of locomotor control to external cues specifically modify correlation pattern of the constrained walking parameter, as suggested by the results of Terrier et al. [1], but this remains to be investigated.\n[SUBTITLE] Correlations between variability indicators [SUBSECTION] While fractal dynamics, local dynamic stability and kinematic variability characterize different features of gait variability, it is not excluded that relationships exists between them.\nJordan et al. [46] recently analyzed fractal dynamics and stability in walking/running transition on treadmill. They observed a positive correlation between λL* and α (r2 = 0.65, N = 12). They also observed that scaling exponent is minimal close to PWS [53] and suggested that \"reduced strength of long range correlations at preferred locomotion speeds is reflective of enhanced stability and adaptability at theses speeds\". Our results, using CCA, did not confirm this suggestion. No evident correlation between scaling exponent and dynamic stability was found. Several differences in the measurement method (trunk accelerometry vs 3D video analysis) and in the experimental design (high speed vs. moderate speed) may explain this divergence.\nPrevious studies have analyzed the relationships between variability (meanSD) and local dynamic stability (λS* and λL*). Dingwell et al. pointed out \"the general lack of correlation between the standard deviation and λ* exponents\" [32]. In contrast, other investigators recently observed significant positive correlation between λS* and MeanSD [54]. The results of the present study showed a counterintuitive negative correlation between λ* and MeanSD: during treadmill walking (but not in OW), higher kinematic variability seemed to be related to higher local stability (i.e. low λ*). As explained above, the use of different methodologies is a potential source of divergence between studies concerning dynamic stability. It is not excluded that a confounding factor, not measured yet, related to both MeanSD and λ* could indirectly explain this correlation. Further investigations are needed to better understand the relationship between these two variability indexes.\nWhile fractal dynamics, local dynamic stability and kinematic variability characterize different features of gait variability, it is not excluded that relationships exists between them.\nJordan et al. [46] recently analyzed fractal dynamics and stability in walking/running transition on treadmill. They observed a positive correlation between λL* and α (r2 = 0.65, N = 12). They also observed that scaling exponent is minimal close to PWS [53] and suggested that \"reduced strength of long range correlations at preferred locomotion speeds is reflective of enhanced stability and adaptability at theses speeds\". Our results, using CCA, did not confirm this suggestion. No evident correlation between scaling exponent and dynamic stability was found. Several differences in the measurement method (trunk accelerometry vs 3D video analysis) and in the experimental design (high speed vs. moderate speed) may explain this divergence.\nPrevious studies have analyzed the relationships between variability (meanSD) and local dynamic stability (λS* and λL*). Dingwell et al. pointed out \"the general lack of correlation between the standard deviation and λ* exponents\" [32]. In contrast, other investigators recently observed significant positive correlation between λS* and MeanSD [54]. The results of the present study showed a counterintuitive negative correlation between λ* and MeanSD: during treadmill walking (but not in OW), higher kinematic variability seemed to be related to higher local stability (i.e. low λ*). As explained above, the use of different methodologies is a potential source of divergence between studies concerning dynamic stability. It is not excluded that a confounding factor, not measured yet, related to both MeanSD and λ* could indirectly explain this correlation. Further investigations are needed to better understand the relationship between these two variability indexes.", "For the present study, portable trunk accelerometry was chosen because it offers the possibility to record long-term free walking. Hence, the results concern the gait stability measured from accelerations of the low-back. Comparisons with other results should take into account that that the different gait stability studies use different kinematic variables (acceleration [7,10], positions [44], angle [8]) and different body location (thorax, head, knee, and ankle) to assess λ*. We found λ* similar to those measured by others [8,32], suggesting that the results are rather independent on the measurements methods.\nIn fractal dynamics studies, the first step is the detection of the periodic pattern of the gait in order to compute time series of stride intervals. Several methodologies have been used to measure long-term time series of stride intervals, such as foot switches [3,5], goniometer [45], video analysis [46], or high accuracy GPS [1]. Because the same variable is used (i.e. time duration of the gait cycle) for DFA analyses, data from different studies are probably comparable.\nIn order to increase the likelihood to point out significant correlations among variability indexes, we designed the experiment to obtain a substantial degree of standardization: we imposed the same speed (1.25 m/s, 4.5 km/h) for all subjects on the treadmill. This speed was chosen on the basis of a previous experiment (partially published yet [34]), which showed that the preferred speed in the same experimental conditions (same room, same treadmill) was 1.26 ± 0.13 m/s (n = 88). Similar values are found in the literature: 1.25 m/s (n = 8) [47], 1.19 m/s (n = 26) [48].\nWalking speed was not standardized between TW and OW, as in other studies [32]. However, by selecting treadmill speed at the same speed of overground preferred speed, the results would be that subjects walk at higher speed than their preferred speed on the treadmill. Several studies showed a substantial difference between both conditions: Dal et al. [48] demonstrated that preferred walking speed determined on a treadmill is slower than overground (21% relative difference); Marsh et al. [49]showed that, when older adults were allowed to choose a preferred walking pace, they walked faster (+61%), used longer strides, and had a faster rate walking overground than when they walked on a treadmill. As a result, speed normalization could introduce unwanted bias. Our experimental design was therefore a compromise, which standardized speed among subjects in TW condition, but also which selected walking speed close to preferred speed, making both OW and TW conditions comparable.\nIn addition, Indirect clues seem to indicate that TW and OW conditions were quite similar: 1) stride time (which is related to walking speed) were close (3% difference, small effect size), 2) stride time variability (CV) was the same (no significant differences), 3) no correlation was observed between stride time and other parameters (results not shown),", "Kinematic variability, fractal dynamics (DFA) and local dynamic stability (Lyapunov exponents) quantify different aspects of locomotor control [32]. Kinematic variability describes the range in which the locomotor system operates. DFA quantify temporal dynamics of discrete events (i.e stride interval) over hundreds of consecutives strides; it assesses the presence of long-range correlations between strides, and hence analyzes the characteristics of feedbacks in locomotor control. Lyapunov exponents quantify the temporal dynamics in continuous time based on the theory of deterministic chaos; it evaluates the degree of divergence in the signal, and hence the resilience of the locomotor system to small perturbations. Therefore, it can be expected that these variability indexes did not react in the same way under various conditions.\nThese assumptions were experimentally verified in various studies that observed changes of λ* and kinematic variability between different experimental conditions or between different populations. For instance it was observed that patients with peripheral neuropathy present altered dynamic stability but normal kinematic variability [33,50]. Other investigators have shown that an exercise training intervention in elderly people could improve dynamic stability but not decrease kinematic variability [19].\nDespite differences in the method of measurement (low-back vs. thorax acceleration) and in the experimental design (speed normalization), our results are generally in accordance with the results of Dingwell et al [32]. They analyzed only 10 healthy individuals, therefore statistical significance for small effects was more difficult to reach than in the present study. They showed a significant treadmill effect in short-term stability (lower λS*). A slight but not significant effect for long-term vertical stability (lower λL*) was found. They observed that kinematic variability (MeanSD) for upper body accelerations was generally greater for OW than TW, but this trend was only significant for antero-posterior accelerations. They explained that underlying causes of differences between TW and OW were unclear: on one hand, the motorized treadmill imposed a constant nominal speed on the subjects and constrained them to walk along a much narrower and straighter path than during OW; but on the other hand, differences may have been induced by intra-stride fluctuations in treadmill belt speed, differences in mechanical compliance between the walking surfaces, and changes in visual and vestibular perceptual information. In light of the results of the present study, we hypothesize that motor control is able to maintain the same range of kinematic variability in both TW and OW conditions (same kinematic variability), probably because of compensating effects: in TW, destabilizing factors (intra-stride belt speed fluctuations, disturbing mechanical compliance, alteration of perceptual information) are balanced by stabilizing factors (constant speed, narrow and straight path). Conversely, motor control strategy adapting the gait to TW seems to specifically alter non-linear dependencies among consecutive strides: the stabilizing factors override the destabilizing ones.\nIn a subsequent study, Dingwell & Marin [51] analyzed speed effect on dynamical stability (λS* and λL*) and kinematic variability (MeanSD). Walking speed was normalized by individual PWS on a treadmill. Speed range was 0.6PW to 1.4PWS by steps of 0.2. They found significant speed effect for both λ* and MeanSD: however the effect was small for 0.8-1.2 PWS. Under our experimental conditions [34], we observed that inter-indivudual variability of PWS on the treadmill was low: 90% of individuals walked in the range of 0.87-1.13 mean PWS. As a result, the speed effect among individuals in the present study was probably low. This is also indirectly confirmed by the low inter-individual variability of stride duration (CV = 6%).\nFractal dynamics of stride intervals has been extensively studied by Hausdorff et al. [36]. Them and other [1,3,52] have observed that constrained walking (paced cadence with a metronome), deeply modified the scaling exponent. By analogy, because treadmill also constraints the gait by imposing a constant speed, a similar effect could be expected. The results of the present study showed, in a majority of subjects, a lowering of scaling exponent to a less correlated pattern. The effect was not as strong as with paced walking [1]. The explanation could be that treadmill constrained walking speed, while metronome constrained walking pace; it could be hypothesized that the adaptation of locomotor control to external cues specifically modify correlation pattern of the constrained walking parameter, as suggested by the results of Terrier et al. [1], but this remains to be investigated.", "While fractal dynamics, local dynamic stability and kinematic variability characterize different features of gait variability, it is not excluded that relationships exists between them.\nJordan et al. [46] recently analyzed fractal dynamics and stability in walking/running transition on treadmill. They observed a positive correlation between λL* and α (r2 = 0.65, N = 12). They also observed that scaling exponent is minimal close to PWS [53] and suggested that \"reduced strength of long range correlations at preferred locomotion speeds is reflective of enhanced stability and adaptability at theses speeds\". Our results, using CCA, did not confirm this suggestion. No evident correlation between scaling exponent and dynamic stability was found. Several differences in the measurement method (trunk accelerometry vs 3D video analysis) and in the experimental design (high speed vs. moderate speed) may explain this divergence.\nPrevious studies have analyzed the relationships between variability (meanSD) and local dynamic stability (λS* and λL*). Dingwell et al. pointed out \"the general lack of correlation between the standard deviation and λ* exponents\" [32]. In contrast, other investigators recently observed significant positive correlation between λS* and MeanSD [54]. The results of the present study showed a counterintuitive negative correlation between λ* and MeanSD: during treadmill walking (but not in OW), higher kinematic variability seemed to be related to higher local stability (i.e. low λ*). As explained above, the use of different methodologies is a potential source of divergence between studies concerning dynamic stability. It is not excluded that a confounding factor, not measured yet, related to both MeanSD and λ* could indirectly explain this correlation. Further investigations are needed to better understand the relationship between these two variability indexes.", "Scaling exponent (α) and maximal Lypunov exponent (λ*) have been advocated as a relevant indicator of neuromuscular control of stability during human locomotion [8,32,36,55]. The results of the present study showed that treadmill modified fractal dynamics (α) and local dynamic stability (λ*) of the gait, but not kinematic variability (MeanSD). This should be kept in mind when using motorized treadmill either for fundamental research or in locomotor therapies.\nWhereas both scaling exponent (α) and maximal Lypunov exponent (λ*) are sensitive enough to identify differences between OW and TW, they seem not correlated together. This suggests that both indexes deserve to be used in conjunction when analyzing long term gait variability, because they describe different locomotor characteristics.", "The authors declare that they have no competing interests.", "PT performed measurements and data analysis, and drafted the manuscript. OD participated in the design and coordination of the study and assisted with drafting the manuscript. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Methods", "Participants", "Apparatus", "Procedures", "Stride intervals and kinematic variability", "Detrended Fluctuation Analysis", "Local dynamic stability", "Statistical analysis", "Results", "Treadmill effect", "Correlations", "Discussion", "Technical issues", "Differences between treadmill and overground walking", "Correlations between variability indicators", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Walking is a repetitive movement which is characterized by a low variability [1]. This motor skill requires not only conscious neuromotor tasks but also complex automated regulation, both interacting to produce steady gait pattern. Classically, gait variability (i.a. kinematic variability) has been assessed from the differences among the strides (Standard Deviation SD, coefficient of variation CV), i.e. each stride considered as an independent event resulting from a random process. However, this approach fails to account for the presence of feedback loops in the motor control of walking: the walking pattern at a given gait cycle may have consequences on subsequent strides. As a result, correlations between consecutive gait cycles and non-linear dependencies are expected.\nDuring the last decades, various new mathematical tools have been used to better characterise the non-linear features of gait variability. With the Detrended Fluctuation Analysis (DFA [2-4]) it has been observed that the stride interval (i.e. time to complete a gait cycle) at any time was related (in a statistical sense) to intervals at relatively remote times (persistent pattern over more than 100 strides). This dependence (memory effect) decayed in a power-law fashion, similar to scale-free, fractal-like phenomena (fractal dynamics [1,3-5]), also known as 1/fβ noise [6]).\nAnother non-linear approach was proposed to characterize the dynamic variability in continuous walking. The sensitivity of a dynamical system to small perturbations can be quantified by the system maximal Lyapunov exponent, which characterizes the average rate of divergence in pseudo-periodic processes [7]. This method allows to evaluate the ability of locomotor system to maintain continuous motion by accommodating infinitesimally small perturbations that occur naturally during walking [8]. This includes external perturbations induced by small variations in the walking surface, as well as internal perturbations resulting from the natural noise present in the neuromuscular system [8].\nMany theoretical questions are still open about the validity and application of these methods. For instance, DFA results are difficult to interpret [9], and no definitive conclusion on the presence of long range correlations should be drawn relying only on it. In addition, the underlying mechanism of long range correlations in stride interval is not fully understood [3,10]. West & Latka suggested that the observed scaling in inter-stride interval data may not be due to long-term memory alone, but may, in fact, be due partly to the statistics [11]. It was also suggested that the use of multi-fractal spectrum could be a better approach than mono-fractal analysis, such as DFA [12,13]. There are also several methodological issues to compute consistent and reliable stability index [14,15].\nIn parallel with the ongoing theoretical research on non-linear analysis of physiological time series, the use of non-linear bio-markers in applied clinical research has been already fruitful. In the field of human locomotion, it has been demonstrated that gait variability could serve as a sensitive and clinically relevant tool in the evaluation of mobility and the response to therapeutic interventions. For instance, gait variability (SD and dynamics) is altered in clinically relevant syndromes, such as falling and neuro-degenerative disease [16,17]. Gait instability measurement apparently predict falls in idiopathic elderly fallers [18]. Improvements in muscle function are associated with enhanced gait stability in elderly [19].\nMotorized treadmills are widely used in biomechanical studies of human locomotion. They allow the documentation of a large number of successive strides under controlled environment, with a selectable steady-state locomotion speed. In the rehabilitation field, treadmill walking is used in locomotor therapy, for instance with partial body weight support in spinal cord injury or stroke rehabilitation [20,21]. Since the classical work of Van Ingen Schenau [22], it is admitted that overground and treadmill locomotion are similar if treadmill belt speed is constant. Nevertheless, both walking types present small differences in kinematics [23,24], kinetics [25] and energetics [26]. It was also observed that treadmill locomotion induced shorter step lengths and higher cadences than walking on the floor at the same speed [26,27]. There is still a matter of debate to interpret such subtle differences [28,29].\nIt is obvious that treadmill walking (TW) induces specific kinaesthetic and perceptual information. Previous studies confirmed that vision plays a central role in the control of locomotion [30,31]. These differences in visual afferences between TW and Overground Walking (OW) may induce a modification in motor control, and consequently in gait variability.\nIn 2000, Dingwell et al. analyzed TW local dynamic stability (maximal Lyapunov exponent) in 10 healthy subjects [8,32]. They highlighted significant differences between TW and OW by evaluating local dynamic stability of lower limbs kinematics [8]. The effect was low in upper body accelerations. Later [32], they calculated more specifically short term stability and found a strong effect of TW in trunk accelerations. On the other hand, they found a greater kinematic variability at the lower limb level in OW as compared to TW, but no significant difference in trunk kinematics.\nIn 2005, Terrier et al. [1], by using high accuracy GPS, described low stride-to-stride variability of speed, step length and step duration in free walking. They observed that the constraint of rhythmical auditory signal (\"metronome walking\") did not alter kinematic variability, but modify the fractal dynamics (DFA) of the stride interval (anti-persistent pattern).\nBased on these previous works, the working hypothesis of the present article is 1) that the constraint of TW (constant speed, narrow pathway) may induce a less persistent pattern in the stride interval, by analogy to the constraint induced by a metronome; 2) that TW may increase the local dynamic stability of walking, due to the diminution of degrees of freedom in the more constrained artificial environment [32,33], 3) that, for the same reasons, TW may slightly reduce kinematic variability [32,33] 4) that no correlation exist between the 3 variability indexes, because they are related to different aspects of the locomotion process.\nThe purpose of the present study was to analyze, by using trunk accelerometry, differences between TW and OW in terms of stride-to-stride kinematic variability (SD), fractal dynamics (by DFA) and local dynamic stability (maximal Lyapunov exponent). In addition, we assessed the strength of the relationships between these variables (canonical correlation analysis).", "[SUBTITLE] Participants [SUBSECTION] Twenty healthy male subjects, with no neurological deficit or orthopaedic impairment, participated to the study. Most of them were recruited among participants of a previous \"treadmill\" study implying only males subjects [34]. Their characteristics were (mean ± SD): age 35 ± 7 yr, body mass 79 ± 10 kg, and height 1.80 ± 0.06 m. All subjects were well trained to walk on a treadmill before the beginning of the study. The experimental protocol was approved by the local ethics committee (commission d'éthique du Valais).\nTwenty healthy male subjects, with no neurological deficit or orthopaedic impairment, participated to the study. Most of them were recruited among participants of a previous \"treadmill\" study implying only males subjects [34]. Their characteristics were (mean ± SD): age 35 ± 7 yr, body mass 79 ± 10 kg, and height 1.80 ± 0.06 m. All subjects were well trained to walk on a treadmill before the beginning of the study. The experimental protocol was approved by the local ethics committee (commission d'éthique du Valais).\n[SUBTITLE] Apparatus [SUBSECTION] The motion sensor (Physilog system, BioAGM, Switzerland [35]) was a triaxial accelerometer connected to a data logger recording body accelerations in medio-lateral (ML), vertical (V) and antero-posterior (AP) directions. The dimensions of the logger were 130 × 68 × 30 mm and the weight was 285 g. The accelerometers are piezoresistive sensors coupled with amplifiers (± 5 g, 500 mV/g) and mounted on a belt. The signals were sampled at 200 Hz with 12-bit resolution. After each experiment, the data were downloaded to a PC computer and converted in earth acceleration units (g) according to a previous calibration. Data analysis was then performed by using Matlab (Mathworks, Natick MA, USA) and Stata 11.0 (StataCorp LP, TX, USA)\nThe motion sensor (Physilog system, BioAGM, Switzerland [35]) was a triaxial accelerometer connected to a data logger recording body accelerations in medio-lateral (ML), vertical (V) and antero-posterior (AP) directions. The dimensions of the logger were 130 × 68 × 30 mm and the weight was 285 g. The accelerometers are piezoresistive sensors coupled with amplifiers (± 5 g, 500 mV/g) and mounted on a belt. The signals were sampled at 200 Hz with 12-bit resolution. After each experiment, the data were downloaded to a PC computer and converted in earth acceleration units (g) according to a previous calibration. Data analysis was then performed by using Matlab (Mathworks, Natick MA, USA) and Stata 11.0 (StataCorp LP, TX, USA)\n[SUBTITLE] Procedures [SUBSECTION] The subjects performed 10 min. treadmill walking (TW) and 10 min. overground walking (OW) in a random order. A rest period of five minutes (sitting still) was imposed between the two trials. The motor-driven treadmill was a Technogym, (Runrace, Italy). The imposed speed was 1.25 m/s (4.5 km/h) for all subjects: in the context of a previous study [34], we assessed average running and walking preferred speed on the same treadmill in 88 male subjects; an average of 1.26 ± 0.13 m/s was observed. A thirty second warm-up was performed before the beginning of the measurement. For the OW test, the subjects walked along a standardized 800 m indoor circuit along hospital corridors and halls. The circuit exhibited only 90° turns. A large part (about 400 m) of the circuit was constituted by a long corridor. Other people working in the hospital were present in the halls. Hence, the OW trials mimicked actual condition of walking. Subects were asked to walk at their Preferred Walking Speed (PWS) with a regular pace. Under both conditions, the accelerometer was attached to the low back (L4-L5 region) with an elastic belt, and the logger was worn on the side of the body. Subjects wore their own low-rise comfortable walking shoes.\nThe subjects performed 10 min. treadmill walking (TW) and 10 min. overground walking (OW) in a random order. A rest period of five minutes (sitting still) was imposed between the two trials. The motor-driven treadmill was a Technogym, (Runrace, Italy). The imposed speed was 1.25 m/s (4.5 km/h) for all subjects: in the context of a previous study [34], we assessed average running and walking preferred speed on the same treadmill in 88 male subjects; an average of 1.26 ± 0.13 m/s was observed. A thirty second warm-up was performed before the beginning of the measurement. For the OW test, the subjects walked along a standardized 800 m indoor circuit along hospital corridors and halls. The circuit exhibited only 90° turns. A large part (about 400 m) of the circuit was constituted by a long corridor. Other people working in the hospital were present in the halls. Hence, the OW trials mimicked actual condition of walking. Subects were asked to walk at their Preferred Walking Speed (PWS) with a regular pace. Under both conditions, the accelerometer was attached to the low back (L4-L5 region) with an elastic belt, and the logger was worn on the side of the body. Subjects wore their own low-rise comfortable walking shoes.\n[SUBTITLE] Stride intervals and kinematic variability [SUBSECTION] Five seconds were removed at the beginning and at the end of the 10 min. acceleration measurements in order to avoid non-stationary periods. Heel strike was detected in the raw acceleration AP signal with a peak detection method designed to minimize the risk of false step detection: first, we generated a low-pass filtered version of the signal (4 order Butterworth, 3 Hz, zero-phase filtering). The time of each local minimum was detected. By superimposing the Filtered Signal (FS) to the original, Unfiltered Signal (US), we tracked the nearest peak in US of each local minimum in FS. US peak time was then chosen as the limit between two steps (Figure 1A). The strides were defined as two consecutive steps. On average, the number of strides was 543 per trial.\nMethod: Step detection, stride intervals and Detrended Fluctuation Analysis. One subject performed 10 min of free walking. A: 2.5s sample of the antero-posterior acceleration signal; red dotted line is a low pass filtered (<3 Hz) version of the raw signal (black continuous line). Cross and black circle indicate how the algorithm specifically detect the heel strike (see method section for further explanation). The duration of two consecutive steps is defined as stride interval. B: Time series of stride intervals during the 10 min walking test. Average stride time (mean) and CV (SD/mean * 100) is also presented. C: Detredend Fluctuation Analysis (DFA). The fractal dynamics of the time series (B) is characterized by the scaling exponent α, computed by comparing the fluctuation (F(n)) at different scales (n) in a log-log plot.\nTime series of the stride intervals were used to compute a traditional variability index (Coefficient of Variation of the stride time, CV = SD/Mean*100, Figure 1B). Moreover, the variability of the acceleration pattern among strides was evaluated as follows (Figure 2): each stride was normalized to 200 sample points by using a polyphase filter implementation (Matlab command Resample); the average stride-to-stride Standard Deviation across all data points ((SD(i) ∀ i ∈ [1 ....200])) was evaluated (MeanSD = 〈SD(i)〉).\nMethod: variability, MeanSD. One subject (same as in Figure 1) performed 10 min of free walking. Each stride (see Figure 1A) was normalized to 200 samples (0% to 100% gait cycle). Top: Average acceleration pattern of the normalized strides (N = 513). Bottom: Standard Deviation (SD) of the normalized strides (N = 513). MeanSD is the average SD of the 200 samples.\nFive seconds were removed at the beginning and at the end of the 10 min. acceleration measurements in order to avoid non-stationary periods. Heel strike was detected in the raw acceleration AP signal with a peak detection method designed to minimize the risk of false step detection: first, we generated a low-pass filtered version of the signal (4 order Butterworth, 3 Hz, zero-phase filtering). The time of each local minimum was detected. By superimposing the Filtered Signal (FS) to the original, Unfiltered Signal (US), we tracked the nearest peak in US of each local minimum in FS. US peak time was then chosen as the limit between two steps (Figure 1A). The strides were defined as two consecutive steps. On average, the number of strides was 543 per trial.\nMethod: Step detection, stride intervals and Detrended Fluctuation Analysis. One subject performed 10 min of free walking. A: 2.5s sample of the antero-posterior acceleration signal; red dotted line is a low pass filtered (<3 Hz) version of the raw signal (black continuous line). Cross and black circle indicate how the algorithm specifically detect the heel strike (see method section for further explanation). The duration of two consecutive steps is defined as stride interval. B: Time series of stride intervals during the 10 min walking test. Average stride time (mean) and CV (SD/mean * 100) is also presented. C: Detredend Fluctuation Analysis (DFA). The fractal dynamics of the time series (B) is characterized by the scaling exponent α, computed by comparing the fluctuation (F(n)) at different scales (n) in a log-log plot.\nTime series of the stride intervals were used to compute a traditional variability index (Coefficient of Variation of the stride time, CV = SD/Mean*100, Figure 1B). Moreover, the variability of the acceleration pattern among strides was evaluated as follows (Figure 2): each stride was normalized to 200 sample points by using a polyphase filter implementation (Matlab command Resample); the average stride-to-stride Standard Deviation across all data points ((SD(i) ∀ i ∈ [1 ....200])) was evaluated (MeanSD = 〈SD(i)〉).\nMethod: variability, MeanSD. One subject (same as in Figure 1) performed 10 min of free walking. Each stride (see Figure 1A) was normalized to 200 samples (0% to 100% gait cycle). Top: Average acceleration pattern of the normalized strides (N = 513). Bottom: Standard Deviation (SD) of the normalized strides (N = 513). MeanSD is the average SD of the 200 samples.\n[SUBTITLE] Detrended Fluctuation Analysis [SUBSECTION] The presence of long range correlations in the time series of stride intervals (fractal dynamics) was assessed by the use of the non-linear DFA method. Strictly speaking, this non-linear method should be used in addition to other statistical tools to definitively conclude that a process is a true 1/fβ noise with power-law decrease of long range auto-correlations [6,9]. However, DFA has been successfully used as relevant biomarker in numerous studies [1,16,17,36,37]. Detrended Fluctuation Analysis is based on a classic root-mean square analysis of a random walk, but is specifically designed to be less likely affected by nonstationarities. Full details of the methodology are published elsewhere [1-4]. In short, the integrated time series of length N is divided into boxes of equal length, n. In each box of length n, a least squares line is fit to the data (representing the trend in that box). The y coordinate of the straight line segments is denoted by yn(k). Next, the integrated time series, y(k), was detrended, by subtracting the local trend, yn(k), in each box. The root-mean-square fluctuation of this integrated and detrended time series is calculated by\n\n\n(1)\n\n\nF\n(\nn\n)\n=\n\n\n\n1\nN\n\n\n\n∑\n\nk\n=\n1\n\nN\n\n\n\n\n[\ny\n(\nk\n)\n−\n\ny\nn\n\n(\nk\n)\n]\n\n2\n\n\n\n\n\n\n\n\n\nThis computation is repeated over all box sizes (from 4 to 200) to characterize the relationship between F(n), the average fluctuation, and the box size, n. The fluctuations can be characterized by the scaling exponent α, which is the slope of the line relating log F(n) to log(n) (F(n) ~ nα), Figure 1C). Long range correlations are present in the original time series when α lies between 0.5 and 1 [3,4].\nIn a finite length time series, an uncorrelated process could exhibit \"by chance\" a scaling exponent different from the theoretical 0.5 value. To statistically differentiate the stride time series from a random uncorrelated process, we applied the surrogate data method [1,3]. This method increases the confidence that the analyzed series exhibits long-range correlation. Twenty different surrogate data sets were generated by shuffling the original time series in a random order. On each data set, DFA analysis was performed to calculate α value. The standard deviation and mean of this sample was calculated and compared to α exponent of the original series. The result is considered significant if the original α is 2 standard deviation away from the mean of the surrogate data set.\nThe presence of long range correlations in the time series of stride intervals (fractal dynamics) was assessed by the use of the non-linear DFA method. Strictly speaking, this non-linear method should be used in addition to other statistical tools to definitively conclude that a process is a true 1/fβ noise with power-law decrease of long range auto-correlations [6,9]. However, DFA has been successfully used as relevant biomarker in numerous studies [1,16,17,36,37]. Detrended Fluctuation Analysis is based on a classic root-mean square analysis of a random walk, but is specifically designed to be less likely affected by nonstationarities. Full details of the methodology are published elsewhere [1-4]. In short, the integrated time series of length N is divided into boxes of equal length, n. In each box of length n, a least squares line is fit to the data (representing the trend in that box). The y coordinate of the straight line segments is denoted by yn(k). Next, the integrated time series, y(k), was detrended, by subtracting the local trend, yn(k), in each box. The root-mean-square fluctuation of this integrated and detrended time series is calculated by\n\n\n(1)\n\n\nF\n(\nn\n)\n=\n\n\n\n1\nN\n\n\n\n∑\n\nk\n=\n1\n\nN\n\n\n\n\n[\ny\n(\nk\n)\n−\n\ny\nn\n\n(\nk\n)\n]\n\n2\n\n\n\n\n\n\n\n\n\nThis computation is repeated over all box sizes (from 4 to 200) to characterize the relationship between F(n), the average fluctuation, and the box size, n. The fluctuations can be characterized by the scaling exponent α, which is the slope of the line relating log F(n) to log(n) (F(n) ~ nα), Figure 1C). Long range correlations are present in the original time series when α lies between 0.5 and 1 [3,4].\nIn a finite length time series, an uncorrelated process could exhibit \"by chance\" a scaling exponent different from the theoretical 0.5 value. To statistically differentiate the stride time series from a random uncorrelated process, we applied the surrogate data method [1,3]. This method increases the confidence that the analyzed series exhibits long-range correlation. Twenty different surrogate data sets were generated by shuffling the original time series in a random order. On each data set, DFA analysis was performed to calculate α value. The standard deviation and mean of this sample was calculated and compared to α exponent of the original series. The result is considered significant if the original α is 2 standard deviation away from the mean of the surrogate data set.\n[SUBTITLE] Local dynamic stability [SUBSECTION] The method for quantifying the local dynamical stability of the gait by using largest Lyapunov exponent has been extensively described in literature [8]. It examines structural characteristics of a time series that is embedded in an appropriately constructed state space. A valid state space contains a sufficient number of independent coordinates to define the state of the system unequivocally [38]. According to the Takens' theorem, an appropriate state space can be reconstructed from a single time series using the original data and its time delayed copies (figure 3A) [38].\nMethod: dynamic stability, maximal Lyapunov exponent. A: Two dimensional state space of the antero-posterior acceleration signal (5s) reconstructed from the original data set and its time delayed copy (Δt = 11 samples). B: Magnification of the state space. An initial local perturbation at dj(0) diverge across i time steps as measured by dj(i). C: Short term (λS*) and long term (λL*) finite-time maximal Lyapunov exponent computed from average logarithmic divergence.\n\n\n(2)\n\n\nX\n(\nt\n)\n=\n[\nx\n(\nt\n)\n,\nx\n(\nt\n+\nT\n)\n,\nx\n(\nt\n+\n2\nT\n)\n,\n…\n,\nx\n(\nt\n+\n(\n\nd\nE\n\n−\n1\n)\nT\n]\n\n\n\n\nWhere X(t) is the dE-dimensional state vector, x(t) are the original data, T is the time delay, and dE is the embedding dimension. The time delays (T) were calculated individually for each of the 120 acceleration data set (3-axis, 2 conditions, and 20 individuals) from the first minimum of the Average Mutual Information (AMI) function [8,39]. Embedding dimensions (dE) were computed from a Global False Nearest Neighbors (GFNN) analysis [8,40]. Because the result was similar for all acceleration time series, we use a constant dimension (dE = 6) [8,32]]. The Lyapunov exponent is the mean exponential rate of divergence of initially nearby points in the reconstructed space (Figure 3B). Because the determination of the maximal Lypunov exponent requires intensive computing power, 7 min of the 10 min walking test (from 1.5 to 8.5 min.) was selected and the raw data were down-sampled to 100 Hz. The determination of the Lyapunov exponent was then achieved by using the algorithm introduced by Rosenstein and colleagues [7], which provided dedicated software to compute divergence as a function of time in finite-time series [41] (Figure 3B). The maximum finite-time Lyapunov exponents (λ*) were estimated from the slopes of linear fits in the divergence diagrams (Figure 3C). Strictly speaking, because divergence diagrams (Figure 3C) are non-linear, multiple slopes could be defined and so no true single maximum Lyapunov exponent exists. The slopes (exponents) quantify local divergence (and hence stability) of the observed dynamics at different time scale, and should not be interpreted as a classical maximal Lyapunov exponent in chaos theory.\nSince each subject exhibited a different average step frequency, the time was normalized by average stride time for each subject and each condition (Figure 3C). As suggested by Dingwell and colleagues [32], we use two different time scales for assessing short-term and long-term dynamic stability: short term exponents (λS*) was computed over the first stride (0 to 1), and long term exponents (λL*) between 4 and 10 strides (Figure 3C).\nThe method for quantifying the local dynamical stability of the gait by using largest Lyapunov exponent has been extensively described in literature [8]. It examines structural characteristics of a time series that is embedded in an appropriately constructed state space. A valid state space contains a sufficient number of independent coordinates to define the state of the system unequivocally [38]. According to the Takens' theorem, an appropriate state space can be reconstructed from a single time series using the original data and its time delayed copies (figure 3A) [38].\nMethod: dynamic stability, maximal Lyapunov exponent. A: Two dimensional state space of the antero-posterior acceleration signal (5s) reconstructed from the original data set and its time delayed copy (Δt = 11 samples). B: Magnification of the state space. An initial local perturbation at dj(0) diverge across i time steps as measured by dj(i). C: Short term (λS*) and long term (λL*) finite-time maximal Lyapunov exponent computed from average logarithmic divergence.\n\n\n(2)\n\n\nX\n(\nt\n)\n=\n[\nx\n(\nt\n)\n,\nx\n(\nt\n+\nT\n)\n,\nx\n(\nt\n+\n2\nT\n)\n,\n…\n,\nx\n(\nt\n+\n(\n\nd\nE\n\n−\n1\n)\nT\n]\n\n\n\n\nWhere X(t) is the dE-dimensional state vector, x(t) are the original data, T is the time delay, and dE is the embedding dimension. The time delays (T) were calculated individually for each of the 120 acceleration data set (3-axis, 2 conditions, and 20 individuals) from the first minimum of the Average Mutual Information (AMI) function [8,39]. Embedding dimensions (dE) were computed from a Global False Nearest Neighbors (GFNN) analysis [8,40]. Because the result was similar for all acceleration time series, we use a constant dimension (dE = 6) [8,32]]. The Lyapunov exponent is the mean exponential rate of divergence of initially nearby points in the reconstructed space (Figure 3B). Because the determination of the maximal Lypunov exponent requires intensive computing power, 7 min of the 10 min walking test (from 1.5 to 8.5 min.) was selected and the raw data were down-sampled to 100 Hz. The determination of the Lyapunov exponent was then achieved by using the algorithm introduced by Rosenstein and colleagues [7], which provided dedicated software to compute divergence as a function of time in finite-time series [41] (Figure 3B). The maximum finite-time Lyapunov exponents (λ*) were estimated from the slopes of linear fits in the divergence diagrams (Figure 3C). Strictly speaking, because divergence diagrams (Figure 3C) are non-linear, multiple slopes could be defined and so no true single maximum Lyapunov exponent exists. The slopes (exponents) quantify local divergence (and hence stability) of the observed dynamics at different time scale, and should not be interpreted as a classical maximal Lyapunov exponent in chaos theory.\nSince each subject exhibited a different average step frequency, the time was normalized by average stride time for each subject and each condition (Figure 3C). As suggested by Dingwell and colleagues [32], we use two different time scales for assessing short-term and long-term dynamic stability: short term exponents (λS*) was computed over the first stride (0 to 1), and long term exponents (λL*) between 4 and 10 strides (Figure 3C).\n[SUBTITLE] Statistical analysis [SUBSECTION] Mean and Standard Deviation (SD) were computed to describe the data (table 1). Ninety-five percent Confidence Intervals (CI) were calculated as ± 1.96 times the Standard Error of the Mean (SEM, N = 20).\nThe effect size of TW as compared to OW was expressed in both absolute (mean difference) and standardized (mean difference divided by SD) terms. The standardized effect size was the Hedge's g, which is a modified version of the Cohen's d for inferential measure [42]. Paired t-tests between OW and TW were performed, and the p-values are shown in the last column of table 1. The precision of the effect sizes was estimated with CI (Figure 4). CI were ± 1.96 times the asymptotic estimates of the standard error (SE) of g [42]. The arbitrary limit of 0.5 was uses to delineate small effect size, as defined by Cohen [42]. The extent of the data (quartiles and median) and individual differences between conditions are shown in Figure 5 for λ*. In order to facilitate results interpretation by reducing the risk of type I statistical error, a Hotelling T2 test was used. This is a multivariate generalization of paired t-test [43]. The null hypothesis is that a vector of p differences is equal to a vector of zeros. Two multivariate sets were tested: meanSD (p = 3) and λ* (p = 6).\nCanonical correlation analyses (CCA, table 2&3) were performed in order to assess the strength of the relationships between different sets of variables [43]. This multivariate method allows one to find linear combinations (variates) in two sets of variables, which have maximum correlation (canonical correlation coefficient or canonical root) with each other. For each condition (OW and TW), two sets of p variables were analyzed: kinematic variability (set#1, p = 3) including MeanSD in ML, V and AP directions, and dynamic stability (set#2, p = 6), including short term and long term lyapunov exponent (λS*, λL*) in ML, V and AP directions. In addition, α scaling exponent was also analyzed with the same method vs. set#1 and set#2. In this case, CCA is equivalent to multiple regression analysis. Significance of the canonical correlations was assessed with the Wilks' lambda statistics.\nTo enhance the interpretation of CCA, different parameters were computed: the standardized canonical weights are the linear coefficients for each set after Z-transform of the variables; canonical loadings are the correlation coefficients between each variable and their respective linear composites; redundancy expresses the amount of variance in one set explained by a linear composite of the other set.\nMean and Standard Deviation (SD) were computed to describe the data (table 1). Ninety-five percent Confidence Intervals (CI) were calculated as ± 1.96 times the Standard Error of the Mean (SEM, N = 20).\nThe effect size of TW as compared to OW was expressed in both absolute (mean difference) and standardized (mean difference divided by SD) terms. The standardized effect size was the Hedge's g, which is a modified version of the Cohen's d for inferential measure [42]. Paired t-tests between OW and TW were performed, and the p-values are shown in the last column of table 1. The precision of the effect sizes was estimated with CI (Figure 4). CI were ± 1.96 times the asymptotic estimates of the standard error (SE) of g [42]. The arbitrary limit of 0.5 was uses to delineate small effect size, as defined by Cohen [42]. The extent of the data (quartiles and median) and individual differences between conditions are shown in Figure 5 for λ*. In order to facilitate results interpretation by reducing the risk of type I statistical error, a Hotelling T2 test was used. This is a multivariate generalization of paired t-test [43]. The null hypothesis is that a vector of p differences is equal to a vector of zeros. Two multivariate sets were tested: meanSD (p = 3) and λ* (p = 6).\nCanonical correlation analyses (CCA, table 2&3) were performed in order to assess the strength of the relationships between different sets of variables [43]. This multivariate method allows one to find linear combinations (variates) in two sets of variables, which have maximum correlation (canonical correlation coefficient or canonical root) with each other. For each condition (OW and TW), two sets of p variables were analyzed: kinematic variability (set#1, p = 3) including MeanSD in ML, V and AP directions, and dynamic stability (set#2, p = 6), including short term and long term lyapunov exponent (λS*, λL*) in ML, V and AP directions. In addition, α scaling exponent was also analyzed with the same method vs. set#1 and set#2. In this case, CCA is equivalent to multiple regression analysis. Significance of the canonical correlations was assessed with the Wilks' lambda statistics.\nTo enhance the interpretation of CCA, different parameters were computed: the standardized canonical weights are the linear coefficients for each set after Z-transform of the variables; canonical loadings are the correlation coefficients between each variable and their respective linear composites; redundancy expresses the amount of variance in one set explained by a linear composite of the other set.", "Twenty healthy male subjects, with no neurological deficit or orthopaedic impairment, participated to the study. Most of them were recruited among participants of a previous \"treadmill\" study implying only males subjects [34]. Their characteristics were (mean ± SD): age 35 ± 7 yr, body mass 79 ± 10 kg, and height 1.80 ± 0.06 m. All subjects were well trained to walk on a treadmill before the beginning of the study. The experimental protocol was approved by the local ethics committee (commission d'éthique du Valais).", "The motion sensor (Physilog system, BioAGM, Switzerland [35]) was a triaxial accelerometer connected to a data logger recording body accelerations in medio-lateral (ML), vertical (V) and antero-posterior (AP) directions. The dimensions of the logger were 130 × 68 × 30 mm and the weight was 285 g. The accelerometers are piezoresistive sensors coupled with amplifiers (± 5 g, 500 mV/g) and mounted on a belt. The signals were sampled at 200 Hz with 12-bit resolution. After each experiment, the data were downloaded to a PC computer and converted in earth acceleration units (g) according to a previous calibration. Data analysis was then performed by using Matlab (Mathworks, Natick MA, USA) and Stata 11.0 (StataCorp LP, TX, USA)", "The subjects performed 10 min. treadmill walking (TW) and 10 min. overground walking (OW) in a random order. A rest period of five minutes (sitting still) was imposed between the two trials. The motor-driven treadmill was a Technogym, (Runrace, Italy). The imposed speed was 1.25 m/s (4.5 km/h) for all subjects: in the context of a previous study [34], we assessed average running and walking preferred speed on the same treadmill in 88 male subjects; an average of 1.26 ± 0.13 m/s was observed. A thirty second warm-up was performed before the beginning of the measurement. For the OW test, the subjects walked along a standardized 800 m indoor circuit along hospital corridors and halls. The circuit exhibited only 90° turns. A large part (about 400 m) of the circuit was constituted by a long corridor. Other people working in the hospital were present in the halls. Hence, the OW trials mimicked actual condition of walking. Subects were asked to walk at their Preferred Walking Speed (PWS) with a regular pace. Under both conditions, the accelerometer was attached to the low back (L4-L5 region) with an elastic belt, and the logger was worn on the side of the body. Subjects wore their own low-rise comfortable walking shoes.", "Five seconds were removed at the beginning and at the end of the 10 min. acceleration measurements in order to avoid non-stationary periods. Heel strike was detected in the raw acceleration AP signal with a peak detection method designed to minimize the risk of false step detection: first, we generated a low-pass filtered version of the signal (4 order Butterworth, 3 Hz, zero-phase filtering). The time of each local minimum was detected. By superimposing the Filtered Signal (FS) to the original, Unfiltered Signal (US), we tracked the nearest peak in US of each local minimum in FS. US peak time was then chosen as the limit between two steps (Figure 1A). The strides were defined as two consecutive steps. On average, the number of strides was 543 per trial.\nMethod: Step detection, stride intervals and Detrended Fluctuation Analysis. One subject performed 10 min of free walking. A: 2.5s sample of the antero-posterior acceleration signal; red dotted line is a low pass filtered (<3 Hz) version of the raw signal (black continuous line). Cross and black circle indicate how the algorithm specifically detect the heel strike (see method section for further explanation). The duration of two consecutive steps is defined as stride interval. B: Time series of stride intervals during the 10 min walking test. Average stride time (mean) and CV (SD/mean * 100) is also presented. C: Detredend Fluctuation Analysis (DFA). The fractal dynamics of the time series (B) is characterized by the scaling exponent α, computed by comparing the fluctuation (F(n)) at different scales (n) in a log-log plot.\nTime series of the stride intervals were used to compute a traditional variability index (Coefficient of Variation of the stride time, CV = SD/Mean*100, Figure 1B). Moreover, the variability of the acceleration pattern among strides was evaluated as follows (Figure 2): each stride was normalized to 200 sample points by using a polyphase filter implementation (Matlab command Resample); the average stride-to-stride Standard Deviation across all data points ((SD(i) ∀ i ∈ [1 ....200])) was evaluated (MeanSD = 〈SD(i)〉).\nMethod: variability, MeanSD. One subject (same as in Figure 1) performed 10 min of free walking. Each stride (see Figure 1A) was normalized to 200 samples (0% to 100% gait cycle). Top: Average acceleration pattern of the normalized strides (N = 513). Bottom: Standard Deviation (SD) of the normalized strides (N = 513). MeanSD is the average SD of the 200 samples.", "The presence of long range correlations in the time series of stride intervals (fractal dynamics) was assessed by the use of the non-linear DFA method. Strictly speaking, this non-linear method should be used in addition to other statistical tools to definitively conclude that a process is a true 1/fβ noise with power-law decrease of long range auto-correlations [6,9]. However, DFA has been successfully used as relevant biomarker in numerous studies [1,16,17,36,37]. Detrended Fluctuation Analysis is based on a classic root-mean square analysis of a random walk, but is specifically designed to be less likely affected by nonstationarities. Full details of the methodology are published elsewhere [1-4]. In short, the integrated time series of length N is divided into boxes of equal length, n. In each box of length n, a least squares line is fit to the data (representing the trend in that box). The y coordinate of the straight line segments is denoted by yn(k). Next, the integrated time series, y(k), was detrended, by subtracting the local trend, yn(k), in each box. The root-mean-square fluctuation of this integrated and detrended time series is calculated by\n\n\n(1)\n\n\nF\n(\nn\n)\n=\n\n\n\n1\nN\n\n\n\n∑\n\nk\n=\n1\n\nN\n\n\n\n\n[\ny\n(\nk\n)\n−\n\ny\nn\n\n(\nk\n)\n]\n\n2\n\n\n\n\n\n\n\n\n\nThis computation is repeated over all box sizes (from 4 to 200) to characterize the relationship between F(n), the average fluctuation, and the box size, n. The fluctuations can be characterized by the scaling exponent α, which is the slope of the line relating log F(n) to log(n) (F(n) ~ nα), Figure 1C). Long range correlations are present in the original time series when α lies between 0.5 and 1 [3,4].\nIn a finite length time series, an uncorrelated process could exhibit \"by chance\" a scaling exponent different from the theoretical 0.5 value. To statistically differentiate the stride time series from a random uncorrelated process, we applied the surrogate data method [1,3]. This method increases the confidence that the analyzed series exhibits long-range correlation. Twenty different surrogate data sets were generated by shuffling the original time series in a random order. On each data set, DFA analysis was performed to calculate α value. The standard deviation and mean of this sample was calculated and compared to α exponent of the original series. The result is considered significant if the original α is 2 standard deviation away from the mean of the surrogate data set.", "The method for quantifying the local dynamical stability of the gait by using largest Lyapunov exponent has been extensively described in literature [8]. It examines structural characteristics of a time series that is embedded in an appropriately constructed state space. A valid state space contains a sufficient number of independent coordinates to define the state of the system unequivocally [38]. According to the Takens' theorem, an appropriate state space can be reconstructed from a single time series using the original data and its time delayed copies (figure 3A) [38].\nMethod: dynamic stability, maximal Lyapunov exponent. A: Two dimensional state space of the antero-posterior acceleration signal (5s) reconstructed from the original data set and its time delayed copy (Δt = 11 samples). B: Magnification of the state space. An initial local perturbation at dj(0) diverge across i time steps as measured by dj(i). C: Short term (λS*) and long term (λL*) finite-time maximal Lyapunov exponent computed from average logarithmic divergence.\n\n\n(2)\n\n\nX\n(\nt\n)\n=\n[\nx\n(\nt\n)\n,\nx\n(\nt\n+\nT\n)\n,\nx\n(\nt\n+\n2\nT\n)\n,\n…\n,\nx\n(\nt\n+\n(\n\nd\nE\n\n−\n1\n)\nT\n]\n\n\n\n\nWhere X(t) is the dE-dimensional state vector, x(t) are the original data, T is the time delay, and dE is the embedding dimension. The time delays (T) were calculated individually for each of the 120 acceleration data set (3-axis, 2 conditions, and 20 individuals) from the first minimum of the Average Mutual Information (AMI) function [8,39]. Embedding dimensions (dE) were computed from a Global False Nearest Neighbors (GFNN) analysis [8,40]. Because the result was similar for all acceleration time series, we use a constant dimension (dE = 6) [8,32]]. The Lyapunov exponent is the mean exponential rate of divergence of initially nearby points in the reconstructed space (Figure 3B). Because the determination of the maximal Lypunov exponent requires intensive computing power, 7 min of the 10 min walking test (from 1.5 to 8.5 min.) was selected and the raw data were down-sampled to 100 Hz. The determination of the Lyapunov exponent was then achieved by using the algorithm introduced by Rosenstein and colleagues [7], which provided dedicated software to compute divergence as a function of time in finite-time series [41] (Figure 3B). The maximum finite-time Lyapunov exponents (λ*) were estimated from the slopes of linear fits in the divergence diagrams (Figure 3C). Strictly speaking, because divergence diagrams (Figure 3C) are non-linear, multiple slopes could be defined and so no true single maximum Lyapunov exponent exists. The slopes (exponents) quantify local divergence (and hence stability) of the observed dynamics at different time scale, and should not be interpreted as a classical maximal Lyapunov exponent in chaos theory.\nSince each subject exhibited a different average step frequency, the time was normalized by average stride time for each subject and each condition (Figure 3C). As suggested by Dingwell and colleagues [32], we use two different time scales for assessing short-term and long-term dynamic stability: short term exponents (λS*) was computed over the first stride (0 to 1), and long term exponents (λL*) between 4 and 10 strides (Figure 3C).", "Mean and Standard Deviation (SD) were computed to describe the data (table 1). Ninety-five percent Confidence Intervals (CI) were calculated as ± 1.96 times the Standard Error of the Mean (SEM, N = 20).\nThe effect size of TW as compared to OW was expressed in both absolute (mean difference) and standardized (mean difference divided by SD) terms. The standardized effect size was the Hedge's g, which is a modified version of the Cohen's d for inferential measure [42]. Paired t-tests between OW and TW were performed, and the p-values are shown in the last column of table 1. The precision of the effect sizes was estimated with CI (Figure 4). CI were ± 1.96 times the asymptotic estimates of the standard error (SE) of g [42]. The arbitrary limit of 0.5 was uses to delineate small effect size, as defined by Cohen [42]. The extent of the data (quartiles and median) and individual differences between conditions are shown in Figure 5 for λ*. In order to facilitate results interpretation by reducing the risk of type I statistical error, a Hotelling T2 test was used. This is a multivariate generalization of paired t-test [43]. The null hypothesis is that a vector of p differences is equal to a vector of zeros. Two multivariate sets were tested: meanSD (p = 3) and λ* (p = 6).\nCanonical correlation analyses (CCA, table 2&3) were performed in order to assess the strength of the relationships between different sets of variables [43]. This multivariate method allows one to find linear combinations (variates) in two sets of variables, which have maximum correlation (canonical correlation coefficient or canonical root) with each other. For each condition (OW and TW), two sets of p variables were analyzed: kinematic variability (set#1, p = 3) including MeanSD in ML, V and AP directions, and dynamic stability (set#2, p = 6), including short term and long term lyapunov exponent (λS*, λL*) in ML, V and AP directions. In addition, α scaling exponent was also analyzed with the same method vs. set#1 and set#2. In this case, CCA is equivalent to multiple regression analysis. Significance of the canonical correlations was assessed with the Wilks' lambda statistics.\nTo enhance the interpretation of CCA, different parameters were computed: the standardized canonical weights are the linear coefficients for each set after Z-transform of the variables; canonical loadings are the correlation coefficients between each variable and their respective linear composites; redundancy expresses the amount of variance in one set explained by a linear composite of the other set.", "[SUBTITLE] Treadmill effect [SUBSECTION] As presented in table 1, TW did not modify the stride-to-stride kinematic variability of normalized acceleration pattern, either considering multivariate T2 statistics (p = 0.87) or individual results for each direction. TW was on average performed at slightly lower cadence than Overground Walking (OW, 3% relative difference). The variability of stride interval was similar under both conditions. DFA of stride intervals revealed that TW changed the fractal dynamics of walking (-11% relative difference). Globally, multivariate analysis showed that the data are compatible with the assumption that TW modified dynamic stability of the gait (T2 (6, 20) p = 0.0002). Five from six particular λ* exponents exhibited significant differences.\nComparison between Overground and Treadmill Walking\nThe Descriptive statistics of variability indexes are expressed as mean, Standard Deviation (SD) and 95% Confidence Interval (mean ± 1.96 times the Standard Error of the Mean). The effect size is given as Absolute (Abs.) and Normalized (Norm.) values, i.e. respectively the difference between Overground (OW) and Treadmill (TW) conditions (Abs.) and the difference normalized by SD (Hedge's g). The t-test column shows the p values of paired t-tests between TW and OW conditions. T2-test is the Hotelling multivariate test by regrouping MeanSD and λ*. Significant results (p < 0.05) are printed in bold. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior, i.e. the 3 directions of the triaxial accelerometer.\nFigure 4 shows the accuracy of the effect size estimation. Non-linear estimators of gait variability (α, λ*) exhibit mostly medium effect size.\nDifferences between overground and treadmill walking. Effect size and confidence intervals. Black circles are the standardized effect size (Hedge's g), as reported in table 1. Horizontal lines are the 95% confidence intervals. The arbitrary limit of 0.5 (vertical dotted line) corresponds to a medium effect as defined by Cohen.\nFigure 5 shows the individual results of the local dynamic stability (λ*). Stability was clearly increased (lower λ*) for a majority of subjects except for long-range Antero-Posterior stability λL*.\nIndividual changes of dynamic stability (λ*). Lyapunov exponent λL* and λS* of the 20 subjects are presented for Overground Walking (OW) and Treadmill Walking (TW). Discontinuous lines join OW and TW results. Boxplots show the quartiles and the median.\nFigure 6 presents the individual results of surrogate testing of fractal dynamics. The response to TW was not homogenous among subjects. Four subjects (20%) exhibited a significant turn of long range correlations to uncorrelated pattern. For ten more subjects (50%), a reduction was observed (more than 0.05), but outside the significant limits.\nDetrended Fluctuation Analysis: surrogate data tests. The time series of stride intervals (Figure 1B) of each subject (#1 to #20) were analyzed by DFA (figure 1C) to determine the scaling exponent α indicating the presence of a long range correlation pattern in stride intervals. Black and white circles are respectively the scaling exponent for Overground Walking (OW) and Treadmill Walking (TW). Each time series was randomly shuffled twenty times to produce 20 surrogate time series. The average of these series is near 0.5 (random process with no correlation). The vertical bars show the extent of 2 times the SD of the 20 surrogate time series. Scaling exponent larger than this value can be considered significantly different from a random uncorrelated series.\nAs presented in table 1, TW did not modify the stride-to-stride kinematic variability of normalized acceleration pattern, either considering multivariate T2 statistics (p = 0.87) or individual results for each direction. TW was on average performed at slightly lower cadence than Overground Walking (OW, 3% relative difference). The variability of stride interval was similar under both conditions. DFA of stride intervals revealed that TW changed the fractal dynamics of walking (-11% relative difference). Globally, multivariate analysis showed that the data are compatible with the assumption that TW modified dynamic stability of the gait (T2 (6, 20) p = 0.0002). Five from six particular λ* exponents exhibited significant differences.\nComparison between Overground and Treadmill Walking\nThe Descriptive statistics of variability indexes are expressed as mean, Standard Deviation (SD) and 95% Confidence Interval (mean ± 1.96 times the Standard Error of the Mean). The effect size is given as Absolute (Abs.) and Normalized (Norm.) values, i.e. respectively the difference between Overground (OW) and Treadmill (TW) conditions (Abs.) and the difference normalized by SD (Hedge's g). The t-test column shows the p values of paired t-tests between TW and OW conditions. T2-test is the Hotelling multivariate test by regrouping MeanSD and λ*. Significant results (p < 0.05) are printed in bold. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior, i.e. the 3 directions of the triaxial accelerometer.\nFigure 4 shows the accuracy of the effect size estimation. Non-linear estimators of gait variability (α, λ*) exhibit mostly medium effect size.\nDifferences between overground and treadmill walking. Effect size and confidence intervals. Black circles are the standardized effect size (Hedge's g), as reported in table 1. Horizontal lines are the 95% confidence intervals. The arbitrary limit of 0.5 (vertical dotted line) corresponds to a medium effect as defined by Cohen.\nFigure 5 shows the individual results of the local dynamic stability (λ*). Stability was clearly increased (lower λ*) for a majority of subjects except for long-range Antero-Posterior stability λL*.\nIndividual changes of dynamic stability (λ*). Lyapunov exponent λL* and λS* of the 20 subjects are presented for Overground Walking (OW) and Treadmill Walking (TW). Discontinuous lines join OW and TW results. Boxplots show the quartiles and the median.\nFigure 6 presents the individual results of surrogate testing of fractal dynamics. The response to TW was not homogenous among subjects. Four subjects (20%) exhibited a significant turn of long range correlations to uncorrelated pattern. For ten more subjects (50%), a reduction was observed (more than 0.05), but outside the significant limits.\nDetrended Fluctuation Analysis: surrogate data tests. The time series of stride intervals (Figure 1B) of each subject (#1 to #20) were analyzed by DFA (figure 1C) to determine the scaling exponent α indicating the presence of a long range correlation pattern in stride intervals. Black and white circles are respectively the scaling exponent for Overground Walking (OW) and Treadmill Walking (TW). Each time series was randomly shuffled twenty times to produce 20 surrogate time series. The average of these series is near 0.5 (random process with no correlation). The vertical bars show the extent of 2 times the SD of the 20 surrogate time series. Scaling exponent larger than this value can be considered significantly different from a random uncorrelated series.\n[SUBTITLE] Correlations [SUBSECTION] Table 2 shows the correlation matrix (Perason's r) of the variables under both conditions. It can be observed that correlations exist between the same variables measured along different axes (for instance MeanSD ML vs. MeanSD V, r = 0.92), what makes difficult the global interpretation of potential correlation among the different variability indexes.\nCorrelation matrix\nPearson's r correlation coefficients between the variables. SD = Mean Standard Deviation (MeanSD). λS* = maximal Lyapunov exponent, short term dynamic stability. λL* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior. Significant correlation are bold printed (p < 0.05).\nIn table 3, the results of 6 CCA are shown in details in order to explore global correlation hypotheses. The data seem compatible with the hypothesis that a negative correlation exists between kinematic variability (MeanSD) and local dynamic stability (λ*) under TW condition. Namely, two significant canonical roots (R2 = 0.88 and 0.62) indicates that the canonical variates share an important variance. In addition, the canonical loadings show that the canonical model extract a substantial portion of the variance from the variables (70% from the set#1 and 27% from the set#2). Finally, the redundancy analysis reveals that at least 70% of the variance of the set#2 (stability) can be explained by the set#1 (kinematic variability). The five other CCA did not produce clear evidence for significant relationship between the analyzed sets of variables. Three CCA showed low and non significant canonical roots. Two CCA exhibited barely significant correlation, but the analysis of loadings showed that the canonical model did not explain a large part of the variances in the sets.\nCanonical Correlation Analysis (CCA)\nCanonical correlation analysis between 6 sets of variables. SD = Mean Standard Deviation. λS* = maximal Lyapunov exponent, short term dynamic stability. λ:* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior.\nTable 2 shows the correlation matrix (Perason's r) of the variables under both conditions. It can be observed that correlations exist between the same variables measured along different axes (for instance MeanSD ML vs. MeanSD V, r = 0.92), what makes difficult the global interpretation of potential correlation among the different variability indexes.\nCorrelation matrix\nPearson's r correlation coefficients between the variables. SD = Mean Standard Deviation (MeanSD). λS* = maximal Lyapunov exponent, short term dynamic stability. λL* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior. Significant correlation are bold printed (p < 0.05).\nIn table 3, the results of 6 CCA are shown in details in order to explore global correlation hypotheses. The data seem compatible with the hypothesis that a negative correlation exists between kinematic variability (MeanSD) and local dynamic stability (λ*) under TW condition. Namely, two significant canonical roots (R2 = 0.88 and 0.62) indicates that the canonical variates share an important variance. In addition, the canonical loadings show that the canonical model extract a substantial portion of the variance from the variables (70% from the set#1 and 27% from the set#2). Finally, the redundancy analysis reveals that at least 70% of the variance of the set#2 (stability) can be explained by the set#1 (kinematic variability). The five other CCA did not produce clear evidence for significant relationship between the analyzed sets of variables. Three CCA showed low and non significant canonical roots. Two CCA exhibited barely significant correlation, but the analysis of loadings showed that the canonical model did not explain a large part of the variances in the sets.\nCanonical Correlation Analysis (CCA)\nCanonical correlation analysis between 6 sets of variables. SD = Mean Standard Deviation. λS* = maximal Lyapunov exponent, short term dynamic stability. λ:* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior.", "As presented in table 1, TW did not modify the stride-to-stride kinematic variability of normalized acceleration pattern, either considering multivariate T2 statistics (p = 0.87) or individual results for each direction. TW was on average performed at slightly lower cadence than Overground Walking (OW, 3% relative difference). The variability of stride interval was similar under both conditions. DFA of stride intervals revealed that TW changed the fractal dynamics of walking (-11% relative difference). Globally, multivariate analysis showed that the data are compatible with the assumption that TW modified dynamic stability of the gait (T2 (6, 20) p = 0.0002). Five from six particular λ* exponents exhibited significant differences.\nComparison between Overground and Treadmill Walking\nThe Descriptive statistics of variability indexes are expressed as mean, Standard Deviation (SD) and 95% Confidence Interval (mean ± 1.96 times the Standard Error of the Mean). The effect size is given as Absolute (Abs.) and Normalized (Norm.) values, i.e. respectively the difference between Overground (OW) and Treadmill (TW) conditions (Abs.) and the difference normalized by SD (Hedge's g). The t-test column shows the p values of paired t-tests between TW and OW conditions. T2-test is the Hotelling multivariate test by regrouping MeanSD and λ*. Significant results (p < 0.05) are printed in bold. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior, i.e. the 3 directions of the triaxial accelerometer.\nFigure 4 shows the accuracy of the effect size estimation. Non-linear estimators of gait variability (α, λ*) exhibit mostly medium effect size.\nDifferences between overground and treadmill walking. Effect size and confidence intervals. Black circles are the standardized effect size (Hedge's g), as reported in table 1. Horizontal lines are the 95% confidence intervals. The arbitrary limit of 0.5 (vertical dotted line) corresponds to a medium effect as defined by Cohen.\nFigure 5 shows the individual results of the local dynamic stability (λ*). Stability was clearly increased (lower λ*) for a majority of subjects except for long-range Antero-Posterior stability λL*.\nIndividual changes of dynamic stability (λ*). Lyapunov exponent λL* and λS* of the 20 subjects are presented for Overground Walking (OW) and Treadmill Walking (TW). Discontinuous lines join OW and TW results. Boxplots show the quartiles and the median.\nFigure 6 presents the individual results of surrogate testing of fractal dynamics. The response to TW was not homogenous among subjects. Four subjects (20%) exhibited a significant turn of long range correlations to uncorrelated pattern. For ten more subjects (50%), a reduction was observed (more than 0.05), but outside the significant limits.\nDetrended Fluctuation Analysis: surrogate data tests. The time series of stride intervals (Figure 1B) of each subject (#1 to #20) were analyzed by DFA (figure 1C) to determine the scaling exponent α indicating the presence of a long range correlation pattern in stride intervals. Black and white circles are respectively the scaling exponent for Overground Walking (OW) and Treadmill Walking (TW). Each time series was randomly shuffled twenty times to produce 20 surrogate time series. The average of these series is near 0.5 (random process with no correlation). The vertical bars show the extent of 2 times the SD of the 20 surrogate time series. Scaling exponent larger than this value can be considered significantly different from a random uncorrelated series.", "Table 2 shows the correlation matrix (Perason's r) of the variables under both conditions. It can be observed that correlations exist between the same variables measured along different axes (for instance MeanSD ML vs. MeanSD V, r = 0.92), what makes difficult the global interpretation of potential correlation among the different variability indexes.\nCorrelation matrix\nPearson's r correlation coefficients between the variables. SD = Mean Standard Deviation (MeanSD). λS* = maximal Lyapunov exponent, short term dynamic stability. λL* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior. Significant correlation are bold printed (p < 0.05).\nIn table 3, the results of 6 CCA are shown in details in order to explore global correlation hypotheses. The data seem compatible with the hypothesis that a negative correlation exists between kinematic variability (MeanSD) and local dynamic stability (λ*) under TW condition. Namely, two significant canonical roots (R2 = 0.88 and 0.62) indicates that the canonical variates share an important variance. In addition, the canonical loadings show that the canonical model extract a substantial portion of the variance from the variables (70% from the set#1 and 27% from the set#2). Finally, the redundancy analysis reveals that at least 70% of the variance of the set#2 (stability) can be explained by the set#1 (kinematic variability). The five other CCA did not produce clear evidence for significant relationship between the analyzed sets of variables. Three CCA showed low and non significant canonical roots. Two CCA exhibited barely significant correlation, but the analysis of loadings showed that the canonical model did not explain a large part of the variances in the sets.\nCanonical Correlation Analysis (CCA)\nCanonical correlation analysis between 6 sets of variables. SD = Mean Standard Deviation. λS* = maximal Lyapunov exponent, short term dynamic stability. λ:* = maximal Lyapunov exponent, long term dynamic stability. α = scaling exponent (Detrended Fluctuation Analysis), fractal dynamics. ML, V and AP stand for respectively Medio-Lateral, Vertical and Antero-posterior.", "The purpose of the present study was to analyze three gait variability indexes under two walking conditions in order to highlight modifications induced by motorized treadmill and to analyze the relationship between the indexes.\nAccording to the working hypothesis, the results are summarized as follows:\n1) As compared to Overground Walking (OW), Treadmill Walking (TW) significantly reduced the average scaling exponent (lower α), but did not reverse the correlated pattern to a random or anti-persistent pattern in a majority of subjects.\n2) TW significantly increased local dynamic stability (lower λ*).\n3) TW did not significantly modify the kinematic variability (MeanSD).\n4) No evident relationship was observed between variability indexes during OW at preferred walking speed, but in TW significant negative correlation was found between kinematic variability (MeanSD) and stability (λ*).\nOverall, Conventional variability analysis (MeanSD) failed to report differences between OW and TW, whereas non-linear approaches were able to show significant changes. The variability indexes were poorly correlated together (with one exception), which might signify that each index was related to a different aspect of motor control.\n[SUBTITLE] Technical issues [SUBSECTION] For the present study, portable trunk accelerometry was chosen because it offers the possibility to record long-term free walking. Hence, the results concern the gait stability measured from accelerations of the low-back. Comparisons with other results should take into account that that the different gait stability studies use different kinematic variables (acceleration [7,10], positions [44], angle [8]) and different body location (thorax, head, knee, and ankle) to assess λ*. We found λ* similar to those measured by others [8,32], suggesting that the results are rather independent on the measurements methods.\nIn fractal dynamics studies, the first step is the detection of the periodic pattern of the gait in order to compute time series of stride intervals. Several methodologies have been used to measure long-term time series of stride intervals, such as foot switches [3,5], goniometer [45], video analysis [46], or high accuracy GPS [1]. Because the same variable is used (i.e. time duration of the gait cycle) for DFA analyses, data from different studies are probably comparable.\nIn order to increase the likelihood to point out significant correlations among variability indexes, we designed the experiment to obtain a substantial degree of standardization: we imposed the same speed (1.25 m/s, 4.5 km/h) for all subjects on the treadmill. This speed was chosen on the basis of a previous experiment (partially published yet [34]), which showed that the preferred speed in the same experimental conditions (same room, same treadmill) was 1.26 ± 0.13 m/s (n = 88). Similar values are found in the literature: 1.25 m/s (n = 8) [47], 1.19 m/s (n = 26) [48].\nWalking speed was not standardized between TW and OW, as in other studies [32]. However, by selecting treadmill speed at the same speed of overground preferred speed, the results would be that subjects walk at higher speed than their preferred speed on the treadmill. Several studies showed a substantial difference between both conditions: Dal et al. [48] demonstrated that preferred walking speed determined on a treadmill is slower than overground (21% relative difference); Marsh et al. [49]showed that, when older adults were allowed to choose a preferred walking pace, they walked faster (+61%), used longer strides, and had a faster rate walking overground than when they walked on a treadmill. As a result, speed normalization could introduce unwanted bias. Our experimental design was therefore a compromise, which standardized speed among subjects in TW condition, but also which selected walking speed close to preferred speed, making both OW and TW conditions comparable.\nIn addition, Indirect clues seem to indicate that TW and OW conditions were quite similar: 1) stride time (which is related to walking speed) were close (3% difference, small effect size), 2) stride time variability (CV) was the same (no significant differences), 3) no correlation was observed between stride time and other parameters (results not shown),\nFor the present study, portable trunk accelerometry was chosen because it offers the possibility to record long-term free walking. Hence, the results concern the gait stability measured from accelerations of the low-back. Comparisons with other results should take into account that that the different gait stability studies use different kinematic variables (acceleration [7,10], positions [44], angle [8]) and different body location (thorax, head, knee, and ankle) to assess λ*. We found λ* similar to those measured by others [8,32], suggesting that the results are rather independent on the measurements methods.\nIn fractal dynamics studies, the first step is the detection of the periodic pattern of the gait in order to compute time series of stride intervals. Several methodologies have been used to measure long-term time series of stride intervals, such as foot switches [3,5], goniometer [45], video analysis [46], or high accuracy GPS [1]. Because the same variable is used (i.e. time duration of the gait cycle) for DFA analyses, data from different studies are probably comparable.\nIn order to increase the likelihood to point out significant correlations among variability indexes, we designed the experiment to obtain a substantial degree of standardization: we imposed the same speed (1.25 m/s, 4.5 km/h) for all subjects on the treadmill. This speed was chosen on the basis of a previous experiment (partially published yet [34]), which showed that the preferred speed in the same experimental conditions (same room, same treadmill) was 1.26 ± 0.13 m/s (n = 88). Similar values are found in the literature: 1.25 m/s (n = 8) [47], 1.19 m/s (n = 26) [48].\nWalking speed was not standardized between TW and OW, as in other studies [32]. However, by selecting treadmill speed at the same speed of overground preferred speed, the results would be that subjects walk at higher speed than their preferred speed on the treadmill. Several studies showed a substantial difference between both conditions: Dal et al. [48] demonstrated that preferred walking speed determined on a treadmill is slower than overground (21% relative difference); Marsh et al. [49]showed that, when older adults were allowed to choose a preferred walking pace, they walked faster (+61%), used longer strides, and had a faster rate walking overground than when they walked on a treadmill. As a result, speed normalization could introduce unwanted bias. Our experimental design was therefore a compromise, which standardized speed among subjects in TW condition, but also which selected walking speed close to preferred speed, making both OW and TW conditions comparable.\nIn addition, Indirect clues seem to indicate that TW and OW conditions were quite similar: 1) stride time (which is related to walking speed) were close (3% difference, small effect size), 2) stride time variability (CV) was the same (no significant differences), 3) no correlation was observed between stride time and other parameters (results not shown),\n[SUBTITLE] Differences between treadmill and overground walking [SUBSECTION] Kinematic variability, fractal dynamics (DFA) and local dynamic stability (Lyapunov exponents) quantify different aspects of locomotor control [32]. Kinematic variability describes the range in which the locomotor system operates. DFA quantify temporal dynamics of discrete events (i.e stride interval) over hundreds of consecutives strides; it assesses the presence of long-range correlations between strides, and hence analyzes the characteristics of feedbacks in locomotor control. Lyapunov exponents quantify the temporal dynamics in continuous time based on the theory of deterministic chaos; it evaluates the degree of divergence in the signal, and hence the resilience of the locomotor system to small perturbations. Therefore, it can be expected that these variability indexes did not react in the same way under various conditions.\nThese assumptions were experimentally verified in various studies that observed changes of λ* and kinematic variability between different experimental conditions or between different populations. For instance it was observed that patients with peripheral neuropathy present altered dynamic stability but normal kinematic variability [33,50]. Other investigators have shown that an exercise training intervention in elderly people could improve dynamic stability but not decrease kinematic variability [19].\nDespite differences in the method of measurement (low-back vs. thorax acceleration) and in the experimental design (speed normalization), our results are generally in accordance with the results of Dingwell et al [32]. They analyzed only 10 healthy individuals, therefore statistical significance for small effects was more difficult to reach than in the present study. They showed a significant treadmill effect in short-term stability (lower λS*). A slight but not significant effect for long-term vertical stability (lower λL*) was found. They observed that kinematic variability (MeanSD) for upper body accelerations was generally greater for OW than TW, but this trend was only significant for antero-posterior accelerations. They explained that underlying causes of differences between TW and OW were unclear: on one hand, the motorized treadmill imposed a constant nominal speed on the subjects and constrained them to walk along a much narrower and straighter path than during OW; but on the other hand, differences may have been induced by intra-stride fluctuations in treadmill belt speed, differences in mechanical compliance between the walking surfaces, and changes in visual and vestibular perceptual information. In light of the results of the present study, we hypothesize that motor control is able to maintain the same range of kinematic variability in both TW and OW conditions (same kinematic variability), probably because of compensating effects: in TW, destabilizing factors (intra-stride belt speed fluctuations, disturbing mechanical compliance, alteration of perceptual information) are balanced by stabilizing factors (constant speed, narrow and straight path). Conversely, motor control strategy adapting the gait to TW seems to specifically alter non-linear dependencies among consecutive strides: the stabilizing factors override the destabilizing ones.\nIn a subsequent study, Dingwell & Marin [51] analyzed speed effect on dynamical stability (λS* and λL*) and kinematic variability (MeanSD). Walking speed was normalized by individual PWS on a treadmill. Speed range was 0.6PW to 1.4PWS by steps of 0.2. They found significant speed effect for both λ* and MeanSD: however the effect was small for 0.8-1.2 PWS. Under our experimental conditions [34], we observed that inter-indivudual variability of PWS on the treadmill was low: 90% of individuals walked in the range of 0.87-1.13 mean PWS. As a result, the speed effect among individuals in the present study was probably low. This is also indirectly confirmed by the low inter-individual variability of stride duration (CV = 6%).\nFractal dynamics of stride intervals has been extensively studied by Hausdorff et al. [36]. Them and other [1,3,52] have observed that constrained walking (paced cadence with a metronome), deeply modified the scaling exponent. By analogy, because treadmill also constraints the gait by imposing a constant speed, a similar effect could be expected. The results of the present study showed, in a majority of subjects, a lowering of scaling exponent to a less correlated pattern. The effect was not as strong as with paced walking [1]. The explanation could be that treadmill constrained walking speed, while metronome constrained walking pace; it could be hypothesized that the adaptation of locomotor control to external cues specifically modify correlation pattern of the constrained walking parameter, as suggested by the results of Terrier et al. [1], but this remains to be investigated.\nKinematic variability, fractal dynamics (DFA) and local dynamic stability (Lyapunov exponents) quantify different aspects of locomotor control [32]. Kinematic variability describes the range in which the locomotor system operates. DFA quantify temporal dynamics of discrete events (i.e stride interval) over hundreds of consecutives strides; it assesses the presence of long-range correlations between strides, and hence analyzes the characteristics of feedbacks in locomotor control. Lyapunov exponents quantify the temporal dynamics in continuous time based on the theory of deterministic chaos; it evaluates the degree of divergence in the signal, and hence the resilience of the locomotor system to small perturbations. Therefore, it can be expected that these variability indexes did not react in the same way under various conditions.\nThese assumptions were experimentally verified in various studies that observed changes of λ* and kinematic variability between different experimental conditions or between different populations. For instance it was observed that patients with peripheral neuropathy present altered dynamic stability but normal kinematic variability [33,50]. Other investigators have shown that an exercise training intervention in elderly people could improve dynamic stability but not decrease kinematic variability [19].\nDespite differences in the method of measurement (low-back vs. thorax acceleration) and in the experimental design (speed normalization), our results are generally in accordance with the results of Dingwell et al [32]. They analyzed only 10 healthy individuals, therefore statistical significance for small effects was more difficult to reach than in the present study. They showed a significant treadmill effect in short-term stability (lower λS*). A slight but not significant effect for long-term vertical stability (lower λL*) was found. They observed that kinematic variability (MeanSD) for upper body accelerations was generally greater for OW than TW, but this trend was only significant for antero-posterior accelerations. They explained that underlying causes of differences between TW and OW were unclear: on one hand, the motorized treadmill imposed a constant nominal speed on the subjects and constrained them to walk along a much narrower and straighter path than during OW; but on the other hand, differences may have been induced by intra-stride fluctuations in treadmill belt speed, differences in mechanical compliance between the walking surfaces, and changes in visual and vestibular perceptual information. In light of the results of the present study, we hypothesize that motor control is able to maintain the same range of kinematic variability in both TW and OW conditions (same kinematic variability), probably because of compensating effects: in TW, destabilizing factors (intra-stride belt speed fluctuations, disturbing mechanical compliance, alteration of perceptual information) are balanced by stabilizing factors (constant speed, narrow and straight path). Conversely, motor control strategy adapting the gait to TW seems to specifically alter non-linear dependencies among consecutive strides: the stabilizing factors override the destabilizing ones.\nIn a subsequent study, Dingwell & Marin [51] analyzed speed effect on dynamical stability (λS* and λL*) and kinematic variability (MeanSD). Walking speed was normalized by individual PWS on a treadmill. Speed range was 0.6PW to 1.4PWS by steps of 0.2. They found significant speed effect for both λ* and MeanSD: however the effect was small for 0.8-1.2 PWS. Under our experimental conditions [34], we observed that inter-indivudual variability of PWS on the treadmill was low: 90% of individuals walked in the range of 0.87-1.13 mean PWS. As a result, the speed effect among individuals in the present study was probably low. This is also indirectly confirmed by the low inter-individual variability of stride duration (CV = 6%).\nFractal dynamics of stride intervals has been extensively studied by Hausdorff et al. [36]. Them and other [1,3,52] have observed that constrained walking (paced cadence with a metronome), deeply modified the scaling exponent. By analogy, because treadmill also constraints the gait by imposing a constant speed, a similar effect could be expected. The results of the present study showed, in a majority of subjects, a lowering of scaling exponent to a less correlated pattern. The effect was not as strong as with paced walking [1]. The explanation could be that treadmill constrained walking speed, while metronome constrained walking pace; it could be hypothesized that the adaptation of locomotor control to external cues specifically modify correlation pattern of the constrained walking parameter, as suggested by the results of Terrier et al. [1], but this remains to be investigated.\n[SUBTITLE] Correlations between variability indicators [SUBSECTION] While fractal dynamics, local dynamic stability and kinematic variability characterize different features of gait variability, it is not excluded that relationships exists between them.\nJordan et al. [46] recently analyzed fractal dynamics and stability in walking/running transition on treadmill. They observed a positive correlation between λL* and α (r2 = 0.65, N = 12). They also observed that scaling exponent is minimal close to PWS [53] and suggested that \"reduced strength of long range correlations at preferred locomotion speeds is reflective of enhanced stability and adaptability at theses speeds\". Our results, using CCA, did not confirm this suggestion. No evident correlation between scaling exponent and dynamic stability was found. Several differences in the measurement method (trunk accelerometry vs 3D video analysis) and in the experimental design (high speed vs. moderate speed) may explain this divergence.\nPrevious studies have analyzed the relationships between variability (meanSD) and local dynamic stability (λS* and λL*). Dingwell et al. pointed out \"the general lack of correlation between the standard deviation and λ* exponents\" [32]. In contrast, other investigators recently observed significant positive correlation between λS* and MeanSD [54]. The results of the present study showed a counterintuitive negative correlation between λ* and MeanSD: during treadmill walking (but not in OW), higher kinematic variability seemed to be related to higher local stability (i.e. low λ*). As explained above, the use of different methodologies is a potential source of divergence between studies concerning dynamic stability. It is not excluded that a confounding factor, not measured yet, related to both MeanSD and λ* could indirectly explain this correlation. Further investigations are needed to better understand the relationship between these two variability indexes.\nWhile fractal dynamics, local dynamic stability and kinematic variability characterize different features of gait variability, it is not excluded that relationships exists between them.\nJordan et al. [46] recently analyzed fractal dynamics and stability in walking/running transition on treadmill. They observed a positive correlation between λL* and α (r2 = 0.65, N = 12). They also observed that scaling exponent is minimal close to PWS [53] and suggested that \"reduced strength of long range correlations at preferred locomotion speeds is reflective of enhanced stability and adaptability at theses speeds\". Our results, using CCA, did not confirm this suggestion. No evident correlation between scaling exponent and dynamic stability was found. Several differences in the measurement method (trunk accelerometry vs 3D video analysis) and in the experimental design (high speed vs. moderate speed) may explain this divergence.\nPrevious studies have analyzed the relationships between variability (meanSD) and local dynamic stability (λS* and λL*). Dingwell et al. pointed out \"the general lack of correlation between the standard deviation and λ* exponents\" [32]. In contrast, other investigators recently observed significant positive correlation between λS* and MeanSD [54]. The results of the present study showed a counterintuitive negative correlation between λ* and MeanSD: during treadmill walking (but not in OW), higher kinematic variability seemed to be related to higher local stability (i.e. low λ*). As explained above, the use of different methodologies is a potential source of divergence between studies concerning dynamic stability. It is not excluded that a confounding factor, not measured yet, related to both MeanSD and λ* could indirectly explain this correlation. Further investigations are needed to better understand the relationship between these two variability indexes.", "For the present study, portable trunk accelerometry was chosen because it offers the possibility to record long-term free walking. Hence, the results concern the gait stability measured from accelerations of the low-back. Comparisons with other results should take into account that that the different gait stability studies use different kinematic variables (acceleration [7,10], positions [44], angle [8]) and different body location (thorax, head, knee, and ankle) to assess λ*. We found λ* similar to those measured by others [8,32], suggesting that the results are rather independent on the measurements methods.\nIn fractal dynamics studies, the first step is the detection of the periodic pattern of the gait in order to compute time series of stride intervals. Several methodologies have been used to measure long-term time series of stride intervals, such as foot switches [3,5], goniometer [45], video analysis [46], or high accuracy GPS [1]. Because the same variable is used (i.e. time duration of the gait cycle) for DFA analyses, data from different studies are probably comparable.\nIn order to increase the likelihood to point out significant correlations among variability indexes, we designed the experiment to obtain a substantial degree of standardization: we imposed the same speed (1.25 m/s, 4.5 km/h) for all subjects on the treadmill. This speed was chosen on the basis of a previous experiment (partially published yet [34]), which showed that the preferred speed in the same experimental conditions (same room, same treadmill) was 1.26 ± 0.13 m/s (n = 88). Similar values are found in the literature: 1.25 m/s (n = 8) [47], 1.19 m/s (n = 26) [48].\nWalking speed was not standardized between TW and OW, as in other studies [32]. However, by selecting treadmill speed at the same speed of overground preferred speed, the results would be that subjects walk at higher speed than their preferred speed on the treadmill. Several studies showed a substantial difference between both conditions: Dal et al. [48] demonstrated that preferred walking speed determined on a treadmill is slower than overground (21% relative difference); Marsh et al. [49]showed that, when older adults were allowed to choose a preferred walking pace, they walked faster (+61%), used longer strides, and had a faster rate walking overground than when they walked on a treadmill. As a result, speed normalization could introduce unwanted bias. Our experimental design was therefore a compromise, which standardized speed among subjects in TW condition, but also which selected walking speed close to preferred speed, making both OW and TW conditions comparable.\nIn addition, Indirect clues seem to indicate that TW and OW conditions were quite similar: 1) stride time (which is related to walking speed) were close (3% difference, small effect size), 2) stride time variability (CV) was the same (no significant differences), 3) no correlation was observed between stride time and other parameters (results not shown),", "Kinematic variability, fractal dynamics (DFA) and local dynamic stability (Lyapunov exponents) quantify different aspects of locomotor control [32]. Kinematic variability describes the range in which the locomotor system operates. DFA quantify temporal dynamics of discrete events (i.e stride interval) over hundreds of consecutives strides; it assesses the presence of long-range correlations between strides, and hence analyzes the characteristics of feedbacks in locomotor control. Lyapunov exponents quantify the temporal dynamics in continuous time based on the theory of deterministic chaos; it evaluates the degree of divergence in the signal, and hence the resilience of the locomotor system to small perturbations. Therefore, it can be expected that these variability indexes did not react in the same way under various conditions.\nThese assumptions were experimentally verified in various studies that observed changes of λ* and kinematic variability between different experimental conditions or between different populations. For instance it was observed that patients with peripheral neuropathy present altered dynamic stability but normal kinematic variability [33,50]. Other investigators have shown that an exercise training intervention in elderly people could improve dynamic stability but not decrease kinematic variability [19].\nDespite differences in the method of measurement (low-back vs. thorax acceleration) and in the experimental design (speed normalization), our results are generally in accordance with the results of Dingwell et al [32]. They analyzed only 10 healthy individuals, therefore statistical significance for small effects was more difficult to reach than in the present study. They showed a significant treadmill effect in short-term stability (lower λS*). A slight but not significant effect for long-term vertical stability (lower λL*) was found. They observed that kinematic variability (MeanSD) for upper body accelerations was generally greater for OW than TW, but this trend was only significant for antero-posterior accelerations. They explained that underlying causes of differences between TW and OW were unclear: on one hand, the motorized treadmill imposed a constant nominal speed on the subjects and constrained them to walk along a much narrower and straighter path than during OW; but on the other hand, differences may have been induced by intra-stride fluctuations in treadmill belt speed, differences in mechanical compliance between the walking surfaces, and changes in visual and vestibular perceptual information. In light of the results of the present study, we hypothesize that motor control is able to maintain the same range of kinematic variability in both TW and OW conditions (same kinematic variability), probably because of compensating effects: in TW, destabilizing factors (intra-stride belt speed fluctuations, disturbing mechanical compliance, alteration of perceptual information) are balanced by stabilizing factors (constant speed, narrow and straight path). Conversely, motor control strategy adapting the gait to TW seems to specifically alter non-linear dependencies among consecutive strides: the stabilizing factors override the destabilizing ones.\nIn a subsequent study, Dingwell & Marin [51] analyzed speed effect on dynamical stability (λS* and λL*) and kinematic variability (MeanSD). Walking speed was normalized by individual PWS on a treadmill. Speed range was 0.6PW to 1.4PWS by steps of 0.2. They found significant speed effect for both λ* and MeanSD: however the effect was small for 0.8-1.2 PWS. Under our experimental conditions [34], we observed that inter-indivudual variability of PWS on the treadmill was low: 90% of individuals walked in the range of 0.87-1.13 mean PWS. As a result, the speed effect among individuals in the present study was probably low. This is also indirectly confirmed by the low inter-individual variability of stride duration (CV = 6%).\nFractal dynamics of stride intervals has been extensively studied by Hausdorff et al. [36]. Them and other [1,3,52] have observed that constrained walking (paced cadence with a metronome), deeply modified the scaling exponent. By analogy, because treadmill also constraints the gait by imposing a constant speed, a similar effect could be expected. The results of the present study showed, in a majority of subjects, a lowering of scaling exponent to a less correlated pattern. The effect was not as strong as with paced walking [1]. The explanation could be that treadmill constrained walking speed, while metronome constrained walking pace; it could be hypothesized that the adaptation of locomotor control to external cues specifically modify correlation pattern of the constrained walking parameter, as suggested by the results of Terrier et al. [1], but this remains to be investigated.", "While fractal dynamics, local dynamic stability and kinematic variability characterize different features of gait variability, it is not excluded that relationships exists between them.\nJordan et al. [46] recently analyzed fractal dynamics and stability in walking/running transition on treadmill. They observed a positive correlation between λL* and α (r2 = 0.65, N = 12). They also observed that scaling exponent is minimal close to PWS [53] and suggested that \"reduced strength of long range correlations at preferred locomotion speeds is reflective of enhanced stability and adaptability at theses speeds\". Our results, using CCA, did not confirm this suggestion. No evident correlation between scaling exponent and dynamic stability was found. Several differences in the measurement method (trunk accelerometry vs 3D video analysis) and in the experimental design (high speed vs. moderate speed) may explain this divergence.\nPrevious studies have analyzed the relationships between variability (meanSD) and local dynamic stability (λS* and λL*). Dingwell et al. pointed out \"the general lack of correlation between the standard deviation and λ* exponents\" [32]. In contrast, other investigators recently observed significant positive correlation between λS* and MeanSD [54]. The results of the present study showed a counterintuitive negative correlation between λ* and MeanSD: during treadmill walking (but not in OW), higher kinematic variability seemed to be related to higher local stability (i.e. low λ*). As explained above, the use of different methodologies is a potential source of divergence between studies concerning dynamic stability. It is not excluded that a confounding factor, not measured yet, related to both MeanSD and λ* could indirectly explain this correlation. Further investigations are needed to better understand the relationship between these two variability indexes.", "Scaling exponent (α) and maximal Lypunov exponent (λ*) have been advocated as a relevant indicator of neuromuscular control of stability during human locomotion [8,32,36,55]. The results of the present study showed that treadmill modified fractal dynamics (α) and local dynamic stability (λ*) of the gait, but not kinematic variability (MeanSD). This should be kept in mind when using motorized treadmill either for fundamental research or in locomotor therapies.\nWhereas both scaling exponent (α) and maximal Lypunov exponent (λ*) are sensitive enough to identify differences between OW and TW, they seem not correlated together. This suggests that both indexes deserve to be used in conjunction when analyzing long term gait variability, because they describe different locomotor characteristics.", "The authors declare that they have no competing interests.", "PT performed measurements and data analysis, and drafted the manuscript. OD participated in the design and coordination of the study and assisted with drafting the manuscript. All authors read and approved the final manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Abortion, Spontaneous", "Adult", "Boston", "Cohort Studies", "Embryo Implantation", "Environmental Exposure", "Environmental Pollutants", "False Positive Reactions", "Female", "Fertilization in Vitro", "Humans", "Polychlorinated Biphenyls", "Pregnancy", "Pregnancy Outcome" ]

[ "Results", "Discussion", "Supplemental Material" ]

[ "Data on a total of 827 IVF or ICSI cycles from 765 women were included in our analysis (Table 1). The women had a mean age of 35.9 years and were primarily Caucasian (91%) nonsmokers (92%). Almost one-quarter of the women (24%) had a previous live birth. Approximately two-thirds of the women had either unexplained infertility (34%) or were in couples diagnosed with male factor infertility (35%), and the remaining one-third of women were diagnosed with either tubal factors (19%) or ovulatory dysfunction (13%). Distributions of serum PCB concentrations are presented as both wet weight (nanograms per gram serum) and as lipid-standardized concentrations (nanograms per gram lipid) in Table 2. As expected, PCB-153 was the congener present in the highest concentration (median = 46.2 ng/g lipid; range 4.0–428 ng/g lipid) and, on average, accounted for 18% of ΣPCBs. There were strong correlations between the three primary congeners of interest, the three congener groupings, and ΣPCBs. Spearman correlation coefficients ranged from 0.62 between PCB-118 and group 1 congeners to 0.95 between PCB-153 and group 3 congeners (data not shown).\nDistribution of demographic, covariate, and outcome\ndata.a\nDistribution of serum PCB concentrations among 765\nwomen undergoing IVF/ICSI.\nTable 3 presents adjusted odds ratios (ORs) for failed implantation, chemical pregnancies, and spontaneous abortions in relation to PCB quartiles. Because of missing data for certain covariates, 774 cycles from 720 women contributed to the multivariate analysis (see Table 3 footnote). PCB-153 and ΣPCBs were associated with significantly elevated odds of failed implantation, and these relationships demonstrated dose-dependent trends. The second through the fourth quartiles were associated with ORs [95% confidence intervals (CIs)] of 1.1 (0.7–1.9), 1.3 (0.8–2.2), and 1.7 (1.0–2.9; p-value for trend = 0.03) for ΣPCBs and 1.6 (1.0–2.7), 1.6 (1.0–2.7), and 2.0 (1.2–3.4; p-value for trend = 0.02) for PCB-153, compared with the lowest quartile. There were suggestive trends for increased odds of implantation failure for PCB-118 (p-value for trend = 0.06) and group 3 congeners (p-value for trend = 0.06), but ORs did not follow monotonic dose–response patterns. None of the congeners or congener groupings were significantly associated with odds of chemical pregnancy or spontaneous abortion.\nAdjusted ORsa for IVF/ICSI failures in\nrelation to serum PCB quartiles (Q) among cycles with an embryo\ntransfer.b\nAs a potentially more clinically relevant measure of effect, we also calculated the odds of a live birth in relation to serum PCB quartiles (Table 4). Quartiles of PCB-153 again followed a dose-related trend, where the odds of live birth were reduced by 12%, 35%, and 41% in quartiles 2–4, respectively, compared with the lowest PCB-153 quartile (p-value for trend = 0.03). The third and fourth ΣPCB quartiles were also associated with 41% reductions in the odds of a live birth (p-value for trend = 0.02). Reductions in the odds of live birth were observed for PCB-118 and for group 2 and group 3 congeners, but these relationships were not statistically significant.\nAdjusted ORs (95% CIs)a for live birth in\nrelation to serum PCB quartiles (Q) among cycles with an embryo\ntransfer.b\nBecause many studies standardize PCB concentrations for lipid content before analyzing the data, we conducted a secondary analysis using lipid-standardized values in the models. Overall results were similar to those presented [see Supplemental Material, Table 1 (doi:10.1289/ehp.1002922)]. The primary differences were that the associations for PCB-118 and group 3 congeners with failed implantation and live birth became slightly stronger, whereas the relationships between ΣPCBs and these outcomes were slightly weakened.", "In our study we report that serum concentrations of PCBs in women may be associated with adverse IVF/ICSI outcomes. In particular, we found that the odds of failed implantation were doubled, and the odds of a live birth were reduced by 41%, among women in the highest serum PCB-153 quartile compared with women in the lowest PCB-153 quartile. Serum ΣPCBs, which was strongly correlated with PCB-153, was also associated with implantation failure and reduced odds of a live birth.\nAs far as we are aware, this is the first human study to assess the relationship between PCB exposure and failed implantation, although two small studies have assessed the relationship between PCBs in follicular fluid and other IVF outcomes (e.g., fertilization rate) (Jirsova et al. 2010; Younglai et al. 2002). Because implantation failure is common and occurs before a pregnancy is recognized among couples conceiving spontaneously (Chard 1991; Norwitz et al. 2001), our findings may be comparable with those in epidemiologic studies of PCB exposure and time to pregnancy (TTP). However, TTP as an outcome lacks specificity, because it may represent one or more aspects of female or male factors. Several studies have reported an association between PCB exposure and increased TTP. A study of 186 women exposed to high concentrations of PCBs in the 1978–1979 Taiwanese cooking oil contamination incident reported a fecundability ratio of 0.90 (95% CI, 0.80–1.00) compared with a group of 226 reference women (Yang et al. 2008). The Collaborative Perinatal Project, which included 390 U.S. women who were pregnant between 1959 and 1965, also reported a reduced fecundability odds ratio (FOR) of 0.65 (95% CI, 0.36–1.18) among women in the highest serum ΣPCB category (> 5 µg/L) compared with women in the lowest exposure category (Law et al. 2005). In the first human study to assess preconception serum PCB concentrations and TTP, Buck Louis et al. (2009) reported FORs of 0.32 (95% CI, 0.03–3.89) and 0.01 (< 0.00–1.99) for estrogenic and antiestrogenic PCB congener groupings, respectively, among a subset of 81 women planning pregnancies who were followed for a total of 444 menstrual cycles in the New York State Angler Cohort Study. A number of earlier studies examined organochlorine exposure through consumption of contaminated fish in relation to TTP, with conflicting results (Arakawa et al. 2006; Axmon et al. 2000b, 2001, 2002, 2004, 2006; Buck et al. 2000). Limitations in many of these studies include the use of exposure surrogates instead of exposure biomarkers, self-recall of TTP, and the inclusion of only women who achieved clinical pregnancy or whose pregnancies resulted in a live birth. Our observation of a relationship between PCBs and failed implantation is consistent with the more recent and rigorously designed TTP studies showing an inverse relationship between PCBs and fecundability. Although the women in those studies had serum PCB concentrations that were much higher than the women in our study, our findings add to those studies by providing insight into a specific aspect and time point within early postconception that contributes to TTP and may be particularly sensitive to PCB exposure.\nIn this study, the lack of association between PCBs and risk of spontaneous abortion was consistent with a recent study of 1,344 pregnancies in Michigan women that reported no association between serum PCB concentrations > 5 ng/g wet weight and risk of miscarriage (Small et al. 2007) but was inconsistent with a recent European study of 1,710 women that reported an association between PCB-153 at serum concentrations of > 200 ng/g lipid and increased risk of fetal loss (miscarriages or stillbirths) (Toft et al. 2010). Several earlier studies also reported evidence for a relationship between PCBs and miscarriage (Bercovici et al. 1983; Leoni et al. 1989; Tsukimori et al. 2008), whereas other earlier studies reported no associations (Axmon et al. 2000a; Dar et al. 1992; Khanjani and Sim 2007; Mendola et al. 1995; Sugiura-Ogasawara et al. 2003). Many of the early studies were limited by small sample sizes or indirect exposure assessments (e.g., consumption of contaminated fish). In addition, study designs and PCB exposure levels between studies have differed greatly. Our data do not support the presence of an association between PCBs at levels similar to those among the general population and fetal loss. However, the overall body of evidence remains inconsistent, and the generalizability of our findings to highly exposed populations or women not undergoing assisted reproduction are unknown, so this relationship may require further inquiry.\nOur observation of an association between PCB exposure and implantation failure is consistent with studies from the experimental literature, many of which were conducted with low and environmentally relevant dose levels (ATSDR 2000). Animal studies have demonstrated that PCBs cause reduced oocyte maturation (reviewed by Pocar et al. 2006), increased embryo degeneration and decreased embryo cell proliferation, blastocyst formation and blastocyst development, and decreased IVF success rates (Campagna et al. 2001, 2002; Kholkute and Dukelow 1997; Kholkute et al. 1994a, 1994b; Krogenaes et al. 1998; Kuchenhoff et al. 1999; Lindenau and Fischer 1996; Pocar et al. 2001), along with increased implantation failures (Linder et al. 1974), increased resorptions (Arnold et al. 1995), and decreased litter production (Jonsson et al. 1975; Seiler et al. 1994). PCB impacts on uterine receptivity may also contribute to an increased risk of implantation failure. For example, consistent with animal research (ATSDR 2000), some epidemiologic studies have reported associations between PCB exposure and altered menstrual cycles (Chao et al. 2007; Cooper et al. 2005; Mendola et al. 1997; Toft et al. 2004, 2008; Yu et al. 2000) and endometriosis (Anger and Foster 2008; Gerhard et al. 1999; Heilier et al. 2008; Porpora et al. 2006, 2009).\nOur study has a number of strengths. First, it was a prospective study with exposure biomarkers and outcome measures collected at or near the likely time window of interest, which represents an improvement over many of the previous studies of PCBs and fertility or pregnancy outcomes that relied on retrospective study designs. For example, some previous studies have used exposure surrogate measures, such as self-reported contaminated fish intake, which may introduce more exposure measurement error than the use of exposure biomarkers. This study used detailed, clinically assessed data on outcome measures, whereas many previous studies relied on participant recall and may have been susceptible to recall bias. Second, we were able to assess early end points that are impossible or difficult to observe in conventional epidemiologic studies. Third, this study was large and used outcome data from repeated IVF cycles, which resulted in good statistical power to detect associations between serum PCB concentrations and failed implantation, chemical pregnancies, and spontaneous abortions. Finally, most previous studies have been conducted among populations with elevated PCB exposures, whereas women in our study had serum PCB concentrations that are more representative of the general population. For example, median (75th percentile) PCB-153 concentrations in serum samples collected between 1994 and 2003 in this study were 46 (70) ng/g lipid compared with 35 (63) ng/g lipid and 24 (47) ng/g lipid among adults ≥ 20 years of age from the National Health and Nutrition Examination Survey (NHANES) 2001–2002 and NHANES 2003–2004, respectively (Centers for Disease Control and Prevention 2009).\nOur study also has a number of limitations. First, as with nearly all environmental epidemiologic studies, there may be unmeasured confounding or coexposures that may explain the observed associations. For example, if fish consumption is the primary source of PCB-153, it is possible that other exposures or activities related to fish consumption are confounding the relationship of PCB-153 with failed implantation. Second, because our study was conducted among women undergoing IVF, the extent to which we can generalize our findings to wider populations is not clear. For our results to lack generalizability, women undergoing IVF would have to differentially respond to PCB exposure compared with women conceiving spontaneously. There is currently no evidence suggesting these women may be more susceptible to PCB exposure, although it is possible that PCB exposure may contribute to a couple’s fecundability and need for IVF treatment. However, even if women undergoing IVF represent a subgroup particularly sensitive to PCBs, our findings could have important implications, because couples seeking fertility treatment represent a large and growing proportion of the population. Nevertheless, the consistency between our findings and human TTP studies suggests that our results may be generalizable. In addition, because our results are consistent with animal studies, they may provide further evidence that couples undergoing IVF provide a unique opportunity to study detailed reproductive end points not normally observable in humans. Another limitation of the present analysis is that we did not include earlier measures that are observable in couples undergoing IVF, such as oocyte quality, failed fertilization, and embryo quality. Because animal studies suggest that these end points may also be adversely affected by PCBs, these outcomes should be assessed in relation to serum PCB concentrations in the future. Studies that assess more detailed outcome measures may also provide additional clues for the potential mechanisms of action involved in the relationships reported here, which currently remain unclear. Finally, we did not have exposure data on male partners. There have been several reports of an association between PCB exposure and reduced semen quality (Meeker and Hauser 2010), which may also contribute to increased TTP and poor embryo quality in relation to PCB exposure.\nIn conclusion, we found that serum PCB concentrations representative of those measured among the U.S. general population were associated with increased odds of failed implantation among women undergoing IVF. These findings may help explain previous reports of reduced fecundability and increased TTP among women exposed to PCBs.", "Click here for additional data file." ]

[ null, null, null ]

[ "Methods", "Results", "Discussion", "Supplemental Material" ]

[ "Study population. Details of the main study within which the present substudy was performed have been described previously (Meeker et al. 2007a, 2007b). Briefly, between August 1994 and June 2003, couples undergoing IVF or ICSI were recruited through three Boston-area clinics to participate in a study of predictors of IVF outcomes. The main study was conducted in two phases (1994–1998 and 1999–2003) corresponding to a 5-year renewal of the study after the completion of the first 5 years. Study protocols were approved by the human research committees at Brigham and Women´s Hospital, Harvard School of Public Health, and the University of Michigan. Study subjects gave written consent before participating in the study. Approximately 65% of couples who were approached agreed to participate in the study. Most women who declined participation cited lack of time or did not want the added stress of participating. Participation did not vary greatly by clinic, and IVF outcomes did not differ by participants and nonparticipants. Couples who required donor gametes, gestational carrier, or gamete intrafallopian transfer were not eligible for enrollment. After these exclusions, 2,350 couples were enrolled in the main study. Many couples underwent multiple IVF/ICSI cycles (up to six), with the mean being two cycles per couple.\nAll IVF outcome variables were defined from data abstracted from the medical record by trained research nurses. Only women who proceeded to embryo transfer were eligible for the present analysis. When at least one embryo was transferred but human chorionic gonadotropin (hCG) levels did not reach 5.0 mIU/mL, the cycle outcome was defined as a failure of implantation. A chemical pregnancy was defined by a luteal hCG measurement of ≥ 5.0 mIU/mL with no further evidence (gestational sac and fetal heartbeat) of a continued pregnancy. Clinical pregnancy was determined by ultrasound visualization of a gestational sac and a fetal heartbeat. Among clinically recognized pregnancies, outcomes included in the analysis were spontaneous abortion (fetal demise before 20 weeks of gestation) and live birth of at least one infant. Numbers for the outcomes of ectopic pregnancy (gestation outside of the uterus), molar pregnancy, and stillbirth (fetal demise at 20 gestational weeks or later) lacked power to investigate independently and were excluded from further consideration in the present analysis.\nBecause of financial constraints, we were unable to measure PCBs in serum from all women and cycles in the study. Thus, to maximize statistical power we based our sampling strategy on the outcomes of interest (i.e., failed implantation, chemical pregnancy, and spontaneous abortion) to improve statistical power. Because failed implantation is a common occurrence, we analyzed serum PCB concentrations for a random sample of 200 women with first-cycle implantation failures. Because chemical pregnancies and spontaneous abortions were less common, all cycles from the entire cohort with these outcomes were selected. A corresponding sample of first-cycle live births was selected using stratified sampling based on age category, study center, and study phase. In addition to these selected women, a random sample of 265 repeated cycles among a subset of 110 women was selected for PCB analysis in a prior analysis of within-woman variability in serum PCB concentrations (Meeker et al. 2009), and outcome data from those cycles were also considered in the present analysis. In total, 765 women, who underwent a total 827 IVF/ICSI cycles, were included in the present analysis. The 827 cycles comprised 229 implantation failures, 177 chemical pregnancies, 124 spontaneous abortions, and 297 live births.\nMeasurement of PCBs in serum. Blood samples were collected from women at their first IVF/ICSI cycle during the follicular phase immediately before hCG administration. Our recent longitudinal analysis showed a very strong correlation between PCB concentrations measured in blood samples collected during different IVF cycles within women (Meeker et al. 2009). Thus, only a single blood sample from the first cycle of each woman was analyzed for PCBs. The serum fraction was separated for all blood samples by centrifugation for 5 min, aliquoted, and stored at –80°C until analysis. Measurement of PCBs in serum was conducted by the organic chemistry analytical laboratory, Harvard School of Public Health, using methods described previously (Hauser et al. 2003; Korrick et al. 2000). In brief, after liquid-liquid extraction and column chromatography cleanup, the samples were analyzed by gas chromatography with dual micro-electron capture detection on two capillary columns of different polarity using two internal standards. Samples were accompanied by the following quality control samples: procedural blanks, matrix spike samples, and laboratory control samples. Each sample was spiked with two surrogate compounds to monitor the efficiency of the extraction procedure. All final results were reported after subtracting the amount of the analyte measured in the procedural blank associated with the analytic batch. Results were not adjusted for surrogate recoveries. Target analytes included 57 individual PCB congeners. Method detection limits (MDLs) were determined as three times the standard deviation obtained from the analysis of the eight aliquots of pooled serum fortified with target analytes at 0.02 ng/g serum (U.S. Environmental Protection Agency 1984). The wet-weight MDL values for all PCB congeners were < 0.05 ng/g, with most of the congeners < 0.01 ng/g.\nBecause PCBs partition according to the lipid content of tissues, and serum lipid levels vary between fasting and nonfasting states, serum lipids also need to be considered for the valid interpretation of serum levels (Phillips et al. 1989a). Serum total cholesterol and triglycerides were measured enzymatically, and total lipids were calculated by Phillips formula (Bernert et al. 2007; Phillips et al. 1989a).\nStatistical analysis. We explored the relationship between IVF outcomes and three individual PCB congeners (PCB congeners 118, 138, and 153), as well as the sum of all measured congeners (ΣPCBs). In addition, an analysis of the relationship between IVF outcomes and groupings of PCBs was conducted based on structural and biological activity, as previously proposed (Wolff et al. 1997). PCBs were grouped as follows: group 1, potentially estrogenic and weak phenobarbitol inducers (congeners 44, 49, 52, 101, 187, 174, 177, 157/201); group 2, potentially antiestrogenic and dioxin-like (congeners 95/66, 74, 77/110, 105/141, 118, 156, 167, 128, 138, 170); and group 3, phenobarbital, cytochrome P450 (CYP)1A, and CYP2B inducers (congeners 99, 153, 180, 196/203, 183). When forming the summed values, congeners with concentrations below the MDL were assigned a value of MDL divided by the square root of 2.0. Distributions of organochlorine concentrations in serum were tabulated and compared between the primary congeners and congener groupings of interest.\nWe used multivariate generalized linear regression models to assess the association of serum levels of PCB congeners and PCB groupings, categorized into quartiles, with early pregnancy failure. The data were structured to accommodate joint models for multiple outcomes and multiple cycles per woman. In each cycle, whenever a woman was at risk of having a failure due to implantation failure, chemical pregnancy, or spontaneous abortion, a binary outcome (Y) was recorded (1 = failure, 0 = not a failure). Thus, a woman could make up to three contributions to risk sets for the possible outcomes within each IVF cycle. For example, if a woman experienced implantation failure in cycle 1, and spontaneous abortion in cycle 2, but had a live birth in cycle 3, then we would consider the woman at risk only for implantation failure in the first cycle (her outcome is failure), at risk for both implantation failure and spontaneous abortion in the second cycle (she did not fail during the first possible failure point and thus continued to be at risk for the second possible failure type), and at risk for all three failure outcomes for the third cycle (but she did not fail at any of these three failure points). The correlation among both multiple failure types and among cycles for the same woman is taken into account by incorporating a random effect for each woman, which adjusts for the correlation between multiple outcomes on each woman.\nWe then used a logistic regression model for each early pregnancy outcome while adjusting for other covariates. To account for the possibility that the relationship between PCBs and different types of failure end point may differ, we included interaction terms between PCB quartile and failure type in the model. Similar interaction terms were also considered for age and failure type to explore the possibility of differing age effects on the various failure end points, although these interactions were not statistically significant and were removed from the final models. To account for dependence between two cycles for a particular woman, we included woman-specific random effects, which were assumed to be Gaussian. Tests for trend were conducted by assigning each PCB quartile an ordinal integer value of 0 (lowest quartile) to 3 (highest quartile).\nPotential confounding variables considered in our analysis included those that may be associated with serum PCB concentrations and/or treatment outcome: serum lipids, site (of the three participating Boston-area clinics), study phase (1994–1998 or 1999–2003), race/ethnicity (Caucasian vs. other), previous live birth (yes/no), maternal age (< 35, 35–37, 38–40, > 40 years), body mass index (BMI), smoking status (never, former, or current), the number of ampules of gonadotropins delivered, ovarian stimulation protocol (downregulation vs. other), ICSI (yes/no), number of embryos transferred, and primary infertility diagnosis (tubal factor, ovulatory dysfunction, male factor, or unexplained). The use of serum levels of lipophilic compounds standardized by serum lipids (by dividing serum concentration of the compound of interest by serum lipids) as an independent variable in multivariate models has been shown to be prone to bias (Schisterman et al. 2005). Thus, we also modeled IVF outcomes in relation to wet-weight serum levels of PCBs and adjusted for serum lipids as a covariate in multivariate models, as recommended by Schisterman et al. (2005) because of the lower bias of this approach in most scenarios. Because many studies report PCB effect estimates based on lipid-standardized concentrations, we also used lipid-standardized concentrations as the exposure variable in the models for comparison in a secondary analysis.\nData analysis was conducted using SAS software (Version 9.1; SAS Institute Inc., Cary, NC). p-Values < 0.05 were considered statistically significant.", "Data on a total of 827 IVF or ICSI cycles from 765 women were included in our analysis (Table 1). The women had a mean age of 35.9 years and were primarily Caucasian (91%) nonsmokers (92%). Almost one-quarter of the women (24%) had a previous live birth. Approximately two-thirds of the women had either unexplained infertility (34%) or were in couples diagnosed with male factor infertility (35%), and the remaining one-third of women were diagnosed with either tubal factors (19%) or ovulatory dysfunction (13%). Distributions of serum PCB concentrations are presented as both wet weight (nanograms per gram serum) and as lipid-standardized concentrations (nanograms per gram lipid) in Table 2. As expected, PCB-153 was the congener present in the highest concentration (median = 46.2 ng/g lipid; range 4.0–428 ng/g lipid) and, on average, accounted for 18% of ΣPCBs. There were strong correlations between the three primary congeners of interest, the three congener groupings, and ΣPCBs. Spearman correlation coefficients ranged from 0.62 between PCB-118 and group 1 congeners to 0.95 between PCB-153 and group 3 congeners (data not shown).\nDistribution of demographic, covariate, and outcome\ndata.a\nDistribution of serum PCB concentrations among 765\nwomen undergoing IVF/ICSI.\nTable 3 presents adjusted odds ratios (ORs) for failed implantation, chemical pregnancies, and spontaneous abortions in relation to PCB quartiles. Because of missing data for certain covariates, 774 cycles from 720 women contributed to the multivariate analysis (see Table 3 footnote). PCB-153 and ΣPCBs were associated with significantly elevated odds of failed implantation, and these relationships demonstrated dose-dependent trends. The second through the fourth quartiles were associated with ORs [95% confidence intervals (CIs)] of 1.1 (0.7–1.9), 1.3 (0.8–2.2), and 1.7 (1.0–2.9; p-value for trend = 0.03) for ΣPCBs and 1.6 (1.0–2.7), 1.6 (1.0–2.7), and 2.0 (1.2–3.4; p-value for trend = 0.02) for PCB-153, compared with the lowest quartile. There were suggestive trends for increased odds of implantation failure for PCB-118 (p-value for trend = 0.06) and group 3 congeners (p-value for trend = 0.06), but ORs did not follow monotonic dose–response patterns. None of the congeners or congener groupings were significantly associated with odds of chemical pregnancy or spontaneous abortion.\nAdjusted ORsa for IVF/ICSI failures in\nrelation to serum PCB quartiles (Q) among cycles with an embryo\ntransfer.b\nAs a potentially more clinically relevant measure of effect, we also calculated the odds of a live birth in relation to serum PCB quartiles (Table 4). Quartiles of PCB-153 again followed a dose-related trend, where the odds of live birth were reduced by 12%, 35%, and 41% in quartiles 2–4, respectively, compared with the lowest PCB-153 quartile (p-value for trend = 0.03). The third and fourth ΣPCB quartiles were also associated with 41% reductions in the odds of a live birth (p-value for trend = 0.02). Reductions in the odds of live birth were observed for PCB-118 and for group 2 and group 3 congeners, but these relationships were not statistically significant.\nAdjusted ORs (95% CIs)a for live birth in\nrelation to serum PCB quartiles (Q) among cycles with an embryo\ntransfer.b\nBecause many studies standardize PCB concentrations for lipid content before analyzing the data, we conducted a secondary analysis using lipid-standardized values in the models. Overall results were similar to those presented [see Supplemental Material, Table 1 (doi:10.1289/ehp.1002922)]. The primary differences were that the associations for PCB-118 and group 3 congeners with failed implantation and live birth became slightly stronger, whereas the relationships between ΣPCBs and these outcomes were slightly weakened.", "In our study we report that serum concentrations of PCBs in women may be associated with adverse IVF/ICSI outcomes. In particular, we found that the odds of failed implantation were doubled, and the odds of a live birth were reduced by 41%, among women in the highest serum PCB-153 quartile compared with women in the lowest PCB-153 quartile. Serum ΣPCBs, which was strongly correlated with PCB-153, was also associated with implantation failure and reduced odds of a live birth.\nAs far as we are aware, this is the first human study to assess the relationship between PCB exposure and failed implantation, although two small studies have assessed the relationship between PCBs in follicular fluid and other IVF outcomes (e.g., fertilization rate) (Jirsova et al. 2010; Younglai et al. 2002). Because implantation failure is common and occurs before a pregnancy is recognized among couples conceiving spontaneously (Chard 1991; Norwitz et al. 2001), our findings may be comparable with those in epidemiologic studies of PCB exposure and time to pregnancy (TTP). However, TTP as an outcome lacks specificity, because it may represent one or more aspects of female or male factors. Several studies have reported an association between PCB exposure and increased TTP. A study of 186 women exposed to high concentrations of PCBs in the 1978–1979 Taiwanese cooking oil contamination incident reported a fecundability ratio of 0.90 (95% CI, 0.80–1.00) compared with a group of 226 reference women (Yang et al. 2008). The Collaborative Perinatal Project, which included 390 U.S. women who were pregnant between 1959 and 1965, also reported a reduced fecundability odds ratio (FOR) of 0.65 (95% CI, 0.36–1.18) among women in the highest serum ΣPCB category (> 5 µg/L) compared with women in the lowest exposure category (Law et al. 2005). In the first human study to assess preconception serum PCB concentrations and TTP, Buck Louis et al. (2009) reported FORs of 0.32 (95% CI, 0.03–3.89) and 0.01 (< 0.00–1.99) for estrogenic and antiestrogenic PCB congener groupings, respectively, among a subset of 81 women planning pregnancies who were followed for a total of 444 menstrual cycles in the New York State Angler Cohort Study. A number of earlier studies examined organochlorine exposure through consumption of contaminated fish in relation to TTP, with conflicting results (Arakawa et al. 2006; Axmon et al. 2000b, 2001, 2002, 2004, 2006; Buck et al. 2000). Limitations in many of these studies include the use of exposure surrogates instead of exposure biomarkers, self-recall of TTP, and the inclusion of only women who achieved clinical pregnancy or whose pregnancies resulted in a live birth. Our observation of a relationship between PCBs and failed implantation is consistent with the more recent and rigorously designed TTP studies showing an inverse relationship between PCBs and fecundability. Although the women in those studies had serum PCB concentrations that were much higher than the women in our study, our findings add to those studies by providing insight into a specific aspect and time point within early postconception that contributes to TTP and may be particularly sensitive to PCB exposure.\nIn this study, the lack of association between PCBs and risk of spontaneous abortion was consistent with a recent study of 1,344 pregnancies in Michigan women that reported no association between serum PCB concentrations > 5 ng/g wet weight and risk of miscarriage (Small et al. 2007) but was inconsistent with a recent European study of 1,710 women that reported an association between PCB-153 at serum concentrations of > 200 ng/g lipid and increased risk of fetal loss (miscarriages or stillbirths) (Toft et al. 2010). Several earlier studies also reported evidence for a relationship between PCBs and miscarriage (Bercovici et al. 1983; Leoni et al. 1989; Tsukimori et al. 2008), whereas other earlier studies reported no associations (Axmon et al. 2000a; Dar et al. 1992; Khanjani and Sim 2007; Mendola et al. 1995; Sugiura-Ogasawara et al. 2003). Many of the early studies were limited by small sample sizes or indirect exposure assessments (e.g., consumption of contaminated fish). In addition, study designs and PCB exposure levels between studies have differed greatly. Our data do not support the presence of an association between PCBs at levels similar to those among the general population and fetal loss. However, the overall body of evidence remains inconsistent, and the generalizability of our findings to highly exposed populations or women not undergoing assisted reproduction are unknown, so this relationship may require further inquiry.\nOur observation of an association between PCB exposure and implantation failure is consistent with studies from the experimental literature, many of which were conducted with low and environmentally relevant dose levels (ATSDR 2000). Animal studies have demonstrated that PCBs cause reduced oocyte maturation (reviewed by Pocar et al. 2006), increased embryo degeneration and decreased embryo cell proliferation, blastocyst formation and blastocyst development, and decreased IVF success rates (Campagna et al. 2001, 2002; Kholkute and Dukelow 1997; Kholkute et al. 1994a, 1994b; Krogenaes et al. 1998; Kuchenhoff et al. 1999; Lindenau and Fischer 1996; Pocar et al. 2001), along with increased implantation failures (Linder et al. 1974), increased resorptions (Arnold et al. 1995), and decreased litter production (Jonsson et al. 1975; Seiler et al. 1994). PCB impacts on uterine receptivity may also contribute to an increased risk of implantation failure. For example, consistent with animal research (ATSDR 2000), some epidemiologic studies have reported associations between PCB exposure and altered menstrual cycles (Chao et al. 2007; Cooper et al. 2005; Mendola et al. 1997; Toft et al. 2004, 2008; Yu et al. 2000) and endometriosis (Anger and Foster 2008; Gerhard et al. 1999; Heilier et al. 2008; Porpora et al. 2006, 2009).\nOur study has a number of strengths. First, it was a prospective study with exposure biomarkers and outcome measures collected at or near the likely time window of interest, which represents an improvement over many of the previous studies of PCBs and fertility or pregnancy outcomes that relied on retrospective study designs. For example, some previous studies have used exposure surrogate measures, such as self-reported contaminated fish intake, which may introduce more exposure measurement error than the use of exposure biomarkers. This study used detailed, clinically assessed data on outcome measures, whereas many previous studies relied on participant recall and may have been susceptible to recall bias. Second, we were able to assess early end points that are impossible or difficult to observe in conventional epidemiologic studies. Third, this study was large and used outcome data from repeated IVF cycles, which resulted in good statistical power to detect associations between serum PCB concentrations and failed implantation, chemical pregnancies, and spontaneous abortions. Finally, most previous studies have been conducted among populations with elevated PCB exposures, whereas women in our study had serum PCB concentrations that are more representative of the general population. For example, median (75th percentile) PCB-153 concentrations in serum samples collected between 1994 and 2003 in this study were 46 (70) ng/g lipid compared with 35 (63) ng/g lipid and 24 (47) ng/g lipid among adults ≥ 20 years of age from the National Health and Nutrition Examination Survey (NHANES) 2001–2002 and NHANES 2003–2004, respectively (Centers for Disease Control and Prevention 2009).\nOur study also has a number of limitations. First, as with nearly all environmental epidemiologic studies, there may be unmeasured confounding or coexposures that may explain the observed associations. For example, if fish consumption is the primary source of PCB-153, it is possible that other exposures or activities related to fish consumption are confounding the relationship of PCB-153 with failed implantation. Second, because our study was conducted among women undergoing IVF, the extent to which we can generalize our findings to wider populations is not clear. For our results to lack generalizability, women undergoing IVF would have to differentially respond to PCB exposure compared with women conceiving spontaneously. There is currently no evidence suggesting these women may be more susceptible to PCB exposure, although it is possible that PCB exposure may contribute to a couple’s fecundability and need for IVF treatment. However, even if women undergoing IVF represent a subgroup particularly sensitive to PCBs, our findings could have important implications, because couples seeking fertility treatment represent a large and growing proportion of the population. Nevertheless, the consistency between our findings and human TTP studies suggests that our results may be generalizable. In addition, because our results are consistent with animal studies, they may provide further evidence that couples undergoing IVF provide a unique opportunity to study detailed reproductive end points not normally observable in humans. Another limitation of the present analysis is that we did not include earlier measures that are observable in couples undergoing IVF, such as oocyte quality, failed fertilization, and embryo quality. Because animal studies suggest that these end points may also be adversely affected by PCBs, these outcomes should be assessed in relation to serum PCB concentrations in the future. Studies that assess more detailed outcome measures may also provide additional clues for the potential mechanisms of action involved in the relationships reported here, which currently remain unclear. Finally, we did not have exposure data on male partners. There have been several reports of an association between PCB exposure and reduced semen quality (Meeker and Hauser 2010), which may also contribute to increased TTP and poor embryo quality in relation to PCB exposure.\nIn conclusion, we found that serum PCB concentrations representative of those measured among the U.S. general population were associated with increased odds of failed implantation among women undergoing IVF. These findings may help explain previous reports of reduced fecundability and increased TTP among women exposed to PCBs.", "Click here for additional data file." ]

[ "methods", null, null, null ]

[ "environment", "epidemiology", "female", "organochlorine", "reproduction" ]

[ "Aged", "Aged, 80 and over", "Aprotinin", "Arthroplasty, Replacement, Hip", "Blood Transfusion", "Compression Bandages", "Double-Blind Method", "Erythrocyte Count", "Female", "Hematoma", "Hemostatics", "Humans", "Male", "Middle Aged", "Postoperative Hemorrhage", "Prospective Studies", "Suction" ]

[ "Introduction", "Patients, materials and methods", "Diagnostic tests", "Loss of blood", "Statistics", "Intraoperative blood loss", "Blood loss through the suction drains", "Soft tissue sonography", "Perioperative blood loss", "Blood transfusion", "Erythrocytes", "Hemoglobin", "Infection rate", "Coagulation", "Platelet concentration", "Adverse events" ]

[ "Surgical dissection of soft tissue and bone in arthroplasty of the hip joint can cause substantial bleeding and is responsible for intraand postoperative blood loss. The intraoperative blood loss can be reduced by using mainly blunt tissue dissection as far as possible, reduction of tissue trauma, local hypothermia and pharmacological influence on the body's blood coagulation mechanism [22,42,56]. Furthermore, the postoperative blood loss is influenced by the effectiveness and durability of the body's own hemostasis, external tissue compression, and particularly by the method of wound drainage. In 1954, the French surgeons Redon, Jost and Torque evacuated glass bottles to promote drainage of wound secretion after surgery. This was the beginning of the era of postoperative suction drains [7,9,10,12].\nThe idea behind wound drainage is an optimized balance of secretion and accumulation of blood and the internal adaptation and stabilization of wounds. Over the years, the drainage technique developed with various modifications [18-21,23,25]. Aiming to avoid postoperative complications and to provide additional positive effects on wound healing, the postoperative secretion drainage using Redon drains is now a standard method for European Orthopaedic-Trauma surgeons [15]. However, there are serious numbers of complications possibly caused by wound drainage [24,26,29,31].\nOne major disadvantage is the risk for retrograde wound contamination with bacteria. Other reasons are the cytoadhesive properties of polymeric drainage systems, increased postoperative blood loss caused by the contact of the drains with tissue surfaces. There are also arguments of injured vessels and impairment of wound healing through foreign body reaction as well as the release of toxic plasticizers, especially when using drainage made of polyvinyl chloride [24,26,31,32]. Concomitantly, numerous studies have addressed the positive effects of a drainage free surgical technique [15,21,28,34]. These results show that wound drainage can be waived for certain surgical procedures without the risk of impaired wound healing. the question is whether the prophylactic use of suction drains for the patient really is an important procedure to prevent complications or whether it increases the risk for postoperative blood loss and wound infections [32].\nAnother cause of increased blood loss in arthroplasty can be an imbalance in the hemostasis system [29,61]. tissue injury induces the release of proteolytic substances which increase fibrinolysis, leading to a relative lack of endogenous proteinase inhibitors like a-antiplasmin and a-macroglobulin. the proteinase inhibitor aprotinin (trasylol® Bayer, Germany) interrupts this mechanism, as it interferes in regulating the mechanisms of bleeding and thereby may reduce perioperative blood loss. Aprotinin, a kallikrein inactivator, was discovered in 1930. It was isolated in 1936 as a trypsin inhibitor by Kunitz and northrop from bovine tissue [4,5,35,40]. these results have been confirmed by many studies showing that aprotinin is able to inhibit a variety of proteinases. Royston and co-workers showed that high doses of aprotinin during cardiac surgery significantly reduced blood loss [50]. Although this effect was achieved through an inhibition of fibrinolysis, there was no increased rate of thrombosis. these results triggered a growing interest in aprotinin and led to a number of studies in cardiac-, vascular-, liverand in orthopedic surgery [3-5,11].\nThe investigation of our study was whether there is a need for suction wound drainage in cement less hip replacement im comparison of wound compression. on the other hand, we sought to analyze whether an intraoperative administration of proteinase inhibitors like aprotinin may lead to a reduction of the perioperative blood loss.", "In a prospective randomized study, patients older than 18 years undergoing elective implantation of cement less arthroplasty of the hip due to manifest coxarthrosis were included (Figure 1). The study was approved by the local Ethics Committee. Patients with a known intolerance to aprotinin, pre-existing coagulation disorders or undergoing drug therapy for regulating coagulation were excluded from the study.\nStudy design.\nThe study was a prospective, randomized doubleblind parallel trial with 4 arms comparing 80 patients (intention to treat) (51-80 years, 43 38 d\"). We compared trasylol® against placebo in one trial and suction drains versus external compression on the other. All patients were exclusively treated by 2 experienced surgeons participating in earlier studies. Drugs consumed by the patients on admission and during the course of the study were recorded as adjunctive therapy in the protocol. Adverse events between consent and beginning of the studyand the causality between event and study procedure were documented. undercurrent diseases and manifestations were considered as \"adverse events\" and were reported. In totally 80 patients could be included for the trials. From this overall population group, 2 × 40 patients were randomized receiving either for receiving intraoperative 250 ml of a solution containing 500 ku of Trasylol® or 250 ml 0.9% NaCl as a placebo. In addition, each group was subdivided into two arms, either with wound drainage by suction drainages, (n = 20) or with a specific external wound compression (n = 20) (Figure 2).\nExternal wound compression in a patient.\nThe implantation of cement less hip arthroplasty type Zweymuller was carried out on the transgluteal access according to the Bauer procedure.\nAt the end of the operation, the surgeon received the randomized information whether the wound should be treated with two (subcutaneous and subfascial located) suction \"Redon\" drains or with external compression (Figure 2). Drains were removed at the 2nd postoperative day. There was a daily inspection and clinical documentation of the wound healing.\n[SUBTITLE] Diagnostic tests [SUBSECTION] [SUBTITLE] Loss of blood [SUBSECTION] A main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\nA main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\n[SUBTITLE] Loss of blood [SUBSECTION] A main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\nA main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\n[SUBTITLE] Statistics [SUBSECTION] The influence of two factors\"drug-\" (aprotinin/ placebo) and \"non drug therapy\" (drainage/compression) were examined.\nData collection was carried out with the spreadsheet program Excel and data processing with the statistical programs SAS and SPSS Version 16. All data were collected in boxplots and the median was described.\nThe significance level was set at p < 0.05. Comparison of the groups was descriptive for important parameters such as age and gender. A two-way analysis of variance was performed for the total blood loss and exploratory at first for hemoglobin and the 2nd postoperative day, respectively. The analysis of variance was tested for interactions.\nThe influence of two factors\"drug-\" (aprotinin/ placebo) and \"non drug therapy\" (drainage/compression) were examined.\nData collection was carried out with the spreadsheet program Excel and data processing with the statistical programs SAS and SPSS Version 16. All data were collected in boxplots and the median was described.\nThe significance level was set at p < 0.05. Comparison of the groups was descriptive for important parameters such as age and gender. A two-way analysis of variance was performed for the total blood loss and exploratory at first for hemoglobin and the 2nd postoperative day, respectively. The analysis of variance was tested for interactions.", "[SUBTITLE] Loss of blood [SUBSECTION] A main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\nA main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.", "A main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.", "The influence of two factors\"drug-\" (aprotinin/ placebo) and \"non drug therapy\" (drainage/compression) were examined.\nData collection was carried out with the spreadsheet program Excel and data processing with the statistical programs SAS and SPSS Version 16. All data were collected in boxplots and the median was described.\nThe significance level was set at p < 0.05. Comparison of the groups was descriptive for important parameters such as age and gender. A two-way analysis of variance was performed for the total blood loss and exploratory at first for hemoglobin and the 2nd postoperative day, respectively. The analysis of variance was tested for interactions.", "The patients in the Trasylol® group (n = 37) reached a median blood loss of 507 ml (123-1141 ml). The median blood loss in the placebo group (n = 40) was 517 ml (132-1226 ml). In comparison of the medians patients with Trasylol® lost 10 ml more blood perioperatively than patients with a NaCl the placebo, the latter had a larger range. These results were however not statistically significant.", "18 patients with intraoperative drainage receiving Trasylol® had a median blood loss of 620 ml (190-1525 ml). 20 patients with intraoperative NaCl placebo lost a median of 615 ml (100-1260 ml) blood, resulting in a median of 5 ml less blood loss in the Trasylol®-treated group.", "In one patient of the Trasylol®/drainage group, a hematoma with the size of 12.3 ml on the 2nd postoperative day was observed, which did not further increase (5.6%, 1/18).\nIn one patient in the Trasylol®/compression group, a hematoma with a size of 32.6 ml on the 2nd postoperative day developed, which was reduced on the 14th postoperative day to 14.4 ml (5%, 1/20).\nIn the placebo/drainage group 4 patients showed hematomas in the operated wound area, 3 hematomas of 10.5 ml, 11.8 ml and 7.4 ml were found on the 2nd postoperative day. One hematoma increased in fourteen days to a size of 32.6 ml. Furthermore, one hematoma was diagnosed on the 8th postoperative day with bleeding after drainage removal at the 2nd postoperative day. One patient, whose examination was without result on the 2nd postoperative day, developed on the 14th postoperative day a hematoma of 11.8 ml. (20%, 4/20).\nFive hematomas (average size 18.7 ml, range 10.0 35.2) in the placebo group/ compression group were diagnosed on the second postoperative day, one of these increased in fourteen days to a size of 14.3 ml. On the 14th postoperative day, bruises of 25.0 ml and 12.5 ml were seen in two patients (35%, 7/20).\nSummarized, the Trasylol®/Redon group showed the lowest number of total hematoma.", "Patients in the Trasylol®/drainage arm (n = 18) showed a median blood loss of 1201 ml (704-1648 ml). Oppositely, the Trasylol®/compression group (n = 20), presented with a median blood loss of 596 ml (245-1141 ml). The blood loss in the placebo/drainage group (n = 20) had a median of 1223 ml (592-1756 ml), while in the placebo/compression group (n = 20) a median blood loss of 554 ml (143-1063 ml) was measured.\nThe blood loss under the influence of the \"drug therapy\" with Trasylol® was not significantly different from placebo (p = 0.7540).\nOn the other hand the „non drug therapy\" showed a significantly higher blood loss in the wound treatment with suction drainage instead of compression (p < 0.001).", "In the Trasylol®/Redon group (n = 18), 10 patients required blood transfusions within 48 hours (56%), while 7 out of 20 patients (35%) from the Trasylol®/ compression group required a blood transfusion (p = 0.02). Similar, in seven out of 20 patients (35%) in the placebo/Redon group, blood transfusion was performed, while nine of 20 patients (45%) in the placebo/compression group needed a blood transfusion. Variance and interaction analysis of variance showed a significant interaction regarding higher blood loss in the \"non-drug therapy\" with suction drainage (p <0.001) in contrast to compression. Through the application of Trasylol® the blood loss did not change significantly (p = 0.75). Interactions were not detectable (p = 0.74). A re-analysis of variance with neglect of the interaction analysis showed a significantly higher blood loss through the use of suction drains (p < 0.001), for the factor 'non-drug therapy' but for the medical treatment no statistically significant change after Trasylol® application was observed\n(p = 0.75).", "The mean of erythrocyte concentration of the Trasylol®/drainage group (n = 18) was preoperatively 4.1 × 106 (3.6 5.0 × 106/μl), decreased on the 1st postoperative day to 3.5 × 106/μl (2.8 4.0 × 106/μl) and remained constant on the 2nd postoperative day with 3.5 × 106/μl (2.8 to 4.1 × 106/μl).\nFor the Trasylol®/compression group (n = 20), the mean of preoperative red blood cell concentration was 4.2 × 106/μl (3,4 5,0 × 106/μl), decreased on the 1st postoperative day to 3.6 × 106/μl (1.8 4.2 × 106/μl) and on the 2nd postoperative day to 3.5 × 106/μl (2.8 to 4.1 × 106/μl).\nPreoperatively, the mean erythrocyte concentration of the group placebo/drainage (n = 20) was 4.4 × 106/μl (3.5 5.0 × 106/μl), decreased to on the 1st postoperative day to 3.3 × 106/μl (2.9 to 4.1 × 106/μl) and remained unchanged on the 2nd postoperative day with 3.3 × 106/μl (2.7 'to 3.9 × 106/μl).\nThe mean of erythrocyte concentration in the placebo group/compression (n = 20) was preoperatively 4.3 × 106/μl (3.3 to 5.2 × 106/μl) and felt at the first postoperative days to 3.4 × 106/μl (2.7 4.0 × 106/μl) and on the 2nd postoperative day to 3.1 × 106/μl (2.6 to 3.9 × 106/μl).\nPatients treated with Trasylol® had a smaller postoperative decrease in red cell concentration than the placebo groups. The highest difference was found on the first postoperative day.", "Figure 4 shows a lower postoperative decline in hemoglobin values in the Trasylol groups. The difference is not significant.\nBox and Whisker Plot of the perioperative blood loss. The upper line of the box shows the 75% quartile of all values, the lower line the 25% quartile. The line in the middle of the box is the median.\nPostoperative decline in hemoglobin values.", "We could not find a difference in the infection rate in all groups.", "The postoperative coagulation values are shown in Figure 5. No significant difference was detected between the Trasylol® and the placebo group.\nPostoperative changes of Quick values of each group in relation to baseline.", "Figure 6 shows the platelet concentration before and after surgery. No significant difference was detected between the Trasylol® and the placebo group.\nPostoperative changes in platelet concentration of each group in relation to baseline.", "In the Trasylol®/drainage and Trasylol®/compression group, one patient each showed an allergic reaction to Trasylol® and was excluded from the study. In the placebo/drainage group, one patient developed a postoperative hematoma which had to be drained surgically on the 8th postoperative day with consequent removal of the prosthesis due to wound infection. The further course of this patient was uneventful. Another patient had the drainage removed due to abundant secretion from the wound. The further postoperative clinical course was uneventful. In the placebo/ compression group, one patient developed a large postoperative hematoma which was growing in size and was initially treated conservatively. Since a dislocation of the prosthesis occurred during mobilization revision of the hip with prosthesis changes was performed at 6th postoperative day. The further course was uneventful. In another patient a bad-fit brace had to be removed after 24 hours. In the further course no additional complications occurred." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Patients, materials and methods", "Diagnostic tests", "Loss of blood", "Statistics", "Results", "Intraoperative blood loss", "Blood loss through the suction drains", "Soft tissue sonography", "Perioperative blood loss", "Blood transfusion", "Erythrocytes", "Hemoglobin", "Infection rate", "Coagulation", "Platelet concentration", "Adverse events", "Discussion", "Conclusion" ]

[ "Surgical dissection of soft tissue and bone in arthroplasty of the hip joint can cause substantial bleeding and is responsible for intraand postoperative blood loss. The intraoperative blood loss can be reduced by using mainly blunt tissue dissection as far as possible, reduction of tissue trauma, local hypothermia and pharmacological influence on the body's blood coagulation mechanism [22,42,56]. Furthermore, the postoperative blood loss is influenced by the effectiveness and durability of the body's own hemostasis, external tissue compression, and particularly by the method of wound drainage. In 1954, the French surgeons Redon, Jost and Torque evacuated glass bottles to promote drainage of wound secretion after surgery. This was the beginning of the era of postoperative suction drains [7,9,10,12].\nThe idea behind wound drainage is an optimized balance of secretion and accumulation of blood and the internal adaptation and stabilization of wounds. Over the years, the drainage technique developed with various modifications [18-21,23,25]. Aiming to avoid postoperative complications and to provide additional positive effects on wound healing, the postoperative secretion drainage using Redon drains is now a standard method for European Orthopaedic-Trauma surgeons [15]. However, there are serious numbers of complications possibly caused by wound drainage [24,26,29,31].\nOne major disadvantage is the risk for retrograde wound contamination with bacteria. Other reasons are the cytoadhesive properties of polymeric drainage systems, increased postoperative blood loss caused by the contact of the drains with tissue surfaces. There are also arguments of injured vessels and impairment of wound healing through foreign body reaction as well as the release of toxic plasticizers, especially when using drainage made of polyvinyl chloride [24,26,31,32]. Concomitantly, numerous studies have addressed the positive effects of a drainage free surgical technique [15,21,28,34]. These results show that wound drainage can be waived for certain surgical procedures without the risk of impaired wound healing. the question is whether the prophylactic use of suction drains for the patient really is an important procedure to prevent complications or whether it increases the risk for postoperative blood loss and wound infections [32].\nAnother cause of increased blood loss in arthroplasty can be an imbalance in the hemostasis system [29,61]. tissue injury induces the release of proteolytic substances which increase fibrinolysis, leading to a relative lack of endogenous proteinase inhibitors like a-antiplasmin and a-macroglobulin. the proteinase inhibitor aprotinin (trasylol® Bayer, Germany) interrupts this mechanism, as it interferes in regulating the mechanisms of bleeding and thereby may reduce perioperative blood loss. Aprotinin, a kallikrein inactivator, was discovered in 1930. It was isolated in 1936 as a trypsin inhibitor by Kunitz and northrop from bovine tissue [4,5,35,40]. these results have been confirmed by many studies showing that aprotinin is able to inhibit a variety of proteinases. Royston and co-workers showed that high doses of aprotinin during cardiac surgery significantly reduced blood loss [50]. Although this effect was achieved through an inhibition of fibrinolysis, there was no increased rate of thrombosis. these results triggered a growing interest in aprotinin and led to a number of studies in cardiac-, vascular-, liverand in orthopedic surgery [3-5,11].\nThe investigation of our study was whether there is a need for suction wound drainage in cement less hip replacement im comparison of wound compression. on the other hand, we sought to analyze whether an intraoperative administration of proteinase inhibitors like aprotinin may lead to a reduction of the perioperative blood loss.", "In a prospective randomized study, patients older than 18 years undergoing elective implantation of cement less arthroplasty of the hip due to manifest coxarthrosis were included (Figure 1). The study was approved by the local Ethics Committee. Patients with a known intolerance to aprotinin, pre-existing coagulation disorders or undergoing drug therapy for regulating coagulation were excluded from the study.\nStudy design.\nThe study was a prospective, randomized doubleblind parallel trial with 4 arms comparing 80 patients (intention to treat) (51-80 years, 43 38 d\"). We compared trasylol® against placebo in one trial and suction drains versus external compression on the other. All patients were exclusively treated by 2 experienced surgeons participating in earlier studies. Drugs consumed by the patients on admission and during the course of the study were recorded as adjunctive therapy in the protocol. Adverse events between consent and beginning of the studyand the causality between event and study procedure were documented. undercurrent diseases and manifestations were considered as \"adverse events\" and were reported. In totally 80 patients could be included for the trials. From this overall population group, 2 × 40 patients were randomized receiving either for receiving intraoperative 250 ml of a solution containing 500 ku of Trasylol® or 250 ml 0.9% NaCl as a placebo. In addition, each group was subdivided into two arms, either with wound drainage by suction drainages, (n = 20) or with a specific external wound compression (n = 20) (Figure 2).\nExternal wound compression in a patient.\nThe implantation of cement less hip arthroplasty type Zweymuller was carried out on the transgluteal access according to the Bauer procedure.\nAt the end of the operation, the surgeon received the randomized information whether the wound should be treated with two (subcutaneous and subfascial located) suction \"Redon\" drains or with external compression (Figure 2). Drains were removed at the 2nd postoperative day. There was a daily inspection and clinical documentation of the wound healing.\n[SUBTITLE] Diagnostic tests [SUBSECTION] [SUBTITLE] Loss of blood [SUBSECTION] A main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\nA main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\n[SUBTITLE] Loss of blood [SUBSECTION] A main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\nA main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\n[SUBTITLE] Statistics [SUBSECTION] The influence of two factors\"drug-\" (aprotinin/ placebo) and \"non drug therapy\" (drainage/compression) were examined.\nData collection was carried out with the spreadsheet program Excel and data processing with the statistical programs SAS and SPSS Version 16. All data were collected in boxplots and the median was described.\nThe significance level was set at p < 0.05. Comparison of the groups was descriptive for important parameters such as age and gender. A two-way analysis of variance was performed for the total blood loss and exploratory at first for hemoglobin and the 2nd postoperative day, respectively. The analysis of variance was tested for interactions.\nThe influence of two factors\"drug-\" (aprotinin/ placebo) and \"non drug therapy\" (drainage/compression) were examined.\nData collection was carried out with the spreadsheet program Excel and data processing with the statistical programs SAS and SPSS Version 16. All data were collected in boxplots and the median was described.\nThe significance level was set at p < 0.05. Comparison of the groups was descriptive for important parameters such as age and gender. A two-way analysis of variance was performed for the total blood loss and exploratory at first for hemoglobin and the 2nd postoperative day, respectively. The analysis of variance was tested for interactions.", "[SUBTITLE] Loss of blood [SUBSECTION] A main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.\nA main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.", "A main objective of this study was to analyze the amount of perioperative blood loss, resulting from the intraoperative blood loss, blood loss via wound drainage and the size of the hematoma constituted at the wound area.\nThe intraoperative blood loss was calculated based on the collected amount of fluid. The intraoperative administered amount of volume of Ringer's solution was subtracted from the total amount of fluid trapped. Therefore all swabs and tissues were weighed preand postoperatively. The difference between the pre-operative and postoperative weight was divided by the specific gravity of blood (1.06 mg/ml). For all patients with suction drains, the entire post-operative blood loss of suction bottle was documented.\nAll patients received ultrasound (US) diagnostic by the same examiner screening for hematoma at the 2nd and 14th postoperative day. The volume of the detected hematomas by US performed was added to the perioperatively counted loss of blood.\nBlood samples were drawn for every patient preoperatively and at the 1st and 2nd postoperative day. Blood transfusions were performed and documented according to the approved study protocol (Hb <8 mg/l). All patients received a duplex ultrasound to exclude or objective venous thrombosis on the 2nd postoperative day.", "The influence of two factors\"drug-\" (aprotinin/ placebo) and \"non drug therapy\" (drainage/compression) were examined.\nData collection was carried out with the spreadsheet program Excel and data processing with the statistical programs SAS and SPSS Version 16. All data were collected in boxplots and the median was described.\nThe significance level was set at p < 0.05. Comparison of the groups was descriptive for important parameters such as age and gender. A two-way analysis of variance was performed for the total blood loss and exploratory at first for hemoglobin and the 2nd postoperative day, respectively. The analysis of variance was tested for interactions.", "Each group of patients was equal concerning the demographic data. Two patients, one in the drainage and one in the compression group developed a severe allergic reaction after Trasylol® medication and were excluded from further analysis, resulting in a final population of 78 patients.\n[SUBTITLE] Intraoperative blood loss [SUBSECTION] The patients in the Trasylol® group (n = 37) reached a median blood loss of 507 ml (123-1141 ml). The median blood loss in the placebo group (n = 40) was 517 ml (132-1226 ml). In comparison of the medians patients with Trasylol® lost 10 ml more blood perioperatively than patients with a NaCl the placebo, the latter had a larger range. These results were however not statistically significant.\nThe patients in the Trasylol® group (n = 37) reached a median blood loss of 507 ml (123-1141 ml). The median blood loss in the placebo group (n = 40) was 517 ml (132-1226 ml). In comparison of the medians patients with Trasylol® lost 10 ml more blood perioperatively than patients with a NaCl the placebo, the latter had a larger range. These results were however not statistically significant.\n[SUBTITLE] Blood loss through the suction drains [SUBSECTION] 18 patients with intraoperative drainage receiving Trasylol® had a median blood loss of 620 ml (190-1525 ml). 20 patients with intraoperative NaCl placebo lost a median of 615 ml (100-1260 ml) blood, resulting in a median of 5 ml less blood loss in the Trasylol®-treated group.\n18 patients with intraoperative drainage receiving Trasylol® had a median blood loss of 620 ml (190-1525 ml). 20 patients with intraoperative NaCl placebo lost a median of 615 ml (100-1260 ml) blood, resulting in a median of 5 ml less blood loss in the Trasylol®-treated group.\n[SUBTITLE] Soft tissue sonography [SUBSECTION] In one patient of the Trasylol®/drainage group, a hematoma with the size of 12.3 ml on the 2nd postoperative day was observed, which did not further increase (5.6%, 1/18).\nIn one patient in the Trasylol®/compression group, a hematoma with a size of 32.6 ml on the 2nd postoperative day developed, which was reduced on the 14th postoperative day to 14.4 ml (5%, 1/20).\nIn the placebo/drainage group 4 patients showed hematomas in the operated wound area, 3 hematomas of 10.5 ml, 11.8 ml and 7.4 ml were found on the 2nd postoperative day. One hematoma increased in fourteen days to a size of 32.6 ml. Furthermore, one hematoma was diagnosed on the 8th postoperative day with bleeding after drainage removal at the 2nd postoperative day. One patient, whose examination was without result on the 2nd postoperative day, developed on the 14th postoperative day a hematoma of 11.8 ml. (20%, 4/20).\nFive hematomas (average size 18.7 ml, range 10.0 35.2) in the placebo group/ compression group were diagnosed on the second postoperative day, one of these increased in fourteen days to a size of 14.3 ml. On the 14th postoperative day, bruises of 25.0 ml and 12.5 ml were seen in two patients (35%, 7/20).\nSummarized, the Trasylol®/Redon group showed the lowest number of total hematoma.\nIn one patient of the Trasylol®/drainage group, a hematoma with the size of 12.3 ml on the 2nd postoperative day was observed, which did not further increase (5.6%, 1/18).\nIn one patient in the Trasylol®/compression group, a hematoma with a size of 32.6 ml on the 2nd postoperative day developed, which was reduced on the 14th postoperative day to 14.4 ml (5%, 1/20).\nIn the placebo/drainage group 4 patients showed hematomas in the operated wound area, 3 hematomas of 10.5 ml, 11.8 ml and 7.4 ml were found on the 2nd postoperative day. One hematoma increased in fourteen days to a size of 32.6 ml. Furthermore, one hematoma was diagnosed on the 8th postoperative day with bleeding after drainage removal at the 2nd postoperative day. One patient, whose examination was without result on the 2nd postoperative day, developed on the 14th postoperative day a hematoma of 11.8 ml. (20%, 4/20).\nFive hematomas (average size 18.7 ml, range 10.0 35.2) in the placebo group/ compression group were diagnosed on the second postoperative day, one of these increased in fourteen days to a size of 14.3 ml. On the 14th postoperative day, bruises of 25.0 ml and 12.5 ml were seen in two patients (35%, 7/20).\nSummarized, the Trasylol®/Redon group showed the lowest number of total hematoma.\n[SUBTITLE] Perioperative blood loss [SUBSECTION] Patients in the Trasylol®/drainage arm (n = 18) showed a median blood loss of 1201 ml (704-1648 ml). Oppositely, the Trasylol®/compression group (n = 20), presented with a median blood loss of 596 ml (245-1141 ml). The blood loss in the placebo/drainage group (n = 20) had a median of 1223 ml (592-1756 ml), while in the placebo/compression group (n = 20) a median blood loss of 554 ml (143-1063 ml) was measured.\nThe blood loss under the influence of the \"drug therapy\" with Trasylol® was not significantly different from placebo (p = 0.7540).\nOn the other hand the „non drug therapy\" showed a significantly higher blood loss in the wound treatment with suction drainage instead of compression (p < 0.001).\nPatients in the Trasylol®/drainage arm (n = 18) showed a median blood loss of 1201 ml (704-1648 ml). Oppositely, the Trasylol®/compression group (n = 20), presented with a median blood loss of 596 ml (245-1141 ml). The blood loss in the placebo/drainage group (n = 20) had a median of 1223 ml (592-1756 ml), while in the placebo/compression group (n = 20) a median blood loss of 554 ml (143-1063 ml) was measured.\nThe blood loss under the influence of the \"drug therapy\" with Trasylol® was not significantly different from placebo (p = 0.7540).\nOn the other hand the „non drug therapy\" showed a significantly higher blood loss in the wound treatment with suction drainage instead of compression (p < 0.001).\n[SUBTITLE] Blood transfusion [SUBSECTION] In the Trasylol®/Redon group (n = 18), 10 patients required blood transfusions within 48 hours (56%), while 7 out of 20 patients (35%) from the Trasylol®/ compression group required a blood transfusion (p = 0.02). Similar, in seven out of 20 patients (35%) in the placebo/Redon group, blood transfusion was performed, while nine of 20 patients (45%) in the placebo/compression group needed a blood transfusion. Variance and interaction analysis of variance showed a significant interaction regarding higher blood loss in the \"non-drug therapy\" with suction drainage (p <0.001) in contrast to compression. Through the application of Trasylol® the blood loss did not change significantly (p = 0.75). Interactions were not detectable (p = 0.74). A re-analysis of variance with neglect of the interaction analysis showed a significantly higher blood loss through the use of suction drains (p < 0.001), for the factor 'non-drug therapy' but for the medical treatment no statistically significant change after Trasylol® application was observed\n(p = 0.75).\nIn the Trasylol®/Redon group (n = 18), 10 patients required blood transfusions within 48 hours (56%), while 7 out of 20 patients (35%) from the Trasylol®/ compression group required a blood transfusion (p = 0.02). Similar, in seven out of 20 patients (35%) in the placebo/Redon group, blood transfusion was performed, while nine of 20 patients (45%) in the placebo/compression group needed a blood transfusion. Variance and interaction analysis of variance showed a significant interaction regarding higher blood loss in the \"non-drug therapy\" with suction drainage (p <0.001) in contrast to compression. Through the application of Trasylol® the blood loss did not change significantly (p = 0.75). Interactions were not detectable (p = 0.74). A re-analysis of variance with neglect of the interaction analysis showed a significantly higher blood loss through the use of suction drains (p < 0.001), for the factor 'non-drug therapy' but for the medical treatment no statistically significant change after Trasylol® application was observed\n(p = 0.75).\n[SUBTITLE] Erythrocytes [SUBSECTION] The mean of erythrocyte concentration of the Trasylol®/drainage group (n = 18) was preoperatively 4.1 × 106 (3.6 5.0 × 106/μl), decreased on the 1st postoperative day to 3.5 × 106/μl (2.8 4.0 × 106/μl) and remained constant on the 2nd postoperative day with 3.5 × 106/μl (2.8 to 4.1 × 106/μl).\nFor the Trasylol®/compression group (n = 20), the mean of preoperative red blood cell concentration was 4.2 × 106/μl (3,4 5,0 × 106/μl), decreased on the 1st postoperative day to 3.6 × 106/μl (1.8 4.2 × 106/μl) and on the 2nd postoperative day to 3.5 × 106/μl (2.8 to 4.1 × 106/μl).\nPreoperatively, the mean erythrocyte concentration of the group placebo/drainage (n = 20) was 4.4 × 106/μl (3.5 5.0 × 106/μl), decreased to on the 1st postoperative day to 3.3 × 106/μl (2.9 to 4.1 × 106/μl) and remained unchanged on the 2nd postoperative day with 3.3 × 106/μl (2.7 'to 3.9 × 106/μl).\nThe mean of erythrocyte concentration in the placebo group/compression (n = 20) was preoperatively 4.3 × 106/μl (3.3 to 5.2 × 106/μl) and felt at the first postoperative days to 3.4 × 106/μl (2.7 4.0 × 106/μl) and on the 2nd postoperative day to 3.1 × 106/μl (2.6 to 3.9 × 106/μl).\nPatients treated with Trasylol® had a smaller postoperative decrease in red cell concentration than the placebo groups. The highest difference was found on the first postoperative day.\nThe mean of erythrocyte concentration of the Trasylol®/drainage group (n = 18) was preoperatively 4.1 × 106 (3.6 5.0 × 106/μl), decreased on the 1st postoperative day to 3.5 × 106/μl (2.8 4.0 × 106/μl) and remained constant on the 2nd postoperative day with 3.5 × 106/μl (2.8 to 4.1 × 106/μl).\nFor the Trasylol®/compression group (n = 20), the mean of preoperative red blood cell concentration was 4.2 × 106/μl (3,4 5,0 × 106/μl), decreased on the 1st postoperative day to 3.6 × 106/μl (1.8 4.2 × 106/μl) and on the 2nd postoperative day to 3.5 × 106/μl (2.8 to 4.1 × 106/μl).\nPreoperatively, the mean erythrocyte concentration of the group placebo/drainage (n = 20) was 4.4 × 106/μl (3.5 5.0 × 106/μl), decreased to on the 1st postoperative day to 3.3 × 106/μl (2.9 to 4.1 × 106/μl) and remained unchanged on the 2nd postoperative day with 3.3 × 106/μl (2.7 'to 3.9 × 106/μl).\nThe mean of erythrocyte concentration in the placebo group/compression (n = 20) was preoperatively 4.3 × 106/μl (3.3 to 5.2 × 106/μl) and felt at the first postoperative days to 3.4 × 106/μl (2.7 4.0 × 106/μl) and on the 2nd postoperative day to 3.1 × 106/μl (2.6 to 3.9 × 106/μl).\nPatients treated with Trasylol® had a smaller postoperative decrease in red cell concentration than the placebo groups. The highest difference was found on the first postoperative day.\n[SUBTITLE] Hemoglobin [SUBSECTION] Figure 4 shows a lower postoperative decline in hemoglobin values in the Trasylol groups. The difference is not significant.\nBox and Whisker Plot of the perioperative blood loss. The upper line of the box shows the 75% quartile of all values, the lower line the 25% quartile. The line in the middle of the box is the median.\nPostoperative decline in hemoglobin values.\nFigure 4 shows a lower postoperative decline in hemoglobin values in the Trasylol groups. The difference is not significant.\nBox and Whisker Plot of the perioperative blood loss. The upper line of the box shows the 75% quartile of all values, the lower line the 25% quartile. The line in the middle of the box is the median.\nPostoperative decline in hemoglobin values.\n[SUBTITLE] Infection rate [SUBSECTION] We could not find a difference in the infection rate in all groups.\nWe could not find a difference in the infection rate in all groups.\n[SUBTITLE] Coagulation [SUBSECTION] The postoperative coagulation values are shown in Figure 5. No significant difference was detected between the Trasylol® and the placebo group.\nPostoperative changes of Quick values of each group in relation to baseline.\nThe postoperative coagulation values are shown in Figure 5. No significant difference was detected between the Trasylol® and the placebo group.\nPostoperative changes of Quick values of each group in relation to baseline.\n[SUBTITLE] Platelet concentration [SUBSECTION] Figure 6 shows the platelet concentration before and after surgery. No significant difference was detected between the Trasylol® and the placebo group.\nPostoperative changes in platelet concentration of each group in relation to baseline.\nFigure 6 shows the platelet concentration before and after surgery. No significant difference was detected between the Trasylol® and the placebo group.\nPostoperative changes in platelet concentration of each group in relation to baseline.\n[SUBTITLE] Adverse events [SUBSECTION] In the Trasylol®/drainage and Trasylol®/compression group, one patient each showed an allergic reaction to Trasylol® and was excluded from the study. In the placebo/drainage group, one patient developed a postoperative hematoma which had to be drained surgically on the 8th postoperative day with consequent removal of the prosthesis due to wound infection. The further course of this patient was uneventful. Another patient had the drainage removed due to abundant secretion from the wound. The further postoperative clinical course was uneventful. In the placebo/ compression group, one patient developed a large postoperative hematoma which was growing in size and was initially treated conservatively. Since a dislocation of the prosthesis occurred during mobilization revision of the hip with prosthesis changes was performed at 6th postoperative day. The further course was uneventful. In another patient a bad-fit brace had to be removed after 24 hours. In the further course no additional complications occurred.\nIn the Trasylol®/drainage and Trasylol®/compression group, one patient each showed an allergic reaction to Trasylol® and was excluded from the study. In the placebo/drainage group, one patient developed a postoperative hematoma which had to be drained surgically on the 8th postoperative day with consequent removal of the prosthesis due to wound infection. The further course of this patient was uneventful. Another patient had the drainage removed due to abundant secretion from the wound. The further postoperative clinical course was uneventful. In the placebo/ compression group, one patient developed a large postoperative hematoma which was growing in size and was initially treated conservatively. Since a dislocation of the prosthesis occurred during mobilization revision of the hip with prosthesis changes was performed at 6th postoperative day. The further course was uneventful. In another patient a bad-fit brace had to be removed after 24 hours. In the further course no additional complications occurred.", "The patients in the Trasylol® group (n = 37) reached a median blood loss of 507 ml (123-1141 ml). The median blood loss in the placebo group (n = 40) was 517 ml (132-1226 ml). In comparison of the medians patients with Trasylol® lost 10 ml more blood perioperatively than patients with a NaCl the placebo, the latter had a larger range. These results were however not statistically significant.", "18 patients with intraoperative drainage receiving Trasylol® had a median blood loss of 620 ml (190-1525 ml). 20 patients with intraoperative NaCl placebo lost a median of 615 ml (100-1260 ml) blood, resulting in a median of 5 ml less blood loss in the Trasylol®-treated group.", "In one patient of the Trasylol®/drainage group, a hematoma with the size of 12.3 ml on the 2nd postoperative day was observed, which did not further increase (5.6%, 1/18).\nIn one patient in the Trasylol®/compression group, a hematoma with a size of 32.6 ml on the 2nd postoperative day developed, which was reduced on the 14th postoperative day to 14.4 ml (5%, 1/20).\nIn the placebo/drainage group 4 patients showed hematomas in the operated wound area, 3 hematomas of 10.5 ml, 11.8 ml and 7.4 ml were found on the 2nd postoperative day. One hematoma increased in fourteen days to a size of 32.6 ml. Furthermore, one hematoma was diagnosed on the 8th postoperative day with bleeding after drainage removal at the 2nd postoperative day. One patient, whose examination was without result on the 2nd postoperative day, developed on the 14th postoperative day a hematoma of 11.8 ml. (20%, 4/20).\nFive hematomas (average size 18.7 ml, range 10.0 35.2) in the placebo group/ compression group were diagnosed on the second postoperative day, one of these increased in fourteen days to a size of 14.3 ml. On the 14th postoperative day, bruises of 25.0 ml and 12.5 ml were seen in two patients (35%, 7/20).\nSummarized, the Trasylol®/Redon group showed the lowest number of total hematoma.", "Patients in the Trasylol®/drainage arm (n = 18) showed a median blood loss of 1201 ml (704-1648 ml). Oppositely, the Trasylol®/compression group (n = 20), presented with a median blood loss of 596 ml (245-1141 ml). The blood loss in the placebo/drainage group (n = 20) had a median of 1223 ml (592-1756 ml), while in the placebo/compression group (n = 20) a median blood loss of 554 ml (143-1063 ml) was measured.\nThe blood loss under the influence of the \"drug therapy\" with Trasylol® was not significantly different from placebo (p = 0.7540).\nOn the other hand the „non drug therapy\" showed a significantly higher blood loss in the wound treatment with suction drainage instead of compression (p < 0.001).", "In the Trasylol®/Redon group (n = 18), 10 patients required blood transfusions within 48 hours (56%), while 7 out of 20 patients (35%) from the Trasylol®/ compression group required a blood transfusion (p = 0.02). Similar, in seven out of 20 patients (35%) in the placebo/Redon group, blood transfusion was performed, while nine of 20 patients (45%) in the placebo/compression group needed a blood transfusion. Variance and interaction analysis of variance showed a significant interaction regarding higher blood loss in the \"non-drug therapy\" with suction drainage (p <0.001) in contrast to compression. Through the application of Trasylol® the blood loss did not change significantly (p = 0.75). Interactions were not detectable (p = 0.74). A re-analysis of variance with neglect of the interaction analysis showed a significantly higher blood loss through the use of suction drains (p < 0.001), for the factor 'non-drug therapy' but for the medical treatment no statistically significant change after Trasylol® application was observed\n(p = 0.75).", "The mean of erythrocyte concentration of the Trasylol®/drainage group (n = 18) was preoperatively 4.1 × 106 (3.6 5.0 × 106/μl), decreased on the 1st postoperative day to 3.5 × 106/μl (2.8 4.0 × 106/μl) and remained constant on the 2nd postoperative day with 3.5 × 106/μl (2.8 to 4.1 × 106/μl).\nFor the Trasylol®/compression group (n = 20), the mean of preoperative red blood cell concentration was 4.2 × 106/μl (3,4 5,0 × 106/μl), decreased on the 1st postoperative day to 3.6 × 106/μl (1.8 4.2 × 106/μl) and on the 2nd postoperative day to 3.5 × 106/μl (2.8 to 4.1 × 106/μl).\nPreoperatively, the mean erythrocyte concentration of the group placebo/drainage (n = 20) was 4.4 × 106/μl (3.5 5.0 × 106/μl), decreased to on the 1st postoperative day to 3.3 × 106/μl (2.9 to 4.1 × 106/μl) and remained unchanged on the 2nd postoperative day with 3.3 × 106/μl (2.7 'to 3.9 × 106/μl).\nThe mean of erythrocyte concentration in the placebo group/compression (n = 20) was preoperatively 4.3 × 106/μl (3.3 to 5.2 × 106/μl) and felt at the first postoperative days to 3.4 × 106/μl (2.7 4.0 × 106/μl) and on the 2nd postoperative day to 3.1 × 106/μl (2.6 to 3.9 × 106/μl).\nPatients treated with Trasylol® had a smaller postoperative decrease in red cell concentration than the placebo groups. The highest difference was found on the first postoperative day.", "Figure 4 shows a lower postoperative decline in hemoglobin values in the Trasylol groups. The difference is not significant.\nBox and Whisker Plot of the perioperative blood loss. The upper line of the box shows the 75% quartile of all values, the lower line the 25% quartile. The line in the middle of the box is the median.\nPostoperative decline in hemoglobin values.", "We could not find a difference in the infection rate in all groups.", "The postoperative coagulation values are shown in Figure 5. No significant difference was detected between the Trasylol® and the placebo group.\nPostoperative changes of Quick values of each group in relation to baseline.", "Figure 6 shows the platelet concentration before and after surgery. No significant difference was detected between the Trasylol® and the placebo group.\nPostoperative changes in platelet concentration of each group in relation to baseline.", "In the Trasylol®/drainage and Trasylol®/compression group, one patient each showed an allergic reaction to Trasylol® and was excluded from the study. In the placebo/drainage group, one patient developed a postoperative hematoma which had to be drained surgically on the 8th postoperative day with consequent removal of the prosthesis due to wound infection. The further course of this patient was uneventful. Another patient had the drainage removed due to abundant secretion from the wound. The further postoperative clinical course was uneventful. In the placebo/ compression group, one patient developed a large postoperative hematoma which was growing in size and was initially treated conservatively. Since a dislocation of the prosthesis occurred during mobilization revision of the hip with prosthesis changes was performed at 6th postoperative day. The further course was uneventful. In another patient a bad-fit brace had to be removed after 24 hours. In the further course no additional complications occurred.", "The extent of perioperative blood loss is undoubtedly an important aspect of surgical quality. The objective of operative drainage is to avoid the loss of blood and wound fluid. Yet, these can be a potential breeding ground for bacteria or inhibit adequate wound healing. In recent years, the routine use of wound drainage was questioned due to a lack of data on the benefit of the drainage of blood or secretions on wound healing. Moreover, several publications suggest that wound drainage can result in an increased hemoglobin drop with a higher rate of blood transfusions, especially in endoprosthetic surgery of the knee or hip joint. In a retrospective study of 364 patients with a hip prosthesis implantation, Hallstrom and Steele compared wound drainage with wound compression and found a significant higher difference in blood loss for the drainage group [21]. Ritter et al. demonstrated that total hip replacement with drainage increased transfusion volume on an average of 95 ml of blood per patient as compared to total hip replacement without drainage [49]. However, no increased transfusion requirements were documented after implantation of knee replacement. A retrospective study by Reilly et al. for knee replacement revealed a significant decline in hemoglobin and a significantly higher transfusion rate in patients with wound drainage [47].\nIn this study, we demonstrated that the blood loss in the suction drainage group was about 500 ml per patient higher than in the compression group. It has to be noted that the blood loss via the drains, underwent a degree of dilution by serous tissue fluid. Determination of hematocrit in the defect is a more accurate detection for blood loss. With regard to systemic hemoglobin, no statistically significant difference between the drainage and compression groups was evident and no hemoglobin decline was found in any group postoperatively. The higher blood loss rate of the drainage group did not translate into hemoglobin decreases. However, this rather reflects the behavior of the hos-\npital transfusion rules than the actual blood loss. In not drained wounds more ecchymoses occur. Hallstrom and Steele reported a hematoma rate in the undrained wounds of 11.4% versus 2.3% in drained wounds and Holt et al. describe 69% versus 39% [21]. In addition, secondary hematomas have been described after removal of drains. Werner et al. attributes this to the destruction of vascular structures, which are attracted to the vacuum in the lumen of the drainage [58]. Kirschner, Wolter and Tittel et. al. showed that the drains under \"high vacuum\" conditions (about 80 kPa) are just of short time use, since the lumen of the drainage tube is quickly occluded by aspirated tissue [31,55,60]. Parker suggests that the hematoma rate is not affected by drainage [43]. Even in our study we found a hemato-seroma one week after the removal of the drainage tube.\nAn additional parameter for measuring the total postoperative blood loss was ultrasound examination of the wound area, which allows a reasonable estimate of any fluid collection. In our study we did not find hematoma as frequently in the compression group as compared to the drainage group.\nBacteria show a high affinity to polymeric materials. In fact, they produce a biofilm that firmly adheres to the plastic structure. Herein, drains serve as a kind of an entry into the wound area, representing a potential risk factor for an increased postoperative infection rate, which we could not find in our study [24-26,31,32]. This reflects the ambivalence between sensible drainage of wound secretion and potential risk of infection which asks for appropriate hygiene when changing wound dressings and limited duration of drainage therapy. Willett et al. found a colonization of the wound fluid at the free end of the drainage tube in 6% of patients (n = 120) [59]. Also Knapp et al. reported a contamination of 14.4% of all drainage tubes, while the aspirated wound fluid was infected in only 7.4% [32]. With regard to the incidence of wound infections in drained and non-drained wounds no significant difference was found in our study. This result correlates with the literature, stating that the perioperative infection risk and wound healing are not affected by dropping the existence of postoperative wound drainage. With sufficient postoperative tissue compression and venous thrombosis prophylaxis (thus increasing venous return and hemostasis) unimpeded wound healing can be expected. This mechanical way of hemostasis counteracts the formation of hematomas, thus making a contribution to an undisturbed wound healing. The compression of the wound increases the local pressure gradient in the tissue and causes a redistribution of water as an anti-edematous effect. However, compression may cause serious complications. Shall et al. conducted that a too tight compression may aggravate chronic arterial occlusive disease up to ischemia, resulting in pain and eventually irreparable tissue damage [58]. Similarly, a localized compression such as a drastic pressure bandage may cause venous stasis, which paradoxically increases the risk of straight bleeding and thromboembolic complications. A correct bandaging technique should achieve a uniformly increasing pressure gradient from proximal to distal on the extremity, resulting in prophylaxis\nof thrombosis, hemostasis and prevention of edema. In our study, we used ProThera-Huftorthesen (PolyMedics, Peer Belgium), which were individually adapted preoperatively to patients' proximal thigh.\nOnly once the brace had to be removed because of pain and a loose fit after 24 hours. In comparison to winding compression bandages, the procedure was perceived by patients as comfortable and gave stabilizing. Even in the absence of drainage therapy, there was no increase in the rate of hematoma. Additionally, there was a reduction of postoperative blood loss.\nFurthermore, we investigated the effect of aprotinin on perioperative blood loss. Recent research results from cardiovascular and orthopedic surgery, a described reduction in perioperative blood loss due to aprotinin medication has been described [4-6]. The antifibrinolytic effect of aprotinin raises the question of favoring intravascular thrombosis. In formal testing, no significantly increased rates of thrombosis have been shown. One explanation for this is that aprotinin causes a stabilization of the platelet membranes. This results in inhibition of thromboxan release and platelet aggregation which counteracts clot formation.\nThe desired hemostasis by aprotinin is therefore not associated with increased thrombogenesis and does not increase the thromboembolic risk. This concept is in line with our study, as we did not find any evidence of any aprotinin induced increase thrombotic events. However, it is important to note that only clinically manifest thromboses were recorded, a targeted search, with the help of a venography was not performed routinely.\nMurkin et al. investigated in several studies the impact and side effects of aprotinin in surgical procedures [39,40]. With bilateral implantation of hip joint prostheses or hip arthrodeses the average blood loss was reduced by Trasylol® from 2098 ml to 1498 ml, without increased incidence of deep venous thrombosis. Also Samama and Leche et al. obtained in their studies a similar result regarding the influence of aprotinin in orthopedic surgery a similar result [51,52]. Similarly Jeserschek et al. could show in a prospective randomized study a significant reduction of blood loss after administration of high dose aprotonin when performing hip and knee replacements and excisions of soft tissue sarcoma [29].\nOn the other hand, Kasper et al. didn't find a reduction in blood loss by aprotinin medication in arthroplasties of the hip [30,35]. Our own study provided a similar result, since with aprotinin treatment the overall blood loss was not affected. The difference in total blood loss between Trasylol® and placebo groups was clinically insignificant and without statistical significance. Aprotinin had no discernible effect on the amount of transfused blood.\nMeanwhile, despite the possibility of a modest reduction in the risk of massive bleeding, the strong and consistent negative mortality trend associated with apro tinin, as compared with the lysine analogues, precludes its use in high-risk cardiac surgery. Following the results of the \"BART trial\" the manufacturer has withdrawn Trasylol® from the market [1,2,8,13,14,16,17,27,33,36-38,41,44-46,48,53,54,57]. Even in our study two patients developed severe allergic side effects.", "Our study shows that the blood loss in hip surgery is not reduced by the use of aprotinin. Furthermore it seems that wound drainage is not necessary when an external compression of the wound is performed." ]

[ null, null, null, null, null, "results", null, null, null, null, null, null, null, null, null, null, null, "discussion", "conclusions" ]

[ "total hip replacement", "wound drainage", "drug reaction", "aprotinin", "Trasylol®", "compression treatment" ]

[ "Adult", "Confidentiality", "Consumer Health Information", "Cross-Sectional Studies", "Female", "Health Services Research", "Humans", "Information Seeking Behavior", "Internet", "Interviews as Topic", "Male", "Middle Aged", "Motivation", "National Health Programs", "Risk Assessment", "Surveys and Questionnaires", "United Kingdom", "Young Adult" ]

[ "Introduction", "Design", "Analysis", "Results", "Questionnaire Survey", "Interview Sample", "Thematic Analysis of Interview Data", "Motivations", "Benefits and Challenges: Convenience, Coverage, and Confidentiality", "Strategy: Navigating the Health Internet", "The NHS Direct Service", "Discussion", "Limitations", "Conclusions" ]

[ "In the United Kingdom, in 2009, 70% of households had Internet access [1]. With this rise has come an increase in health-related Internet use. The proportion of UK Internet users using online information on health matters increased from 37% in 2005 to 68% in 2009 [1,2]. A survey in seven European countries in 2005 found that 71% of all adults reported using the Internet for health information [3], and in the United States this figure is 61% [4]. Furthermore, public perceptions of the importance of the Internet as a source of health information have risen dramatically [5].\nThe National Health Service (NHS) Direct website is the main health advice and information website for patients and the public in the United Kingdom. It was launched in December 1999, and in 2009, there were 18 million visits to the website, compared with 1.5 million visits in 2001 [6]. At the time of this research it provided health advice (through symptom checkers, a health encyclopedia, and an online enquiry service), information on local health services, articles regarding healthy living and fitness, and many other features, including a pregnancy planner, support for long-term conditions, and access to information about health care abroad. Since mid-2009 it has worked with the NHS Choices service, and some of the features described above have migrated to other parts of the NHS Choices platform. \nThere is limited information describing how and why people use online health information, or the effect of this on health status, although this literature base is growing [6-11]. In this study we used a mixed-methods approach to investigate the characteristics and motivations of online health information seekers in England. In theory, the Internet offers certain advantages as a health information resource. In particular, it provides convenient and anonymous access at any time, from any location, to a wide range of expert sources; and through virtual communities it can provide peer support and social interaction [12]. Health-related use of the Internet has been hailed as a tool to support the emergence of the informed and empowered health consumer, and a shift in the balance of power between patient and professional [13]. At the same time, concerns have been raised about the quality of information, the potential for unhelpful peer-to-peer interactions, and the exclusion of individuals who experience barriers to access [14,15]. In Table 1 we have summarized the main characteristics and potential public health benefits and challenges that have been proposed for health-related Internet use [12,16,17]. In this paper, by exploring the expressed reasons for seeking online health information, we hope to assess the extent to which the theoretical benefits and challenges of the health Internet are being realized in practice.\nTheoretical characteristics and potential public health benefits and challenges of health-related Internet use\nVast quantity of information\nUnregulated\nAlways on\nAccessible from anywhere\nInteractive\nInformation can be captured, archived, and retrieved\nContent from both expert sources and user-generated sources\nContent can be free or paid for\nUsers can organize in virtual communities\nPublic education\nPublic empowerment supporting informed consumers engaged in their own care\nConnect people with others who have similar problems\nOnline social support\nReduce barriers (time, location, and cost) to accessing information and services\nAvoid the stigma of real-world consultation for certain problems\nDeliver interactive interventions, as well as information\nIntegrated health services such as shared electronic records\nReduced travel and carbon emissions\nMisinformation leading to harm\nMisuse of accurate information or services such as e-pharmacy\nExacerbation of inequalities in health caused by the digital divide\nChallenges to the authority of health professionals\nDisruptive behavior in virtual communities\nSocial isolation of users\nInternet addiction of users\nErgonomic effects of computer use and reduced physical activity", "A combination of questionnaire survey and semistructured interviews was used. A self-administered, open, cross-sectional survey of visitors to the NHS Direct website was undertaken using a link placed on the home page of the website. It was therefore a web-based opt-in survey of a convenience sample. Cookies prevented multiple submissions from one computer. Consent was given by participants entering an email address to be sent the link to the survey. There were no incentives to participation. The questionnaire had two aims: first, to identify the characteristics and motivations of users of the website; and second, to recruit potential participants for a qualitative interview study. The questionnaire included 15 questions (one question per screen) covering demographic and health status characteristics, reasons for using the website, and questions related to information-seeking behavior. There was no adaptive questioning or manipulation of item order. In the final part of the survey, the respondents were asked for consent to be contacted at a later date for an interview. The questionnaire was developed based on previous work. The instrument is included as an online appendix.\nInterview participants were selected by maximum variation sampling with respect to demographic and health status characteristics of gender, age, ethnicity, those seeking help for acute and chronic illnesses, and those seeking help for themselves or for others. To minimize recall bias, while allowing sufficient time for participants to act on the information they had found, interviews were conducted within 1 to 2 weeks of use of the site. These semistructured interviews were undertaken via telephone, email, or instant messaging. Interviews were conducted by two interviewers, who were members of the research team (JP and NI). Anonymized interview transcripts were used for analysis. We used open-ended questioning and determined the order of questioning by the direction taken by each interview participant. Each interview usually began with a brief description of the interviewee’s last visit to the NHS Direct website and went on to explore their online health-related information needs and their information-seeking behavior under the following headings: motivation for using the NHS Direct website, the NHS Direct website itself, facilitators and barriers to online health seeking, role of the Internet compared with other sources of information, and consequences of using online health information. Demographic characteristics of each participant were also recorded (age, gender, ethnicity, educational attainment, and current or most recent occupation). The interview topic guide is included as an online appendix. Survey and interview data were held in password-protected files on password-protected computers. NHS ethics committee approval was obtained for both elements of this study. ", "Questionnaire results were analyzed to provide summary descriptive statistics and cross-tabulations for which chi-square statistics were calculated to examine differences in proportions by demographic characteristics. No statistical corrections such as weighting were used, but nonresponders to individual questions were excluded from the analysis of those questions. All interview transcripts were read by three investigators (JP, NL, and JR), who familiarized themselves with the data through reading and reflection, and each independently undertook open coding of all transcripts [18]. Constant comparison was used to refine emerging conceptual categories, including a search for deviant cases. The investigators met to agree on a series of thematic codes that described a number of categories and subcategories. These agreed-on codes were reapplied to the transcripts, using NVivo software (version 8, QSR International Pty Ltd, Southport, UK).", "[SUBTITLE] Questionnaire Survey [SUBSECTION] A total of 792 respondents completed at least part of the survey accessed via the homepage link. Results are presented as a proportion of the total number of responses to each question (the denominator therefore varies according to the individual question response rate). As can be seen from Table 2, 71.2% (534/750) of respondents were aged under 45 years, 67.4% (511/758) were female, and 37.7% (286/759) had a university degree or higher qualification. With respect to personal general health, 61.7% (474/768) rated it as good or very good (compared with a general population figure of 76%) [19]; 42.6% (322/755) reported having a long-standing illness, disability, or infirmity (very similar to the general population figure of 42%) [19]. \nRegarding reasons for seeking help, 69.8% (545/781) reported looking for information for their own health issue, while 22.0% (172/781) reported looking for information for someone else (8.2% were looking for both, 64/781). These proportions differed by gender, with women more likely than men to report seeking help for someone else (χ2\n2 = 6.35, P = .04; 33% of women, 168/509, versus 25.2% of men, 61/242, reported seeking information for someone else or for both themselves and someone else). There was a significant difference by age group for women, which was predominantly due to a higher number of women in the 56- to 65-year-old age group reporting looking for help for someone else compared with women in other age bands (χ2\n7 = 22.89, P = .002). There was no difference by age group for men (χ2\n7 = 10.65, P = .15). \nOf all respondents, 47.5% (366/770) reported seeking help for a new health issue, while 19.6% (151/770) reported seeking help for a long-standing issue; 17.1% (132/770) reported seeking help for both new and long-standing issues, and 15.7% (121/770) for neither. The commonest category of user was a person who reported seeking help for a new health issue, regarding their own health (257/770, 33.4%). A total of 44.9% (346/770) reported having already consulted a health professional (such as a general practitioner or nurse) about the problem for which they were using the NHS Direct website, and 6.1% (47/770) had previously consulted the NHS Direct telephone service about the issue they were currently looking up online. There were no significant differences in this previous consultation behavior by gender (χ2\n1 = 0.625, P = .43). Users in younger age groups (<36 years) were less likely to report having had prior consultation with a health professional before using the website (χ2\n1 = 24.22, P < .001); 35.9% (143/398) of those aged up to 35 reported having consulted prior to using the website, compared with 53.9% (186/345) of those over 35. \nSurvey responses by gender (missing data reported because partially completed surveys were included in the analysis). Total respondents N = 792\nA total of 792 respondents completed at least part of the survey accessed via the homepage link. Results are presented as a proportion of the total number of responses to each question (the denominator therefore varies according to the individual question response rate). As can be seen from Table 2, 71.2% (534/750) of respondents were aged under 45 years, 67.4% (511/758) were female, and 37.7% (286/759) had a university degree or higher qualification. With respect to personal general health, 61.7% (474/768) rated it as good or very good (compared with a general population figure of 76%) [19]; 42.6% (322/755) reported having a long-standing illness, disability, or infirmity (very similar to the general population figure of 42%) [19]. \nRegarding reasons for seeking help, 69.8% (545/781) reported looking for information for their own health issue, while 22.0% (172/781) reported looking for information for someone else (8.2% were looking for both, 64/781). These proportions differed by gender, with women more likely than men to report seeking help for someone else (χ2\n2 = 6.35, P = .04; 33% of women, 168/509, versus 25.2% of men, 61/242, reported seeking information for someone else or for both themselves and someone else). There was a significant difference by age group for women, which was predominantly due to a higher number of women in the 56- to 65-year-old age group reporting looking for help for someone else compared with women in other age bands (χ2\n7 = 22.89, P = .002). There was no difference by age group for men (χ2\n7 = 10.65, P = .15). \nOf all respondents, 47.5% (366/770) reported seeking help for a new health issue, while 19.6% (151/770) reported seeking help for a long-standing issue; 17.1% (132/770) reported seeking help for both new and long-standing issues, and 15.7% (121/770) for neither. The commonest category of user was a person who reported seeking help for a new health issue, regarding their own health (257/770, 33.4%). A total of 44.9% (346/770) reported having already consulted a health professional (such as a general practitioner or nurse) about the problem for which they were using the NHS Direct website, and 6.1% (47/770) had previously consulted the NHS Direct telephone service about the issue they were currently looking up online. There were no significant differences in this previous consultation behavior by gender (χ2\n1 = 0.625, P = .43). Users in younger age groups (<36 years) were less likely to report having had prior consultation with a health professional before using the website (χ2\n1 = 24.22, P < .001); 35.9% (143/398) of those aged up to 35 reported having consulted prior to using the website, compared with 53.9% (186/345) of those over 35. \nSurvey responses by gender (missing data reported because partially completed surveys were included in the analysis). Total respondents N = 792\n[SUBTITLE] Interview Sample [SUBSECTION] Twenty-six (20 female, 6 male) participants aged 16 to 75 years were recruited from a total of 265 who had indicated their willingness to take part in an interview on the questionnaire. They were interviewed either by telephone (n = 23) or by instant messaging/email (n = 3). Twenty-one described their ethnicity as white British, and five belonged to other ethnic groups. At the time of recruitment the participants were looking up information for themselves (n = 15), for someone else (n = 10), or for both themselves and someone else (n = 1). The participants rated their health as very good (n = 2), good (n = 6), fair (n = 14), bad (n = 1), or very bad (n = 3). Participant characteristics are summarized in Table 3.\nThe range of health topics searched for was broad, with the most popular being musculoskeletal problems (6/26), mental health problems (3/26), and dermatological problems (3/26).\nCharacteristics of interview participants (n = 26)\nTwenty-six (20 female, 6 male) participants aged 16 to 75 years were recruited from a total of 265 who had indicated their willingness to take part in an interview on the questionnaire. They were interviewed either by telephone (n = 23) or by instant messaging/email (n = 3). Twenty-one described their ethnicity as white British, and five belonged to other ethnic groups. At the time of recruitment the participants were looking up information for themselves (n = 15), for someone else (n = 10), or for both themselves and someone else (n = 1). The participants rated their health as very good (n = 2), good (n = 6), fair (n = 14), bad (n = 1), or very bad (n = 3). Participant characteristics are summarized in Table 3.\nThe range of health topics searched for was broad, with the most popular being musculoskeletal problems (6/26), mental health problems (3/26), and dermatological problems (3/26).\nCharacteristics of interview participants (n = 26)\n[SUBTITLE] Thematic Analysis of Interview Data [SUBSECTION] Themes were identified under the following headings, framed by the questions used in the interview topic guide: the motivations for seeking help online; the benefits of seeking help in this way, and some of the challenges faced; the strategies employed in navigating online health information provision and determining what information to use and to trust; and finally, specific comments regarding the NHS Direct website service. These will be discussed in turn.\n[SUBTITLE] Motivations [SUBSECTION] Within the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\nWithin the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\n[SUBTITLE] Benefits and Challenges: Convenience, Coverage, and Confidentiality [SUBSECTION] Interviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\nInterviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\n[SUBTITLE] Strategy: Navigating the Health Internet [SUBSECTION] Throughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\nThroughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\n[SUBTITLE] The NHS Direct Service [SUBSECTION] As would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.\nAs would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.\nThemes were identified under the following headings, framed by the questions used in the interview topic guide: the motivations for seeking help online; the benefits of seeking help in this way, and some of the challenges faced; the strategies employed in navigating online health information provision and determining what information to use and to trust; and finally, specific comments regarding the NHS Direct website service. These will be discussed in turn.\n[SUBTITLE] Motivations [SUBSECTION] Within the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\nWithin the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\n[SUBTITLE] Benefits and Challenges: Convenience, Coverage, and Confidentiality [SUBSECTION] Interviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\nInterviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\n[SUBTITLE] Strategy: Navigating the Health Internet [SUBSECTION] Throughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\nThroughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\n[SUBTITLE] The NHS Direct Service [SUBSECTION] As would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.\nAs would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.", "A total of 792 respondents completed at least part of the survey accessed via the homepage link. Results are presented as a proportion of the total number of responses to each question (the denominator therefore varies according to the individual question response rate). As can be seen from Table 2, 71.2% (534/750) of respondents were aged under 45 years, 67.4% (511/758) were female, and 37.7% (286/759) had a university degree or higher qualification. With respect to personal general health, 61.7% (474/768) rated it as good or very good (compared with a general population figure of 76%) [19]; 42.6% (322/755) reported having a long-standing illness, disability, or infirmity (very similar to the general population figure of 42%) [19]. \nRegarding reasons for seeking help, 69.8% (545/781) reported looking for information for their own health issue, while 22.0% (172/781) reported looking for information for someone else (8.2% were looking for both, 64/781). These proportions differed by gender, with women more likely than men to report seeking help for someone else (χ2\n2 = 6.35, P = .04; 33% of women, 168/509, versus 25.2% of men, 61/242, reported seeking information for someone else or for both themselves and someone else). There was a significant difference by age group for women, which was predominantly due to a higher number of women in the 56- to 65-year-old age group reporting looking for help for someone else compared with women in other age bands (χ2\n7 = 22.89, P = .002). There was no difference by age group for men (χ2\n7 = 10.65, P = .15). \nOf all respondents, 47.5% (366/770) reported seeking help for a new health issue, while 19.6% (151/770) reported seeking help for a long-standing issue; 17.1% (132/770) reported seeking help for both new and long-standing issues, and 15.7% (121/770) for neither. The commonest category of user was a person who reported seeking help for a new health issue, regarding their own health (257/770, 33.4%). A total of 44.9% (346/770) reported having already consulted a health professional (such as a general practitioner or nurse) about the problem for which they were using the NHS Direct website, and 6.1% (47/770) had previously consulted the NHS Direct telephone service about the issue they were currently looking up online. There were no significant differences in this previous consultation behavior by gender (χ2\n1 = 0.625, P = .43). Users in younger age groups (<36 years) were less likely to report having had prior consultation with a health professional before using the website (χ2\n1 = 24.22, P < .001); 35.9% (143/398) of those aged up to 35 reported having consulted prior to using the website, compared with 53.9% (186/345) of those over 35. \nSurvey responses by gender (missing data reported because partially completed surveys were included in the analysis). Total respondents N = 792", "Twenty-six (20 female, 6 male) participants aged 16 to 75 years were recruited from a total of 265 who had indicated their willingness to take part in an interview on the questionnaire. They were interviewed either by telephone (n = 23) or by instant messaging/email (n = 3). Twenty-one described their ethnicity as white British, and five belonged to other ethnic groups. At the time of recruitment the participants were looking up information for themselves (n = 15), for someone else (n = 10), or for both themselves and someone else (n = 1). The participants rated their health as very good (n = 2), good (n = 6), fair (n = 14), bad (n = 1), or very bad (n = 3). Participant characteristics are summarized in Table 3.\nThe range of health topics searched for was broad, with the most popular being musculoskeletal problems (6/26), mental health problems (3/26), and dermatological problems (3/26).\nCharacteristics of interview participants (n = 26)", "Themes were identified under the following headings, framed by the questions used in the interview topic guide: the motivations for seeking help online; the benefits of seeking help in this way, and some of the challenges faced; the strategies employed in navigating online health information provision and determining what information to use and to trust; and finally, specific comments regarding the NHS Direct website service. These will be discussed in turn.\n[SUBTITLE] Motivations [SUBSECTION] Within the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\nWithin the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\n[SUBTITLE] Benefits and Challenges: Convenience, Coverage, and Confidentiality [SUBSECTION] Interviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\nInterviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\n[SUBTITLE] Strategy: Navigating the Health Internet [SUBSECTION] Throughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\nThroughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\n[SUBTITLE] The NHS Direct Service [SUBSECTION] As would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.\nAs would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.", "Within the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.", "Interviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.", "Throughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.", "As would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.", "The survey findings with respect to the age, gender, and educational status of online health seekers add to the accumulated evidence of several studies over the last decade that have shown that being female, being younger, and having a higher level of educational attainment are all associated with more frequent health-related Internet use [11,19-23]. The reported health status profile of our survey participants appeared to be very close to that of the general population. Work in other countries has sometimes shown a tendency to overrepresent people with chronic illness among health Internet users [7,8,24,25], while others have not found this association [11,22]. The NHS Direct website is used for both acute and chronic problems, mild or serious, and is used by individuals for themselves and on behalf of family members, especially by women. It is therefore perhaps not surprising that the health status of the users in our study was similar to that of the general population. The majority of those surveyed were seeking information for themselves, which is consistent with the findings of others [11]. Furthermore, a large proportion of users (over 40%) had already sought help from a health professional for the same health issue prior to accessing the website.\nHaving established the characteristics of the users of the site, we undertook in-depth qualitative work to explore in detail their motivations and attitudes. In 2003, the lead author (JP) wrote a review paper that summarized the benefits and challenges of health-related Internet use [12]. In Table 1 we integrated these with the thoughts of other authors in this field [16,17]. The analysis of our interviews supports most of the theoretical benefits discussed in the literature, and indicates that the health Internet is delivering on its potential benefits, while at the same time presenting some challenges to health professionals. Participants’ responses indicated that the Internet was being used as a tool to educate and reassure, and to sometimes challenge information received by health professionals. Previous work on the sociology of health-related Internet use has invoked theories of empowerment, democratization, and the challenge to health professional power [13,15,26] Most empirical work has indicated that, while these processes are taking place, the change is more subtle than many theorists have predicted, with the ongoing predominance of a biomedical model in the context of more-informed health consumers [27-29]. By sampling users of the NHS website it is perhaps not surprising that our findings support this model of evolutionary rather than revolutionary change, with health e-consumers seeking to become more informed through authoritative advice from official websites. Health-related Internet use was seen by most of our participants as a supplement to existing health service provision rather than a replacement for it [4,30]. The motivations of reassurance and of seeking greater understanding can be seen in this context. Even the motivation to find a second opinion to challenge other information was within the context of a model of biomedical authority. Our findings support the idea that online health resources are enmeshed with other (offline) approaches to seeking help [26], and that health-related Internet use is now embedded in everyday health practices [31]. \nThe majority of online events were related to real-world consultations, whether as preparation for them or as a search for further information afterward [23,32,33]. There were few examples of demand management occurring in practice, in terms of reducing the need for consultations, but our findings do support the idea that a health website can lead to more appropriate use of other services. Peer-to-peer interaction was not a focus of this study, as this is not provided on the NHS Direct website, and the number of participants in our qualitative sample reporting use of online support groups for health conditions was not high. Nevertheless, some participants did discuss the value of online interaction with others with similar problems, in particular the reassurance of knowing they were not alone, as found in previous work [34], while others expressed concerns about the trustworthiness of peer-to-peer sources. Consumer access to poor-quality information on the Internet has been a long-standing concern in the eHealth literature [15]. We found that, in avoiding misinformation and identifying which information to trust, participants put great emphasis on recognition of brands such as the NHS, which were trusted in the non-Internet world, together with using common sense approaches to navigate the health Internet. The reported value of official branding of health websites in determining trustworthiness is supported by previous work [35]. The online benefits of convenience and anonymity are well established [36] and were widely reported, as was the expectation that online health services would be fully integrated with their real-world counterparts, something that remains an aspiration for the NHS in the United Kingdom but is not yet a reality.\nSome theoretical benefits and challenges were not prominent in these interviews. The “green” potential of the Internet to reduce travel, but at the same time possibly reduce physical activity and lead to social isolation or depression [37], was not discussed by our participants. Nor were any ergonomic effects of computer use, or the problem of Internet addiction. The issue of the digital divide, although mentioned by three participants, did not emerge as a consistent theme. However, given that this sample were all Internet users, this was perhaps not surprising.\n[SUBTITLE] Limitations [SUBSECTION] Because this was an opt-in survey accessed via a weblink, it was not possible to calculate a response rate for the questionnaire. There were also design issues with the website during the survey period, which meant that the link to the survey was not always clearly visible to users. This affected the overall response and the number of individuals consenting to be interviewed in the second stage. To minimize social desirability bias, the researchers made it clear to interviewees that the researchers were independent of the NHS Direct organization, but those volunteering for interviews may still have been a particular population who wanted to relate their experiences with NHS Direct, good or bad. More women than men were interviewed due to having very few male volunteers. Interview methodology of this type, asking people to report how and why they used a particular source, may reflect attitudes rather than actual behavior, for which direct observation may be preferred. Nevertheless, questions were designed to focus on the most recent actual use of the Internet for health, rather than rely on hypothetical questioning. The participants were users of the NHS Direct website and were therefore not necessarily representative of the overall population of online health information seekers in the United Kingdom. However, the NHS site is the most popular health information site in the United Kingdom, and the demographic profile of respondents was similar to that of non-UK-based studies.\nBecause this was an opt-in survey accessed via a weblink, it was not possible to calculate a response rate for the questionnaire. There were also design issues with the website during the survey period, which meant that the link to the survey was not always clearly visible to users. This affected the overall response and the number of individuals consenting to be interviewed in the second stage. To minimize social desirability bias, the researchers made it clear to interviewees that the researchers were independent of the NHS Direct organization, but those volunteering for interviews may still have been a particular population who wanted to relate their experiences with NHS Direct, good or bad. More women than men were interviewed due to having very few male volunteers. Interview methodology of this type, asking people to report how and why they used a particular source, may reflect attitudes rather than actual behavior, for which direct observation may be preferred. Nevertheless, questions were designed to focus on the most recent actual use of the Internet for health, rather than rely on hypothetical questioning. The participants were users of the NHS Direct website and were therefore not necessarily representative of the overall population of online health information seekers in the United Kingdom. However, the NHS site is the most popular health information site in the United Kingdom, and the demographic profile of respondents was similar to that of non-UK-based studies.\n[SUBTITLE] Conclusions [SUBSECTION] Given increasing resource constraints, the health care community needs to seek ways of promoting efficient and appropriate health care use, which should include consideration of how Internet health information is provided and used, and how traditional NHS services and online services can be best integrated. The study findings support a model of evolutionary rather than revolutionary change in health information use, with real-world trusted brands being used online, in conjunction with traditional consultations. It will be interesting to see whether in time, particularly as the younger “Internet generation” ages and eHealth literacy increases in all age groups [38], Internet health information will be trusted enough to be used as an alternative, as opposed to an adjunct, to other types of health-seeking activities, and by individuals of broader demographic profiles. Our findings fit with a “shared decision-making” model [39], where individuals seek information to help the decision-making process and confirm what they are being told, rather than seeking to become independent experts. One of the primary motivations was the seeking of reassurance, and the value of this in terms of health or social benefit or more appropriate service use needs to be further explored. The relationship between Internet use and health outcomes is an area for research development, including examination of the role of user empowerment. Health service providers should aim to harness the potential benefits of health-related Internet use, rather than see it as a burden or challenge.\nGiven increasing resource constraints, the health care community needs to seek ways of promoting efficient and appropriate health care use, which should include consideration of how Internet health information is provided and used, and how traditional NHS services and online services can be best integrated. The study findings support a model of evolutionary rather than revolutionary change in health information use, with real-world trusted brands being used online, in conjunction with traditional consultations. It will be interesting to see whether in time, particularly as the younger “Internet generation” ages and eHealth literacy increases in all age groups [38], Internet health information will be trusted enough to be used as an alternative, as opposed to an adjunct, to other types of health-seeking activities, and by individuals of broader demographic profiles. Our findings fit with a “shared decision-making” model [39], where individuals seek information to help the decision-making process and confirm what they are being told, rather than seeking to become independent experts. One of the primary motivations was the seeking of reassurance, and the value of this in terms of health or social benefit or more appropriate service use needs to be further explored. The relationship between Internet use and health outcomes is an area for research development, including examination of the role of user empowerment. Health service providers should aim to harness the potential benefits of health-related Internet use, rather than see it as a burden or challenge.", "Because this was an opt-in survey accessed via a weblink, it was not possible to calculate a response rate for the questionnaire. There were also design issues with the website during the survey period, which meant that the link to the survey was not always clearly visible to users. This affected the overall response and the number of individuals consenting to be interviewed in the second stage. To minimize social desirability bias, the researchers made it clear to interviewees that the researchers were independent of the NHS Direct organization, but those volunteering for interviews may still have been a particular population who wanted to relate their experiences with NHS Direct, good or bad. More women than men were interviewed due to having very few male volunteers. Interview methodology of this type, asking people to report how and why they used a particular source, may reflect attitudes rather than actual behavior, for which direct observation may be preferred. Nevertheless, questions were designed to focus on the most recent actual use of the Internet for health, rather than rely on hypothetical questioning. The participants were users of the NHS Direct website and were therefore not necessarily representative of the overall population of online health information seekers in the United Kingdom. However, the NHS site is the most popular health information site in the United Kingdom, and the demographic profile of respondents was similar to that of non-UK-based studies.", "Given increasing resource constraints, the health care community needs to seek ways of promoting efficient and appropriate health care use, which should include consideration of how Internet health information is provided and used, and how traditional NHS services and online services can be best integrated. The study findings support a model of evolutionary rather than revolutionary change in health information use, with real-world trusted brands being used online, in conjunction with traditional consultations. It will be interesting to see whether in time, particularly as the younger “Internet generation” ages and eHealth literacy increases in all age groups [38], Internet health information will be trusted enough to be used as an alternative, as opposed to an adjunct, to other types of health-seeking activities, and by individuals of broader demographic profiles. Our findings fit with a “shared decision-making” model [39], where individuals seek information to help the decision-making process and confirm what they are being told, rather than seeking to become independent experts. One of the primary motivations was the seeking of reassurance, and the value of this in terms of health or social benefit or more appropriate service use needs to be further explored. The relationship between Internet use and health outcomes is an area for research development, including examination of the role of user empowerment. Health service providers should aim to harness the potential benefits of health-related Internet use, rather than see it as a burden or challenge." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Methods", "Design", "Analysis", "Results", "Questionnaire Survey", "Interview Sample", "Thematic Analysis of Interview Data", "Motivations", "Benefits and Challenges: Convenience, Coverage, and Confidentiality", "Strategy: Navigating the Health Internet", "The NHS Direct Service", "Discussion", "Limitations", "Conclusions" ]

[ "In the United Kingdom, in 2009, 70% of households had Internet access [1]. With this rise has come an increase in health-related Internet use. The proportion of UK Internet users using online information on health matters increased from 37% in 2005 to 68% in 2009 [1,2]. A survey in seven European countries in 2005 found that 71% of all adults reported using the Internet for health information [3], and in the United States this figure is 61% [4]. Furthermore, public perceptions of the importance of the Internet as a source of health information have risen dramatically [5].\nThe National Health Service (NHS) Direct website is the main health advice and information website for patients and the public in the United Kingdom. It was launched in December 1999, and in 2009, there were 18 million visits to the website, compared with 1.5 million visits in 2001 [6]. At the time of this research it provided health advice (through symptom checkers, a health encyclopedia, and an online enquiry service), information on local health services, articles regarding healthy living and fitness, and many other features, including a pregnancy planner, support for long-term conditions, and access to information about health care abroad. Since mid-2009 it has worked with the NHS Choices service, and some of the features described above have migrated to other parts of the NHS Choices platform. \nThere is limited information describing how and why people use online health information, or the effect of this on health status, although this literature base is growing [6-11]. In this study we used a mixed-methods approach to investigate the characteristics and motivations of online health information seekers in England. In theory, the Internet offers certain advantages as a health information resource. In particular, it provides convenient and anonymous access at any time, from any location, to a wide range of expert sources; and through virtual communities it can provide peer support and social interaction [12]. Health-related use of the Internet has been hailed as a tool to support the emergence of the informed and empowered health consumer, and a shift in the balance of power between patient and professional [13]. At the same time, concerns have been raised about the quality of information, the potential for unhelpful peer-to-peer interactions, and the exclusion of individuals who experience barriers to access [14,15]. In Table 1 we have summarized the main characteristics and potential public health benefits and challenges that have been proposed for health-related Internet use [12,16,17]. In this paper, by exploring the expressed reasons for seeking online health information, we hope to assess the extent to which the theoretical benefits and challenges of the health Internet are being realized in practice.\nTheoretical characteristics and potential public health benefits and challenges of health-related Internet use\nVast quantity of information\nUnregulated\nAlways on\nAccessible from anywhere\nInteractive\nInformation can be captured, archived, and retrieved\nContent from both expert sources and user-generated sources\nContent can be free or paid for\nUsers can organize in virtual communities\nPublic education\nPublic empowerment supporting informed consumers engaged in their own care\nConnect people with others who have similar problems\nOnline social support\nReduce barriers (time, location, and cost) to accessing information and services\nAvoid the stigma of real-world consultation for certain problems\nDeliver interactive interventions, as well as information\nIntegrated health services such as shared electronic records\nReduced travel and carbon emissions\nMisinformation leading to harm\nMisuse of accurate information or services such as e-pharmacy\nExacerbation of inequalities in health caused by the digital divide\nChallenges to the authority of health professionals\nDisruptive behavior in virtual communities\nSocial isolation of users\nInternet addiction of users\nErgonomic effects of computer use and reduced physical activity", "[SUBTITLE] Design [SUBSECTION] A combination of questionnaire survey and semistructured interviews was used. A self-administered, open, cross-sectional survey of visitors to the NHS Direct website was undertaken using a link placed on the home page of the website. It was therefore a web-based opt-in survey of a convenience sample. Cookies prevented multiple submissions from one computer. Consent was given by participants entering an email address to be sent the link to the survey. There were no incentives to participation. The questionnaire had two aims: first, to identify the characteristics and motivations of users of the website; and second, to recruit potential participants for a qualitative interview study. The questionnaire included 15 questions (one question per screen) covering demographic and health status characteristics, reasons for using the website, and questions related to information-seeking behavior. There was no adaptive questioning or manipulation of item order. In the final part of the survey, the respondents were asked for consent to be contacted at a later date for an interview. The questionnaire was developed based on previous work. The instrument is included as an online appendix.\nInterview participants were selected by maximum variation sampling with respect to demographic and health status characteristics of gender, age, ethnicity, those seeking help for acute and chronic illnesses, and those seeking help for themselves or for others. To minimize recall bias, while allowing sufficient time for participants to act on the information they had found, interviews were conducted within 1 to 2 weeks of use of the site. These semistructured interviews were undertaken via telephone, email, or instant messaging. Interviews were conducted by two interviewers, who were members of the research team (JP and NI). Anonymized interview transcripts were used for analysis. We used open-ended questioning and determined the order of questioning by the direction taken by each interview participant. Each interview usually began with a brief description of the interviewee’s last visit to the NHS Direct website and went on to explore their online health-related information needs and their information-seeking behavior under the following headings: motivation for using the NHS Direct website, the NHS Direct website itself, facilitators and barriers to online health seeking, role of the Internet compared with other sources of information, and consequences of using online health information. Demographic characteristics of each participant were also recorded (age, gender, ethnicity, educational attainment, and current or most recent occupation). The interview topic guide is included as an online appendix. Survey and interview data were held in password-protected files on password-protected computers. NHS ethics committee approval was obtained for both elements of this study. \nA combination of questionnaire survey and semistructured interviews was used. A self-administered, open, cross-sectional survey of visitors to the NHS Direct website was undertaken using a link placed on the home page of the website. It was therefore a web-based opt-in survey of a convenience sample. Cookies prevented multiple submissions from one computer. Consent was given by participants entering an email address to be sent the link to the survey. There were no incentives to participation. The questionnaire had two aims: first, to identify the characteristics and motivations of users of the website; and second, to recruit potential participants for a qualitative interview study. The questionnaire included 15 questions (one question per screen) covering demographic and health status characteristics, reasons for using the website, and questions related to information-seeking behavior. There was no adaptive questioning or manipulation of item order. In the final part of the survey, the respondents were asked for consent to be contacted at a later date for an interview. The questionnaire was developed based on previous work. The instrument is included as an online appendix.\nInterview participants were selected by maximum variation sampling with respect to demographic and health status characteristics of gender, age, ethnicity, those seeking help for acute and chronic illnesses, and those seeking help for themselves or for others. To minimize recall bias, while allowing sufficient time for participants to act on the information they had found, interviews were conducted within 1 to 2 weeks of use of the site. These semistructured interviews were undertaken via telephone, email, or instant messaging. Interviews were conducted by two interviewers, who were members of the research team (JP and NI). Anonymized interview transcripts were used for analysis. We used open-ended questioning and determined the order of questioning by the direction taken by each interview participant. Each interview usually began with a brief description of the interviewee’s last visit to the NHS Direct website and went on to explore their online health-related information needs and their information-seeking behavior under the following headings: motivation for using the NHS Direct website, the NHS Direct website itself, facilitators and barriers to online health seeking, role of the Internet compared with other sources of information, and consequences of using online health information. Demographic characteristics of each participant were also recorded (age, gender, ethnicity, educational attainment, and current or most recent occupation). The interview topic guide is included as an online appendix. Survey and interview data were held in password-protected files on password-protected computers. NHS ethics committee approval was obtained for both elements of this study. \n[SUBTITLE] Analysis [SUBSECTION] Questionnaire results were analyzed to provide summary descriptive statistics and cross-tabulations for which chi-square statistics were calculated to examine differences in proportions by demographic characteristics. No statistical corrections such as weighting were used, but nonresponders to individual questions were excluded from the analysis of those questions. All interview transcripts were read by three investigators (JP, NL, and JR), who familiarized themselves with the data through reading and reflection, and each independently undertook open coding of all transcripts [18]. Constant comparison was used to refine emerging conceptual categories, including a search for deviant cases. The investigators met to agree on a series of thematic codes that described a number of categories and subcategories. These agreed-on codes were reapplied to the transcripts, using NVivo software (version 8, QSR International Pty Ltd, Southport, UK).\nQuestionnaire results were analyzed to provide summary descriptive statistics and cross-tabulations for which chi-square statistics were calculated to examine differences in proportions by demographic characteristics. No statistical corrections such as weighting were used, but nonresponders to individual questions were excluded from the analysis of those questions. All interview transcripts were read by three investigators (JP, NL, and JR), who familiarized themselves with the data through reading and reflection, and each independently undertook open coding of all transcripts [18]. Constant comparison was used to refine emerging conceptual categories, including a search for deviant cases. The investigators met to agree on a series of thematic codes that described a number of categories and subcategories. These agreed-on codes were reapplied to the transcripts, using NVivo software (version 8, QSR International Pty Ltd, Southport, UK).", "A combination of questionnaire survey and semistructured interviews was used. A self-administered, open, cross-sectional survey of visitors to the NHS Direct website was undertaken using a link placed on the home page of the website. It was therefore a web-based opt-in survey of a convenience sample. Cookies prevented multiple submissions from one computer. Consent was given by participants entering an email address to be sent the link to the survey. There were no incentives to participation. The questionnaire had two aims: first, to identify the characteristics and motivations of users of the website; and second, to recruit potential participants for a qualitative interview study. The questionnaire included 15 questions (one question per screen) covering demographic and health status characteristics, reasons for using the website, and questions related to information-seeking behavior. There was no adaptive questioning or manipulation of item order. In the final part of the survey, the respondents were asked for consent to be contacted at a later date for an interview. The questionnaire was developed based on previous work. The instrument is included as an online appendix.\nInterview participants were selected by maximum variation sampling with respect to demographic and health status characteristics of gender, age, ethnicity, those seeking help for acute and chronic illnesses, and those seeking help for themselves or for others. To minimize recall bias, while allowing sufficient time for participants to act on the information they had found, interviews were conducted within 1 to 2 weeks of use of the site. These semistructured interviews were undertaken via telephone, email, or instant messaging. Interviews were conducted by two interviewers, who were members of the research team (JP and NI). Anonymized interview transcripts were used for analysis. We used open-ended questioning and determined the order of questioning by the direction taken by each interview participant. Each interview usually began with a brief description of the interviewee’s last visit to the NHS Direct website and went on to explore their online health-related information needs and their information-seeking behavior under the following headings: motivation for using the NHS Direct website, the NHS Direct website itself, facilitators and barriers to online health seeking, role of the Internet compared with other sources of information, and consequences of using online health information. Demographic characteristics of each participant were also recorded (age, gender, ethnicity, educational attainment, and current or most recent occupation). The interview topic guide is included as an online appendix. Survey and interview data were held in password-protected files on password-protected computers. NHS ethics committee approval was obtained for both elements of this study. ", "Questionnaire results were analyzed to provide summary descriptive statistics and cross-tabulations for which chi-square statistics were calculated to examine differences in proportions by demographic characteristics. No statistical corrections such as weighting were used, but nonresponders to individual questions were excluded from the analysis of those questions. All interview transcripts were read by three investigators (JP, NL, and JR), who familiarized themselves with the data through reading and reflection, and each independently undertook open coding of all transcripts [18]. Constant comparison was used to refine emerging conceptual categories, including a search for deviant cases. The investigators met to agree on a series of thematic codes that described a number of categories and subcategories. These agreed-on codes were reapplied to the transcripts, using NVivo software (version 8, QSR International Pty Ltd, Southport, UK).", "[SUBTITLE] Questionnaire Survey [SUBSECTION] A total of 792 respondents completed at least part of the survey accessed via the homepage link. Results are presented as a proportion of the total number of responses to each question (the denominator therefore varies according to the individual question response rate). As can be seen from Table 2, 71.2% (534/750) of respondents were aged under 45 years, 67.4% (511/758) were female, and 37.7% (286/759) had a university degree or higher qualification. With respect to personal general health, 61.7% (474/768) rated it as good or very good (compared with a general population figure of 76%) [19]; 42.6% (322/755) reported having a long-standing illness, disability, or infirmity (very similar to the general population figure of 42%) [19]. \nRegarding reasons for seeking help, 69.8% (545/781) reported looking for information for their own health issue, while 22.0% (172/781) reported looking for information for someone else (8.2% were looking for both, 64/781). These proportions differed by gender, with women more likely than men to report seeking help for someone else (χ2\n2 = 6.35, P = .04; 33% of women, 168/509, versus 25.2% of men, 61/242, reported seeking information for someone else or for both themselves and someone else). There was a significant difference by age group for women, which was predominantly due to a higher number of women in the 56- to 65-year-old age group reporting looking for help for someone else compared with women in other age bands (χ2\n7 = 22.89, P = .002). There was no difference by age group for men (χ2\n7 = 10.65, P = .15). \nOf all respondents, 47.5% (366/770) reported seeking help for a new health issue, while 19.6% (151/770) reported seeking help for a long-standing issue; 17.1% (132/770) reported seeking help for both new and long-standing issues, and 15.7% (121/770) for neither. The commonest category of user was a person who reported seeking help for a new health issue, regarding their own health (257/770, 33.4%). A total of 44.9% (346/770) reported having already consulted a health professional (such as a general practitioner or nurse) about the problem for which they were using the NHS Direct website, and 6.1% (47/770) had previously consulted the NHS Direct telephone service about the issue they were currently looking up online. There were no significant differences in this previous consultation behavior by gender (χ2\n1 = 0.625, P = .43). Users in younger age groups (<36 years) were less likely to report having had prior consultation with a health professional before using the website (χ2\n1 = 24.22, P < .001); 35.9% (143/398) of those aged up to 35 reported having consulted prior to using the website, compared with 53.9% (186/345) of those over 35. \nSurvey responses by gender (missing data reported because partially completed surveys were included in the analysis). Total respondents N = 792\nA total of 792 respondents completed at least part of the survey accessed via the homepage link. Results are presented as a proportion of the total number of responses to each question (the denominator therefore varies according to the individual question response rate). As can be seen from Table 2, 71.2% (534/750) of respondents were aged under 45 years, 67.4% (511/758) were female, and 37.7% (286/759) had a university degree or higher qualification. With respect to personal general health, 61.7% (474/768) rated it as good or very good (compared with a general population figure of 76%) [19]; 42.6% (322/755) reported having a long-standing illness, disability, or infirmity (very similar to the general population figure of 42%) [19]. \nRegarding reasons for seeking help, 69.8% (545/781) reported looking for information for their own health issue, while 22.0% (172/781) reported looking for information for someone else (8.2% were looking for both, 64/781). These proportions differed by gender, with women more likely than men to report seeking help for someone else (χ2\n2 = 6.35, P = .04; 33% of women, 168/509, versus 25.2% of men, 61/242, reported seeking information for someone else or for both themselves and someone else). There was a significant difference by age group for women, which was predominantly due to a higher number of women in the 56- to 65-year-old age group reporting looking for help for someone else compared with women in other age bands (χ2\n7 = 22.89, P = .002). There was no difference by age group for men (χ2\n7 = 10.65, P = .15). \nOf all respondents, 47.5% (366/770) reported seeking help for a new health issue, while 19.6% (151/770) reported seeking help for a long-standing issue; 17.1% (132/770) reported seeking help for both new and long-standing issues, and 15.7% (121/770) for neither. The commonest category of user was a person who reported seeking help for a new health issue, regarding their own health (257/770, 33.4%). A total of 44.9% (346/770) reported having already consulted a health professional (such as a general practitioner or nurse) about the problem for which they were using the NHS Direct website, and 6.1% (47/770) had previously consulted the NHS Direct telephone service about the issue they were currently looking up online. There were no significant differences in this previous consultation behavior by gender (χ2\n1 = 0.625, P = .43). Users in younger age groups (<36 years) were less likely to report having had prior consultation with a health professional before using the website (χ2\n1 = 24.22, P < .001); 35.9% (143/398) of those aged up to 35 reported having consulted prior to using the website, compared with 53.9% (186/345) of those over 35. \nSurvey responses by gender (missing data reported because partially completed surveys were included in the analysis). Total respondents N = 792\n[SUBTITLE] Interview Sample [SUBSECTION] Twenty-six (20 female, 6 male) participants aged 16 to 75 years were recruited from a total of 265 who had indicated their willingness to take part in an interview on the questionnaire. They were interviewed either by telephone (n = 23) or by instant messaging/email (n = 3). Twenty-one described their ethnicity as white British, and five belonged to other ethnic groups. At the time of recruitment the participants were looking up information for themselves (n = 15), for someone else (n = 10), or for both themselves and someone else (n = 1). The participants rated their health as very good (n = 2), good (n = 6), fair (n = 14), bad (n = 1), or very bad (n = 3). Participant characteristics are summarized in Table 3.\nThe range of health topics searched for was broad, with the most popular being musculoskeletal problems (6/26), mental health problems (3/26), and dermatological problems (3/26).\nCharacteristics of interview participants (n = 26)\nTwenty-six (20 female, 6 male) participants aged 16 to 75 years were recruited from a total of 265 who had indicated their willingness to take part in an interview on the questionnaire. They were interviewed either by telephone (n = 23) or by instant messaging/email (n = 3). Twenty-one described their ethnicity as white British, and five belonged to other ethnic groups. At the time of recruitment the participants were looking up information for themselves (n = 15), for someone else (n = 10), or for both themselves and someone else (n = 1). The participants rated their health as very good (n = 2), good (n = 6), fair (n = 14), bad (n = 1), or very bad (n = 3). Participant characteristics are summarized in Table 3.\nThe range of health topics searched for was broad, with the most popular being musculoskeletal problems (6/26), mental health problems (3/26), and dermatological problems (3/26).\nCharacteristics of interview participants (n = 26)\n[SUBTITLE] Thematic Analysis of Interview Data [SUBSECTION] Themes were identified under the following headings, framed by the questions used in the interview topic guide: the motivations for seeking help online; the benefits of seeking help in this way, and some of the challenges faced; the strategies employed in navigating online health information provision and determining what information to use and to trust; and finally, specific comments regarding the NHS Direct website service. These will be discussed in turn.\n[SUBTITLE] Motivations [SUBSECTION] Within the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\nWithin the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\n[SUBTITLE] Benefits and Challenges: Convenience, Coverage, and Confidentiality [SUBSECTION] Interviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\nInterviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\n[SUBTITLE] Strategy: Navigating the Health Internet [SUBSECTION] Throughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\nThroughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\n[SUBTITLE] The NHS Direct Service [SUBSECTION] As would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.\nAs would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.\nThemes were identified under the following headings, framed by the questions used in the interview topic guide: the motivations for seeking help online; the benefits of seeking help in this way, and some of the challenges faced; the strategies employed in navigating online health information provision and determining what information to use and to trust; and finally, specific comments regarding the NHS Direct website service. These will be discussed in turn.\n[SUBTITLE] Motivations [SUBSECTION] Within the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\nWithin the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\n[SUBTITLE] Benefits and Challenges: Convenience, Coverage, and Confidentiality [SUBSECTION] Interviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\nInterviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\n[SUBTITLE] Strategy: Navigating the Health Internet [SUBSECTION] Throughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\nThroughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\n[SUBTITLE] The NHS Direct Service [SUBSECTION] As would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.\nAs would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.", "A total of 792 respondents completed at least part of the survey accessed via the homepage link. Results are presented as a proportion of the total number of responses to each question (the denominator therefore varies according to the individual question response rate). As can be seen from Table 2, 71.2% (534/750) of respondents were aged under 45 years, 67.4% (511/758) were female, and 37.7% (286/759) had a university degree or higher qualification. With respect to personal general health, 61.7% (474/768) rated it as good or very good (compared with a general population figure of 76%) [19]; 42.6% (322/755) reported having a long-standing illness, disability, or infirmity (very similar to the general population figure of 42%) [19]. \nRegarding reasons for seeking help, 69.8% (545/781) reported looking for information for their own health issue, while 22.0% (172/781) reported looking for information for someone else (8.2% were looking for both, 64/781). These proportions differed by gender, with women more likely than men to report seeking help for someone else (χ2\n2 = 6.35, P = .04; 33% of women, 168/509, versus 25.2% of men, 61/242, reported seeking information for someone else or for both themselves and someone else). There was a significant difference by age group for women, which was predominantly due to a higher number of women in the 56- to 65-year-old age group reporting looking for help for someone else compared with women in other age bands (χ2\n7 = 22.89, P = .002). There was no difference by age group for men (χ2\n7 = 10.65, P = .15). \nOf all respondents, 47.5% (366/770) reported seeking help for a new health issue, while 19.6% (151/770) reported seeking help for a long-standing issue; 17.1% (132/770) reported seeking help for both new and long-standing issues, and 15.7% (121/770) for neither. The commonest category of user was a person who reported seeking help for a new health issue, regarding their own health (257/770, 33.4%). A total of 44.9% (346/770) reported having already consulted a health professional (such as a general practitioner or nurse) about the problem for which they were using the NHS Direct website, and 6.1% (47/770) had previously consulted the NHS Direct telephone service about the issue they were currently looking up online. There were no significant differences in this previous consultation behavior by gender (χ2\n1 = 0.625, P = .43). Users in younger age groups (<36 years) were less likely to report having had prior consultation with a health professional before using the website (χ2\n1 = 24.22, P < .001); 35.9% (143/398) of those aged up to 35 reported having consulted prior to using the website, compared with 53.9% (186/345) of those over 35. \nSurvey responses by gender (missing data reported because partially completed surveys were included in the analysis). Total respondents N = 792", "Twenty-six (20 female, 6 male) participants aged 16 to 75 years were recruited from a total of 265 who had indicated their willingness to take part in an interview on the questionnaire. They were interviewed either by telephone (n = 23) or by instant messaging/email (n = 3). Twenty-one described their ethnicity as white British, and five belonged to other ethnic groups. At the time of recruitment the participants were looking up information for themselves (n = 15), for someone else (n = 10), or for both themselves and someone else (n = 1). The participants rated their health as very good (n = 2), good (n = 6), fair (n = 14), bad (n = 1), or very bad (n = 3). Participant characteristics are summarized in Table 3.\nThe range of health topics searched for was broad, with the most popular being musculoskeletal problems (6/26), mental health problems (3/26), and dermatological problems (3/26).\nCharacteristics of interview participants (n = 26)", "Themes were identified under the following headings, framed by the questions used in the interview topic guide: the motivations for seeking help online; the benefits of seeking help in this way, and some of the challenges faced; the strategies employed in navigating online health information provision and determining what information to use and to trust; and finally, specific comments regarding the NHS Direct website service. These will be discussed in turn.\n[SUBTITLE] Motivations [SUBSECTION] Within the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\nWithin the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.\n[SUBTITLE] Benefits and Challenges: Convenience, Coverage, and Confidentiality [SUBSECTION] Interviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\nInterviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.\n[SUBTITLE] Strategy: Navigating the Health Internet [SUBSECTION] Throughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\nThroughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.\n[SUBTITLE] The NHS Direct Service [SUBSECTION] As would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.\nAs would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.", "Within the category of motivations, four concepts emerged through the thematic analysis: the desire for reassurance; the desire for a second opinion to challenge other information; the desire for greater understanding to supplement other information; and perceived external barriers to accessing information through traditional sources (including the desire to avoid “bothering” their health care provider).\nOne prominent reported motivation for seeking online health information was reassurance, often at the time symptoms appeared and prior to consultation with a health professional. As one participant stated “sometimes you just want your fears eased” [Interviewee 25, a 41-year-old woman]. In general, this online search for reassurance and relief from anxiety was not said to replace other forms of seeking help; rather, it was seen as an adjunct to other sources of help and information, not a substitute for them, thereby providing an extra layer of information but not necessarily altering consulting behaviors. As Interviewee 19 put it:\nI think I probably followed a course of action I would have taken anyway.\nFor the most part, the interviewees in this study reported seeking health information from official health websites that gave authoritative health information, and not from other users. This was not surprising given the route of recruitment through their use of the NHS Direct website. Even so, among the participants in this sample a few participants did report seeking nonprofessional “peer-to-peer” information. In these cases the motivation was again reassurance – wanting to know that the person was not alone in what they were experiencing. This was illustrated by Interviewee 17:\nI’ve gone to a menopause site that specializes in that [peer-to-peer interaction]. There are lots and lots of contributors over several years and I search for that, have a read and see what other people think and then I’ve posted on that and said“Look, I don’t have what people classically call flushes”... I just have it at night. What do other people think? And then lots of people come on and say“Yes, that’s perfectly normal. That’s what I’ve experienced and this is not unusual. People do have...”...so quite often I get reassurance that I’m not an odd one out from this.\nA few examples of the Internet providing “demand management” for primary care or emergency services were identified. Sometimes participants described the motivation of not wanting to “bother” their doctor with a problem that might be trivial. Interviewee 13 described how reassurance over a bloodshot eye eliminated the need for a general practice appointment.\nIt looks as if you should do something about it. I looked it up again on the website and it said “there’s nothing to worry about,” you know...it normally goes off on its own and it doesn’t need a trip to the doctor’s so again, it saved the trip to the doctor and it saved a lot of worrying.\nHowever, while 13 interviewees talked about accessing the NHS Direct website as one of their first actions to find out information about symptoms, in most cases they did report going on to seek help from traditional health service sources, albeit sometimes with less urgency or less anxiety. This was the case whether they reported seeking help for themselves or for someone else. Furthermore, half the interviewees also stated that they would tend to see a professional as a first point of call if they had a health problem\nAs I say, it didn’t make me do anything that I wasn’t already contemplating, but I guess it gave me the answer to the question that I was looking for and then the peace of mind that it wasn’t becoming an emergency I guess and that we should stick with it.\nWhere demand management appeared to be occurring in practice, this was usually explained by the avoidance of barriers to accessing traditional health services, such as difficulties in getting an appointment or in travelling to one.\nIt used to take two hours and two bus journeys to get to the doctors...It’s easier to use than to get down to the doctor’s...It’s the time and the money...you know – those kinds of factors and then all the problems with getting appointments as well.\nThe desire for a second opinion following initial advice received from a heath professional was another reported motivation. Participants described the Internet as a way of accessing specialist knowledge, which they could use to challenge the advice given during their consultation. As in the example below from Interviewee 21, this challenge was an explicit response to not believing the health professional. At other times, illustrated by a quote from Interviewee 5, this was more about becoming fully informed on the range of divergent professional opinion.\nThey were telling me that treatment would be such and such and I thought“well I don’t believe that and I’ll use NHS Direct to see whether they can give me some information.”\nWell, when you go to the doctor you’ve only got his opinion haven’t you? I mean I’m sure he’s basing it on knowledge and research and things like that, but I just wanted to see other opinions about it.\nThe next motivation described by participants was also related to researching information prior to or following a consultation, but was not motivated by a desire to challenge. Instead the reported motivation was to seek clarity and confirmatory information in greater depth. This could be characterized as “homework” to support informed decision making, which could be done at the individual’s own pace. As Interviewee 20 explained:\nIt [the Internet] is excellent for a slower time study of information that my doctor hasn’t fully explained.\nI was really looking to substantiate a little bit more about the treatment options that I was given by the GP.", "Interviewees volunteered a range of benefits of online health information seeking. These reported benefits clustered around three theme areas: convenience, coverage, and anonymity. The convenience of online health encompassed the ease and speed of access, at any time, and from any location, especially from home. Access could be “in your own time\n...\nat your own pace” [Interviewee 19, a 48-year-old man]. This was contrasted with issues in accessing traditional health services.\nYou can go on it any time of the day quite honestly, whereas you can’t get your doctor any time of the day, or they have to ring you back and then you’re sitting either waiting or...you know.\nIt’s quick and it’s direct. It’s there in front of me every day and every evening.\nIn their responses, interviewees recognized that the Internet played a role in allowing them to become informed consumers, better able to share decisions with their health care provider. For example, Interviewee 18 (a 46-year-old woman) reported being able to “make an informed decision” in conjunction with her specialist, about treatment for fibroids, having sought information online. The benefit of coverage related to the wide range and depth of health information available on the Internet, and access to specialist medical knowledge. This access to esoteric medical knowledge was highlighted by Interviewee 4 (a 59-year-old woman):\nIt’s [the Internet] the perfect tool for finding out something that you need to know about and you probably don’t have the information unless you’re a medic.\nConfidentiality was the third area that emerged from the interviews as a key valued benefit for Internet health users. This encapsulated both the anonymous nature of online identity and the ability to use the Internet privately from any location. This could be of particular value for conditions that were more personal or stigmatizing.\nConfidentiality...you’re not speaking to someone about health issues. I mean for someone that has a lot more of a personal problem and they didn’t really want to discuss it with someone it’s ideal.\nInterviewees also discussed some of the challenges of health-related Internet use but no prominent theme emerged. The issues raised included (1) inaccurate information leading to harmful health decisions, which was reported as more of a theoretical problem, rather than by anyone with direct experience; (2) misuse of accurate information, leading to inappropriate self-diagnosis: “there’s always the worry of misdiagnosing something, or reading something into it” [Interviewee 19, a 48-year-old man]; (3) confusion caused by the sheer volume of information, which was sometimes perceived as being “confusing” and “daunting”; and (4) sometimes criticism of health-related Internet use for its “impersonal” nature, which lacked the quality of face-to-face contact, and which could not replace a real consultation.", "Throughout the interviews, participants explained the strategies employed in navigating online health provision and determining what information to use and to trust. They reported finding NHS Direct Online either by using a search engine (almost always this was Google), or by going directly to the uniform resource locator (URL) once they were familiar with it. Choosing a site was regarded as a common sense activity by interviewees. They were well aware that the Internet could be a source of misleading information – “there’s a lot of crap on the Internet\n,\n” as Interviewee 21 (a 59-year-old man) put it – but they used common sense to avoid this. For Interviewee 1, the Internet provided her with:\nA vast range of information from the idiotic to the academic, so I’ve got a vast range of information and it’s on tap so to speak. It’s up to the individual to adjudicate whether the information is relevant, or whether it’s valid...I do have enough common sense to evaluate what I see...I keep repeating that you need to use your discretion when you read...I would say that it could be a dangerous thing, on the other hand but I think the majority of people do have common sense.\nSeveral participants expressed negative views regarding peer-to-peer sources of health information, related to concerns about its trustworthiness. For example, Interviewee 4 (a 59-year-old woman) was concerned that the information may be written by “wild women from Minnesota\n,\n” and Interviewee 5 explained:\nHow do you know if they’re trustworthy? No, I wouldn’t do that at all. I don’t particularly like these chat rooms anyway...Because I don’t know them. No, I don’t want to talk to people I don’t know about things really. I think it could be quite dangerous...Perhaps I’m being cautious but that’s how I think.\n“Brand recognition” was reported as very important to the interviewees in navigating the health Internet. Interviewees reported choosing sites that had “real-world” branding, that is, an identity that they recognized from their offline experiences. The importance of the brand in establishing that a site was trustworthy was a very strong theme across the interviews. The NHS brand in particular was seen by respondents as giving the website the valued qualities of being impartial, reliable, and up-to-date. As one interviewee put it “You tend to trust the NHS don’t you?” [Interviewee 11, a 38-year-old woman]. Interviewees often contrasted this inherent trustworthiness of the NHS brand with their views on commercial health websites, particularly those produced by pharmaceutical companies:\nI thought the NHS one probably has no axe to grind...Whereas if it’s related to a drug company or somebody with herbal medicine and all of this sort of thing, I think they tend to be more biased, whereas the NHS is not trying to sell you something.\nA further interesting finding regarding which sites were valued and used was the low esteem in which North American sites were sometimes held as reported by our British participants. This was partly due to a perception that US sites had commercial aims and were therefore seen as “trying to sell something” and partly the lack of local or cultural relevance for some of the information.\nThe thing about the NHS online is that you know you’re looking at genuine stuff. The answers that you’re going to get are absolutely spot on and you can rely and trust them, whereas if I just Google something I may end up on an American site or something. I wouldn’t feel confident that the information I was looking at was absolutely right.", "As would be expected given the route for recruitment, respondents made many comments about specific aspects of the NHS Direct website; for example, wanting further information on specific topics. Over and above these individual remarks, two broad issues concerning the NHS Internet service were present across the interviews and had generic relevance to health-related Internet provision. The first of these was the perception that the NHS Direct website was integrated with the real-world NHS service. As Interviewee 4 (a 59-year-old woman) put it “I would hope that it ties in with the NHS generally, so it seemed to be the sensible place to go\n.\n” Interviewees had an expectation that there would be some connection between their use of a virtual health service and the care they received from the physical counterpart. Furthermore, they felt that, because they were NHS patients, using the NHS website was the “right thing to do,” because of this perceived integration across online and traditional services.\nSince we are under the NHS system, it would be logical for me to go first of all to the NHS Direct to see what the NHS’s take was...As far as I’m concerned, if the NHS Direct website is offering this information then it should make it uniform all through the NHS...I specifically used the NHS website because we live in an NHS world and where better than to get it straight from the horse’s mouth?\nThe second broad issue of generic relevance related to feedback about the clarity and simplicity of design of the NHS Direct website, in terms of the language used and the architecture of the site.\nIt’s got clear information and there’s enough there, but not like reels and reels that you get that you struggle to understand it and it’s very well broken down into sections as well I think, so like very specific for children and adults.\nTogether with the issue of the NHS brand, which was described above under navigation strategy, the clarity of the site and the perceived link with the real-world service were the principal reasons reported for valuing the NHS Direct site in particular as a source of online health information.", "The survey findings with respect to the age, gender, and educational status of online health seekers add to the accumulated evidence of several studies over the last decade that have shown that being female, being younger, and having a higher level of educational attainment are all associated with more frequent health-related Internet use [11,19-23]. The reported health status profile of our survey participants appeared to be very close to that of the general population. Work in other countries has sometimes shown a tendency to overrepresent people with chronic illness among health Internet users [7,8,24,25], while others have not found this association [11,22]. The NHS Direct website is used for both acute and chronic problems, mild or serious, and is used by individuals for themselves and on behalf of family members, especially by women. It is therefore perhaps not surprising that the health status of the users in our study was similar to that of the general population. The majority of those surveyed were seeking information for themselves, which is consistent with the findings of others [11]. Furthermore, a large proportion of users (over 40%) had already sought help from a health professional for the same health issue prior to accessing the website.\nHaving established the characteristics of the users of the site, we undertook in-depth qualitative work to explore in detail their motivations and attitudes. In 2003, the lead author (JP) wrote a review paper that summarized the benefits and challenges of health-related Internet use [12]. In Table 1 we integrated these with the thoughts of other authors in this field [16,17]. The analysis of our interviews supports most of the theoretical benefits discussed in the literature, and indicates that the health Internet is delivering on its potential benefits, while at the same time presenting some challenges to health professionals. Participants’ responses indicated that the Internet was being used as a tool to educate and reassure, and to sometimes challenge information received by health professionals. Previous work on the sociology of health-related Internet use has invoked theories of empowerment, democratization, and the challenge to health professional power [13,15,26] Most empirical work has indicated that, while these processes are taking place, the change is more subtle than many theorists have predicted, with the ongoing predominance of a biomedical model in the context of more-informed health consumers [27-29]. By sampling users of the NHS website it is perhaps not surprising that our findings support this model of evolutionary rather than revolutionary change, with health e-consumers seeking to become more informed through authoritative advice from official websites. Health-related Internet use was seen by most of our participants as a supplement to existing health service provision rather than a replacement for it [4,30]. The motivations of reassurance and of seeking greater understanding can be seen in this context. Even the motivation to find a second opinion to challenge other information was within the context of a model of biomedical authority. Our findings support the idea that online health resources are enmeshed with other (offline) approaches to seeking help [26], and that health-related Internet use is now embedded in everyday health practices [31]. \nThe majority of online events were related to real-world consultations, whether as preparation for them or as a search for further information afterward [23,32,33]. There were few examples of demand management occurring in practice, in terms of reducing the need for consultations, but our findings do support the idea that a health website can lead to more appropriate use of other services. Peer-to-peer interaction was not a focus of this study, as this is not provided on the NHS Direct website, and the number of participants in our qualitative sample reporting use of online support groups for health conditions was not high. Nevertheless, some participants did discuss the value of online interaction with others with similar problems, in particular the reassurance of knowing they were not alone, as found in previous work [34], while others expressed concerns about the trustworthiness of peer-to-peer sources. Consumer access to poor-quality information on the Internet has been a long-standing concern in the eHealth literature [15]. We found that, in avoiding misinformation and identifying which information to trust, participants put great emphasis on recognition of brands such as the NHS, which were trusted in the non-Internet world, together with using common sense approaches to navigate the health Internet. The reported value of official branding of health websites in determining trustworthiness is supported by previous work [35]. The online benefits of convenience and anonymity are well established [36] and were widely reported, as was the expectation that online health services would be fully integrated with their real-world counterparts, something that remains an aspiration for the NHS in the United Kingdom but is not yet a reality.\nSome theoretical benefits and challenges were not prominent in these interviews. The “green” potential of the Internet to reduce travel, but at the same time possibly reduce physical activity and lead to social isolation or depression [37], was not discussed by our participants. Nor were any ergonomic effects of computer use, or the problem of Internet addiction. The issue of the digital divide, although mentioned by three participants, did not emerge as a consistent theme. However, given that this sample were all Internet users, this was perhaps not surprising.\n[SUBTITLE] Limitations [SUBSECTION] Because this was an opt-in survey accessed via a weblink, it was not possible to calculate a response rate for the questionnaire. There were also design issues with the website during the survey period, which meant that the link to the survey was not always clearly visible to users. This affected the overall response and the number of individuals consenting to be interviewed in the second stage. To minimize social desirability bias, the researchers made it clear to interviewees that the researchers were independent of the NHS Direct organization, but those volunteering for interviews may still have been a particular population who wanted to relate their experiences with NHS Direct, good or bad. More women than men were interviewed due to having very few male volunteers. Interview methodology of this type, asking people to report how and why they used a particular source, may reflect attitudes rather than actual behavior, for which direct observation may be preferred. Nevertheless, questions were designed to focus on the most recent actual use of the Internet for health, rather than rely on hypothetical questioning. The participants were users of the NHS Direct website and were therefore not necessarily representative of the overall population of online health information seekers in the United Kingdom. However, the NHS site is the most popular health information site in the United Kingdom, and the demographic profile of respondents was similar to that of non-UK-based studies.\nBecause this was an opt-in survey accessed via a weblink, it was not possible to calculate a response rate for the questionnaire. There were also design issues with the website during the survey period, which meant that the link to the survey was not always clearly visible to users. This affected the overall response and the number of individuals consenting to be interviewed in the second stage. To minimize social desirability bias, the researchers made it clear to interviewees that the researchers were independent of the NHS Direct organization, but those volunteering for interviews may still have been a particular population who wanted to relate their experiences with NHS Direct, good or bad. More women than men were interviewed due to having very few male volunteers. Interview methodology of this type, asking people to report how and why they used a particular source, may reflect attitudes rather than actual behavior, for which direct observation may be preferred. Nevertheless, questions were designed to focus on the most recent actual use of the Internet for health, rather than rely on hypothetical questioning. The participants were users of the NHS Direct website and were therefore not necessarily representative of the overall population of online health information seekers in the United Kingdom. However, the NHS site is the most popular health information site in the United Kingdom, and the demographic profile of respondents was similar to that of non-UK-based studies.\n[SUBTITLE] Conclusions [SUBSECTION] Given increasing resource constraints, the health care community needs to seek ways of promoting efficient and appropriate health care use, which should include consideration of how Internet health information is provided and used, and how traditional NHS services and online services can be best integrated. The study findings support a model of evolutionary rather than revolutionary change in health information use, with real-world trusted brands being used online, in conjunction with traditional consultations. It will be interesting to see whether in time, particularly as the younger “Internet generation” ages and eHealth literacy increases in all age groups [38], Internet health information will be trusted enough to be used as an alternative, as opposed to an adjunct, to other types of health-seeking activities, and by individuals of broader demographic profiles. Our findings fit with a “shared decision-making” model [39], where individuals seek information to help the decision-making process and confirm what they are being told, rather than seeking to become independent experts. One of the primary motivations was the seeking of reassurance, and the value of this in terms of health or social benefit or more appropriate service use needs to be further explored. The relationship between Internet use and health outcomes is an area for research development, including examination of the role of user empowerment. Health service providers should aim to harness the potential benefits of health-related Internet use, rather than see it as a burden or challenge.\nGiven increasing resource constraints, the health care community needs to seek ways of promoting efficient and appropriate health care use, which should include consideration of how Internet health information is provided and used, and how traditional NHS services and online services can be best integrated. The study findings support a model of evolutionary rather than revolutionary change in health information use, with real-world trusted brands being used online, in conjunction with traditional consultations. It will be interesting to see whether in time, particularly as the younger “Internet generation” ages and eHealth literacy increases in all age groups [38], Internet health information will be trusted enough to be used as an alternative, as opposed to an adjunct, to other types of health-seeking activities, and by individuals of broader demographic profiles. Our findings fit with a “shared decision-making” model [39], where individuals seek information to help the decision-making process and confirm what they are being told, rather than seeking to become independent experts. One of the primary motivations was the seeking of reassurance, and the value of this in terms of health or social benefit or more appropriate service use needs to be further explored. The relationship between Internet use and health outcomes is an area for research development, including examination of the role of user empowerment. Health service providers should aim to harness the potential benefits of health-related Internet use, rather than see it as a burden or challenge.", "Because this was an opt-in survey accessed via a weblink, it was not possible to calculate a response rate for the questionnaire. There were also design issues with the website during the survey period, which meant that the link to the survey was not always clearly visible to users. This affected the overall response and the number of individuals consenting to be interviewed in the second stage. To minimize social desirability bias, the researchers made it clear to interviewees that the researchers were independent of the NHS Direct organization, but those volunteering for interviews may still have been a particular population who wanted to relate their experiences with NHS Direct, good or bad. More women than men were interviewed due to having very few male volunteers. Interview methodology of this type, asking people to report how and why they used a particular source, may reflect attitudes rather than actual behavior, for which direct observation may be preferred. Nevertheless, questions were designed to focus on the most recent actual use of the Internet for health, rather than rely on hypothetical questioning. The participants were users of the NHS Direct website and were therefore not necessarily representative of the overall population of online health information seekers in the United Kingdom. However, the NHS site is the most popular health information site in the United Kingdom, and the demographic profile of respondents was similar to that of non-UK-based studies.", "Given increasing resource constraints, the health care community needs to seek ways of promoting efficient and appropriate health care use, which should include consideration of how Internet health information is provided and used, and how traditional NHS services and online services can be best integrated. The study findings support a model of evolutionary rather than revolutionary change in health information use, with real-world trusted brands being used online, in conjunction with traditional consultations. It will be interesting to see whether in time, particularly as the younger “Internet generation” ages and eHealth literacy increases in all age groups [38], Internet health information will be trusted enough to be used as an alternative, as opposed to an adjunct, to other types of health-seeking activities, and by individuals of broader demographic profiles. Our findings fit with a “shared decision-making” model [39], where individuals seek information to help the decision-making process and confirm what they are being told, rather than seeking to become independent experts. One of the primary motivations was the seeking of reassurance, and the value of this in terms of health or social benefit or more appropriate service use needs to be further explored. The relationship between Internet use and health outcomes is an area for research development, including examination of the role of user empowerment. Health service providers should aim to harness the potential benefits of health-related Internet use, rather than see it as a burden or challenge." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "information seeking behaviour", "information seeking behaviour", "internet", "patient-provider communication", "health services research" ]

[ "Administration, Oral", "Adult", "Aged", "Antimetabolites, Antineoplastic", "Drug Combinations", "Female", "Humans", "Middle Aged", "Neoplasm Grading", "Neoplasm Staging", "Oxonic Acid", "Recurrence", "Tegafur", "Uterine Cervical Neoplasms" ]

[ "introduction", "eligibility criteria", "treatment schedule", "response and toxicity evaluation", "statistical consideration", "results", "patient population", "antitumor activity", "safety", "discussion", "funding", "disclosure" ]

[ "Cancer of the uterine cervix is the main cause of death from gynecologic malignancy in emerging countries. In the developed world as well, a third of women with cervical cancer die of uncontrolled disease. Although a number of chemotherapeutic agents have been investigated in patients with advanced or recurrent cervical cancer, the prognosis of those patients remains poor. Identification of new agents with activity in cervical cancer is needed.\nS-1 (TS-1; Taiho Pharmaceutical, Tokyo, Japan) is an oral fluoropyrimidine consisting of tegafur [a prodrug that is metabolized to 5-fluorouracil (5-FU) in blood, largely by the cytochrome P450 system in the liver], gimeracil (an inhibitor of dihydropyrimidine dehydrogenase, which degrades fluorouracil), and oteracil (which inhibits the phosphorylation of fluorouracil in the gastrointestinal tract, thereby reducing the gastrointestinal toxic effects of fluorouracil) in a molar ratio of 1 : 0.4 : 1 [1]. S-1 is known to be active against gastric, head and neck, colorectal, lung, breast, pancreatic, and biliary tract cancers [2–9]. This phase II study was designed to evaluate the efficacy and safety of S-1 in patients with uterine cervical cancer and is the first exploration of S-1 for the treatment of any gynecologic cancer. S-1 has also shown activity for cervical cancer in preclinical study (data are available only in investigator’s brochure); phase II study of S-1 in patients with cervical cancer has been launched to evaluate the usefulness of S-1 in those patients.", "Eligible patients were aged between 20 and 74 years, had Eastern Cooperative Oncology Group performance status of zero or one, and had histological documented primary stage IVB or recurrent cervical carcinoma. All patients had measurable disease according to the RECIST [10]. Measurable lesions defined unit dimensionally were ≥20 mm using conventional imaging or ≥10 mm with spiral computed topographic scan. Patients had not received more than one prior chemotherapy regimen since diagnosis of metastatic or recurrent disease. Patients who were administered in conjunction with radiation were not counted under prior chemotherapy. Four weeks from prior chemotherapy or radiotherapy were required before study entry. Adequate organ function was required for study entry: neutrophil count ≥2000/μl; platelet count ≥100 000/μl; hemoglobin ≥8.0 g/dl; serum bilirubin level ≤1.5 times upper limit of the institutional normal (ULN); asparate aminotransferase, alanine aminotransferase, and alkaline phosphatase levels ≤2.5 times ULN; and serum creatinine level ≤ ULN. Only patients who could swallow tablets were eligible. Patients with any of the following conditions were excluded from the study: active infection, severe heart disease, interstitial pneumonitis, history of hypersensitivity, malignant or benign effusions requiring drainage, active brain metastasis, or active concomitant malignancy. Patients receiving drugs with potential interactions with S-1 (flucytosine, warfarin, and phenytoin) were excluded. All patients gave informed consent before entering this study, which was approved by the institutional review boards at all participating institutions.", "Patients received two oral doses of S-1 35 mg/m2 daily for 4 weeks of a 6-week cycle. As S-1 is provided in 20 or 25 mg tablets, the actual dosage of S-1 was decided according to the patient’s body surface area as follows: patients with a body surface area of less than 1.25 m2 received 40 mg; those with a body surface area of 1.25–1.5 m2 received 50 mg; and those with a body surface area of more than 1.5 m2 received 60 mg. The schedule was repeated until the occurrence of disease progression, unacceptable toxic effects, or patient’s refusal. If a grade 3 or higher hematological toxicity or a grade 2 or higher nonhematological toxicity was observed, the dose was reduced from 60 to 50 mg, 50 to 40 mg, or temporary interruption of S-1 administration was recommended. Patients whose toxic effects necessitated a rest period of >4 weeks were withdrawn from treatment. When initial dose was 40, 50, or 60 mg, dose escalation could be allowed to 50, 60, and 75 mg for subsequent cycles, unless adverse events were observed.", "The tumor response was assessed according to the guidelines of RECIST. Target lesions included all measurable lesions up to a maximum of five lesions per organ and 10 lesions in total. Target lesions were included the lesions with previously irradiated area. Complete response (CR) was defined as the complete disappearance of all target and nontarget lesions, with no development of new disease. Partial response (PR) was defined as a reduction by ≥30% in the sum of the longest diameter of target lesions. CRs or PRs were confirmed by repeat assessments carried out no <4 weeks after the criteria for response were first met. Progressive disease (PD) was defined as an increase ≥20% in the sum of the longest diameter of all target lesions or the appearance of one or more new lesions and/or unequivocal progression of existing nontarget lesions. Stable disease (SD) was defined as neither sufficient lesion shrinkage to qualify for PR nor sufficient increase to qualify for PD. Best response was defined as the most CR achieved by a patient (thus, each patient had a single best response: CR, PR, SD, or PD), and the date of best response was the date it was first detected. Radiological studies were repeated every two cycles. If a patient was documented as having a CR or a PR, the response was confirmed at least 4 weeks after the first evidence of response. An independent response review committee (IRRC) evaluated all tumor responses after the investigators had completed their judgment.\nToxic effects were evaluated with respect to incidence and severity using Common Terminology Criteria of Adverse Events (version 3.0) (www.cancer.gov/).", "The primary end point of this study was to assess the overall response rate determined by the IRRC. The secondary end points were to assess duration of response, time to response, time to progression (TTP), overall survival, and adverse events. Assuming a response rate of 20%, the study was designed with 80% power such that the lower limit of the 95% confidence interval (CI) for the estimate of the response rate was >0.05. A sample size of 32 assessable patients was required. The Kaplan–Meier method was used to determine the TTP and median survival time (MST) in the assessable population. TTP was defined as the time from the first medication to the date of a PD event or death (due to cervical cancer or study drugs).", "[SUBTITLE] patient population [SUBSECTION] A total of 37 patients were entered into the study from July 2005 to September 2007 and 36 patients were eligible and assessable. One patient had a lack of absolute neutrophil count for eligibility criteria. All 37 patients were evaluated for safety. Patient characteristics are listed in Table 1. More than half of the patients had distant diseases. Seventeen patients (8 for neoadjuvant chemotherapy, 3 for metastatic disease, and 6 for both) received prior chemotherapy (not including chemoradiotherapy): 14 received platinum-containing regimen and 3 received oral 5-FU derivative drug alone. Thirteen patients (36%) received chemoradiotherapy. Prior platinum therapy including chemotherapy or chemoradiotherapy was administered for 22 patients.\nPatient characteristics\nNot included chemoradiotherapy.\nA total of 167 treatment cycles (median 4, range 1–19) were administered. Nineteen patients (53%) were subjected to dose reduction owing to adverse events. The median relative dose intensity was 0.83 (range 0.45–1.04).\nA total of 37 patients were entered into the study from July 2005 to September 2007 and 36 patients were eligible and assessable. One patient had a lack of absolute neutrophil count for eligibility criteria. All 37 patients were evaluated for safety. Patient characteristics are listed in Table 1. More than half of the patients had distant diseases. Seventeen patients (8 for neoadjuvant chemotherapy, 3 for metastatic disease, and 6 for both) received prior chemotherapy (not including chemoradiotherapy): 14 received platinum-containing regimen and 3 received oral 5-FU derivative drug alone. Thirteen patients (36%) received chemoradiotherapy. Prior platinum therapy including chemotherapy or chemoradiotherapy was administered for 22 patients.\nPatient characteristics\nNot included chemoradiotherapy.\nA total of 167 treatment cycles (median 4, range 1–19) were administered. Nineteen patients (53%) were subjected to dose reduction owing to adverse events. The median relative dose intensity was 0.83 (range 0.45–1.04).\n[SUBTITLE] antitumor activity [SUBSECTION] Table 2 describes the response assessment. The objective response rate assessed by IRRC was 30.6% (95% CI 15.5% to 45.6%). The median duration of response was 134 days (range 73–553 days). The investigators identified one CR and nine PRs. One clinical responded patient who had CR was downgraded to PR, two clinical responded patients who had PR were downgraded to SD and PD, respectively, and three patients who had SD were upgraded to PR by the judgment of IRRC. Therefore, a total of 11 patients were judged PR. Responses according to prior therapy are listed in Table 2. Patients who received chemotherapy alone had a response of 17.6%, patients who received chemoradiotherapy 53.8%, and patients who received platinum-containing chemotherapy or chemoradiotherapy 31.8%. Eighteen patients had target lesions with previously irradiated area and five (27.8%) of them were responded.\nResponses to S-1 according to the patient characteristics\nCR, complete response; PR, partial response; SD, stable disease; PD, Progressive disease; CI, confidential interval.\nAfter a median follow-up duration of 25 months, the median TTP was 5.2 months (95% CI 4.5–6.6 months; Figure 1) and the MST was 15.4 months (95% CI 11.5–17.8 months; Figure 2). One-year survival was 58.3%.\nKaplan–Meier plot for time to progression (TTP; n = 36). CI, confidence interval.\nKaplan–Meier plot for overall survival (n = 36). CI, confidence interval.\nTable 2 describes the response assessment. The objective response rate assessed by IRRC was 30.6% (95% CI 15.5% to 45.6%). The median duration of response was 134 days (range 73–553 days). The investigators identified one CR and nine PRs. One clinical responded patient who had CR was downgraded to PR, two clinical responded patients who had PR were downgraded to SD and PD, respectively, and three patients who had SD were upgraded to PR by the judgment of IRRC. Therefore, a total of 11 patients were judged PR. Responses according to prior therapy are listed in Table 2. Patients who received chemotherapy alone had a response of 17.6%, patients who received chemoradiotherapy 53.8%, and patients who received platinum-containing chemotherapy or chemoradiotherapy 31.8%. Eighteen patients had target lesions with previously irradiated area and five (27.8%) of them were responded.\nResponses to S-1 according to the patient characteristics\nCR, complete response; PR, partial response; SD, stable disease; PD, Progressive disease; CI, confidential interval.\nAfter a median follow-up duration of 25 months, the median TTP was 5.2 months (95% CI 4.5–6.6 months; Figure 1) and the MST was 15.4 months (95% CI 11.5–17.8 months; Figure 2). One-year survival was 58.3%.\nKaplan–Meier plot for time to progression (TTP; n = 36). CI, confidence interval.\nKaplan–Meier plot for overall survival (n = 36). CI, confidence interval.\n[SUBTITLE] safety [SUBSECTION] All 37 patients were assessed for safety. Four patients were discontinued due to toxic effects. Adverse events are listed in Table 3. Grade 3 or 4 hematologic toxic effects were anemia (16%), neutropenia (8%), and thrombocytopenia (5%). Among grade 3 or 4 nonhematologic toxic effects, the most frequent were anorexia (16%) and diarrhea (22%). All other grade 3 or 4 toxic effects were recorded in <10% of patients.\nAdverse events (n = 37)\nAll 37 patients were assessed for safety. Four patients were discontinued due to toxic effects. Adverse events are listed in Table 3. Grade 3 or 4 hematologic toxic effects were anemia (16%), neutropenia (8%), and thrombocytopenia (5%). Among grade 3 or 4 nonhematologic toxic effects, the most frequent were anorexia (16%) and diarrhea (22%). All other grade 3 or 4 toxic effects were recorded in <10% of patients.\nAdverse events (n = 37)", "A total of 37 patients were entered into the study from July 2005 to September 2007 and 36 patients were eligible and assessable. One patient had a lack of absolute neutrophil count for eligibility criteria. All 37 patients were evaluated for safety. Patient characteristics are listed in Table 1. More than half of the patients had distant diseases. Seventeen patients (8 for neoadjuvant chemotherapy, 3 for metastatic disease, and 6 for both) received prior chemotherapy (not including chemoradiotherapy): 14 received platinum-containing regimen and 3 received oral 5-FU derivative drug alone. Thirteen patients (36%) received chemoradiotherapy. Prior platinum therapy including chemotherapy or chemoradiotherapy was administered for 22 patients.\nPatient characteristics\nNot included chemoradiotherapy.\nA total of 167 treatment cycles (median 4, range 1–19) were administered. Nineteen patients (53%) were subjected to dose reduction owing to adverse events. The median relative dose intensity was 0.83 (range 0.45–1.04).", "Table 2 describes the response assessment. The objective response rate assessed by IRRC was 30.6% (95% CI 15.5% to 45.6%). The median duration of response was 134 days (range 73–553 days). The investigators identified one CR and nine PRs. One clinical responded patient who had CR was downgraded to PR, two clinical responded patients who had PR were downgraded to SD and PD, respectively, and three patients who had SD were upgraded to PR by the judgment of IRRC. Therefore, a total of 11 patients were judged PR. Responses according to prior therapy are listed in Table 2. Patients who received chemotherapy alone had a response of 17.6%, patients who received chemoradiotherapy 53.8%, and patients who received platinum-containing chemotherapy or chemoradiotherapy 31.8%. Eighteen patients had target lesions with previously irradiated area and five (27.8%) of them were responded.\nResponses to S-1 according to the patient characteristics\nCR, complete response; PR, partial response; SD, stable disease; PD, Progressive disease; CI, confidential interval.\nAfter a median follow-up duration of 25 months, the median TTP was 5.2 months (95% CI 4.5–6.6 months; Figure 1) and the MST was 15.4 months (95% CI 11.5–17.8 months; Figure 2). One-year survival was 58.3%.\nKaplan–Meier plot for time to progression (TTP; n = 36). CI, confidence interval.\nKaplan–Meier plot for overall survival (n = 36). CI, confidence interval.", "All 37 patients were assessed for safety. Four patients were discontinued due to toxic effects. Adverse events are listed in Table 3. Grade 3 or 4 hematologic toxic effects were anemia (16%), neutropenia (8%), and thrombocytopenia (5%). Among grade 3 or 4 nonhematologic toxic effects, the most frequent were anorexia (16%) and diarrhea (22%). All other grade 3 or 4 toxic effects were recorded in <10% of patients.\nAdverse events (n = 37)", "The prognosis of patients with advanced or recurrent cervical cancer remains poor and there is an urgent need for novel therapeutic agents. This current study was designed to determine the efficacy and tolerability of an oral agent of S-1 for advanced or recurrent cervical cancer and demonstrated a higher response rate of 30.6% with modest toxic effects: grade 3 or 4 anemia (16%), anorexia (16%), and diarrhea (22%).\nThe most extensively studied agent in the treatment of advanced cervical cancer is cisplatin, which has been used as a single agent, in combination chemotherapy, or with radiotherapy. The eligibility criteria of our study included patients with prior chemotherapy or chemoradiotherapy. Twenty-two of the 36 patients (61%) had previously received platinum therapy including chemoradiotherapy. There may be drug resistance to cisplatin in such patients; however, objective responses were seen in patients who had received prior platinum therapy. Therefore, it is suggested that S-1 is a noncross resistant drug for cisplatin.\nSeveral non-platinum agents, such as paclitaxel [11–13], topotecan [14, 15], irinotecan [16, 17], vinorelbine [18–20], capecitabine [21, 22], and ifosphamide [23–25] were found to have moderate activity in patients with metastatic cervical cancer. However, none of the previously reported phase II studies of non-platinum single-agent chemotherapy for patients with advanced cervical cancer have reported >30% response rate, except paclitaxel and ifosphamide [26]. Paclitaxel is an active agent for cervical cancer and has been evaluated in randomized trial. GOG 0204 compared doublets of paclitaxel, vinorelbine, and gemcitabine plus cisplatin with the combination of topotecan plus cisplatin, and there was a trend favoring treatment with cisplatin/paclitaxel for response rate, progression-free survival (PFS), overall survival, and quality of life [27]. Ifosphamide in combination with cisplatin was tested in randomized trial comparing cisplatin alone and showed a better response rate and PFS but not overall survival and including severe toxic effects. Although our study examined a small number of patients and the CI was wide, notable objective responses were achieved in this single-agent chemotherapy.\nCombinations of 5-FU and cisplatin yield synergistic in preclinical studies [28, 29]. A combination therapy of S-1 and cisplatin has been studied in other malignancies, including gastric cancer, lung cancer, and head and neck cancer [30–32]. Phase III trial comparing S-1 in combination with cisplatin versus S-1 alone in advanced gastric cancer demonstrated a significant benefit for combined S-1 plus cisplatin in response rate, PFS, and overall survival [33]. Based on the promising activity of S-1 in the present phase II study, and the experience with S-1 plus cisplatin in other malignancies, we have started phase III trial of S-1 plus cisplatin compared with single-agent cisplatin for metastatic cervical cancer in an Asian trial, including Japan, Korea, and Taiwan.\nIn conclusion, S-1 is active in patients with metastatic cervical cancer and well tolerated. S-1 plus cisplatin has now entered a prospective randomized phase III trial.", "Taiho Pharmaceutical, Tokyo, Japan.", "Dr Kamiura has reported honoraria for Taiho Pharmaceutical. Dr Ochiai has reported consultant or advisory role for Taiho Pharmaceutical, and honoraria for Taiho Pharmaceutical, Bristol Myers Squibb, and Sanofi Aventis; he has received research support from Taiho Pharmaceutical, and Bristol Myers Squibb. The other authors have not reported any conflicts of interest." ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "introduction", "patients and methods", "eligibility criteria", "treatment schedule", "response and toxicity evaluation", "statistical consideration", "results", "patient population", "antitumor activity", "safety", "discussion", "funding", "disclosure" ]

[ "Cancer of the uterine cervix is the main cause of death from gynecologic malignancy in emerging countries. In the developed world as well, a third of women with cervical cancer die of uncontrolled disease. Although a number of chemotherapeutic agents have been investigated in patients with advanced or recurrent cervical cancer, the prognosis of those patients remains poor. Identification of new agents with activity in cervical cancer is needed.\nS-1 (TS-1; Taiho Pharmaceutical, Tokyo, Japan) is an oral fluoropyrimidine consisting of tegafur [a prodrug that is metabolized to 5-fluorouracil (5-FU) in blood, largely by the cytochrome P450 system in the liver], gimeracil (an inhibitor of dihydropyrimidine dehydrogenase, which degrades fluorouracil), and oteracil (which inhibits the phosphorylation of fluorouracil in the gastrointestinal tract, thereby reducing the gastrointestinal toxic effects of fluorouracil) in a molar ratio of 1 : 0.4 : 1 [1]. S-1 is known to be active against gastric, head and neck, colorectal, lung, breast, pancreatic, and biliary tract cancers [2–9]. This phase II study was designed to evaluate the efficacy and safety of S-1 in patients with uterine cervical cancer and is the first exploration of S-1 for the treatment of any gynecologic cancer. S-1 has also shown activity for cervical cancer in preclinical study (data are available only in investigator’s brochure); phase II study of S-1 in patients with cervical cancer has been launched to evaluate the usefulness of S-1 in those patients.", "[SUBTITLE] eligibility criteria [SUBSECTION] Eligible patients were aged between 20 and 74 years, had Eastern Cooperative Oncology Group performance status of zero or one, and had histological documented primary stage IVB or recurrent cervical carcinoma. All patients had measurable disease according to the RECIST [10]. Measurable lesions defined unit dimensionally were ≥20 mm using conventional imaging or ≥10 mm with spiral computed topographic scan. Patients had not received more than one prior chemotherapy regimen since diagnosis of metastatic or recurrent disease. Patients who were administered in conjunction with radiation were not counted under prior chemotherapy. Four weeks from prior chemotherapy or radiotherapy were required before study entry. Adequate organ function was required for study entry: neutrophil count ≥2000/μl; platelet count ≥100 000/μl; hemoglobin ≥8.0 g/dl; serum bilirubin level ≤1.5 times upper limit of the institutional normal (ULN); asparate aminotransferase, alanine aminotransferase, and alkaline phosphatase levels ≤2.5 times ULN; and serum creatinine level ≤ ULN. Only patients who could swallow tablets were eligible. Patients with any of the following conditions were excluded from the study: active infection, severe heart disease, interstitial pneumonitis, history of hypersensitivity, malignant or benign effusions requiring drainage, active brain metastasis, or active concomitant malignancy. Patients receiving drugs with potential interactions with S-1 (flucytosine, warfarin, and phenytoin) were excluded. All patients gave informed consent before entering this study, which was approved by the institutional review boards at all participating institutions.\nEligible patients were aged between 20 and 74 years, had Eastern Cooperative Oncology Group performance status of zero or one, and had histological documented primary stage IVB or recurrent cervical carcinoma. All patients had measurable disease according to the RECIST [10]. Measurable lesions defined unit dimensionally were ≥20 mm using conventional imaging or ≥10 mm with spiral computed topographic scan. Patients had not received more than one prior chemotherapy regimen since diagnosis of metastatic or recurrent disease. Patients who were administered in conjunction with radiation were not counted under prior chemotherapy. Four weeks from prior chemotherapy or radiotherapy were required before study entry. Adequate organ function was required for study entry: neutrophil count ≥2000/μl; platelet count ≥100 000/μl; hemoglobin ≥8.0 g/dl; serum bilirubin level ≤1.5 times upper limit of the institutional normal (ULN); asparate aminotransferase, alanine aminotransferase, and alkaline phosphatase levels ≤2.5 times ULN; and serum creatinine level ≤ ULN. Only patients who could swallow tablets were eligible. Patients with any of the following conditions were excluded from the study: active infection, severe heart disease, interstitial pneumonitis, history of hypersensitivity, malignant or benign effusions requiring drainage, active brain metastasis, or active concomitant malignancy. Patients receiving drugs with potential interactions with S-1 (flucytosine, warfarin, and phenytoin) were excluded. All patients gave informed consent before entering this study, which was approved by the institutional review boards at all participating institutions.\n[SUBTITLE] treatment schedule [SUBSECTION] Patients received two oral doses of S-1 35 mg/m2 daily for 4 weeks of a 6-week cycle. As S-1 is provided in 20 or 25 mg tablets, the actual dosage of S-1 was decided according to the patient’s body surface area as follows: patients with a body surface area of less than 1.25 m2 received 40 mg; those with a body surface area of 1.25–1.5 m2 received 50 mg; and those with a body surface area of more than 1.5 m2 received 60 mg. The schedule was repeated until the occurrence of disease progression, unacceptable toxic effects, or patient’s refusal. If a grade 3 or higher hematological toxicity or a grade 2 or higher nonhematological toxicity was observed, the dose was reduced from 60 to 50 mg, 50 to 40 mg, or temporary interruption of S-1 administration was recommended. Patients whose toxic effects necessitated a rest period of >4 weeks were withdrawn from treatment. When initial dose was 40, 50, or 60 mg, dose escalation could be allowed to 50, 60, and 75 mg for subsequent cycles, unless adverse events were observed.\nPatients received two oral doses of S-1 35 mg/m2 daily for 4 weeks of a 6-week cycle. As S-1 is provided in 20 or 25 mg tablets, the actual dosage of S-1 was decided according to the patient’s body surface area as follows: patients with a body surface area of less than 1.25 m2 received 40 mg; those with a body surface area of 1.25–1.5 m2 received 50 mg; and those with a body surface area of more than 1.5 m2 received 60 mg. The schedule was repeated until the occurrence of disease progression, unacceptable toxic effects, or patient’s refusal. If a grade 3 or higher hematological toxicity or a grade 2 or higher nonhematological toxicity was observed, the dose was reduced from 60 to 50 mg, 50 to 40 mg, or temporary interruption of S-1 administration was recommended. Patients whose toxic effects necessitated a rest period of >4 weeks were withdrawn from treatment. When initial dose was 40, 50, or 60 mg, dose escalation could be allowed to 50, 60, and 75 mg for subsequent cycles, unless adverse events were observed.\n[SUBTITLE] response and toxicity evaluation [SUBSECTION] The tumor response was assessed according to the guidelines of RECIST. Target lesions included all measurable lesions up to a maximum of five lesions per organ and 10 lesions in total. Target lesions were included the lesions with previously irradiated area. Complete response (CR) was defined as the complete disappearance of all target and nontarget lesions, with no development of new disease. Partial response (PR) was defined as a reduction by ≥30% in the sum of the longest diameter of target lesions. CRs or PRs were confirmed by repeat assessments carried out no <4 weeks after the criteria for response were first met. Progressive disease (PD) was defined as an increase ≥20% in the sum of the longest diameter of all target lesions or the appearance of one or more new lesions and/or unequivocal progression of existing nontarget lesions. Stable disease (SD) was defined as neither sufficient lesion shrinkage to qualify for PR nor sufficient increase to qualify for PD. Best response was defined as the most CR achieved by a patient (thus, each patient had a single best response: CR, PR, SD, or PD), and the date of best response was the date it was first detected. Radiological studies were repeated every two cycles. If a patient was documented as having a CR or a PR, the response was confirmed at least 4 weeks after the first evidence of response. An independent response review committee (IRRC) evaluated all tumor responses after the investigators had completed their judgment.\nToxic effects were evaluated with respect to incidence and severity using Common Terminology Criteria of Adverse Events (version 3.0) (www.cancer.gov/).\nThe tumor response was assessed according to the guidelines of RECIST. Target lesions included all measurable lesions up to a maximum of five lesions per organ and 10 lesions in total. Target lesions were included the lesions with previously irradiated area. Complete response (CR) was defined as the complete disappearance of all target and nontarget lesions, with no development of new disease. Partial response (PR) was defined as a reduction by ≥30% in the sum of the longest diameter of target lesions. CRs or PRs were confirmed by repeat assessments carried out no <4 weeks after the criteria for response were first met. Progressive disease (PD) was defined as an increase ≥20% in the sum of the longest diameter of all target lesions or the appearance of one or more new lesions and/or unequivocal progression of existing nontarget lesions. Stable disease (SD) was defined as neither sufficient lesion shrinkage to qualify for PR nor sufficient increase to qualify for PD. Best response was defined as the most CR achieved by a patient (thus, each patient had a single best response: CR, PR, SD, or PD), and the date of best response was the date it was first detected. Radiological studies were repeated every two cycles. If a patient was documented as having a CR or a PR, the response was confirmed at least 4 weeks after the first evidence of response. An independent response review committee (IRRC) evaluated all tumor responses after the investigators had completed their judgment.\nToxic effects were evaluated with respect to incidence and severity using Common Terminology Criteria of Adverse Events (version 3.0) (www.cancer.gov/).\n[SUBTITLE] statistical consideration [SUBSECTION] The primary end point of this study was to assess the overall response rate determined by the IRRC. The secondary end points were to assess duration of response, time to response, time to progression (TTP), overall survival, and adverse events. Assuming a response rate of 20%, the study was designed with 80% power such that the lower limit of the 95% confidence interval (CI) for the estimate of the response rate was >0.05. A sample size of 32 assessable patients was required. The Kaplan–Meier method was used to determine the TTP and median survival time (MST) in the assessable population. TTP was defined as the time from the first medication to the date of a PD event or death (due to cervical cancer or study drugs).\nThe primary end point of this study was to assess the overall response rate determined by the IRRC. The secondary end points were to assess duration of response, time to response, time to progression (TTP), overall survival, and adverse events. Assuming a response rate of 20%, the study was designed with 80% power such that the lower limit of the 95% confidence interval (CI) for the estimate of the response rate was >0.05. A sample size of 32 assessable patients was required. The Kaplan–Meier method was used to determine the TTP and median survival time (MST) in the assessable population. TTP was defined as the time from the first medication to the date of a PD event or death (due to cervical cancer or study drugs).", "Eligible patients were aged between 20 and 74 years, had Eastern Cooperative Oncology Group performance status of zero or one, and had histological documented primary stage IVB or recurrent cervical carcinoma. All patients had measurable disease according to the RECIST [10]. Measurable lesions defined unit dimensionally were ≥20 mm using conventional imaging or ≥10 mm with spiral computed topographic scan. Patients had not received more than one prior chemotherapy regimen since diagnosis of metastatic or recurrent disease. Patients who were administered in conjunction with radiation were not counted under prior chemotherapy. Four weeks from prior chemotherapy or radiotherapy were required before study entry. Adequate organ function was required for study entry: neutrophil count ≥2000/μl; platelet count ≥100 000/μl; hemoglobin ≥8.0 g/dl; serum bilirubin level ≤1.5 times upper limit of the institutional normal (ULN); asparate aminotransferase, alanine aminotransferase, and alkaline phosphatase levels ≤2.5 times ULN; and serum creatinine level ≤ ULN. Only patients who could swallow tablets were eligible. Patients with any of the following conditions were excluded from the study: active infection, severe heart disease, interstitial pneumonitis, history of hypersensitivity, malignant or benign effusions requiring drainage, active brain metastasis, or active concomitant malignancy. Patients receiving drugs with potential interactions with S-1 (flucytosine, warfarin, and phenytoin) were excluded. All patients gave informed consent before entering this study, which was approved by the institutional review boards at all participating institutions.", "Patients received two oral doses of S-1 35 mg/m2 daily for 4 weeks of a 6-week cycle. As S-1 is provided in 20 or 25 mg tablets, the actual dosage of S-1 was decided according to the patient’s body surface area as follows: patients with a body surface area of less than 1.25 m2 received 40 mg; those with a body surface area of 1.25–1.5 m2 received 50 mg; and those with a body surface area of more than 1.5 m2 received 60 mg. The schedule was repeated until the occurrence of disease progression, unacceptable toxic effects, or patient’s refusal. If a grade 3 or higher hematological toxicity or a grade 2 or higher nonhematological toxicity was observed, the dose was reduced from 60 to 50 mg, 50 to 40 mg, or temporary interruption of S-1 administration was recommended. Patients whose toxic effects necessitated a rest period of >4 weeks were withdrawn from treatment. When initial dose was 40, 50, or 60 mg, dose escalation could be allowed to 50, 60, and 75 mg for subsequent cycles, unless adverse events were observed.", "The tumor response was assessed according to the guidelines of RECIST. Target lesions included all measurable lesions up to a maximum of five lesions per organ and 10 lesions in total. Target lesions were included the lesions with previously irradiated area. Complete response (CR) was defined as the complete disappearance of all target and nontarget lesions, with no development of new disease. Partial response (PR) was defined as a reduction by ≥30% in the sum of the longest diameter of target lesions. CRs or PRs were confirmed by repeat assessments carried out no <4 weeks after the criteria for response were first met. Progressive disease (PD) was defined as an increase ≥20% in the sum of the longest diameter of all target lesions or the appearance of one or more new lesions and/or unequivocal progression of existing nontarget lesions. Stable disease (SD) was defined as neither sufficient lesion shrinkage to qualify for PR nor sufficient increase to qualify for PD. Best response was defined as the most CR achieved by a patient (thus, each patient had a single best response: CR, PR, SD, or PD), and the date of best response was the date it was first detected. Radiological studies were repeated every two cycles. If a patient was documented as having a CR or a PR, the response was confirmed at least 4 weeks after the first evidence of response. An independent response review committee (IRRC) evaluated all tumor responses after the investigators had completed their judgment.\nToxic effects were evaluated with respect to incidence and severity using Common Terminology Criteria of Adverse Events (version 3.0) (www.cancer.gov/).", "The primary end point of this study was to assess the overall response rate determined by the IRRC. The secondary end points were to assess duration of response, time to response, time to progression (TTP), overall survival, and adverse events. Assuming a response rate of 20%, the study was designed with 80% power such that the lower limit of the 95% confidence interval (CI) for the estimate of the response rate was >0.05. A sample size of 32 assessable patients was required. The Kaplan–Meier method was used to determine the TTP and median survival time (MST) in the assessable population. TTP was defined as the time from the first medication to the date of a PD event or death (due to cervical cancer or study drugs).", "[SUBTITLE] patient population [SUBSECTION] A total of 37 patients were entered into the study from July 2005 to September 2007 and 36 patients were eligible and assessable. One patient had a lack of absolute neutrophil count for eligibility criteria. All 37 patients were evaluated for safety. Patient characteristics are listed in Table 1. More than half of the patients had distant diseases. Seventeen patients (8 for neoadjuvant chemotherapy, 3 for metastatic disease, and 6 for both) received prior chemotherapy (not including chemoradiotherapy): 14 received platinum-containing regimen and 3 received oral 5-FU derivative drug alone. Thirteen patients (36%) received chemoradiotherapy. Prior platinum therapy including chemotherapy or chemoradiotherapy was administered for 22 patients.\nPatient characteristics\nNot included chemoradiotherapy.\nA total of 167 treatment cycles (median 4, range 1–19) were administered. Nineteen patients (53%) were subjected to dose reduction owing to adverse events. The median relative dose intensity was 0.83 (range 0.45–1.04).\nA total of 37 patients were entered into the study from July 2005 to September 2007 and 36 patients were eligible and assessable. One patient had a lack of absolute neutrophil count for eligibility criteria. All 37 patients were evaluated for safety. Patient characteristics are listed in Table 1. More than half of the patients had distant diseases. Seventeen patients (8 for neoadjuvant chemotherapy, 3 for metastatic disease, and 6 for both) received prior chemotherapy (not including chemoradiotherapy): 14 received platinum-containing regimen and 3 received oral 5-FU derivative drug alone. Thirteen patients (36%) received chemoradiotherapy. Prior platinum therapy including chemotherapy or chemoradiotherapy was administered for 22 patients.\nPatient characteristics\nNot included chemoradiotherapy.\nA total of 167 treatment cycles (median 4, range 1–19) were administered. Nineteen patients (53%) were subjected to dose reduction owing to adverse events. The median relative dose intensity was 0.83 (range 0.45–1.04).\n[SUBTITLE] antitumor activity [SUBSECTION] Table 2 describes the response assessment. The objective response rate assessed by IRRC was 30.6% (95% CI 15.5% to 45.6%). The median duration of response was 134 days (range 73–553 days). The investigators identified one CR and nine PRs. One clinical responded patient who had CR was downgraded to PR, two clinical responded patients who had PR were downgraded to SD and PD, respectively, and three patients who had SD were upgraded to PR by the judgment of IRRC. Therefore, a total of 11 patients were judged PR. Responses according to prior therapy are listed in Table 2. Patients who received chemotherapy alone had a response of 17.6%, patients who received chemoradiotherapy 53.8%, and patients who received platinum-containing chemotherapy or chemoradiotherapy 31.8%. Eighteen patients had target lesions with previously irradiated area and five (27.8%) of them were responded.\nResponses to S-1 according to the patient characteristics\nCR, complete response; PR, partial response; SD, stable disease; PD, Progressive disease; CI, confidential interval.\nAfter a median follow-up duration of 25 months, the median TTP was 5.2 months (95% CI 4.5–6.6 months; Figure 1) and the MST was 15.4 months (95% CI 11.5–17.8 months; Figure 2). One-year survival was 58.3%.\nKaplan–Meier plot for time to progression (TTP; n = 36). CI, confidence interval.\nKaplan–Meier plot for overall survival (n = 36). CI, confidence interval.\nTable 2 describes the response assessment. The objective response rate assessed by IRRC was 30.6% (95% CI 15.5% to 45.6%). The median duration of response was 134 days (range 73–553 days). The investigators identified one CR and nine PRs. One clinical responded patient who had CR was downgraded to PR, two clinical responded patients who had PR were downgraded to SD and PD, respectively, and three patients who had SD were upgraded to PR by the judgment of IRRC. Therefore, a total of 11 patients were judged PR. Responses according to prior therapy are listed in Table 2. Patients who received chemotherapy alone had a response of 17.6%, patients who received chemoradiotherapy 53.8%, and patients who received platinum-containing chemotherapy or chemoradiotherapy 31.8%. Eighteen patients had target lesions with previously irradiated area and five (27.8%) of them were responded.\nResponses to S-1 according to the patient characteristics\nCR, complete response; PR, partial response; SD, stable disease; PD, Progressive disease; CI, confidential interval.\nAfter a median follow-up duration of 25 months, the median TTP was 5.2 months (95% CI 4.5–6.6 months; Figure 1) and the MST was 15.4 months (95% CI 11.5–17.8 months; Figure 2). One-year survival was 58.3%.\nKaplan–Meier plot for time to progression (TTP; n = 36). CI, confidence interval.\nKaplan–Meier plot for overall survival (n = 36). CI, confidence interval.\n[SUBTITLE] safety [SUBSECTION] All 37 patients were assessed for safety. Four patients were discontinued due to toxic effects. Adverse events are listed in Table 3. Grade 3 or 4 hematologic toxic effects were anemia (16%), neutropenia (8%), and thrombocytopenia (5%). Among grade 3 or 4 nonhematologic toxic effects, the most frequent were anorexia (16%) and diarrhea (22%). All other grade 3 or 4 toxic effects were recorded in <10% of patients.\nAdverse events (n = 37)\nAll 37 patients were assessed for safety. Four patients were discontinued due to toxic effects. Adverse events are listed in Table 3. Grade 3 or 4 hematologic toxic effects were anemia (16%), neutropenia (8%), and thrombocytopenia (5%). Among grade 3 or 4 nonhematologic toxic effects, the most frequent were anorexia (16%) and diarrhea (22%). All other grade 3 or 4 toxic effects were recorded in <10% of patients.\nAdverse events (n = 37)", "A total of 37 patients were entered into the study from July 2005 to September 2007 and 36 patients were eligible and assessable. One patient had a lack of absolute neutrophil count for eligibility criteria. All 37 patients were evaluated for safety. Patient characteristics are listed in Table 1. More than half of the patients had distant diseases. Seventeen patients (8 for neoadjuvant chemotherapy, 3 for metastatic disease, and 6 for both) received prior chemotherapy (not including chemoradiotherapy): 14 received platinum-containing regimen and 3 received oral 5-FU derivative drug alone. Thirteen patients (36%) received chemoradiotherapy. Prior platinum therapy including chemotherapy or chemoradiotherapy was administered for 22 patients.\nPatient characteristics\nNot included chemoradiotherapy.\nA total of 167 treatment cycles (median 4, range 1–19) were administered. Nineteen patients (53%) were subjected to dose reduction owing to adverse events. The median relative dose intensity was 0.83 (range 0.45–1.04).", "Table 2 describes the response assessment. The objective response rate assessed by IRRC was 30.6% (95% CI 15.5% to 45.6%). The median duration of response was 134 days (range 73–553 days). The investigators identified one CR and nine PRs. One clinical responded patient who had CR was downgraded to PR, two clinical responded patients who had PR were downgraded to SD and PD, respectively, and three patients who had SD were upgraded to PR by the judgment of IRRC. Therefore, a total of 11 patients were judged PR. Responses according to prior therapy are listed in Table 2. Patients who received chemotherapy alone had a response of 17.6%, patients who received chemoradiotherapy 53.8%, and patients who received platinum-containing chemotherapy or chemoradiotherapy 31.8%. Eighteen patients had target lesions with previously irradiated area and five (27.8%) of them were responded.\nResponses to S-1 according to the patient characteristics\nCR, complete response; PR, partial response; SD, stable disease; PD, Progressive disease; CI, confidential interval.\nAfter a median follow-up duration of 25 months, the median TTP was 5.2 months (95% CI 4.5–6.6 months; Figure 1) and the MST was 15.4 months (95% CI 11.5–17.8 months; Figure 2). One-year survival was 58.3%.\nKaplan–Meier plot for time to progression (TTP; n = 36). CI, confidence interval.\nKaplan–Meier plot for overall survival (n = 36). CI, confidence interval.", "All 37 patients were assessed for safety. Four patients were discontinued due to toxic effects. Adverse events are listed in Table 3. Grade 3 or 4 hematologic toxic effects were anemia (16%), neutropenia (8%), and thrombocytopenia (5%). Among grade 3 or 4 nonhematologic toxic effects, the most frequent were anorexia (16%) and diarrhea (22%). All other grade 3 or 4 toxic effects were recorded in <10% of patients.\nAdverse events (n = 37)", "The prognosis of patients with advanced or recurrent cervical cancer remains poor and there is an urgent need for novel therapeutic agents. This current study was designed to determine the efficacy and tolerability of an oral agent of S-1 for advanced or recurrent cervical cancer and demonstrated a higher response rate of 30.6% with modest toxic effects: grade 3 or 4 anemia (16%), anorexia (16%), and diarrhea (22%).\nThe most extensively studied agent in the treatment of advanced cervical cancer is cisplatin, which has been used as a single agent, in combination chemotherapy, or with radiotherapy. The eligibility criteria of our study included patients with prior chemotherapy or chemoradiotherapy. Twenty-two of the 36 patients (61%) had previously received platinum therapy including chemoradiotherapy. There may be drug resistance to cisplatin in such patients; however, objective responses were seen in patients who had received prior platinum therapy. Therefore, it is suggested that S-1 is a noncross resistant drug for cisplatin.\nSeveral non-platinum agents, such as paclitaxel [11–13], topotecan [14, 15], irinotecan [16, 17], vinorelbine [18–20], capecitabine [21, 22], and ifosphamide [23–25] were found to have moderate activity in patients with metastatic cervical cancer. However, none of the previously reported phase II studies of non-platinum single-agent chemotherapy for patients with advanced cervical cancer have reported >30% response rate, except paclitaxel and ifosphamide [26]. Paclitaxel is an active agent for cervical cancer and has been evaluated in randomized trial. GOG 0204 compared doublets of paclitaxel, vinorelbine, and gemcitabine plus cisplatin with the combination of topotecan plus cisplatin, and there was a trend favoring treatment with cisplatin/paclitaxel for response rate, progression-free survival (PFS), overall survival, and quality of life [27]. Ifosphamide in combination with cisplatin was tested in randomized trial comparing cisplatin alone and showed a better response rate and PFS but not overall survival and including severe toxic effects. Although our study examined a small number of patients and the CI was wide, notable objective responses were achieved in this single-agent chemotherapy.\nCombinations of 5-FU and cisplatin yield synergistic in preclinical studies [28, 29]. A combination therapy of S-1 and cisplatin has been studied in other malignancies, including gastric cancer, lung cancer, and head and neck cancer [30–32]. Phase III trial comparing S-1 in combination with cisplatin versus S-1 alone in advanced gastric cancer demonstrated a significant benefit for combined S-1 plus cisplatin in response rate, PFS, and overall survival [33]. Based on the promising activity of S-1 in the present phase II study, and the experience with S-1 plus cisplatin in other malignancies, we have started phase III trial of S-1 plus cisplatin compared with single-agent cisplatin for metastatic cervical cancer in an Asian trial, including Japan, Korea, and Taiwan.\nIn conclusion, S-1 is active in patients with metastatic cervical cancer and well tolerated. S-1 plus cisplatin has now entered a prospective randomized phase III trial.", "Taiho Pharmaceutical, Tokyo, Japan.", "Dr Kamiura has reported honoraria for Taiho Pharmaceutical. Dr Ochiai has reported consultant or advisory role for Taiho Pharmaceutical, and honoraria for Taiho Pharmaceutical, Bristol Myers Squibb, and Sanofi Aventis; he has received research support from Taiho Pharmaceutical, and Bristol Myers Squibb. The other authors have not reported any conflicts of interest." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null ]

[ "cervical cancer", "chemotherapy", "phase II trial", "relapse", "S-1" ]

[ "Adiponectin", "Adipose Tissue", "Adult", "Asian People", "Biomarkers", "Body Fat Distribution", "Body Mass Index", "Female", "Humans", "Leptin", "Male", "Metabolic Syndrome", "Middle Aged", "Molecular Weight", "Receptors, Leptin", "Risk Factors", "Solubility" ]

[ "Introduction", "Study design and subjects", "Laboratory measurements", "Ascertainment of metabolic syndrome", "Statistical analysis", "Results", "Risk of metabolic syndrome", "Stratified analyses", "Discussion", "HMW-adiponectin and metabolic syndrome", "Leptin, soluble leptin receptor and metabolic syndrome", "Strengths and Limitations" ]

[ "Adipose tissue is an endocrine organ playing a pivotal role in the pathogenesis of metabolic diseases. Several adipose-derived adipokines have been demonstrated as mediators linking obesity to insulin resistance, dyslipidemia and inflammation [1]. Adiponectin, one of the most abundant adipokines in circulation, exhibits insulin-sensitizing, fat-burning and anti-inflammatory properties [2], [3]. Hypoadiponectinmia frequently appeared in Individuals with obesity, metabolic syndrome (MetS) and type 2 diabetes [4], [5]. However, there are at least 3 isomers of adiponectin, i.e., trimer, hexamer and high-molecular-weight (HMW-) multimer [2]. Accumulating evidence suggests that HMW-adiponectin is the most physiologically active form related to glucose tolerance [6], insulin sensitivity [7], central fat distribution and multiple metabolic disorders [8]. Previous studies also indicated that HMW-adiponectin could be a useful marker to evaluate risk of MetS or type 2 diabetes in elderly Japanese [9], Japanese-Americans [10] or Caucasian women [11].\nAnother major and well studied adipokine in blood stream is leptin which could suppress food intake and stimulate energy expenditure [12]. However, rather than leptin deficient, obese persons often have hyperleptinemia specified as “selective leptin resistance” associated with MetS, type 2 diabetes and cardiovascular disease (CVD) [13], [14]. Circulating leptin is either in a free form which could trigger downstream signaling or in a binding form with its soluble receptor (sOB-R). Recently, a large prospective cohort reported a strong inverse association between sOB-R and diabetes, independent of BMI, leptin and adiponectin [15]. However, data regarding the associations of leptin and sOB-R and MetS, the important risk factor of diabetes and CVD, is rather limited and inconsistent [14]–[19].\nAdipokines dysregulation has been considered as one of the major mechanisms mediating adverse effects of excess body fat on metabolic abnormalities. However, most studies so far have used BMI to evaluate adiposity and it remains unclear how fat mass or fat distribution per se influences the associations between the adipokines and metabolic disorders. Moreover, body composition may also vary according to ethnical background. Compared to Caucasians, Asians are more likely to have abdominal obesity under “normal BMI” [20] and prone to type 2 diabetes at lower BMI [21]. Meanwhile, accompanying rapid diet and lifestyle transition, epidemic trend of metabolic diseases has become a major public health problem in China [22], which is estimated to have 92.4 million adults with diabetes and another 148.2 million people with prediabetes [23]. Obviously, understanding the roles of these adipokines and also their modifying factors could be critical for metabolic disease control and prevention in countries like China.\nTherefore, our primary aim was to examine the associations of plasma HMW-adiponectin, leptin and sOB-R with risk of MetS and its components. Meanwhile, we also evaluated how these associations were modified by BMI, inflammatory markers, body fat mass or trunk fat percentage and other established risk factors in 1055 middle-aged Chinese men and women.", "The study population was non-institutionalized residents from The Gut Microbiota and Obesity Study, a case-control study of normal weight (18≤ BMI <24.0 kg/m2) and overweight/obesity (BMI ≥24.0 kg/m2) participants [22] aged from 35 to 54 years living in Shanghai for at least 10 years. Detailed study design and inclusion/exclusion criteria were described elsewhere [24]. Four men with missing values for adipokines were excluded. The final analytical sample comprised 557 overweight/obese and 498 normal-weight subjects. The protocol was approved by the institutional review board of the Institute for nutritional sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. Written informed consents were obtained from all participants.\nInformation of demographics, health status, diet and lifestyles was collected using a standardized questionnaire during home interview. Dietary intake was assessed with a modified food frequency questionnaire used in the National Survey on the Status of Nutrition and Health of the Chinese People in 2002 [25]. Smoking and drinking were defined as “yes” or “no”. Family history of chronic diseases was defined as one of the parents or siblings having CVD, stroke or type 2 diabetes. Educational attainment was categorized according to self-reported school years. Levels of physical activity were calculated as a sum of metabolic equivalent (MET)-minute/week score [26] and then classified as low, moderate and high. Sleeping was assessed by average daily sleep time.\nAll participants had a physical examination after overnight fasting. Body weight, height, waist circumference, blood pressure were measured according to a standardized protocol described elsewhere [27]. A dual-energy X-ray absorptiometry scan (DEXA, QDR-4500, Hologic, Waltham, MA, USA) was conducted in 956 (90.6%) individuals and no significant difference in characteristics was found between those with and without DEXA. Values for total fat mass, trunk fat mass and trunk mass (kg) were obtained by using software built into the scanner (version 11.2.1) and daily quality control was performed using phantom #12447 the same as previous study [28].", "The blood processing procedure and assays for fasting plasma glucose, insulin, triglycerides, HDL cholesterol (HDL-C), high-sensitive C-reactive protein (hsCRP) and IL-6 were described in a previous study [29].\nHMW-adiponectin was assessed using an ELISA kit (Millipore, St. Charles, MO, USA). Leptin and sOB-R were determined also by ELISA (R&D Systems Inc., Minneapolis, MN, USA). The average intra-assay and inter-assay coefficients of variation (CVs) were <10%.", "Metabolic syndrome was defined according to the updated National Cholesterol Education Program Adult Treatment Panel III criteria for Asian-Americans [30], which includes at least three of the following components: 1) waist circumferences ≥90 cm in men or ≥80 cm in women; 2) triglycerides ≥1.7 mmol/L; 3) HDL-C <1.03 mmol/L in men or <1.30 mmol/L in women; 4) blood pressure ≥130/85 mmHg, or current use of anti-hypertensive medications; 5) fasting plasma glucose ≥5.6 mmol/L.", "Fat mass index (FMI) was calculated based upon DEXA data as total fat mass (kg)/height (m)2\n[28]. Trunk fat percentage was the ratio of trunk fat mass (kg) to total trunk mass (kg). Homeostasis model assessment of insulin resistance (HOMA-IR) was computed using updated homeostasis model assessment methods (http://www.dtu.ox.ac.uk/).\nAll continuous variables were log-transformed to improve normality. General linear model was applied for the comparisons between non-MetS and MetS (Table S1). Spearman partial correlation coefficients were estimated after adjustment for age, sex and other covariates (Table 1). Because of gender differences in the distributions of adipokines, sex-specific quartile cut-points were used. Multivariate logistic regression models were used to estimate the odds ratios, adjusting for age (continuous), sex, newly diagnosed diabetes (fasting plasma glucose ≥7.0 mmol/L or 2-h post load plasma glucose ≥11.1 mmol/L), smoke (past/current or not), current alcohol use (yes or not), family history of chronic diseases (yes or not), education (0∼9, 10∼12 or >12 years), physical activity (low, moderate or high), sleep (<7, 7∼9 or ≥9 h/d), total energy intake (kcal/d), BMI or FMI or trunk fat percentage, hsCRP, IL-6 and adipokines (log-transformed, continuous). All statistical analyses were performed using Stata 9.2 (Stata, College Station, TX, USA). Two-sided P<0.05 was considered statistically significant.\nAnalyses were conducted in 956 participants. All coefficients were significant unless indicated “ns”. Adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and fat mass index (FMI).\nAdjusted all abovementioned variables except FMI.\nAnalyses were conducted in 1055 participants.", "Individuals with MetS were older, and had higher BMI, waist circumference, blood pressure, fasting glucose, insulin, HOMA-IR, triglycerides, inflammatory markers (hsCRP and IL-6), fat mass index and trunk fat percentage, but lower HDL-C. Meanwhile, they also exhibited lower concentrations of HMW-adiponectin and sOB-R, but greater leptin compared with those without MetS (Table S1).\nHMW-adiponectin was inversely associated with BMI (Table 1, r = −0.31) and FMI (r = −0.27). Controlling for FMI, HMW-adiponectin was still associated negatively with trunk fat percentage (r = −0.18), waist circumference (r = −0.19), insulin (r = −0.22), HOMA-IR (r = −0.22), triglycerides (r = −0.28) and IL-6 (r = −0.08), while positively with HDL-C (r = 0.29) and sOB-R (r = 0.17). In comparison, sOB-R showed somewhat weaker associations: r = −0.13 with waist circumference, r = −0.09 with triglycerides, r = 0.17 with HDL-C and no relation to trunk fat percentage (r = −0.05).\nIn contrast, leptin was strongly correlated with BMI (r = 0.66) and FMI (r = 0.75). After adjustment for FMI, only the correlations with insulin (r = 0.13), HOMA-IR (r = 0.13), triglycerides (r = 0.09), remained statistically significant.\n[SUBTITLE] Risk of metabolic syndrome [SUBSECTION] HMW-adiponectin and sOB-R decreased (Figure 1, A and C) while leptin increased (Figure 1B) with an increased number of MetS components in both men and women (all P<.0001).\nBlack bars  =  Men; white bars  =  Women. All P values were <.0001.\nIn multivariate logistic regression analyses, both HMW-adiponectin and sOB-R were negatively, whereas leptin was positively, associated with the risk of MetS independent of BMI and inflammatory markers (Table 2, Model 1 and 2). The odds ratios (ORs) comparing the highest with the lowest quartile were 0.34 (95%CI 0.20∼0.58, P\ntrend<.0001) for HMW-adiponectin, 0.68 (95%CI 0.41∼1.13, P\ntrend = 0.05) for sOB-R and 2.60 (95%CI 1.34∼5.05, P\ntrend = .005) for leptin. Further adjustments of leptin and sOB-R showed little impact on the association between HMW-adiponectin and MetS. Similarly, adjustment for sOB-R and HMW-adiponectin did not affect the association between leptin and MetS. However, the significant association between sOB-R and MetS disappeared after HMW-adiponectin was included in the Model 3 (P\ntrend = 0.15). Replacing BMI with FMI (Model 4) did not substantially change the significant associations for both HMW-adiponectin and sOB-R (ORs in the highest quartile were 0.30, 95%CI 0.18∼0.50, P\ntrend<.0001 and 0.61, 95%CI 0.36∼1.02, P\ntrend = 0.02, accordingly); but abolished the significance between leptin and the risk of MetS (P\ntrend = 0.23).\nModel 1: adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and log-transformed BMI.\nModel 2: model 1+ inflammatory factors (hsCRP and IL-6).\nModel 3: model 2+ other adipokines (leptin and sOB-R or HMW-adiponectin).\nModel 4: adjusted for log-transformed FMI instead of BMI in Model 1.\nQuartiles of HMW-adiponectin (µg/mL) were <1.08, 1.08–1.84, 1.84–3.16, >3.16 for men and <1.86, 1.86–3.15, 3.15–5.33, >5.33 for women. Quartiles of leptin (ng/mL) were <1.95, 1.95–3.22, 3.22–5.58, >5.58 for men and <5.92, 5.92–9.64, 9.64–15.21, >15.21 for women. Quartiles of sOB-R (ng/mL) were <15.16, 15.16–18.19, 18.19–22.15, >22.15 for men and <15.54, 15.54–18.45, 18.45–21.82, >21.82 for women.\nData were available for 956 participants.\nWith respect to individual components of MetS (Table 3), HMW-adiponectin was strongly associated with a decreased risk of hypertriglyceridemia (OR in the highest quartile  = 0.26, 95%CI 0.16∼0.42, P\ntrend<.0001) and low HDL-C (OR in the highest quartile  = 0.22, 95%CI 0.14∼0.35, P\ntrend<.0001), while marginally associated with central obesity (OR in the highest quartile  = 0.49, 95%CI 0.25∼0.97, P\ntrend = 0.06) in the FMI-adjusted model. Furthermore, trunk fat percentage adjustment did not affect these associations substantially (data not shown). Meanwhile, sOB-R was negatively associated with abdominal fat (OR in the highest quartile  = 0.36, 95%CI 0.18∼0.73, P\ntrend = 0.002), high triglycerides (OR in the highest quartile  = 0.64, 95%CI 0.40∼1.01, P\ntrend = 0.02) and low HDL-C (OR in the highest quartile  = 0.47, 95%CI 0.31∼0.73, P\ntrend<.0001). In contrast, leptin was only positively associated with hypertriglyceridemia after adjustment for FMI (OR in the highest quartile  = 2.90, 95%CI 1.46∼5.77, P\ntrend = 0.006).\nNumber of cases: central obesity (496), hyperglycemia (616), elevated blood pressure (395), hypertriglyceridemia (305), reduced HDL-C (331).\nAdjusted for the same variables as Model 4 in Table 2, including FMI. Analyses were conducted in 956 participants.\nHMW-adiponectin and sOB-R decreased (Figure 1, A and C) while leptin increased (Figure 1B) with an increased number of MetS components in both men and women (all P<.0001).\nBlack bars  =  Men; white bars  =  Women. All P values were <.0001.\nIn multivariate logistic regression analyses, both HMW-adiponectin and sOB-R were negatively, whereas leptin was positively, associated with the risk of MetS independent of BMI and inflammatory markers (Table 2, Model 1 and 2). The odds ratios (ORs) comparing the highest with the lowest quartile were 0.34 (95%CI 0.20∼0.58, P\ntrend<.0001) for HMW-adiponectin, 0.68 (95%CI 0.41∼1.13, P\ntrend = 0.05) for sOB-R and 2.60 (95%CI 1.34∼5.05, P\ntrend = .005) for leptin. Further adjustments of leptin and sOB-R showed little impact on the association between HMW-adiponectin and MetS. Similarly, adjustment for sOB-R and HMW-adiponectin did not affect the association between leptin and MetS. However, the significant association between sOB-R and MetS disappeared after HMW-adiponectin was included in the Model 3 (P\ntrend = 0.15). Replacing BMI with FMI (Model 4) did not substantially change the significant associations for both HMW-adiponectin and sOB-R (ORs in the highest quartile were 0.30, 95%CI 0.18∼0.50, P\ntrend<.0001 and 0.61, 95%CI 0.36∼1.02, P\ntrend = 0.02, accordingly); but abolished the significance between leptin and the risk of MetS (P\ntrend = 0.23).\nModel 1: adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and log-transformed BMI.\nModel 2: model 1+ inflammatory factors (hsCRP and IL-6).\nModel 3: model 2+ other adipokines (leptin and sOB-R or HMW-adiponectin).\nModel 4: adjusted for log-transformed FMI instead of BMI in Model 1.\nQuartiles of HMW-adiponectin (µg/mL) were <1.08, 1.08–1.84, 1.84–3.16, >3.16 for men and <1.86, 1.86–3.15, 3.15–5.33, >5.33 for women. Quartiles of leptin (ng/mL) were <1.95, 1.95–3.22, 3.22–5.58, >5.58 for men and <5.92, 5.92–9.64, 9.64–15.21, >15.21 for women. Quartiles of sOB-R (ng/mL) were <15.16, 15.16–18.19, 18.19–22.15, >22.15 for men and <15.54, 15.54–18.45, 18.45–21.82, >21.82 for women.\nData were available for 956 participants.\nWith respect to individual components of MetS (Table 3), HMW-adiponectin was strongly associated with a decreased risk of hypertriglyceridemia (OR in the highest quartile  = 0.26, 95%CI 0.16∼0.42, P\ntrend<.0001) and low HDL-C (OR in the highest quartile  = 0.22, 95%CI 0.14∼0.35, P\ntrend<.0001), while marginally associated with central obesity (OR in the highest quartile  = 0.49, 95%CI 0.25∼0.97, P\ntrend = 0.06) in the FMI-adjusted model. Furthermore, trunk fat percentage adjustment did not affect these associations substantially (data not shown). Meanwhile, sOB-R was negatively associated with abdominal fat (OR in the highest quartile  = 0.36, 95%CI 0.18∼0.73, P\ntrend = 0.002), high triglycerides (OR in the highest quartile  = 0.64, 95%CI 0.40∼1.01, P\ntrend = 0.02) and low HDL-C (OR in the highest quartile  = 0.47, 95%CI 0.31∼0.73, P\ntrend<.0001). In contrast, leptin was only positively associated with hypertriglyceridemia after adjustment for FMI (OR in the highest quartile  = 2.90, 95%CI 1.46∼5.77, P\ntrend = 0.006).\nNumber of cases: central obesity (496), hyperglycemia (616), elevated blood pressure (395), hypertriglyceridemia (305), reduced HDL-C (331).\nAdjusted for the same variables as Model 4 in Table 2, including FMI. Analyses were conducted in 956 participants.\n[SUBTITLE] Stratified analyses [SUBSECTION] Considering the obesity case-control design of this study and to explore how obesity, total fat mass and trunk fat percentage influenced the observed relationships, we conducted BMI-stratified analyses. In both normal-weight and overweight group, HMW-adiponectin showed strong inverse associations with modified MetS, regardless whether BMI, FMI or trunk fat percentage was adjusted (Table S2, Model 1 to 3). However, leptin was not significantly associated with modified MetS under control of total or abdominal adiposity. A negative association between sOB-R and modified MetS was observed in normal weight individuals only (Model 1), and the significant association disappeared following adjustments of FMI or trunk fat percentage (Model 2 and 3). The results remained essentially the same and no significant interaction was found in consequent subgroup analyses according to gender, FMI, hsCRP and HOMA-IR (Table S3).\nConsidering the obesity case-control design of this study and to explore how obesity, total fat mass and trunk fat percentage influenced the observed relationships, we conducted BMI-stratified analyses. In both normal-weight and overweight group, HMW-adiponectin showed strong inverse associations with modified MetS, regardless whether BMI, FMI or trunk fat percentage was adjusted (Table S2, Model 1 to 3). However, leptin was not significantly associated with modified MetS under control of total or abdominal adiposity. A negative association between sOB-R and modified MetS was observed in normal weight individuals only (Model 1), and the significant association disappeared following adjustments of FMI or trunk fat percentage (Model 2 and 3). The results remained essentially the same and no significant interaction was found in consequent subgroup analyses according to gender, FMI, hsCRP and HOMA-IR (Table S3).", "HMW-adiponectin and sOB-R decreased (Figure 1, A and C) while leptin increased (Figure 1B) with an increased number of MetS components in both men and women (all P<.0001).\nBlack bars  =  Men; white bars  =  Women. All P values were <.0001.\nIn multivariate logistic regression analyses, both HMW-adiponectin and sOB-R were negatively, whereas leptin was positively, associated with the risk of MetS independent of BMI and inflammatory markers (Table 2, Model 1 and 2). The odds ratios (ORs) comparing the highest with the lowest quartile were 0.34 (95%CI 0.20∼0.58, P\ntrend<.0001) for HMW-adiponectin, 0.68 (95%CI 0.41∼1.13, P\ntrend = 0.05) for sOB-R and 2.60 (95%CI 1.34∼5.05, P\ntrend = .005) for leptin. Further adjustments of leptin and sOB-R showed little impact on the association between HMW-adiponectin and MetS. Similarly, adjustment for sOB-R and HMW-adiponectin did not affect the association between leptin and MetS. However, the significant association between sOB-R and MetS disappeared after HMW-adiponectin was included in the Model 3 (P\ntrend = 0.15). Replacing BMI with FMI (Model 4) did not substantially change the significant associations for both HMW-adiponectin and sOB-R (ORs in the highest quartile were 0.30, 95%CI 0.18∼0.50, P\ntrend<.0001 and 0.61, 95%CI 0.36∼1.02, P\ntrend = 0.02, accordingly); but abolished the significance between leptin and the risk of MetS (P\ntrend = 0.23).\nModel 1: adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and log-transformed BMI.\nModel 2: model 1+ inflammatory factors (hsCRP and IL-6).\nModel 3: model 2+ other adipokines (leptin and sOB-R or HMW-adiponectin).\nModel 4: adjusted for log-transformed FMI instead of BMI in Model 1.\nQuartiles of HMW-adiponectin (µg/mL) were <1.08, 1.08–1.84, 1.84–3.16, >3.16 for men and <1.86, 1.86–3.15, 3.15–5.33, >5.33 for women. Quartiles of leptin (ng/mL) were <1.95, 1.95–3.22, 3.22–5.58, >5.58 for men and <5.92, 5.92–9.64, 9.64–15.21, >15.21 for women. Quartiles of sOB-R (ng/mL) were <15.16, 15.16–18.19, 18.19–22.15, >22.15 for men and <15.54, 15.54–18.45, 18.45–21.82, >21.82 for women.\nData were available for 956 participants.\nWith respect to individual components of MetS (Table 3), HMW-adiponectin was strongly associated with a decreased risk of hypertriglyceridemia (OR in the highest quartile  = 0.26, 95%CI 0.16∼0.42, P\ntrend<.0001) and low HDL-C (OR in the highest quartile  = 0.22, 95%CI 0.14∼0.35, P\ntrend<.0001), while marginally associated with central obesity (OR in the highest quartile  = 0.49, 95%CI 0.25∼0.97, P\ntrend = 0.06) in the FMI-adjusted model. Furthermore, trunk fat percentage adjustment did not affect these associations substantially (data not shown). Meanwhile, sOB-R was negatively associated with abdominal fat (OR in the highest quartile  = 0.36, 95%CI 0.18∼0.73, P\ntrend = 0.002), high triglycerides (OR in the highest quartile  = 0.64, 95%CI 0.40∼1.01, P\ntrend = 0.02) and low HDL-C (OR in the highest quartile  = 0.47, 95%CI 0.31∼0.73, P\ntrend<.0001). In contrast, leptin was only positively associated with hypertriglyceridemia after adjustment for FMI (OR in the highest quartile  = 2.90, 95%CI 1.46∼5.77, P\ntrend = 0.006).\nNumber of cases: central obesity (496), hyperglycemia (616), elevated blood pressure (395), hypertriglyceridemia (305), reduced HDL-C (331).\nAdjusted for the same variables as Model 4 in Table 2, including FMI. Analyses were conducted in 956 participants.", "Considering the obesity case-control design of this study and to explore how obesity, total fat mass and trunk fat percentage influenced the observed relationships, we conducted BMI-stratified analyses. In both normal-weight and overweight group, HMW-adiponectin showed strong inverse associations with modified MetS, regardless whether BMI, FMI or trunk fat percentage was adjusted (Table S2, Model 1 to 3). However, leptin was not significantly associated with modified MetS under control of total or abdominal adiposity. A negative association between sOB-R and modified MetS was observed in normal weight individuals only (Model 1), and the significant association disappeared following adjustments of FMI or trunk fat percentage (Model 2 and 3). The results remained essentially the same and no significant interaction was found in consequent subgroup analyses according to gender, FMI, hsCRP and HOMA-IR (Table S3).", "Among 1055 middle-aged Chinese men and women, we observed that reduced plasma HMW-adiponectin and sOB-R, and elevated leptin level were significantly associated with an increased risk of MetS and some of its components independent of multiple confounders including BMI and inflammatory markers. However, body fat mass or adipokines showed different modifying effects on these associations. Unlike the association between HMW-adiponectin and MetS which was unaffected by adjusting for fat depots or other adipokines, the positive associations between leptin and MetS were mainly explained by total fat mass, while the associations between sOB-R and MetS were largely influenced by HMW-adiponectin.\n[SUBTITLE] HMW-adiponectin and metabolic syndrome [SUBSECTION] Our study indicated that low plasma HMW-adiponectin was a strong and independent risk factor related to MetS in Chinese when diet, lifestyles, adiposity, inflammatory factors leptin and sOB-R were extensively controlled. The inverse associations of adiponectin with metabolic diseases and type 2 diabetes have been well established [4], [5]. However, most studies only measured total adiponectin rather than its high-molecular-weight multimer (12–18mer), which may be more biologically active than trimer and hexamer [6], [7], [8]. Evidence regarding the associations between HMW-adiponectin and MetS was sparse and most of them were limited by small sample size and residual confounding. In one relatively large-scale study, Tabar et al. [9] reported inverse associations between HMW-adiponectin and MetS and its components, except high blood pressure in middle-aged to elderly Japanese. Likewise, our study also supported such associations in Chinese population. On the other hand, we found that low HMW-adiponectin concentration was only significantly associated with high triglyceride and low HDL-C, but not with elevated glucose and central obesity. This minor discrepancy could be due to the different definitions for MetS between two studies. It is also plausible that more confounders were controlled in our analyses.\nOne interesting observation in current study is that associations between HMW-adiponectin and MetS were independent of adiposity measured by BMI, FMI, or trunk fat percentage. In the Nurses' Health Study, Heidemann et al. reported that higher HMW-adiponectin was associated with lower insulin and diabetes risk independent of BMI or waist circumference [11]. While our data demonstrated that significant negative correlations between HMW-adiponectin and insulin or HOMA-IR (both r = −0.22) were independent of FMI. Moreover, adjustment for IL-6 and hsCRP had little effect on the association between HMW-adiponectin and MetS (Table 2). The mechanism underlining adiponectin regulation and signaling pathway is rather complex and non-adipose factors might also be involved. Besides fat cells, adiponectin could be secreted by skeletal muscle, cardiac myocytes and endothelial cells as well [2]. Moreover, two receptors, namely AdipoR1 and AdipoR2, operate closely with AMPK or PPAR-α pathway to enhance glucose uptake and utilization in muscle, promote lipid oxidation in liver, and improve systemic insulin sensitivity [2], [31]. In addition, insulin resistance and inflammatory factors, hallmarks of MetS, are proposed to down regulate adiponectin production [2]. Previous small-scale infusion studies suggested that correlations of adiponectin with insulin sensitivity were independent of BMI [32], total adiposity measured by DEXA, visceral fat measured by magnetic resonance imaging and intramyocellular lipid measured by 1H-magnetic resonance spectroscopy [33]. Indeed, both enlarged size and increased number of subcutaneous adipocytes could lead to accumulation of fat in non-adipose tissue, such as skeletal muscle, liver and heart. These ectopic fats are closely related to hypoadiponectinemia and might modify its link with insulin resistance [34]. Due to the limitation of DEXA method in current study, it is not possible to exclude the effects of visceral adiposity, intramyocellular or intrahepatic lipid content or adipocytes size by adjustment of FMI, although BMI were already partitioned by fat mass and fat free mass,respectively [35]. Collectively, data from our study and others supported that HMW-adiponectin was an important biomarker for MetS in addition to many established risk factors.\nOur study indicated that low plasma HMW-adiponectin was a strong and independent risk factor related to MetS in Chinese when diet, lifestyles, adiposity, inflammatory factors leptin and sOB-R were extensively controlled. The inverse associations of adiponectin with metabolic diseases and type 2 diabetes have been well established [4], [5]. However, most studies only measured total adiponectin rather than its high-molecular-weight multimer (12–18mer), which may be more biologically active than trimer and hexamer [6], [7], [8]. Evidence regarding the associations between HMW-adiponectin and MetS was sparse and most of them were limited by small sample size and residual confounding. In one relatively large-scale study, Tabar et al. [9] reported inverse associations between HMW-adiponectin and MetS and its components, except high blood pressure in middle-aged to elderly Japanese. Likewise, our study also supported such associations in Chinese population. On the other hand, we found that low HMW-adiponectin concentration was only significantly associated with high triglyceride and low HDL-C, but not with elevated glucose and central obesity. This minor discrepancy could be due to the different definitions for MetS between two studies. It is also plausible that more confounders were controlled in our analyses.\nOne interesting observation in current study is that associations between HMW-adiponectin and MetS were independent of adiposity measured by BMI, FMI, or trunk fat percentage. In the Nurses' Health Study, Heidemann et al. reported that higher HMW-adiponectin was associated with lower insulin and diabetes risk independent of BMI or waist circumference [11]. While our data demonstrated that significant negative correlations between HMW-adiponectin and insulin or HOMA-IR (both r = −0.22) were independent of FMI. Moreover, adjustment for IL-6 and hsCRP had little effect on the association between HMW-adiponectin and MetS (Table 2). The mechanism underlining adiponectin regulation and signaling pathway is rather complex and non-adipose factors might also be involved. Besides fat cells, adiponectin could be secreted by skeletal muscle, cardiac myocytes and endothelial cells as well [2]. Moreover, two receptors, namely AdipoR1 and AdipoR2, operate closely with AMPK or PPAR-α pathway to enhance glucose uptake and utilization in muscle, promote lipid oxidation in liver, and improve systemic insulin sensitivity [2], [31]. In addition, insulin resistance and inflammatory factors, hallmarks of MetS, are proposed to down regulate adiponectin production [2]. Previous small-scale infusion studies suggested that correlations of adiponectin with insulin sensitivity were independent of BMI [32], total adiposity measured by DEXA, visceral fat measured by magnetic resonance imaging and intramyocellular lipid measured by 1H-magnetic resonance spectroscopy [33]. Indeed, both enlarged size and increased number of subcutaneous adipocytes could lead to accumulation of fat in non-adipose tissue, such as skeletal muscle, liver and heart. These ectopic fats are closely related to hypoadiponectinemia and might modify its link with insulin resistance [34]. Due to the limitation of DEXA method in current study, it is not possible to exclude the effects of visceral adiposity, intramyocellular or intrahepatic lipid content or adipocytes size by adjustment of FMI, although BMI were already partitioned by fat mass and fat free mass,respectively [35]. Collectively, data from our study and others supported that HMW-adiponectin was an important biomarker for MetS in addition to many established risk factors.\n[SUBTITLE] Leptin, soluble leptin receptor and metabolic syndrome [SUBSECTION] In this study, we also found strong associations between leptin and risk of MetS with adjusting BMI, inflammation and several known confounders (Table 2, Model 1 to 3). Interestingly, in contrast to the case of HMW-adiponectin, these associations could be diminished by adjusting FMI (Table 2, Model 4 and Table 3). Previously, data from a cohort study in Caucasian population suggested that leptin could predict future MetS independent of baseline BMI [14]. While leptin was associated with MetS risk in older Chinese women, but not in men after adjustment for BMI [36]. However, it was noteworthy that most existing studies utilized BMI to evaluate adiposity status, containing both fat and fat-free mass, which have different influences on metabolic disorders [28]. As an adipose-derived hormone, leptin plays a critical role in regulating energy homeostasis [13]. However, it is still not clear whether leptin resistance is one of the causes or the consequences of obesity, or the “vicious cycle” of them might be the culprit of metabolic disorders. In this study, we provide direct evidence highlighting the key role of fat mass in hyperleptinemia associated metabolic abnormalities. Our finding might be particularly important for Chinese who are characterized to have more body fat under given BMI [20], [21].\nIn addition to hyperleptinemia, our data revealed that reduced soluble leptin receptor (sOB-R), another component of leptin resistance, was significantly associated with an increased MetS risk independent of fat mass. sOB-R, formed by cleaving the ectodomain of membrane-anchored leptin receptors, represents the major leptin binding fraction in blood [37]. It regulates the available leptin pool and downstream signaling [38]. Few epidemiological studies have simultaneously investigated the associations of both plasma leptin and sOB-R with MetS. In a study from Framingham third generation participants (n = 362), leptin and free leptin index (molar ratio of leptin and sOB-R) were positively associated with severity of MetS after adjustment for age and sex [16]. With larger sample size, we further adjusted for diet, lifestyle, BMI/FMI, inflammatory markers and adipokines. However, our data indicated that leptin and sOB-R had unique properties and might reflect the different aspects of MetS. For instance, sOB-R was negatively associated with central obesity, elevated triglyceride and reduced HDL-C, whereas leptin was only associated with hypertriglyceridemia in multivariate regression including FMI (Table 3). Given the wide distribution in tissues containing membrane leptin receptors, it is reasonable to speculate that non-adipose mechanism(s) such as tissue specific effect might also play a role in sOB-R regulation. On the other hand, leptin was almost exclusively secreted by adipocytes, particularly subcutaneous fat. Hypertriglyceridemia might result from the limited capacity of adipose tissue to store excessive fat or up-regulated hepatic triglyceride secretion and down-regulated plasma triglyceride degradation [39], [40]. However, adjustment for HMW-adiponectin could abolish significance between sOB-R and MetS but not that between leptin and MetS (Table 2, Model 3). Indeed, findings from the Nurses' Health Study showed an inverse association between sOB-R and diabetes independent of BMI and HMW-adiponectin [15]. Kim et al. found if leptin-deficient mice have over-expressed adiponectin would result in improvements of glucose and lipid metabolism and inflammatory profile [41]. Obviously, further studies are warranted to clarify whether there are functional and/or signaling crosstalk among adiponectin, leptin and its receptor (s).\nIn this study, we also found strong associations between leptin and risk of MetS with adjusting BMI, inflammation and several known confounders (Table 2, Model 1 to 3). Interestingly, in contrast to the case of HMW-adiponectin, these associations could be diminished by adjusting FMI (Table 2, Model 4 and Table 3). Previously, data from a cohort study in Caucasian population suggested that leptin could predict future MetS independent of baseline BMI [14]. While leptin was associated with MetS risk in older Chinese women, but not in men after adjustment for BMI [36]. However, it was noteworthy that most existing studies utilized BMI to evaluate adiposity status, containing both fat and fat-free mass, which have different influences on metabolic disorders [28]. As an adipose-derived hormone, leptin plays a critical role in regulating energy homeostasis [13]. However, it is still not clear whether leptin resistance is one of the causes or the consequences of obesity, or the “vicious cycle” of them might be the culprit of metabolic disorders. In this study, we provide direct evidence highlighting the key role of fat mass in hyperleptinemia associated metabolic abnormalities. Our finding might be particularly important for Chinese who are characterized to have more body fat under given BMI [20], [21].\nIn addition to hyperleptinemia, our data revealed that reduced soluble leptin receptor (sOB-R), another component of leptin resistance, was significantly associated with an increased MetS risk independent of fat mass. sOB-R, formed by cleaving the ectodomain of membrane-anchored leptin receptors, represents the major leptin binding fraction in blood [37]. It regulates the available leptin pool and downstream signaling [38]. Few epidemiological studies have simultaneously investigated the associations of both plasma leptin and sOB-R with MetS. In a study from Framingham third generation participants (n = 362), leptin and free leptin index (molar ratio of leptin and sOB-R) were positively associated with severity of MetS after adjustment for age and sex [16]. With larger sample size, we further adjusted for diet, lifestyle, BMI/FMI, inflammatory markers and adipokines. However, our data indicated that leptin and sOB-R had unique properties and might reflect the different aspects of MetS. For instance, sOB-R was negatively associated with central obesity, elevated triglyceride and reduced HDL-C, whereas leptin was only associated with hypertriglyceridemia in multivariate regression including FMI (Table 3). Given the wide distribution in tissues containing membrane leptin receptors, it is reasonable to speculate that non-adipose mechanism(s) such as tissue specific effect might also play a role in sOB-R regulation. On the other hand, leptin was almost exclusively secreted by adipocytes, particularly subcutaneous fat. Hypertriglyceridemia might result from the limited capacity of adipose tissue to store excessive fat or up-regulated hepatic triglyceride secretion and down-regulated plasma triglyceride degradation [39], [40]. However, adjustment for HMW-adiponectin could abolish significance between sOB-R and MetS but not that between leptin and MetS (Table 2, Model 3). Indeed, findings from the Nurses' Health Study showed an inverse association between sOB-R and diabetes independent of BMI and HMW-adiponectin [15]. Kim et al. found if leptin-deficient mice have over-expressed adiponectin would result in improvements of glucose and lipid metabolism and inflammatory profile [41]. Obviously, further studies are warranted to clarify whether there are functional and/or signaling crosstalk among adiponectin, leptin and its receptor (s).\n[SUBTITLE] Strengths and Limitations [SUBSECTION] This is the first study that systematically investigates the associations of HMW-adiponectin, leptin, sOB-R, body fat mass measured by DEXA, along with a wide range of inflammatory and metabolic parameters with MetS and its features in a relatively large population with both sexes. Meanwhile, we acknowledge some limitations. Because of the cross-sectional nature, no causal relationship could be established. Also, despite of the originally obesity case-control design, we adapted a cross-sectional approach in data analyses in order to enhance statistical power. Nevertheless, similar associations of HMW-adiponectin, leptin with MetS were observed in both normal and overweight/obese individuals as well as in further subgroup analyses. Obviously, our findings need to be confirmed prospectively in different populations.\nIn conclusion, we found strong inverse associations between HMW-adiponectin and MetS independent of adiposity, inflammatory statuses, leptin and sOB-R. Similar associations between sOB-R and MetS were also evidenced, but were weak and attenuated by HMW-adiponectin adjustment. In contrast, leptin showed strong positive associations with the risk of MetS which could be mainly explained by body fat mass. Overall, current study provides more mechanistic insights into the linkage of adipose tissue, adipokines and inflammation to metabolic syndrome and also more evidences for detection of decreased HMW-adiponectin and leptin resistance in clinical settings to prevent metabolic disorders and future diabetic and cardiovascular outcomes.\nThis is the first study that systematically investigates the associations of HMW-adiponectin, leptin, sOB-R, body fat mass measured by DEXA, along with a wide range of inflammatory and metabolic parameters with MetS and its features in a relatively large population with both sexes. Meanwhile, we acknowledge some limitations. Because of the cross-sectional nature, no causal relationship could be established. Also, despite of the originally obesity case-control design, we adapted a cross-sectional approach in data analyses in order to enhance statistical power. Nevertheless, similar associations of HMW-adiponectin, leptin with MetS were observed in both normal and overweight/obese individuals as well as in further subgroup analyses. Obviously, our findings need to be confirmed prospectively in different populations.\nIn conclusion, we found strong inverse associations between HMW-adiponectin and MetS independent of adiposity, inflammatory statuses, leptin and sOB-R. Similar associations between sOB-R and MetS were also evidenced, but were weak and attenuated by HMW-adiponectin adjustment. In contrast, leptin showed strong positive associations with the risk of MetS which could be mainly explained by body fat mass. Overall, current study provides more mechanistic insights into the linkage of adipose tissue, adipokines and inflammation to metabolic syndrome and also more evidences for detection of decreased HMW-adiponectin and leptin resistance in clinical settings to prevent metabolic disorders and future diabetic and cardiovascular outcomes.", "Our study indicated that low plasma HMW-adiponectin was a strong and independent risk factor related to MetS in Chinese when diet, lifestyles, adiposity, inflammatory factors leptin and sOB-R were extensively controlled. The inverse associations of adiponectin with metabolic diseases and type 2 diabetes have been well established [4], [5]. However, most studies only measured total adiponectin rather than its high-molecular-weight multimer (12–18mer), which may be more biologically active than trimer and hexamer [6], [7], [8]. Evidence regarding the associations between HMW-adiponectin and MetS was sparse and most of them were limited by small sample size and residual confounding. In one relatively large-scale study, Tabar et al. [9] reported inverse associations between HMW-adiponectin and MetS and its components, except high blood pressure in middle-aged to elderly Japanese. Likewise, our study also supported such associations in Chinese population. On the other hand, we found that low HMW-adiponectin concentration was only significantly associated with high triglyceride and low HDL-C, but not with elevated glucose and central obesity. This minor discrepancy could be due to the different definitions for MetS between two studies. It is also plausible that more confounders were controlled in our analyses.\nOne interesting observation in current study is that associations between HMW-adiponectin and MetS were independent of adiposity measured by BMI, FMI, or trunk fat percentage. In the Nurses' Health Study, Heidemann et al. reported that higher HMW-adiponectin was associated with lower insulin and diabetes risk independent of BMI or waist circumference [11]. While our data demonstrated that significant negative correlations between HMW-adiponectin and insulin or HOMA-IR (both r = −0.22) were independent of FMI. Moreover, adjustment for IL-6 and hsCRP had little effect on the association between HMW-adiponectin and MetS (Table 2). The mechanism underlining adiponectin regulation and signaling pathway is rather complex and non-adipose factors might also be involved. Besides fat cells, adiponectin could be secreted by skeletal muscle, cardiac myocytes and endothelial cells as well [2]. Moreover, two receptors, namely AdipoR1 and AdipoR2, operate closely with AMPK or PPAR-α pathway to enhance glucose uptake and utilization in muscle, promote lipid oxidation in liver, and improve systemic insulin sensitivity [2], [31]. In addition, insulin resistance and inflammatory factors, hallmarks of MetS, are proposed to down regulate adiponectin production [2]. Previous small-scale infusion studies suggested that correlations of adiponectin with insulin sensitivity were independent of BMI [32], total adiposity measured by DEXA, visceral fat measured by magnetic resonance imaging and intramyocellular lipid measured by 1H-magnetic resonance spectroscopy [33]. Indeed, both enlarged size and increased number of subcutaneous adipocytes could lead to accumulation of fat in non-adipose tissue, such as skeletal muscle, liver and heart. These ectopic fats are closely related to hypoadiponectinemia and might modify its link with insulin resistance [34]. Due to the limitation of DEXA method in current study, it is not possible to exclude the effects of visceral adiposity, intramyocellular or intrahepatic lipid content or adipocytes size by adjustment of FMI, although BMI were already partitioned by fat mass and fat free mass,respectively [35]. Collectively, data from our study and others supported that HMW-adiponectin was an important biomarker for MetS in addition to many established risk factors.", "In this study, we also found strong associations between leptin and risk of MetS with adjusting BMI, inflammation and several known confounders (Table 2, Model 1 to 3). Interestingly, in contrast to the case of HMW-adiponectin, these associations could be diminished by adjusting FMI (Table 2, Model 4 and Table 3). Previously, data from a cohort study in Caucasian population suggested that leptin could predict future MetS independent of baseline BMI [14]. While leptin was associated with MetS risk in older Chinese women, but not in men after adjustment for BMI [36]. However, it was noteworthy that most existing studies utilized BMI to evaluate adiposity status, containing both fat and fat-free mass, which have different influences on metabolic disorders [28]. As an adipose-derived hormone, leptin plays a critical role in regulating energy homeostasis [13]. However, it is still not clear whether leptin resistance is one of the causes or the consequences of obesity, or the “vicious cycle” of them might be the culprit of metabolic disorders. In this study, we provide direct evidence highlighting the key role of fat mass in hyperleptinemia associated metabolic abnormalities. Our finding might be particularly important for Chinese who are characterized to have more body fat under given BMI [20], [21].\nIn addition to hyperleptinemia, our data revealed that reduced soluble leptin receptor (sOB-R), another component of leptin resistance, was significantly associated with an increased MetS risk independent of fat mass. sOB-R, formed by cleaving the ectodomain of membrane-anchored leptin receptors, represents the major leptin binding fraction in blood [37]. It regulates the available leptin pool and downstream signaling [38]. Few epidemiological studies have simultaneously investigated the associations of both plasma leptin and sOB-R with MetS. In a study from Framingham third generation participants (n = 362), leptin and free leptin index (molar ratio of leptin and sOB-R) were positively associated with severity of MetS after adjustment for age and sex [16]. With larger sample size, we further adjusted for diet, lifestyle, BMI/FMI, inflammatory markers and adipokines. However, our data indicated that leptin and sOB-R had unique properties and might reflect the different aspects of MetS. For instance, sOB-R was negatively associated with central obesity, elevated triglyceride and reduced HDL-C, whereas leptin was only associated with hypertriglyceridemia in multivariate regression including FMI (Table 3). Given the wide distribution in tissues containing membrane leptin receptors, it is reasonable to speculate that non-adipose mechanism(s) such as tissue specific effect might also play a role in sOB-R regulation. On the other hand, leptin was almost exclusively secreted by adipocytes, particularly subcutaneous fat. Hypertriglyceridemia might result from the limited capacity of adipose tissue to store excessive fat or up-regulated hepatic triglyceride secretion and down-regulated plasma triglyceride degradation [39], [40]. However, adjustment for HMW-adiponectin could abolish significance between sOB-R and MetS but not that between leptin and MetS (Table 2, Model 3). Indeed, findings from the Nurses' Health Study showed an inverse association between sOB-R and diabetes independent of BMI and HMW-adiponectin [15]. Kim et al. found if leptin-deficient mice have over-expressed adiponectin would result in improvements of glucose and lipid metabolism and inflammatory profile [41]. Obviously, further studies are warranted to clarify whether there are functional and/or signaling crosstalk among adiponectin, leptin and its receptor (s).", "This is the first study that systematically investigates the associations of HMW-adiponectin, leptin, sOB-R, body fat mass measured by DEXA, along with a wide range of inflammatory and metabolic parameters with MetS and its features in a relatively large population with both sexes. Meanwhile, we acknowledge some limitations. Because of the cross-sectional nature, no causal relationship could be established. Also, despite of the originally obesity case-control design, we adapted a cross-sectional approach in data analyses in order to enhance statistical power. Nevertheless, similar associations of HMW-adiponectin, leptin with MetS were observed in both normal and overweight/obese individuals as well as in further subgroup analyses. Obviously, our findings need to be confirmed prospectively in different populations.\nIn conclusion, we found strong inverse associations between HMW-adiponectin and MetS independent of adiposity, inflammatory statuses, leptin and sOB-R. Similar associations between sOB-R and MetS were also evidenced, but were weak and attenuated by HMW-adiponectin adjustment. In contrast, leptin showed strong positive associations with the risk of MetS which could be mainly explained by body fat mass. Overall, current study provides more mechanistic insights into the linkage of adipose tissue, adipokines and inflammation to metabolic syndrome and also more evidences for detection of decreased HMW-adiponectin and leptin resistance in clinical settings to prevent metabolic disorders and future diabetic and cardiovascular outcomes." ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Methods", "Study design and subjects", "Laboratory measurements", "Ascertainment of metabolic syndrome", "Statistical analysis", "Results", "Risk of metabolic syndrome", "Stratified analyses", "Discussion", "HMW-adiponectin and metabolic syndrome", "Leptin, soluble leptin receptor and metabolic syndrome", "Strengths and Limitations", "Supporting Information" ]

[ "Adipose tissue is an endocrine organ playing a pivotal role in the pathogenesis of metabolic diseases. Several adipose-derived adipokines have been demonstrated as mediators linking obesity to insulin resistance, dyslipidemia and inflammation [1]. Adiponectin, one of the most abundant adipokines in circulation, exhibits insulin-sensitizing, fat-burning and anti-inflammatory properties [2], [3]. Hypoadiponectinmia frequently appeared in Individuals with obesity, metabolic syndrome (MetS) and type 2 diabetes [4], [5]. However, there are at least 3 isomers of adiponectin, i.e., trimer, hexamer and high-molecular-weight (HMW-) multimer [2]. Accumulating evidence suggests that HMW-adiponectin is the most physiologically active form related to glucose tolerance [6], insulin sensitivity [7], central fat distribution and multiple metabolic disorders [8]. Previous studies also indicated that HMW-adiponectin could be a useful marker to evaluate risk of MetS or type 2 diabetes in elderly Japanese [9], Japanese-Americans [10] or Caucasian women [11].\nAnother major and well studied adipokine in blood stream is leptin which could suppress food intake and stimulate energy expenditure [12]. However, rather than leptin deficient, obese persons often have hyperleptinemia specified as “selective leptin resistance” associated with MetS, type 2 diabetes and cardiovascular disease (CVD) [13], [14]. Circulating leptin is either in a free form which could trigger downstream signaling or in a binding form with its soluble receptor (sOB-R). Recently, a large prospective cohort reported a strong inverse association between sOB-R and diabetes, independent of BMI, leptin and adiponectin [15]. However, data regarding the associations of leptin and sOB-R and MetS, the important risk factor of diabetes and CVD, is rather limited and inconsistent [14]–[19].\nAdipokines dysregulation has been considered as one of the major mechanisms mediating adverse effects of excess body fat on metabolic abnormalities. However, most studies so far have used BMI to evaluate adiposity and it remains unclear how fat mass or fat distribution per se influences the associations between the adipokines and metabolic disorders. Moreover, body composition may also vary according to ethnical background. Compared to Caucasians, Asians are more likely to have abdominal obesity under “normal BMI” [20] and prone to type 2 diabetes at lower BMI [21]. Meanwhile, accompanying rapid diet and lifestyle transition, epidemic trend of metabolic diseases has become a major public health problem in China [22], which is estimated to have 92.4 million adults with diabetes and another 148.2 million people with prediabetes [23]. Obviously, understanding the roles of these adipokines and also their modifying factors could be critical for metabolic disease control and prevention in countries like China.\nTherefore, our primary aim was to examine the associations of plasma HMW-adiponectin, leptin and sOB-R with risk of MetS and its components. Meanwhile, we also evaluated how these associations were modified by BMI, inflammatory markers, body fat mass or trunk fat percentage and other established risk factors in 1055 middle-aged Chinese men and women.", "[SUBTITLE] Study design and subjects [SUBSECTION] The study population was non-institutionalized residents from The Gut Microbiota and Obesity Study, a case-control study of normal weight (18≤ BMI <24.0 kg/m2) and overweight/obesity (BMI ≥24.0 kg/m2) participants [22] aged from 35 to 54 years living in Shanghai for at least 10 years. Detailed study design and inclusion/exclusion criteria were described elsewhere [24]. Four men with missing values for adipokines were excluded. The final analytical sample comprised 557 overweight/obese and 498 normal-weight subjects. The protocol was approved by the institutional review board of the Institute for nutritional sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. Written informed consents were obtained from all participants.\nInformation of demographics, health status, diet and lifestyles was collected using a standardized questionnaire during home interview. Dietary intake was assessed with a modified food frequency questionnaire used in the National Survey on the Status of Nutrition and Health of the Chinese People in 2002 [25]. Smoking and drinking were defined as “yes” or “no”. Family history of chronic diseases was defined as one of the parents or siblings having CVD, stroke or type 2 diabetes. Educational attainment was categorized according to self-reported school years. Levels of physical activity were calculated as a sum of metabolic equivalent (MET)-minute/week score [26] and then classified as low, moderate and high. Sleeping was assessed by average daily sleep time.\nAll participants had a physical examination after overnight fasting. Body weight, height, waist circumference, blood pressure were measured according to a standardized protocol described elsewhere [27]. A dual-energy X-ray absorptiometry scan (DEXA, QDR-4500, Hologic, Waltham, MA, USA) was conducted in 956 (90.6%) individuals and no significant difference in characteristics was found between those with and without DEXA. Values for total fat mass, trunk fat mass and trunk mass (kg) were obtained by using software built into the scanner (version 11.2.1) and daily quality control was performed using phantom #12447 the same as previous study [28].\nThe study population was non-institutionalized residents from The Gut Microbiota and Obesity Study, a case-control study of normal weight (18≤ BMI <24.0 kg/m2) and overweight/obesity (BMI ≥24.0 kg/m2) participants [22] aged from 35 to 54 years living in Shanghai for at least 10 years. Detailed study design and inclusion/exclusion criteria were described elsewhere [24]. Four men with missing values for adipokines were excluded. The final analytical sample comprised 557 overweight/obese and 498 normal-weight subjects. The protocol was approved by the institutional review board of the Institute for nutritional sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. Written informed consents were obtained from all participants.\nInformation of demographics, health status, diet and lifestyles was collected using a standardized questionnaire during home interview. Dietary intake was assessed with a modified food frequency questionnaire used in the National Survey on the Status of Nutrition and Health of the Chinese People in 2002 [25]. Smoking and drinking were defined as “yes” or “no”. Family history of chronic diseases was defined as one of the parents or siblings having CVD, stroke or type 2 diabetes. Educational attainment was categorized according to self-reported school years. Levels of physical activity were calculated as a sum of metabolic equivalent (MET)-minute/week score [26] and then classified as low, moderate and high. Sleeping was assessed by average daily sleep time.\nAll participants had a physical examination after overnight fasting. Body weight, height, waist circumference, blood pressure were measured according to a standardized protocol described elsewhere [27]. A dual-energy X-ray absorptiometry scan (DEXA, QDR-4500, Hologic, Waltham, MA, USA) was conducted in 956 (90.6%) individuals and no significant difference in characteristics was found between those with and without DEXA. Values for total fat mass, trunk fat mass and trunk mass (kg) were obtained by using software built into the scanner (version 11.2.1) and daily quality control was performed using phantom #12447 the same as previous study [28].\n[SUBTITLE] Laboratory measurements [SUBSECTION] The blood processing procedure and assays for fasting plasma glucose, insulin, triglycerides, HDL cholesterol (HDL-C), high-sensitive C-reactive protein (hsCRP) and IL-6 were described in a previous study [29].\nHMW-adiponectin was assessed using an ELISA kit (Millipore, St. Charles, MO, USA). Leptin and sOB-R were determined also by ELISA (R&D Systems Inc., Minneapolis, MN, USA). The average intra-assay and inter-assay coefficients of variation (CVs) were <10%.\nThe blood processing procedure and assays for fasting plasma glucose, insulin, triglycerides, HDL cholesterol (HDL-C), high-sensitive C-reactive protein (hsCRP) and IL-6 were described in a previous study [29].\nHMW-adiponectin was assessed using an ELISA kit (Millipore, St. Charles, MO, USA). Leptin and sOB-R were determined also by ELISA (R&D Systems Inc., Minneapolis, MN, USA). The average intra-assay and inter-assay coefficients of variation (CVs) were <10%.\n[SUBTITLE] Ascertainment of metabolic syndrome [SUBSECTION] Metabolic syndrome was defined according to the updated National Cholesterol Education Program Adult Treatment Panel III criteria for Asian-Americans [30], which includes at least three of the following components: 1) waist circumferences ≥90 cm in men or ≥80 cm in women; 2) triglycerides ≥1.7 mmol/L; 3) HDL-C <1.03 mmol/L in men or <1.30 mmol/L in women; 4) blood pressure ≥130/85 mmHg, or current use of anti-hypertensive medications; 5) fasting plasma glucose ≥5.6 mmol/L.\nMetabolic syndrome was defined according to the updated National Cholesterol Education Program Adult Treatment Panel III criteria for Asian-Americans [30], which includes at least three of the following components: 1) waist circumferences ≥90 cm in men or ≥80 cm in women; 2) triglycerides ≥1.7 mmol/L; 3) HDL-C <1.03 mmol/L in men or <1.30 mmol/L in women; 4) blood pressure ≥130/85 mmHg, or current use of anti-hypertensive medications; 5) fasting plasma glucose ≥5.6 mmol/L.\n[SUBTITLE] Statistical analysis [SUBSECTION] Fat mass index (FMI) was calculated based upon DEXA data as total fat mass (kg)/height (m)2\n[28]. Trunk fat percentage was the ratio of trunk fat mass (kg) to total trunk mass (kg). Homeostasis model assessment of insulin resistance (HOMA-IR) was computed using updated homeostasis model assessment methods (http://www.dtu.ox.ac.uk/).\nAll continuous variables were log-transformed to improve normality. General linear model was applied for the comparisons between non-MetS and MetS (Table S1). Spearman partial correlation coefficients were estimated after adjustment for age, sex and other covariates (Table 1). Because of gender differences in the distributions of adipokines, sex-specific quartile cut-points were used. Multivariate logistic regression models were used to estimate the odds ratios, adjusting for age (continuous), sex, newly diagnosed diabetes (fasting plasma glucose ≥7.0 mmol/L or 2-h post load plasma glucose ≥11.1 mmol/L), smoke (past/current or not), current alcohol use (yes or not), family history of chronic diseases (yes or not), education (0∼9, 10∼12 or >12 years), physical activity (low, moderate or high), sleep (<7, 7∼9 or ≥9 h/d), total energy intake (kcal/d), BMI or FMI or trunk fat percentage, hsCRP, IL-6 and adipokines (log-transformed, continuous). All statistical analyses were performed using Stata 9.2 (Stata, College Station, TX, USA). Two-sided P<0.05 was considered statistically significant.\nAnalyses were conducted in 956 participants. All coefficients were significant unless indicated “ns”. Adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and fat mass index (FMI).\nAdjusted all abovementioned variables except FMI.\nAnalyses were conducted in 1055 participants.\nFat mass index (FMI) was calculated based upon DEXA data as total fat mass (kg)/height (m)2\n[28]. Trunk fat percentage was the ratio of trunk fat mass (kg) to total trunk mass (kg). Homeostasis model assessment of insulin resistance (HOMA-IR) was computed using updated homeostasis model assessment methods (http://www.dtu.ox.ac.uk/).\nAll continuous variables were log-transformed to improve normality. General linear model was applied for the comparisons between non-MetS and MetS (Table S1). Spearman partial correlation coefficients were estimated after adjustment for age, sex and other covariates (Table 1). Because of gender differences in the distributions of adipokines, sex-specific quartile cut-points were used. Multivariate logistic regression models were used to estimate the odds ratios, adjusting for age (continuous), sex, newly diagnosed diabetes (fasting plasma glucose ≥7.0 mmol/L or 2-h post load plasma glucose ≥11.1 mmol/L), smoke (past/current or not), current alcohol use (yes or not), family history of chronic diseases (yes or not), education (0∼9, 10∼12 or >12 years), physical activity (low, moderate or high), sleep (<7, 7∼9 or ≥9 h/d), total energy intake (kcal/d), BMI or FMI or trunk fat percentage, hsCRP, IL-6 and adipokines (log-transformed, continuous). All statistical analyses were performed using Stata 9.2 (Stata, College Station, TX, USA). Two-sided P<0.05 was considered statistically significant.\nAnalyses were conducted in 956 participants. All coefficients were significant unless indicated “ns”. Adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and fat mass index (FMI).\nAdjusted all abovementioned variables except FMI.\nAnalyses were conducted in 1055 participants.", "The study population was non-institutionalized residents from The Gut Microbiota and Obesity Study, a case-control study of normal weight (18≤ BMI <24.0 kg/m2) and overweight/obesity (BMI ≥24.0 kg/m2) participants [22] aged from 35 to 54 years living in Shanghai for at least 10 years. Detailed study design and inclusion/exclusion criteria were described elsewhere [24]. Four men with missing values for adipokines were excluded. The final analytical sample comprised 557 overweight/obese and 498 normal-weight subjects. The protocol was approved by the institutional review board of the Institute for nutritional sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. Written informed consents were obtained from all participants.\nInformation of demographics, health status, diet and lifestyles was collected using a standardized questionnaire during home interview. Dietary intake was assessed with a modified food frequency questionnaire used in the National Survey on the Status of Nutrition and Health of the Chinese People in 2002 [25]. Smoking and drinking were defined as “yes” or “no”. Family history of chronic diseases was defined as one of the parents or siblings having CVD, stroke or type 2 diabetes. Educational attainment was categorized according to self-reported school years. Levels of physical activity were calculated as a sum of metabolic equivalent (MET)-minute/week score [26] and then classified as low, moderate and high. Sleeping was assessed by average daily sleep time.\nAll participants had a physical examination after overnight fasting. Body weight, height, waist circumference, blood pressure were measured according to a standardized protocol described elsewhere [27]. A dual-energy X-ray absorptiometry scan (DEXA, QDR-4500, Hologic, Waltham, MA, USA) was conducted in 956 (90.6%) individuals and no significant difference in characteristics was found between those with and without DEXA. Values for total fat mass, trunk fat mass and trunk mass (kg) were obtained by using software built into the scanner (version 11.2.1) and daily quality control was performed using phantom #12447 the same as previous study [28].", "The blood processing procedure and assays for fasting plasma glucose, insulin, triglycerides, HDL cholesterol (HDL-C), high-sensitive C-reactive protein (hsCRP) and IL-6 were described in a previous study [29].\nHMW-adiponectin was assessed using an ELISA kit (Millipore, St. Charles, MO, USA). Leptin and sOB-R were determined also by ELISA (R&D Systems Inc., Minneapolis, MN, USA). The average intra-assay and inter-assay coefficients of variation (CVs) were <10%.", "Metabolic syndrome was defined according to the updated National Cholesterol Education Program Adult Treatment Panel III criteria for Asian-Americans [30], which includes at least three of the following components: 1) waist circumferences ≥90 cm in men or ≥80 cm in women; 2) triglycerides ≥1.7 mmol/L; 3) HDL-C <1.03 mmol/L in men or <1.30 mmol/L in women; 4) blood pressure ≥130/85 mmHg, or current use of anti-hypertensive medications; 5) fasting plasma glucose ≥5.6 mmol/L.", "Fat mass index (FMI) was calculated based upon DEXA data as total fat mass (kg)/height (m)2\n[28]. Trunk fat percentage was the ratio of trunk fat mass (kg) to total trunk mass (kg). Homeostasis model assessment of insulin resistance (HOMA-IR) was computed using updated homeostasis model assessment methods (http://www.dtu.ox.ac.uk/).\nAll continuous variables were log-transformed to improve normality. General linear model was applied for the comparisons between non-MetS and MetS (Table S1). Spearman partial correlation coefficients were estimated after adjustment for age, sex and other covariates (Table 1). Because of gender differences in the distributions of adipokines, sex-specific quartile cut-points were used. Multivariate logistic regression models were used to estimate the odds ratios, adjusting for age (continuous), sex, newly diagnosed diabetes (fasting plasma glucose ≥7.0 mmol/L or 2-h post load plasma glucose ≥11.1 mmol/L), smoke (past/current or not), current alcohol use (yes or not), family history of chronic diseases (yes or not), education (0∼9, 10∼12 or >12 years), physical activity (low, moderate or high), sleep (<7, 7∼9 or ≥9 h/d), total energy intake (kcal/d), BMI or FMI or trunk fat percentage, hsCRP, IL-6 and adipokines (log-transformed, continuous). All statistical analyses were performed using Stata 9.2 (Stata, College Station, TX, USA). Two-sided P<0.05 was considered statistically significant.\nAnalyses were conducted in 956 participants. All coefficients were significant unless indicated “ns”. Adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and fat mass index (FMI).\nAdjusted all abovementioned variables except FMI.\nAnalyses were conducted in 1055 participants.", "Individuals with MetS were older, and had higher BMI, waist circumference, blood pressure, fasting glucose, insulin, HOMA-IR, triglycerides, inflammatory markers (hsCRP and IL-6), fat mass index and trunk fat percentage, but lower HDL-C. Meanwhile, they also exhibited lower concentrations of HMW-adiponectin and sOB-R, but greater leptin compared with those without MetS (Table S1).\nHMW-adiponectin was inversely associated with BMI (Table 1, r = −0.31) and FMI (r = −0.27). Controlling for FMI, HMW-adiponectin was still associated negatively with trunk fat percentage (r = −0.18), waist circumference (r = −0.19), insulin (r = −0.22), HOMA-IR (r = −0.22), triglycerides (r = −0.28) and IL-6 (r = −0.08), while positively with HDL-C (r = 0.29) and sOB-R (r = 0.17). In comparison, sOB-R showed somewhat weaker associations: r = −0.13 with waist circumference, r = −0.09 with triglycerides, r = 0.17 with HDL-C and no relation to trunk fat percentage (r = −0.05).\nIn contrast, leptin was strongly correlated with BMI (r = 0.66) and FMI (r = 0.75). After adjustment for FMI, only the correlations with insulin (r = 0.13), HOMA-IR (r = 0.13), triglycerides (r = 0.09), remained statistically significant.\n[SUBTITLE] Risk of metabolic syndrome [SUBSECTION] HMW-adiponectin and sOB-R decreased (Figure 1, A and C) while leptin increased (Figure 1B) with an increased number of MetS components in both men and women (all P<.0001).\nBlack bars  =  Men; white bars  =  Women. All P values were <.0001.\nIn multivariate logistic regression analyses, both HMW-adiponectin and sOB-R were negatively, whereas leptin was positively, associated with the risk of MetS independent of BMI and inflammatory markers (Table 2, Model 1 and 2). The odds ratios (ORs) comparing the highest with the lowest quartile were 0.34 (95%CI 0.20∼0.58, P\ntrend<.0001) for HMW-adiponectin, 0.68 (95%CI 0.41∼1.13, P\ntrend = 0.05) for sOB-R and 2.60 (95%CI 1.34∼5.05, P\ntrend = .005) for leptin. Further adjustments of leptin and sOB-R showed little impact on the association between HMW-adiponectin and MetS. Similarly, adjustment for sOB-R and HMW-adiponectin did not affect the association between leptin and MetS. However, the significant association between sOB-R and MetS disappeared after HMW-adiponectin was included in the Model 3 (P\ntrend = 0.15). Replacing BMI with FMI (Model 4) did not substantially change the significant associations for both HMW-adiponectin and sOB-R (ORs in the highest quartile were 0.30, 95%CI 0.18∼0.50, P\ntrend<.0001 and 0.61, 95%CI 0.36∼1.02, P\ntrend = 0.02, accordingly); but abolished the significance between leptin and the risk of MetS (P\ntrend = 0.23).\nModel 1: adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and log-transformed BMI.\nModel 2: model 1+ inflammatory factors (hsCRP and IL-6).\nModel 3: model 2+ other adipokines (leptin and sOB-R or HMW-adiponectin).\nModel 4: adjusted for log-transformed FMI instead of BMI in Model 1.\nQuartiles of HMW-adiponectin (µg/mL) were <1.08, 1.08–1.84, 1.84–3.16, >3.16 for men and <1.86, 1.86–3.15, 3.15–5.33, >5.33 for women. Quartiles of leptin (ng/mL) were <1.95, 1.95–3.22, 3.22–5.58, >5.58 for men and <5.92, 5.92–9.64, 9.64–15.21, >15.21 for women. Quartiles of sOB-R (ng/mL) were <15.16, 15.16–18.19, 18.19–22.15, >22.15 for men and <15.54, 15.54–18.45, 18.45–21.82, >21.82 for women.\nData were available for 956 participants.\nWith respect to individual components of MetS (Table 3), HMW-adiponectin was strongly associated with a decreased risk of hypertriglyceridemia (OR in the highest quartile  = 0.26, 95%CI 0.16∼0.42, P\ntrend<.0001) and low HDL-C (OR in the highest quartile  = 0.22, 95%CI 0.14∼0.35, P\ntrend<.0001), while marginally associated with central obesity (OR in the highest quartile  = 0.49, 95%CI 0.25∼0.97, P\ntrend = 0.06) in the FMI-adjusted model. Furthermore, trunk fat percentage adjustment did not affect these associations substantially (data not shown). Meanwhile, sOB-R was negatively associated with abdominal fat (OR in the highest quartile  = 0.36, 95%CI 0.18∼0.73, P\ntrend = 0.002), high triglycerides (OR in the highest quartile  = 0.64, 95%CI 0.40∼1.01, P\ntrend = 0.02) and low HDL-C (OR in the highest quartile  = 0.47, 95%CI 0.31∼0.73, P\ntrend<.0001). In contrast, leptin was only positively associated with hypertriglyceridemia after adjustment for FMI (OR in the highest quartile  = 2.90, 95%CI 1.46∼5.77, P\ntrend = 0.006).\nNumber of cases: central obesity (496), hyperglycemia (616), elevated blood pressure (395), hypertriglyceridemia (305), reduced HDL-C (331).\nAdjusted for the same variables as Model 4 in Table 2, including FMI. Analyses were conducted in 956 participants.\nHMW-adiponectin and sOB-R decreased (Figure 1, A and C) while leptin increased (Figure 1B) with an increased number of MetS components in both men and women (all P<.0001).\nBlack bars  =  Men; white bars  =  Women. All P values were <.0001.\nIn multivariate logistic regression analyses, both HMW-adiponectin and sOB-R were negatively, whereas leptin was positively, associated with the risk of MetS independent of BMI and inflammatory markers (Table 2, Model 1 and 2). The odds ratios (ORs) comparing the highest with the lowest quartile were 0.34 (95%CI 0.20∼0.58, P\ntrend<.0001) for HMW-adiponectin, 0.68 (95%CI 0.41∼1.13, P\ntrend = 0.05) for sOB-R and 2.60 (95%CI 1.34∼5.05, P\ntrend = .005) for leptin. Further adjustments of leptin and sOB-R showed little impact on the association between HMW-adiponectin and MetS. Similarly, adjustment for sOB-R and HMW-adiponectin did not affect the association between leptin and MetS. However, the significant association between sOB-R and MetS disappeared after HMW-adiponectin was included in the Model 3 (P\ntrend = 0.15). Replacing BMI with FMI (Model 4) did not substantially change the significant associations for both HMW-adiponectin and sOB-R (ORs in the highest quartile were 0.30, 95%CI 0.18∼0.50, P\ntrend<.0001 and 0.61, 95%CI 0.36∼1.02, P\ntrend = 0.02, accordingly); but abolished the significance between leptin and the risk of MetS (P\ntrend = 0.23).\nModel 1: adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and log-transformed BMI.\nModel 2: model 1+ inflammatory factors (hsCRP and IL-6).\nModel 3: model 2+ other adipokines (leptin and sOB-R or HMW-adiponectin).\nModel 4: adjusted for log-transformed FMI instead of BMI in Model 1.\nQuartiles of HMW-adiponectin (µg/mL) were <1.08, 1.08–1.84, 1.84–3.16, >3.16 for men and <1.86, 1.86–3.15, 3.15–5.33, >5.33 for women. Quartiles of leptin (ng/mL) were <1.95, 1.95–3.22, 3.22–5.58, >5.58 for men and <5.92, 5.92–9.64, 9.64–15.21, >15.21 for women. Quartiles of sOB-R (ng/mL) were <15.16, 15.16–18.19, 18.19–22.15, >22.15 for men and <15.54, 15.54–18.45, 18.45–21.82, >21.82 for women.\nData were available for 956 participants.\nWith respect to individual components of MetS (Table 3), HMW-adiponectin was strongly associated with a decreased risk of hypertriglyceridemia (OR in the highest quartile  = 0.26, 95%CI 0.16∼0.42, P\ntrend<.0001) and low HDL-C (OR in the highest quartile  = 0.22, 95%CI 0.14∼0.35, P\ntrend<.0001), while marginally associated with central obesity (OR in the highest quartile  = 0.49, 95%CI 0.25∼0.97, P\ntrend = 0.06) in the FMI-adjusted model. Furthermore, trunk fat percentage adjustment did not affect these associations substantially (data not shown). Meanwhile, sOB-R was negatively associated with abdominal fat (OR in the highest quartile  = 0.36, 95%CI 0.18∼0.73, P\ntrend = 0.002), high triglycerides (OR in the highest quartile  = 0.64, 95%CI 0.40∼1.01, P\ntrend = 0.02) and low HDL-C (OR in the highest quartile  = 0.47, 95%CI 0.31∼0.73, P\ntrend<.0001). In contrast, leptin was only positively associated with hypertriglyceridemia after adjustment for FMI (OR in the highest quartile  = 2.90, 95%CI 1.46∼5.77, P\ntrend = 0.006).\nNumber of cases: central obesity (496), hyperglycemia (616), elevated blood pressure (395), hypertriglyceridemia (305), reduced HDL-C (331).\nAdjusted for the same variables as Model 4 in Table 2, including FMI. Analyses were conducted in 956 participants.\n[SUBTITLE] Stratified analyses [SUBSECTION] Considering the obesity case-control design of this study and to explore how obesity, total fat mass and trunk fat percentage influenced the observed relationships, we conducted BMI-stratified analyses. In both normal-weight and overweight group, HMW-adiponectin showed strong inverse associations with modified MetS, regardless whether BMI, FMI or trunk fat percentage was adjusted (Table S2, Model 1 to 3). However, leptin was not significantly associated with modified MetS under control of total or abdominal adiposity. A negative association between sOB-R and modified MetS was observed in normal weight individuals only (Model 1), and the significant association disappeared following adjustments of FMI or trunk fat percentage (Model 2 and 3). The results remained essentially the same and no significant interaction was found in consequent subgroup analyses according to gender, FMI, hsCRP and HOMA-IR (Table S3).\nConsidering the obesity case-control design of this study and to explore how obesity, total fat mass and trunk fat percentage influenced the observed relationships, we conducted BMI-stratified analyses. In both normal-weight and overweight group, HMW-adiponectin showed strong inverse associations with modified MetS, regardless whether BMI, FMI or trunk fat percentage was adjusted (Table S2, Model 1 to 3). However, leptin was not significantly associated with modified MetS under control of total or abdominal adiposity. A negative association between sOB-R and modified MetS was observed in normal weight individuals only (Model 1), and the significant association disappeared following adjustments of FMI or trunk fat percentage (Model 2 and 3). The results remained essentially the same and no significant interaction was found in consequent subgroup analyses according to gender, FMI, hsCRP and HOMA-IR (Table S3).", "HMW-adiponectin and sOB-R decreased (Figure 1, A and C) while leptin increased (Figure 1B) with an increased number of MetS components in both men and women (all P<.0001).\nBlack bars  =  Men; white bars  =  Women. All P values were <.0001.\nIn multivariate logistic regression analyses, both HMW-adiponectin and sOB-R were negatively, whereas leptin was positively, associated with the risk of MetS independent of BMI and inflammatory markers (Table 2, Model 1 and 2). The odds ratios (ORs) comparing the highest with the lowest quartile were 0.34 (95%CI 0.20∼0.58, P\ntrend<.0001) for HMW-adiponectin, 0.68 (95%CI 0.41∼1.13, P\ntrend = 0.05) for sOB-R and 2.60 (95%CI 1.34∼5.05, P\ntrend = .005) for leptin. Further adjustments of leptin and sOB-R showed little impact on the association between HMW-adiponectin and MetS. Similarly, adjustment for sOB-R and HMW-adiponectin did not affect the association between leptin and MetS. However, the significant association between sOB-R and MetS disappeared after HMW-adiponectin was included in the Model 3 (P\ntrend = 0.15). Replacing BMI with FMI (Model 4) did not substantially change the significant associations for both HMW-adiponectin and sOB-R (ORs in the highest quartile were 0.30, 95%CI 0.18∼0.50, P\ntrend<.0001 and 0.61, 95%CI 0.36∼1.02, P\ntrend = 0.02, accordingly); but abolished the significance between leptin and the risk of MetS (P\ntrend = 0.23).\nModel 1: adjusted for age, sex, diabetes, smoke, alcohol, family history of chronic diseases, education, physical activity, sleep, total energy intake and log-transformed BMI.\nModel 2: model 1+ inflammatory factors (hsCRP and IL-6).\nModel 3: model 2+ other adipokines (leptin and sOB-R or HMW-adiponectin).\nModel 4: adjusted for log-transformed FMI instead of BMI in Model 1.\nQuartiles of HMW-adiponectin (µg/mL) were <1.08, 1.08–1.84, 1.84–3.16, >3.16 for men and <1.86, 1.86–3.15, 3.15–5.33, >5.33 for women. Quartiles of leptin (ng/mL) were <1.95, 1.95–3.22, 3.22–5.58, >5.58 for men and <5.92, 5.92–9.64, 9.64–15.21, >15.21 for women. Quartiles of sOB-R (ng/mL) were <15.16, 15.16–18.19, 18.19–22.15, >22.15 for men and <15.54, 15.54–18.45, 18.45–21.82, >21.82 for women.\nData were available for 956 participants.\nWith respect to individual components of MetS (Table 3), HMW-adiponectin was strongly associated with a decreased risk of hypertriglyceridemia (OR in the highest quartile  = 0.26, 95%CI 0.16∼0.42, P\ntrend<.0001) and low HDL-C (OR in the highest quartile  = 0.22, 95%CI 0.14∼0.35, P\ntrend<.0001), while marginally associated with central obesity (OR in the highest quartile  = 0.49, 95%CI 0.25∼0.97, P\ntrend = 0.06) in the FMI-adjusted model. Furthermore, trunk fat percentage adjustment did not affect these associations substantially (data not shown). Meanwhile, sOB-R was negatively associated with abdominal fat (OR in the highest quartile  = 0.36, 95%CI 0.18∼0.73, P\ntrend = 0.002), high triglycerides (OR in the highest quartile  = 0.64, 95%CI 0.40∼1.01, P\ntrend = 0.02) and low HDL-C (OR in the highest quartile  = 0.47, 95%CI 0.31∼0.73, P\ntrend<.0001). In contrast, leptin was only positively associated with hypertriglyceridemia after adjustment for FMI (OR in the highest quartile  = 2.90, 95%CI 1.46∼5.77, P\ntrend = 0.006).\nNumber of cases: central obesity (496), hyperglycemia (616), elevated blood pressure (395), hypertriglyceridemia (305), reduced HDL-C (331).\nAdjusted for the same variables as Model 4 in Table 2, including FMI. Analyses were conducted in 956 participants.", "Considering the obesity case-control design of this study and to explore how obesity, total fat mass and trunk fat percentage influenced the observed relationships, we conducted BMI-stratified analyses. In both normal-weight and overweight group, HMW-adiponectin showed strong inverse associations with modified MetS, regardless whether BMI, FMI or trunk fat percentage was adjusted (Table S2, Model 1 to 3). However, leptin was not significantly associated with modified MetS under control of total or abdominal adiposity. A negative association between sOB-R and modified MetS was observed in normal weight individuals only (Model 1), and the significant association disappeared following adjustments of FMI or trunk fat percentage (Model 2 and 3). The results remained essentially the same and no significant interaction was found in consequent subgroup analyses according to gender, FMI, hsCRP and HOMA-IR (Table S3).", "Among 1055 middle-aged Chinese men and women, we observed that reduced plasma HMW-adiponectin and sOB-R, and elevated leptin level were significantly associated with an increased risk of MetS and some of its components independent of multiple confounders including BMI and inflammatory markers. However, body fat mass or adipokines showed different modifying effects on these associations. Unlike the association between HMW-adiponectin and MetS which was unaffected by adjusting for fat depots or other adipokines, the positive associations between leptin and MetS were mainly explained by total fat mass, while the associations between sOB-R and MetS were largely influenced by HMW-adiponectin.\n[SUBTITLE] HMW-adiponectin and metabolic syndrome [SUBSECTION] Our study indicated that low plasma HMW-adiponectin was a strong and independent risk factor related to MetS in Chinese when diet, lifestyles, adiposity, inflammatory factors leptin and sOB-R were extensively controlled. The inverse associations of adiponectin with metabolic diseases and type 2 diabetes have been well established [4], [5]. However, most studies only measured total adiponectin rather than its high-molecular-weight multimer (12–18mer), which may be more biologically active than trimer and hexamer [6], [7], [8]. Evidence regarding the associations between HMW-adiponectin and MetS was sparse and most of them were limited by small sample size and residual confounding. In one relatively large-scale study, Tabar et al. [9] reported inverse associations between HMW-adiponectin and MetS and its components, except high blood pressure in middle-aged to elderly Japanese. Likewise, our study also supported such associations in Chinese population. On the other hand, we found that low HMW-adiponectin concentration was only significantly associated with high triglyceride and low HDL-C, but not with elevated glucose and central obesity. This minor discrepancy could be due to the different definitions for MetS between two studies. It is also plausible that more confounders were controlled in our analyses.\nOne interesting observation in current study is that associations between HMW-adiponectin and MetS were independent of adiposity measured by BMI, FMI, or trunk fat percentage. In the Nurses' Health Study, Heidemann et al. reported that higher HMW-adiponectin was associated with lower insulin and diabetes risk independent of BMI or waist circumference [11]. While our data demonstrated that significant negative correlations between HMW-adiponectin and insulin or HOMA-IR (both r = −0.22) were independent of FMI. Moreover, adjustment for IL-6 and hsCRP had little effect on the association between HMW-adiponectin and MetS (Table 2). The mechanism underlining adiponectin regulation and signaling pathway is rather complex and non-adipose factors might also be involved. Besides fat cells, adiponectin could be secreted by skeletal muscle, cardiac myocytes and endothelial cells as well [2]. Moreover, two receptors, namely AdipoR1 and AdipoR2, operate closely with AMPK or PPAR-α pathway to enhance glucose uptake and utilization in muscle, promote lipid oxidation in liver, and improve systemic insulin sensitivity [2], [31]. In addition, insulin resistance and inflammatory factors, hallmarks of MetS, are proposed to down regulate adiponectin production [2]. Previous small-scale infusion studies suggested that correlations of adiponectin with insulin sensitivity were independent of BMI [32], total adiposity measured by DEXA, visceral fat measured by magnetic resonance imaging and intramyocellular lipid measured by 1H-magnetic resonance spectroscopy [33]. Indeed, both enlarged size and increased number of subcutaneous adipocytes could lead to accumulation of fat in non-adipose tissue, such as skeletal muscle, liver and heart. These ectopic fats are closely related to hypoadiponectinemia and might modify its link with insulin resistance [34]. Due to the limitation of DEXA method in current study, it is not possible to exclude the effects of visceral adiposity, intramyocellular or intrahepatic lipid content or adipocytes size by adjustment of FMI, although BMI were already partitioned by fat mass and fat free mass,respectively [35]. Collectively, data from our study and others supported that HMW-adiponectin was an important biomarker for MetS in addition to many established risk factors.\nOur study indicated that low plasma HMW-adiponectin was a strong and independent risk factor related to MetS in Chinese when diet, lifestyles, adiposity, inflammatory factors leptin and sOB-R were extensively controlled. The inverse associations of adiponectin with metabolic diseases and type 2 diabetes have been well established [4], [5]. However, most studies only measured total adiponectin rather than its high-molecular-weight multimer (12–18mer), which may be more biologically active than trimer and hexamer [6], [7], [8]. Evidence regarding the associations between HMW-adiponectin and MetS was sparse and most of them were limited by small sample size and residual confounding. In one relatively large-scale study, Tabar et al. [9] reported inverse associations between HMW-adiponectin and MetS and its components, except high blood pressure in middle-aged to elderly Japanese. Likewise, our study also supported such associations in Chinese population. On the other hand, we found that low HMW-adiponectin concentration was only significantly associated with high triglyceride and low HDL-C, but not with elevated glucose and central obesity. This minor discrepancy could be due to the different definitions for MetS between two studies. It is also plausible that more confounders were controlled in our analyses.\nOne interesting observation in current study is that associations between HMW-adiponectin and MetS were independent of adiposity measured by BMI, FMI, or trunk fat percentage. In the Nurses' Health Study, Heidemann et al. reported that higher HMW-adiponectin was associated with lower insulin and diabetes risk independent of BMI or waist circumference [11]. While our data demonstrated that significant negative correlations between HMW-adiponectin and insulin or HOMA-IR (both r = −0.22) were independent of FMI. Moreover, adjustment for IL-6 and hsCRP had little effect on the association between HMW-adiponectin and MetS (Table 2). The mechanism underlining adiponectin regulation and signaling pathway is rather complex and non-adipose factors might also be involved. Besides fat cells, adiponectin could be secreted by skeletal muscle, cardiac myocytes and endothelial cells as well [2]. Moreover, two receptors, namely AdipoR1 and AdipoR2, operate closely with AMPK or PPAR-α pathway to enhance glucose uptake and utilization in muscle, promote lipid oxidation in liver, and improve systemic insulin sensitivity [2], [31]. In addition, insulin resistance and inflammatory factors, hallmarks of MetS, are proposed to down regulate adiponectin production [2]. Previous small-scale infusion studies suggested that correlations of adiponectin with insulin sensitivity were independent of BMI [32], total adiposity measured by DEXA, visceral fat measured by magnetic resonance imaging and intramyocellular lipid measured by 1H-magnetic resonance spectroscopy [33]. Indeed, both enlarged size and increased number of subcutaneous adipocytes could lead to accumulation of fat in non-adipose tissue, such as skeletal muscle, liver and heart. These ectopic fats are closely related to hypoadiponectinemia and might modify its link with insulin resistance [34]. Due to the limitation of DEXA method in current study, it is not possible to exclude the effects of visceral adiposity, intramyocellular or intrahepatic lipid content or adipocytes size by adjustment of FMI, although BMI were already partitioned by fat mass and fat free mass,respectively [35]. Collectively, data from our study and others supported that HMW-adiponectin was an important biomarker for MetS in addition to many established risk factors.\n[SUBTITLE] Leptin, soluble leptin receptor and metabolic syndrome [SUBSECTION] In this study, we also found strong associations between leptin and risk of MetS with adjusting BMI, inflammation and several known confounders (Table 2, Model 1 to 3). Interestingly, in contrast to the case of HMW-adiponectin, these associations could be diminished by adjusting FMI (Table 2, Model 4 and Table 3). Previously, data from a cohort study in Caucasian population suggested that leptin could predict future MetS independent of baseline BMI [14]. While leptin was associated with MetS risk in older Chinese women, but not in men after adjustment for BMI [36]. However, it was noteworthy that most existing studies utilized BMI to evaluate adiposity status, containing both fat and fat-free mass, which have different influences on metabolic disorders [28]. As an adipose-derived hormone, leptin plays a critical role in regulating energy homeostasis [13]. However, it is still not clear whether leptin resistance is one of the causes or the consequences of obesity, or the “vicious cycle” of them might be the culprit of metabolic disorders. In this study, we provide direct evidence highlighting the key role of fat mass in hyperleptinemia associated metabolic abnormalities. Our finding might be particularly important for Chinese who are characterized to have more body fat under given BMI [20], [21].\nIn addition to hyperleptinemia, our data revealed that reduced soluble leptin receptor (sOB-R), another component of leptin resistance, was significantly associated with an increased MetS risk independent of fat mass. sOB-R, formed by cleaving the ectodomain of membrane-anchored leptin receptors, represents the major leptin binding fraction in blood [37]. It regulates the available leptin pool and downstream signaling [38]. Few epidemiological studies have simultaneously investigated the associations of both plasma leptin and sOB-R with MetS. In a study from Framingham third generation participants (n = 362), leptin and free leptin index (molar ratio of leptin and sOB-R) were positively associated with severity of MetS after adjustment for age and sex [16]. With larger sample size, we further adjusted for diet, lifestyle, BMI/FMI, inflammatory markers and adipokines. However, our data indicated that leptin and sOB-R had unique properties and might reflect the different aspects of MetS. For instance, sOB-R was negatively associated with central obesity, elevated triglyceride and reduced HDL-C, whereas leptin was only associated with hypertriglyceridemia in multivariate regression including FMI (Table 3). Given the wide distribution in tissues containing membrane leptin receptors, it is reasonable to speculate that non-adipose mechanism(s) such as tissue specific effect might also play a role in sOB-R regulation. On the other hand, leptin was almost exclusively secreted by adipocytes, particularly subcutaneous fat. Hypertriglyceridemia might result from the limited capacity of adipose tissue to store excessive fat or up-regulated hepatic triglyceride secretion and down-regulated plasma triglyceride degradation [39], [40]. However, adjustment for HMW-adiponectin could abolish significance between sOB-R and MetS but not that between leptin and MetS (Table 2, Model 3). Indeed, findings from the Nurses' Health Study showed an inverse association between sOB-R and diabetes independent of BMI and HMW-adiponectin [15]. Kim et al. found if leptin-deficient mice have over-expressed adiponectin would result in improvements of glucose and lipid metabolism and inflammatory profile [41]. Obviously, further studies are warranted to clarify whether there are functional and/or signaling crosstalk among adiponectin, leptin and its receptor (s).\nIn this study, we also found strong associations between leptin and risk of MetS with adjusting BMI, inflammation and several known confounders (Table 2, Model 1 to 3). Interestingly, in contrast to the case of HMW-adiponectin, these associations could be diminished by adjusting FMI (Table 2, Model 4 and Table 3). Previously, data from a cohort study in Caucasian population suggested that leptin could predict future MetS independent of baseline BMI [14]. While leptin was associated with MetS risk in older Chinese women, but not in men after adjustment for BMI [36]. However, it was noteworthy that most existing studies utilized BMI to evaluate adiposity status, containing both fat and fat-free mass, which have different influences on metabolic disorders [28]. As an adipose-derived hormone, leptin plays a critical role in regulating energy homeostasis [13]. However, it is still not clear whether leptin resistance is one of the causes or the consequences of obesity, or the “vicious cycle” of them might be the culprit of metabolic disorders. In this study, we provide direct evidence highlighting the key role of fat mass in hyperleptinemia associated metabolic abnormalities. Our finding might be particularly important for Chinese who are characterized to have more body fat under given BMI [20], [21].\nIn addition to hyperleptinemia, our data revealed that reduced soluble leptin receptor (sOB-R), another component of leptin resistance, was significantly associated with an increased MetS risk independent of fat mass. sOB-R, formed by cleaving the ectodomain of membrane-anchored leptin receptors, represents the major leptin binding fraction in blood [37]. It regulates the available leptin pool and downstream signaling [38]. Few epidemiological studies have simultaneously investigated the associations of both plasma leptin and sOB-R with MetS. In a study from Framingham third generation participants (n = 362), leptin and free leptin index (molar ratio of leptin and sOB-R) were positively associated with severity of MetS after adjustment for age and sex [16]. With larger sample size, we further adjusted for diet, lifestyle, BMI/FMI, inflammatory markers and adipokines. However, our data indicated that leptin and sOB-R had unique properties and might reflect the different aspects of MetS. For instance, sOB-R was negatively associated with central obesity, elevated triglyceride and reduced HDL-C, whereas leptin was only associated with hypertriglyceridemia in multivariate regression including FMI (Table 3). Given the wide distribution in tissues containing membrane leptin receptors, it is reasonable to speculate that non-adipose mechanism(s) such as tissue specific effect might also play a role in sOB-R regulation. On the other hand, leptin was almost exclusively secreted by adipocytes, particularly subcutaneous fat. Hypertriglyceridemia might result from the limited capacity of adipose tissue to store excessive fat or up-regulated hepatic triglyceride secretion and down-regulated plasma triglyceride degradation [39], [40]. However, adjustment for HMW-adiponectin could abolish significance between sOB-R and MetS but not that between leptin and MetS (Table 2, Model 3). Indeed, findings from the Nurses' Health Study showed an inverse association between sOB-R and diabetes independent of BMI and HMW-adiponectin [15]. Kim et al. found if leptin-deficient mice have over-expressed adiponectin would result in improvements of glucose and lipid metabolism and inflammatory profile [41]. Obviously, further studies are warranted to clarify whether there are functional and/or signaling crosstalk among adiponectin, leptin and its receptor (s).\n[SUBTITLE] Strengths and Limitations [SUBSECTION] This is the first study that systematically investigates the associations of HMW-adiponectin, leptin, sOB-R, body fat mass measured by DEXA, along with a wide range of inflammatory and metabolic parameters with MetS and its features in a relatively large population with both sexes. Meanwhile, we acknowledge some limitations. Because of the cross-sectional nature, no causal relationship could be established. Also, despite of the originally obesity case-control design, we adapted a cross-sectional approach in data analyses in order to enhance statistical power. Nevertheless, similar associations of HMW-adiponectin, leptin with MetS were observed in both normal and overweight/obese individuals as well as in further subgroup analyses. Obviously, our findings need to be confirmed prospectively in different populations.\nIn conclusion, we found strong inverse associations between HMW-adiponectin and MetS independent of adiposity, inflammatory statuses, leptin and sOB-R. Similar associations between sOB-R and MetS were also evidenced, but were weak and attenuated by HMW-adiponectin adjustment. In contrast, leptin showed strong positive associations with the risk of MetS which could be mainly explained by body fat mass. Overall, current study provides more mechanistic insights into the linkage of adipose tissue, adipokines and inflammation to metabolic syndrome and also more evidences for detection of decreased HMW-adiponectin and leptin resistance in clinical settings to prevent metabolic disorders and future diabetic and cardiovascular outcomes.\nThis is the first study that systematically investigates the associations of HMW-adiponectin, leptin, sOB-R, body fat mass measured by DEXA, along with a wide range of inflammatory and metabolic parameters with MetS and its features in a relatively large population with both sexes. Meanwhile, we acknowledge some limitations. Because of the cross-sectional nature, no causal relationship could be established. Also, despite of the originally obesity case-control design, we adapted a cross-sectional approach in data analyses in order to enhance statistical power. Nevertheless, similar associations of HMW-adiponectin, leptin with MetS were observed in both normal and overweight/obese individuals as well as in further subgroup analyses. Obviously, our findings need to be confirmed prospectively in different populations.\nIn conclusion, we found strong inverse associations between HMW-adiponectin and MetS independent of adiposity, inflammatory statuses, leptin and sOB-R. Similar associations between sOB-R and MetS were also evidenced, but were weak and attenuated by HMW-adiponectin adjustment. In contrast, leptin showed strong positive associations with the risk of MetS which could be mainly explained by body fat mass. Overall, current study provides more mechanistic insights into the linkage of adipose tissue, adipokines and inflammation to metabolic syndrome and also more evidences for detection of decreased HMW-adiponectin and leptin resistance in clinical settings to prevent metabolic disorders and future diabetic and cardiovascular outcomes.", "Our study indicated that low plasma HMW-adiponectin was a strong and independent risk factor related to MetS in Chinese when diet, lifestyles, adiposity, inflammatory factors leptin and sOB-R were extensively controlled. The inverse associations of adiponectin with metabolic diseases and type 2 diabetes have been well established [4], [5]. However, most studies only measured total adiponectin rather than its high-molecular-weight multimer (12–18mer), which may be more biologically active than trimer and hexamer [6], [7], [8]. Evidence regarding the associations between HMW-adiponectin and MetS was sparse and most of them were limited by small sample size and residual confounding. In one relatively large-scale study, Tabar et al. [9] reported inverse associations between HMW-adiponectin and MetS and its components, except high blood pressure in middle-aged to elderly Japanese. Likewise, our study also supported such associations in Chinese population. On the other hand, we found that low HMW-adiponectin concentration was only significantly associated with high triglyceride and low HDL-C, but not with elevated glucose and central obesity. This minor discrepancy could be due to the different definitions for MetS between two studies. It is also plausible that more confounders were controlled in our analyses.\nOne interesting observation in current study is that associations between HMW-adiponectin and MetS were independent of adiposity measured by BMI, FMI, or trunk fat percentage. In the Nurses' Health Study, Heidemann et al. reported that higher HMW-adiponectin was associated with lower insulin and diabetes risk independent of BMI or waist circumference [11]. While our data demonstrated that significant negative correlations between HMW-adiponectin and insulin or HOMA-IR (both r = −0.22) were independent of FMI. Moreover, adjustment for IL-6 and hsCRP had little effect on the association between HMW-adiponectin and MetS (Table 2). The mechanism underlining adiponectin regulation and signaling pathway is rather complex and non-adipose factors might also be involved. Besides fat cells, adiponectin could be secreted by skeletal muscle, cardiac myocytes and endothelial cells as well [2]. Moreover, two receptors, namely AdipoR1 and AdipoR2, operate closely with AMPK or PPAR-α pathway to enhance glucose uptake and utilization in muscle, promote lipid oxidation in liver, and improve systemic insulin sensitivity [2], [31]. In addition, insulin resistance and inflammatory factors, hallmarks of MetS, are proposed to down regulate adiponectin production [2]. Previous small-scale infusion studies suggested that correlations of adiponectin with insulin sensitivity were independent of BMI [32], total adiposity measured by DEXA, visceral fat measured by magnetic resonance imaging and intramyocellular lipid measured by 1H-magnetic resonance spectroscopy [33]. Indeed, both enlarged size and increased number of subcutaneous adipocytes could lead to accumulation of fat in non-adipose tissue, such as skeletal muscle, liver and heart. These ectopic fats are closely related to hypoadiponectinemia and might modify its link with insulin resistance [34]. Due to the limitation of DEXA method in current study, it is not possible to exclude the effects of visceral adiposity, intramyocellular or intrahepatic lipid content or adipocytes size by adjustment of FMI, although BMI were already partitioned by fat mass and fat free mass,respectively [35]. Collectively, data from our study and others supported that HMW-adiponectin was an important biomarker for MetS in addition to many established risk factors.", "In this study, we also found strong associations between leptin and risk of MetS with adjusting BMI, inflammation and several known confounders (Table 2, Model 1 to 3). Interestingly, in contrast to the case of HMW-adiponectin, these associations could be diminished by adjusting FMI (Table 2, Model 4 and Table 3). Previously, data from a cohort study in Caucasian population suggested that leptin could predict future MetS independent of baseline BMI [14]. While leptin was associated with MetS risk in older Chinese women, but not in men after adjustment for BMI [36]. However, it was noteworthy that most existing studies utilized BMI to evaluate adiposity status, containing both fat and fat-free mass, which have different influences on metabolic disorders [28]. As an adipose-derived hormone, leptin plays a critical role in regulating energy homeostasis [13]. However, it is still not clear whether leptin resistance is one of the causes or the consequences of obesity, or the “vicious cycle” of them might be the culprit of metabolic disorders. In this study, we provide direct evidence highlighting the key role of fat mass in hyperleptinemia associated metabolic abnormalities. Our finding might be particularly important for Chinese who are characterized to have more body fat under given BMI [20], [21].\nIn addition to hyperleptinemia, our data revealed that reduced soluble leptin receptor (sOB-R), another component of leptin resistance, was significantly associated with an increased MetS risk independent of fat mass. sOB-R, formed by cleaving the ectodomain of membrane-anchored leptin receptors, represents the major leptin binding fraction in blood [37]. It regulates the available leptin pool and downstream signaling [38]. Few epidemiological studies have simultaneously investigated the associations of both plasma leptin and sOB-R with MetS. In a study from Framingham third generation participants (n = 362), leptin and free leptin index (molar ratio of leptin and sOB-R) were positively associated with severity of MetS after adjustment for age and sex [16]. With larger sample size, we further adjusted for diet, lifestyle, BMI/FMI, inflammatory markers and adipokines. However, our data indicated that leptin and sOB-R had unique properties and might reflect the different aspects of MetS. For instance, sOB-R was negatively associated with central obesity, elevated triglyceride and reduced HDL-C, whereas leptin was only associated with hypertriglyceridemia in multivariate regression including FMI (Table 3). Given the wide distribution in tissues containing membrane leptin receptors, it is reasonable to speculate that non-adipose mechanism(s) such as tissue specific effect might also play a role in sOB-R regulation. On the other hand, leptin was almost exclusively secreted by adipocytes, particularly subcutaneous fat. Hypertriglyceridemia might result from the limited capacity of adipose tissue to store excessive fat or up-regulated hepatic triglyceride secretion and down-regulated plasma triglyceride degradation [39], [40]. However, adjustment for HMW-adiponectin could abolish significance between sOB-R and MetS but not that between leptin and MetS (Table 2, Model 3). Indeed, findings from the Nurses' Health Study showed an inverse association between sOB-R and diabetes independent of BMI and HMW-adiponectin [15]. Kim et al. found if leptin-deficient mice have over-expressed adiponectin would result in improvements of glucose and lipid metabolism and inflammatory profile [41]. Obviously, further studies are warranted to clarify whether there are functional and/or signaling crosstalk among adiponectin, leptin and its receptor (s).", "This is the first study that systematically investigates the associations of HMW-adiponectin, leptin, sOB-R, body fat mass measured by DEXA, along with a wide range of inflammatory and metabolic parameters with MetS and its features in a relatively large population with both sexes. Meanwhile, we acknowledge some limitations. Because of the cross-sectional nature, no causal relationship could be established. Also, despite of the originally obesity case-control design, we adapted a cross-sectional approach in data analyses in order to enhance statistical power. Nevertheless, similar associations of HMW-adiponectin, leptin with MetS were observed in both normal and overweight/obese individuals as well as in further subgroup analyses. Obviously, our findings need to be confirmed prospectively in different populations.\nIn conclusion, we found strong inverse associations between HMW-adiponectin and MetS independent of adiposity, inflammatory statuses, leptin and sOB-R. Similar associations between sOB-R and MetS were also evidenced, but were weak and attenuated by HMW-adiponectin adjustment. In contrast, leptin showed strong positive associations with the risk of MetS which could be mainly explained by body fat mass. Overall, current study provides more mechanistic insights into the linkage of adipose tissue, adipokines and inflammation to metabolic syndrome and also more evidences for detection of decreased HMW-adiponectin and leptin resistance in clinical settings to prevent metabolic disorders and future diabetic and cardiovascular outcomes.", "\nCharacteristics of participants 1. Abbreviations: MetS  =  metabolic syndrome, hsCRP  =  high sensitive C-reactive protein, HMW-adiponectin  =  high-molecular-weight adiponectin, sOB-R  =  soluble leptin receptor. 1 Data were Mean ± SD or Median (IQR) or Number (percentage). 2 Adjusted for age and sex. 3 Data were available for 956 participants.\n(DOC)\nClick here for additional data file.\n\nOdds ratio (95% CI) of modified metabolic syndrome according to sex-specific tertile of HMW-adiponectin, leptin and sOB-R in BMI-stratified analyses 1.\n1 Definition of metabolic syndrome was modified: having 2 or more components of metabolic syndrome without central obesity. Tertiles were based on sex-specific levels in BMI-stratified subgroup. NO. of cases/control were 160/338 in normal weight group and 414/143 in overweight group. 2 Adjusted for the same variables in Table 2. Model 1: including BMI; Model 2: including FMI; Model 3: adjustment for log-transformed trunk fat percentage. 3 Data were available for 956 participants. NO. of cases/control were 147/307 in normal weight group and 369/133 in overweight group.\n(DOC)\nClick here for additional data file.\n\nOdds ratio (95% CI) of metabolic syndrome according to sex-specific tertile of HMW-adiponectin, leptin and sOB-R in subgroup analyses 1.\n1 Tertiles were based on sex-specific levels in each subgroup. Adjusted for the same variables as Model 4 in Table 2, including FMI. Data were available for 956 participants. 2 Median concentration of FMI was 5.87 for men and 7.79 for women. Definition of metabolic syndrome was modified: having 2 or more components of metabolic syndrome without central obesity. No adjustment for FMI. 3 Median concentration of hsCRP was 0.87 mg/L. 4 Median level of HOMA-IR was 1.08.\n(DOC)\nClick here for additional data file." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Disease Outbreaks", "Hand, Foot and Mouth Disease", "Rain", "Risk", "Singapore", "Temperature", "Time Factors", "Weather" ]

[ "Introduction", "Results", "Discussion", "Conclusion" ]

[ "Hand foot and mouth disease (HFMD) is caused by a number of different enteroviruses, among which are enterovirus 71 (EV71) and coxsackie A16 (CA16). The disease is transmitted through direct contact with respiratory droplets, feces, and blister fluid of infective patients or through contact with contaminated environment such as water, food, or surface [1], [2]. HFMD inflicts mainly children with mild clinical symptoms include fever, blisters and sores in mouth, palms and soles following 3–7 days of incubation period and a patient generally recovers in 7–10 days. Nevertheless, severe health consequences or death may occur owing to complications such as encephalitis, aseptic meningitis, and acute flaccid paralysis mainly following EV71 infection. HFMD can be asymptomatic and it is possible for a recovered person to be infected again by different serotypes of enteroviruses [1]. There is no specific treatment or vaccine available; therefore, preventive measures such as avoid direct contact with infective patients, disinfection of viral contaminated items or premises, and good personal hygiene practices remain the only effective methods to disrupt disease transmission.\nIn recent decade, Asian countries have experienced increasing trend of HFMD outbreak with deaths among children due to severe complications [3], [4], [5]. HFMD has raised public health concerns in Asia following severe outbreaks in Malaysia and Taiwan in 1997 and 1998, respectively. About 6000 young children were infected with 42 deaths during the HFMD outbreak in Malaysia [6]. Whereas a total of 129,106 cases were reported in Taiwan with 405 children suffered complications that led to 78 deaths [7], [8]. In the first half of 2010, mainland China experienced 72% increase in HFMD cases compared with 2009; resulting in approximately 1.3 million reported cases [5]. Also, 5454 severe cases that caused 260 deaths were reported in mainland China between January to the first week of May [5]. HFMD outbreaks generally occur in 2–3 years cyclical pattern in endemic countries across the Western Pacific Region and EV71-related HFMD outbreak frequency is expected to increase in the region partly owing to complex factors including continued evolution and emergence of novel recombinants of EV71, inadequate healthcare capacity, and a lack of effective surveillance system in some countries [9].\nHFMD is endemic in Singapore with EV71 and CA16 as two main dominant circulating strains. About 90% of the reported cases were children below 10 years old. The number of HFMD outbreaks reported from childcare centers, kindergartens, and preschools had escalated from about 167 in year 2001 to more than 1700 in year 2007. In the same period, the incidence rate among children aged 0–4 years surged from approximately 16 to 60 per 1000 populations and from 3 to 21 per 1000 populations among children 5–9 years old [10].\nHFMD is endemic in the tropical and subtropical countries with tendency of higher number of cases in wet season depending on geographical locations [5], [11]. Whereas outbreaks usually occur in summer or early fall in temperate countries. Enterovirus surveillance in USA for the period 1970–2005 showed that EV71 and CA16 had endemic circulation pattern and that around 70% of cases were reported during warmer seasons between June–October [12]. Considering the seasonal pattern of HFMD outbreaks, we hypothesized that short-term changes in weather can influence the transmission dynamic of HFMD. This study aims to establish a relationship, analyze the effects and estimate potential thresholds of weekly temperature and rainfall associated with the risk of HFMD outbreaks in Singapore.", "During the study period, Singapore experienced nationwide HFMD outbreaks in March-May in year 2002 and 2005–2008. Bimodal outbreaks occurred yearly with second outbreak in August-October during 2005 to 2008.\nMinimum and maximum temperature during the study period ranged from 22.8–27.6°C and 27.7–34.6°C, respectively. The highest weekly maximum temperature and lowest minimum temperature was recorded in 2005 and 2008 respectively. Overall, Singapore experienced temperature above 32°C in one-third of the study period with a total of 30 weeks in 2002 and about 6 weeks in 2008 (Figure 1). A total of 30 weeks with temperature difference (Tp) above 7°C was observed in year 2005, and highest weekly Tp of 8.8°C was recorded in 2006 and 2008. The weekly cumulative rainfall during the study period ranged from 0–388 mm with higher amount of rainfall recorded between October and January. About 80% of the weekly cumulative rainfall was below 75 mm.\nThe HFMD incidence was significantly associated with short term variability of weekly temperature difference (Tp), minimum temperature, maximum temperature, and cumulative rainfall at time lag of 1–2 weeks. Figure 2a depicts a wide J-shaped relationship between HFMD incidence and Tp with significant increase in the risk of HFMD incidence when Tp is above 7°C. Likewise, the risk of HFMD incidence rises steeply when maximum temperature is above 32°C; while low maximum temperature poses negligible effects on HFMD (Figure 2c). Inverse relationship is observed between minimum temperature and HFMD incidence (Figure 2d). Similar results are obtained when we test minimum and maximum temperature independently, except that minimum temperature is not statistically significant (Figure 2e & 2f). Figure 2b indicates the relative risk of HFMD incidence increases linearly with weekly cumulative rainfall between 0–75 mm; whereas every unit increases beyond 75 mm reduces the risk of HFMD incidence. During the study period, Tp>7°C mainly coincided with maximum temperature above 32°C and low rainfall, except year 2008.\nRelative risks of HFMD incidence as function of weekly (2a) temperature difference (Tp) & (2b) weekly cumulative rainfall at time lag of 1–2 weeks using model A; (2c) maximum temperature & (2d) minimum temperature at time lag of 1–2 weeks using model B; (2e) maximum temperature tested singly & (2f) minimum temperature tested singly at time lag of 1–2 weeks using model B (Shaded area: confidence interval).\nIn the piecewise linear Poisson regression function, we used breakpoints of 32°C for maximum temperature, 25°C for minimum temperature, 7°C for Tp, and 75 mm for rainfall. Table 1 shows every 1°C increases in maximum temperature above 32°C and in weekly Tp above 7°C elevates the risks of HFMD incidence significantly by 36% and 41%, respectively. The risk, however, declines by 17% with each degree increases in minimum temperature above 25°C. Additionally, a unit (mm) increases in weekly cumulative rainfall between 0–75 mm elevates risk of HFMD incidence by 0.3%.\nObserved and predicted time series of HFMD using both models produced similar results. Both model A and B explained about 87% of variations among HFMD cases using temperature, rainfall, and trend parameters. The residuals plots indicated model residuals did not violate the statistical modeling assumptions. Furthermore, the PACF showed no indication of further non-adjusted residual autocorrelation.", "To date, very few studies document the link between weather and HFMD and none (that the authors are aware of) have previously shown that there exists an explicit relationship with threshold effects between short-term changes in weather and the incidence of HFMD. Our findings showed that high weekly maximum temperature and a large difference of minimum and maximum temperature increased the risk of HFMD incidence in the subsequent 1–2 weeks. The risk functions of minimum and maximum temperature provide more insights into the J-shaped curve of the relation between temperature difference and HFMD cases. Moreover, moderate weekly cumulative rainfall partly sustained the HFMD endemic during the study period. A study on the effects of climate on Herpangina and HFMD in Japan indicates that higher temperature could have influenced the increase of Herpangina & HFMD incidence [20].\nExact reasons for the relationship between weather and HFMD are not known. During the study period, high temperature in Singapore generally accompanied with rainfall below 75 mm and thus provided warm and damp environment which was conducive for enteroviruses viability and HFMD transmission. Rainfall levels below 75 mm were more prevalent during the annual outbreak transmission season of HFMD, while rainfall above this threshold dominated in the wet season. This could potentially explain the contrasting risk ratios. On the other hand, lower ambient temperature and heavy downpour could help to interrupt the disease transmission partly through serving as a barrier for social gathering activity or contact with other children in public. Together with multiple risk factors including continued evolution and introduction of novel strains of EV71, the anticipated increasing extreme weather events or extreme temperature as a consequence of climate change may amplify the risks of HFMD outbreaks in the future.\nEnteroviruses can tolerate temperature fluctuation and survive well in sewage treatment and water environment [2]. Thermal effect on enteroviruses vary depending on temperature sensitive or resistant strains of serotypes as well as the types, pH, evaporation rate, and water content of viral environment [21], [22], [23], [24]. A laboratory study on recovery of enteroviruses in various soil conditions shows that thermal effect plays an important role on the infectivity or plague forming units of enteroviruses; the study indicates enteroviruses in sand amended with septic tank liquor can be recovered in up to 36 and 11 days at 22°C and 37°C, respectively [22]. Other studies have also revealed that temperature resistant strains of EV71 that link with fatal complications are not inactivated at 40°C [24], [25].\nThe insignificant result of minimum temperature when tested alone indicated that minimum temperature was not sufficient by itself in predicting risk of HFMD. Though maximum temperature performed well in predicting HFMD incidence, inclusion of minimum temperature could strengthen the predictive power of the model. This was likely due to difference in maximum and minimum temperature was influential on HFMD incidence as shown in figure 2a. Also, temperature difference was likely more influential on HFMD incidence in year 2008 as maximum temperature was below 32°C almost throughout the year.\nOur approach controls for non-climatic determinants of HFMD such as infectious disease control exercises, good hygiene practices, school holiday, and dominant strains of circulating enterovirus in the trends function. Since 1998, Singapore set up surveillance system and formed multi-sector HFMD Task Force to plan, monitor, and manage outbreaks in Singapore [15]. Public health measures that aimed to prevent HFMD outbreaks included thorough disinfection of virus contaminated premises and temporary closure of educational institutions such as childcare centers/preschools/kindergartens where disease transmission took place for more than 15 days [26]. Different dominating serotypes of enterovirus had been alternating during outbreaks in the period 2001 to 2008. CA16 was the dominant circulating virus during the outbreaks in year 2002, 2005, and 2007; whereas EV71 was the predominating strain for the epidemics in 2006 and 2008 [10], [26]. It was reported that EV71 affected significantly more cases during 2008 outbreaks, compared with 2006 [26].\nHFMD cases were lower during school holiday in June, November, and December. Separation of children during the school holiday reduced social contacts; thus, interrupted transmission in the childcare centers and led to reducing HFMD incidence. During the outbreak of deadly episodes of severe acute respiratory syndrome (SARS) in year 2003, school and childcare centers were closed for 2 additional weeks in March in response to concerned parents. Stringent preventive measures included daily body temperature taking in all schools across the nation and children with high body temperature were sent home. Additionally, children who had contact with suspected SARS patients or as siblings of contacts were advised to stay at home for minimum 10 days. Massive public education was launched to caution community to observe strict hygiene practices, avoid crowded areas, and other precautionary measures to prevent and disrupt the disease transmission of SARS. Though the outbreak of SARS was recorded between the months from March to May, the ripple effects of the preventive measures and the alertness among populations could partly be the reasons for lower HFMD incidence in this year.\nThe bimodal outbreaks in years 2005–2008 coincided mostly with weekly maximum temperature greater than 32°C and temperature difference above 7°C. Nonetheless, the second peak which occurred between 2–6 months after the first outbreak could also possibly be influenced by the extension of disease transmission from the first outbreak due to continual presence of viruses in the environment [26]. Enterovirus can remain in an infected person's feces for several weeks after onset of symptoms and also possibly persist for days or weeks on materials found in domestic or institution environment [18], [23], [27], [28]. The high population density in Singapore could also compound the disease transmission rate and sustainability of outbreak. Furthermore, the infective dose of coxsackie viruses requires for human is low; therefore, transmission of HFMD persists even with the presence of low amount of shed virus [23].\nAsymptomatic and unreported cases are common limitations encountered in the study of infectious diseases epidemiology. It could limit and bias the analysis as actual cases could be many times higher. The disease transmission pattern is also less clearly defined as asymptomatic patients represent an undetected source of infection. Furthermore, heavy traffic flow among populations and trade between Singapore and neighbor countries could also influence the size and duration of outbreaks in Singapore during study period.\nIn view of the links between warmer season and HFMD cases, climate change and global warming may increase susceptibility of areas or regions to the transmission of HFMD, which may possibly be another important emerging infectious disease affecting millions of life. It was reported that seasonal patterns of enteroviruses differed by geographical localities and that some tropical and subtropical countries experienced more outbreaks in the rainy season [29]. Thus, we encourage similar studies in diverse geographical areas to increase understanding of the impacts of weather on HFMD cases, as well as studies to establish the causal associations between meteorological determinants and pathways of HFMD infection.\n[SUBTITLE] Conclusion [SUBSECTION] We found a strong relationship between HFMD incidence and the preceding 1–2 weeks weather parameters after adjusting for other time varying factors. A maximum daily temperature above 32°C and rainfall up to 75 mm is expected to increase the HFMD incidence in the subsequent 1–2 weeks. These findings suggest that weather parameters can be used as early risk indicators for potential HFMD outbreaks. Implementing a simple weather-based early warning could help 1) local authority to heighten alert, intensify surveillance, and activate infection control measures to prevent or curb disease outbreaks; and 2) community to exercise vigilance and take precautionary actions such as good hygiene practices and isolation to disrupt HFMD transmission chain. However, future studies are required to confirm this relationship in other regions. Subsequent studies need to elucidate the chain of events following temperature and rainfall changes and their pathways to the increase of HFMD incidence.\nWe found a strong relationship between HFMD incidence and the preceding 1–2 weeks weather parameters after adjusting for other time varying factors. A maximum daily temperature above 32°C and rainfall up to 75 mm is expected to increase the HFMD incidence in the subsequent 1–2 weeks. These findings suggest that weather parameters can be used as early risk indicators for potential HFMD outbreaks. Implementing a simple weather-based early warning could help 1) local authority to heighten alert, intensify surveillance, and activate infection control measures to prevent or curb disease outbreaks; and 2) community to exercise vigilance and take precautionary actions such as good hygiene practices and isolation to disrupt HFMD transmission chain. However, future studies are required to confirm this relationship in other regions. Subsequent studies need to elucidate the chain of events following temperature and rainfall changes and their pathways to the increase of HFMD incidence.", "We found a strong relationship between HFMD incidence and the preceding 1–2 weeks weather parameters after adjusting for other time varying factors. A maximum daily temperature above 32°C and rainfall up to 75 mm is expected to increase the HFMD incidence in the subsequent 1–2 weeks. These findings suggest that weather parameters can be used as early risk indicators for potential HFMD outbreaks. Implementing a simple weather-based early warning could help 1) local authority to heighten alert, intensify surveillance, and activate infection control measures to prevent or curb disease outbreaks; and 2) community to exercise vigilance and take precautionary actions such as good hygiene practices and isolation to disrupt HFMD transmission chain. However, future studies are required to confirm this relationship in other regions. Subsequent studies need to elucidate the chain of events following temperature and rainfall changes and their pathways to the increase of HFMD incidence." ]

[ null, null, null, null ]

[ "Introduction", "Methods", "Results", "Discussion", "Conclusion" ]

[ "Hand foot and mouth disease (HFMD) is caused by a number of different enteroviruses, among which are enterovirus 71 (EV71) and coxsackie A16 (CA16). The disease is transmitted through direct contact with respiratory droplets, feces, and blister fluid of infective patients or through contact with contaminated environment such as water, food, or surface [1], [2]. HFMD inflicts mainly children with mild clinical symptoms include fever, blisters and sores in mouth, palms and soles following 3–7 days of incubation period and a patient generally recovers in 7–10 days. Nevertheless, severe health consequences or death may occur owing to complications such as encephalitis, aseptic meningitis, and acute flaccid paralysis mainly following EV71 infection. HFMD can be asymptomatic and it is possible for a recovered person to be infected again by different serotypes of enteroviruses [1]. There is no specific treatment or vaccine available; therefore, preventive measures such as avoid direct contact with infective patients, disinfection of viral contaminated items or premises, and good personal hygiene practices remain the only effective methods to disrupt disease transmission.\nIn recent decade, Asian countries have experienced increasing trend of HFMD outbreak with deaths among children due to severe complications [3], [4], [5]. HFMD has raised public health concerns in Asia following severe outbreaks in Malaysia and Taiwan in 1997 and 1998, respectively. About 6000 young children were infected with 42 deaths during the HFMD outbreak in Malaysia [6]. Whereas a total of 129,106 cases were reported in Taiwan with 405 children suffered complications that led to 78 deaths [7], [8]. In the first half of 2010, mainland China experienced 72% increase in HFMD cases compared with 2009; resulting in approximately 1.3 million reported cases [5]. Also, 5454 severe cases that caused 260 deaths were reported in mainland China between January to the first week of May [5]. HFMD outbreaks generally occur in 2–3 years cyclical pattern in endemic countries across the Western Pacific Region and EV71-related HFMD outbreak frequency is expected to increase in the region partly owing to complex factors including continued evolution and emergence of novel recombinants of EV71, inadequate healthcare capacity, and a lack of effective surveillance system in some countries [9].\nHFMD is endemic in Singapore with EV71 and CA16 as two main dominant circulating strains. About 90% of the reported cases were children below 10 years old. The number of HFMD outbreaks reported from childcare centers, kindergartens, and preschools had escalated from about 167 in year 2001 to more than 1700 in year 2007. In the same period, the incidence rate among children aged 0–4 years surged from approximately 16 to 60 per 1000 populations and from 3 to 21 per 1000 populations among children 5–9 years old [10].\nHFMD is endemic in the tropical and subtropical countries with tendency of higher number of cases in wet season depending on geographical locations [5], [11]. Whereas outbreaks usually occur in summer or early fall in temperate countries. Enterovirus surveillance in USA for the period 1970–2005 showed that EV71 and CA16 had endemic circulation pattern and that around 70% of cases were reported during warmer seasons between June–October [12]. Considering the seasonal pattern of HFMD outbreaks, we hypothesized that short-term changes in weather can influence the transmission dynamic of HFMD. This study aims to establish a relationship, analyze the effects and estimate potential thresholds of weekly temperature and rainfall associated with the risk of HFMD outbreaks in Singapore.", "Singapore is an island state nation with land size of approximately 710 km2 and population density of 7000 persons per km2\n[13]. The island experiences tropical climate with high temperature, humidity, and rainfall.\nWeekly cases of HFMD for the period 2001–2008 were obtained from the Weekly Infectious Diseases Bulletins of Communicable Diseases Division, Ministry of Health Singapore [14]. Report of HFMD cases from physicians, education institutions, and laboratories was mandatory in Singapore since October 2000 [15], [16]. Data of daily temperature and rainfall were retrieved from National Climatic Data Center, National Oceanic and Atmospheric Administration (NOAA), USA [17]. Weekly average temperature and cumulative rainfall were computed or aggregated from daily weather data. Weekly temperature difference (Tp) was computed as the difference between weekly average maximum and minimum temperature.\nWe established time series Poisson regression models to analyze the relationship between weather and HFMD cases adjusting for long-term time trends and seasonality. Time trends, driven by other factors such as circulating virus serotypes, disease control measures, and social behavior, could confound the relationship between temperature (e.g. seasonality) and HFMD. Therefore, trend and seasonality were adjusted to account for time varying factors that were influential on weekly HFMD incidence during the study period. We modeled the seasonality and long-term time trends in one function allowing the seasonality (driven by unknown factors) of HFMD to change between years as that allows more flexible adjustment for the potential confounding. We used smooth functions of natural cubic splines allowing 12 degrees of freedom (df) to adjust for seasonality and long-term time trends. Sensitivity to the effect estimates the flexibility of the smooth function of time trends was further tested using df of 8, 10, 15, and 20. The sensitivity tests did not show significantly different results of the estimated risk functions. We tested the lag time between temperature, rainfall, and HFMD by including lag terms 1–2 and 3–4 weeks using a backward stepwise model fitting procedure. The tests indicated insignificant results for lag term 3–4 weeks; thus, only lag term 1–2 weeks was included in the model. We analyzed the risk of HFMD as function of temperature and rainfall using 2 different models. In Model A, we included temperature difference (Tp) as the only temperature parameter; whereas both weekly minimum and maximum temperature were included in Model B. Model A was included to demonstrate risk function without potential co-linearity bias due to correlation between minimum and maximum temperature. To test the sensitivity of Model B for the co-linearity between minimum and maximum temperature, we repeated analysis using either minimum or maximum temperature parameter in each test. At a first stage, we allowed a non-linear exposure-response relationship using natural cubic splines with 4 df. In a second stage, we estimated the risk ratio of the relationship between weather parameters and HFMD incidence using piecewise linear spline functions where such approximation appeared applicable. One characteristic of infectious disease is the serial correlation between past and current incidence. Examination of the time series using autocorrelation function (ACF) indicated serial correlation for consecutive lags of HFMD cases; thus, we included autoregressive term of time lag 1–2 weeks based on average infectious and recovery period of a patient [1], [18].\nModel A =  model with temperature differences and rainfall\n\n\nModel B  =  model with minimum temperature, maximum temperature, and rainfall\n\nWhere Log (µ(t)) is the mean predicted weekly cases of HFMD; t equals the week in year 2001–2008; β0 represents the intercept; tmaxt represents maximum temperature; tmint represents minimum temperature; Tpt means temperature difference; S denotes a cubic spline function with corresponding degrees of freedom (df); raint is the cumulative rainfall; trend equals to week number running from 1 in the first week of year 2001; and hfmdt is an autoregressive term of HFMD.\nThe optimal model and parameters were selected and validated using post estimation residuals and the Akaike's information criterion (AIC). The post estimation was performed by plotting predicted residuals against observed data in addition to the partial autocorrelation function (PACF), scatter plots, normality tests and histogram of residuals. The risk ratio of HFMD incidence was presented as function of weather in relation to a minimum point of the curve or as corresponding to a unit increase of a particular weather predictor. All estimates were presented with corresponding 95% confidence intervals. Statistical analyses were conducted using R 2.10.1 [19] and STATA 11.1 (StataCorp, USA).", "During the study period, Singapore experienced nationwide HFMD outbreaks in March-May in year 2002 and 2005–2008. Bimodal outbreaks occurred yearly with second outbreak in August-October during 2005 to 2008.\nMinimum and maximum temperature during the study period ranged from 22.8–27.6°C and 27.7–34.6°C, respectively. The highest weekly maximum temperature and lowest minimum temperature was recorded in 2005 and 2008 respectively. Overall, Singapore experienced temperature above 32°C in one-third of the study period with a total of 30 weeks in 2002 and about 6 weeks in 2008 (Figure 1). A total of 30 weeks with temperature difference (Tp) above 7°C was observed in year 2005, and highest weekly Tp of 8.8°C was recorded in 2006 and 2008. The weekly cumulative rainfall during the study period ranged from 0–388 mm with higher amount of rainfall recorded between October and January. About 80% of the weekly cumulative rainfall was below 75 mm.\nThe HFMD incidence was significantly associated with short term variability of weekly temperature difference (Tp), minimum temperature, maximum temperature, and cumulative rainfall at time lag of 1–2 weeks. Figure 2a depicts a wide J-shaped relationship between HFMD incidence and Tp with significant increase in the risk of HFMD incidence when Tp is above 7°C. Likewise, the risk of HFMD incidence rises steeply when maximum temperature is above 32°C; while low maximum temperature poses negligible effects on HFMD (Figure 2c). Inverse relationship is observed between minimum temperature and HFMD incidence (Figure 2d). Similar results are obtained when we test minimum and maximum temperature independently, except that minimum temperature is not statistically significant (Figure 2e & 2f). Figure 2b indicates the relative risk of HFMD incidence increases linearly with weekly cumulative rainfall between 0–75 mm; whereas every unit increases beyond 75 mm reduces the risk of HFMD incidence. During the study period, Tp>7°C mainly coincided with maximum temperature above 32°C and low rainfall, except year 2008.\nRelative risks of HFMD incidence as function of weekly (2a) temperature difference (Tp) & (2b) weekly cumulative rainfall at time lag of 1–2 weeks using model A; (2c) maximum temperature & (2d) minimum temperature at time lag of 1–2 weeks using model B; (2e) maximum temperature tested singly & (2f) minimum temperature tested singly at time lag of 1–2 weeks using model B (Shaded area: confidence interval).\nIn the piecewise linear Poisson regression function, we used breakpoints of 32°C for maximum temperature, 25°C for minimum temperature, 7°C for Tp, and 75 mm for rainfall. Table 1 shows every 1°C increases in maximum temperature above 32°C and in weekly Tp above 7°C elevates the risks of HFMD incidence significantly by 36% and 41%, respectively. The risk, however, declines by 17% with each degree increases in minimum temperature above 25°C. Additionally, a unit (mm) increases in weekly cumulative rainfall between 0–75 mm elevates risk of HFMD incidence by 0.3%.\nObserved and predicted time series of HFMD using both models produced similar results. Both model A and B explained about 87% of variations among HFMD cases using temperature, rainfall, and trend parameters. The residuals plots indicated model residuals did not violate the statistical modeling assumptions. Furthermore, the PACF showed no indication of further non-adjusted residual autocorrelation.", "To date, very few studies document the link between weather and HFMD and none (that the authors are aware of) have previously shown that there exists an explicit relationship with threshold effects between short-term changes in weather and the incidence of HFMD. Our findings showed that high weekly maximum temperature and a large difference of minimum and maximum temperature increased the risk of HFMD incidence in the subsequent 1–2 weeks. The risk functions of minimum and maximum temperature provide more insights into the J-shaped curve of the relation between temperature difference and HFMD cases. Moreover, moderate weekly cumulative rainfall partly sustained the HFMD endemic during the study period. A study on the effects of climate on Herpangina and HFMD in Japan indicates that higher temperature could have influenced the increase of Herpangina & HFMD incidence [20].\nExact reasons for the relationship between weather and HFMD are not known. During the study period, high temperature in Singapore generally accompanied with rainfall below 75 mm and thus provided warm and damp environment which was conducive for enteroviruses viability and HFMD transmission. Rainfall levels below 75 mm were more prevalent during the annual outbreak transmission season of HFMD, while rainfall above this threshold dominated in the wet season. This could potentially explain the contrasting risk ratios. On the other hand, lower ambient temperature and heavy downpour could help to interrupt the disease transmission partly through serving as a barrier for social gathering activity or contact with other children in public. Together with multiple risk factors including continued evolution and introduction of novel strains of EV71, the anticipated increasing extreme weather events or extreme temperature as a consequence of climate change may amplify the risks of HFMD outbreaks in the future.\nEnteroviruses can tolerate temperature fluctuation and survive well in sewage treatment and water environment [2]. Thermal effect on enteroviruses vary depending on temperature sensitive or resistant strains of serotypes as well as the types, pH, evaporation rate, and water content of viral environment [21], [22], [23], [24]. A laboratory study on recovery of enteroviruses in various soil conditions shows that thermal effect plays an important role on the infectivity or plague forming units of enteroviruses; the study indicates enteroviruses in sand amended with septic tank liquor can be recovered in up to 36 and 11 days at 22°C and 37°C, respectively [22]. Other studies have also revealed that temperature resistant strains of EV71 that link with fatal complications are not inactivated at 40°C [24], [25].\nThe insignificant result of minimum temperature when tested alone indicated that minimum temperature was not sufficient by itself in predicting risk of HFMD. Though maximum temperature performed well in predicting HFMD incidence, inclusion of minimum temperature could strengthen the predictive power of the model. This was likely due to difference in maximum and minimum temperature was influential on HFMD incidence as shown in figure 2a. Also, temperature difference was likely more influential on HFMD incidence in year 2008 as maximum temperature was below 32°C almost throughout the year.\nOur approach controls for non-climatic determinants of HFMD such as infectious disease control exercises, good hygiene practices, school holiday, and dominant strains of circulating enterovirus in the trends function. Since 1998, Singapore set up surveillance system and formed multi-sector HFMD Task Force to plan, monitor, and manage outbreaks in Singapore [15]. Public health measures that aimed to prevent HFMD outbreaks included thorough disinfection of virus contaminated premises and temporary closure of educational institutions such as childcare centers/preschools/kindergartens where disease transmission took place for more than 15 days [26]. Different dominating serotypes of enterovirus had been alternating during outbreaks in the period 2001 to 2008. CA16 was the dominant circulating virus during the outbreaks in year 2002, 2005, and 2007; whereas EV71 was the predominating strain for the epidemics in 2006 and 2008 [10], [26]. It was reported that EV71 affected significantly more cases during 2008 outbreaks, compared with 2006 [26].\nHFMD cases were lower during school holiday in June, November, and December. Separation of children during the school holiday reduced social contacts; thus, interrupted transmission in the childcare centers and led to reducing HFMD incidence. During the outbreak of deadly episodes of severe acute respiratory syndrome (SARS) in year 2003, school and childcare centers were closed for 2 additional weeks in March in response to concerned parents. Stringent preventive measures included daily body temperature taking in all schools across the nation and children with high body temperature were sent home. Additionally, children who had contact with suspected SARS patients or as siblings of contacts were advised to stay at home for minimum 10 days. Massive public education was launched to caution community to observe strict hygiene practices, avoid crowded areas, and other precautionary measures to prevent and disrupt the disease transmission of SARS. Though the outbreak of SARS was recorded between the months from March to May, the ripple effects of the preventive measures and the alertness among populations could partly be the reasons for lower HFMD incidence in this year.\nThe bimodal outbreaks in years 2005–2008 coincided mostly with weekly maximum temperature greater than 32°C and temperature difference above 7°C. Nonetheless, the second peak which occurred between 2–6 months after the first outbreak could also possibly be influenced by the extension of disease transmission from the first outbreak due to continual presence of viruses in the environment [26]. Enterovirus can remain in an infected person's feces for several weeks after onset of symptoms and also possibly persist for days or weeks on materials found in domestic or institution environment [18], [23], [27], [28]. The high population density in Singapore could also compound the disease transmission rate and sustainability of outbreak. Furthermore, the infective dose of coxsackie viruses requires for human is low; therefore, transmission of HFMD persists even with the presence of low amount of shed virus [23].\nAsymptomatic and unreported cases are common limitations encountered in the study of infectious diseases epidemiology. It could limit and bias the analysis as actual cases could be many times higher. The disease transmission pattern is also less clearly defined as asymptomatic patients represent an undetected source of infection. Furthermore, heavy traffic flow among populations and trade between Singapore and neighbor countries could also influence the size and duration of outbreaks in Singapore during study period.\nIn view of the links between warmer season and HFMD cases, climate change and global warming may increase susceptibility of areas or regions to the transmission of HFMD, which may possibly be another important emerging infectious disease affecting millions of life. It was reported that seasonal patterns of enteroviruses differed by geographical localities and that some tropical and subtropical countries experienced more outbreaks in the rainy season [29]. Thus, we encourage similar studies in diverse geographical areas to increase understanding of the impacts of weather on HFMD cases, as well as studies to establish the causal associations between meteorological determinants and pathways of HFMD infection.\n[SUBTITLE] Conclusion [SUBSECTION] We found a strong relationship between HFMD incidence and the preceding 1–2 weeks weather parameters after adjusting for other time varying factors. A maximum daily temperature above 32°C and rainfall up to 75 mm is expected to increase the HFMD incidence in the subsequent 1–2 weeks. These findings suggest that weather parameters can be used as early risk indicators for potential HFMD outbreaks. Implementing a simple weather-based early warning could help 1) local authority to heighten alert, intensify surveillance, and activate infection control measures to prevent or curb disease outbreaks; and 2) community to exercise vigilance and take precautionary actions such as good hygiene practices and isolation to disrupt HFMD transmission chain. However, future studies are required to confirm this relationship in other regions. Subsequent studies need to elucidate the chain of events following temperature and rainfall changes and their pathways to the increase of HFMD incidence.\nWe found a strong relationship between HFMD incidence and the preceding 1–2 weeks weather parameters after adjusting for other time varying factors. A maximum daily temperature above 32°C and rainfall up to 75 mm is expected to increase the HFMD incidence in the subsequent 1–2 weeks. These findings suggest that weather parameters can be used as early risk indicators for potential HFMD outbreaks. Implementing a simple weather-based early warning could help 1) local authority to heighten alert, intensify surveillance, and activate infection control measures to prevent or curb disease outbreaks; and 2) community to exercise vigilance and take precautionary actions such as good hygiene practices and isolation to disrupt HFMD transmission chain. However, future studies are required to confirm this relationship in other regions. Subsequent studies need to elucidate the chain of events following temperature and rainfall changes and their pathways to the increase of HFMD incidence.", "We found a strong relationship between HFMD incidence and the preceding 1–2 weeks weather parameters after adjusting for other time varying factors. A maximum daily temperature above 32°C and rainfall up to 75 mm is expected to increase the HFMD incidence in the subsequent 1–2 weeks. These findings suggest that weather parameters can be used as early risk indicators for potential HFMD outbreaks. Implementing a simple weather-based early warning could help 1) local authority to heighten alert, intensify surveillance, and activate infection control measures to prevent or curb disease outbreaks; and 2) community to exercise vigilance and take precautionary actions such as good hygiene practices and isolation to disrupt HFMD transmission chain. However, future studies are required to confirm this relationship in other regions. Subsequent studies need to elucidate the chain of events following temperature and rainfall changes and their pathways to the increase of HFMD incidence." ]

[ null, "methods", null, null, null ]

[]

[ "Cost of Illness", "Geography", "HIV Infections", "Humans", "Public Health", "Public Policy", "Publishing", "Randomized Controlled Trials as Topic", "Research", "Smoking" ]

[ "Introduction", "I. Classification of countries", "II. Where research is conducted", "a. Identification of RCTs included in Cochrane systematic reviews", "b. Identification of ongoing RCTs", "c. Data collection", "III. Where research is needed", "1. Prevalence of tobacco use and HIV infection", "2. Mortality and global burden of disease (GBD)", "IV. Geographic representation", "Statistical analysis", "Results", "Where research is conducted", "Geographic representation", "Tobacco use trials", "HIV infection trials", "Discussion" ]

[ "The main challenge for public health policies is to reduce mortality and the burden of disease. This aim implies having adequate evidence to guide health care providers and policymakers for implementing the most efficient and cost-effective interventions. More than two-thirds of the world's population live in low- and middle-income countries [1], [2], and 93% of the burden of preventable mortality occurs in these countries [3]. The shortage of resources in low- and middle-income countries paradoxically increases the need for reliable healthcare evidence to prioritize the use of these scarce resources [4], [5]. Such evidence is essential to determine the prevention and therapeutic strategies that work best but also under which circumstances they work and how best to deliver them [6].\nIn this context, it could be useful to tailor research to the needs of the particular population and to the context (socio-cultural, access to services, etc.). Addressing context-specific questions is fundamental to designing interventions that improve health [7], [8]. Results of research performed in high-income countries cannot be easily transposed to low- and middle-income countries. The extent to which questions addressed by researchers in high-income countries are relevant to patients and physicians in low-income countries is largely unknown.\nThis study aimed to compare where research is conducted to where research is needed according to the attributable burden of disease, mortality and prevalence. We compared high-income countries to low- or middle-income countries defined by the 2008 World Bank classification. We considered both current available evidence (i.e., randomized controlled trials [RCTs] included in systematic reviews) and awaiting evidence (i.e., RCTs registered on the World Health Organization's International Clinical Trials Registry Platform [WHO-ICTRP]) that provide data for current clinical decision making and future public health policies. We focused on 2 public-health priorities, tobacco use and HIV infection, because they are ranked among the top 5 leading causes of mortality in the world [9].", "We classified the countries according to the 2008 World Bank classification that divides the world into income categories (low, middle, and high) according to economies in terms of gross region national income per capita and in geographic regions for low- and middle-income economies only. We considered trials performed in high-income countries (country with a per capita income of ≥$11 906 in 2008 [e.g., USA, European Union [EU], Canada, Australia and other]) as one group and trials performed in low- and middle-income countries as another group [9], [10]. In addition, we grouped low- and middle-income countries into 6 geographic regions according to the 2008 World Bank classification (i.e., East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa).", "To describe the context (i.e., geographic region and socioeconomic situation of the countries where the trial took place) of research performed in these fields, we focused on the context of currently available evidence and awaiting evidence that could be used for clinical decision making. Because RCTs are the gold standard to assess the effectiveness of interventions, we evaluated the context of all RCTs included in Cochrane systematic reviews (i.e., context of current evidence) and all ongoing RCTs in the registries of the WHO-ICTRP (i.e., context of awaiting evidence). We focused on systematic reviews performed by the Cochrane collaboration because this international, not-for-profit organisation is a source of high-quality, reliable health information providing up-to-date knowledge about the effects of health care. Further, an important goal of the Cochrane collaboration is the dissemination of information to low- and middle-income countries. The Cochrane Library is freely available to all residents of low-income countries.\n[SUBTITLE] a. Identification of RCTs included in Cochrane systematic reviews [SUBSECTION] We identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).\nWe identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).\n[SUBTITLE] b. Identification of ongoing RCTs [SUBSECTION] In January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.\nIn January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.\n[SUBTITLE] c. Data collection [SUBSECTION] One of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.\nOne of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.", "We identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).", "In January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.", "One of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.", "The following indicators were elaborated:\n[SUBTITLE] 1. Prevalence of tobacco use and HIV infection [SUBSECTION] To obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].\nTo obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].\n[SUBTITLE] 2. Mortality and global burden of disease (GBD) [SUBSECTION] The global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].\nThe global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].", "To obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].", "The global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].", "For an adequate representation of the contrast between the area where trials are needed and the area where trials are conducted, we constructed an area cartogram according to the method developed by Gastner and colleagues [17]. Area cartograms are maps in which the sizes of geographic regions such as countries are not proportional to the area on the ground but, rather, appear in proportion to their population or some other demographic feature such as disease prevalence. We used such maps to compare the area for number of people smoking and with HIV infection to where RCTs have been and are being conducted.", "We used descriptive statistics; categorical variables are described with frequencies and percentages. All data analyses involved use of SAS for Windows, Release 9.1 (SAS Inst., Cary, NC).\nCartographic representation was developed with use of Migratio 8.0 (Stéphane Le Rouzic, Migratio.fr, Rennes, France) and ScapeToad cartography software (Chôros Laboratory, EPFL-ENAC-INTER, Lausanne, Switzerland).", "[SUBTITLE] Where research is conducted [SUBSECTION] The flow of RCTs included in the selected systematic reviews and ongoing RCTs through the study is in Figures 1 and 2, respectively. The electronic search yielded 118 systematic reviews, of which 57 (28 of tobacco use, 29 of HIV infection) were selected. From these systematic reviews, data for 1 012 trials were retrieved; 143 duplicate reports and reports of 129 nonrandomised trials were excluded. We included in the final analysis data for 740 RCTs (556 of tobacco use, 184 of HIV infection). Of the 940 ongoing RCTs (266 of tobacco use, 674 of HIV infection) identified in the international clinical trial registries, we included in the final analysis data for 346 (112 of tobacco use, 234 of HIV infection).\nThe characteristics of the RCTs and ongoing RCTs are in Table 1.\n* International clinical trial registries: Clinicaltrials.gov, Australian New Zealand Clinical Trials Registry, Chinese Clinical Trial Register, Clinical Trials Registry - India, German Clinical Trials Register, Iranian Registry of Clinical Trials, Japan Primary Registries Network, International Standard Randomised Controlled Trials Number, The Netherlands National Trial Register, Sri Lanka Clinical Trials Registry.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention.\nThe flow of RCTs included in the selected systematic reviews and ongoing RCTs through the study is in Figures 1 and 2, respectively. The electronic search yielded 118 systematic reviews, of which 57 (28 of tobacco use, 29 of HIV infection) were selected. From these systematic reviews, data for 1 012 trials were retrieved; 143 duplicate reports and reports of 129 nonrandomised trials were excluded. We included in the final analysis data for 740 RCTs (556 of tobacco use, 184 of HIV infection). Of the 940 ongoing RCTs (266 of tobacco use, 674 of HIV infection) identified in the international clinical trial registries, we included in the final analysis data for 346 (112 of tobacco use, 234 of HIV infection).\nThe characteristics of the RCTs and ongoing RCTs are in Table 1.\n* International clinical trial registries: Clinicaltrials.gov, Australian New Zealand Clinical Trials Registry, Chinese Clinical Trial Register, Clinical Trials Registry - India, German Clinical Trials Register, Iranian Registry of Clinical Trials, Japan Primary Registries Network, International Standard Randomised Controlled Trials Number, The Netherlands National Trial Register, Sri Lanka Clinical Trials Registry.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention.\n[SUBTITLE] Geographic representation [SUBSECTION] [SUBTITLE] Tobacco use trials [SUBSECTION] The geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nThe geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\n[SUBTITLE] HIV infection trials [SUBSECTION] The geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nThe geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\n[SUBTITLE] Tobacco use trials [SUBSECTION] The geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nThe geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\n[SUBTITLE] HIV infection trials [SUBSECTION] The geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nThe geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.", "The flow of RCTs included in the selected systematic reviews and ongoing RCTs through the study is in Figures 1 and 2, respectively. The electronic search yielded 118 systematic reviews, of which 57 (28 of tobacco use, 29 of HIV infection) were selected. From these systematic reviews, data for 1 012 trials were retrieved; 143 duplicate reports and reports of 129 nonrandomised trials were excluded. We included in the final analysis data for 740 RCTs (556 of tobacco use, 184 of HIV infection). Of the 940 ongoing RCTs (266 of tobacco use, 674 of HIV infection) identified in the international clinical trial registries, we included in the final analysis data for 346 (112 of tobacco use, 234 of HIV infection).\nThe characteristics of the RCTs and ongoing RCTs are in Table 1.\n* International clinical trial registries: Clinicaltrials.gov, Australian New Zealand Clinical Trials Registry, Chinese Clinical Trial Register, Clinical Trials Registry - India, German Clinical Trials Register, Iranian Registry of Clinical Trials, Japan Primary Registries Network, International Standard Randomised Controlled Trials Number, The Netherlands National Trial Register, Sri Lanka Clinical Trials Registry.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention.", "[SUBTITLE] Tobacco use trials [SUBSECTION] The geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nThe geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\n[SUBTITLE] HIV infection trials [SUBSECTION] The geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nThe geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.", "The geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.", "The geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.", "This study evaluated the context of clinical research (i.e., where research is performed) for 2 public health priorities: tobacco use and HIV infection. We described the location of 740 RCTs included in Cochrane systematic reviews and 346 ongoing RCTs identified through international clinical trial registries. Our results highlight a gap between where trials are conducted and where research is needed in terms of GBD, mortality and prevalence of tobacco use and HIV infection. Most ongoing and RCTs were performed in high-income countries, even though most of the mortality and GBD concerns low- and middle-income countries. Only 4% and 32% of the total and ongoing RCTs performed in the fields of tobacco use and HIV infection, respectively, were conducted in low- and middle-income countries. However, low- and middle-income countries represent 70% and 99% of the mortality attributable to tobacco use and HIV infection, respectively [13], [18].\nThe number of trials of tobacco use performed in low- and middle-income countries was particularly low (4%) as compared with trials of HIV infection performed in these countries (about one-third).\nPrevious studies have highlighted the underrepresentation of research addressing priority issues for low- and middle-income countries [19]. Some authors highlighted the poor association of GBD and reports of RCTs published in high-impact-factor journals [19], [20]. Swingler et al. [21] analysed nearly 3 000 systematic reviews and assessed the correlation between the number of systematic reviews undertaken and the burden of disease: only a few systematic reviews focused on some diseases affecting a large number of the world's population. Sheriff showed that only 3% of mental disease research is performed in low- and middle-income countries [22], but the religious and cultural individuality of these countries could greatly hamper implementation of any mental health interventions recommended by research performed in high-income countries. These low research percentages are probably related to the limited financial investment in research for low- and middle-income countries because of the phenomenon of the “10/90 gap”. The Global Forum has highlighted that of the US$73 billion invested annually in global health research, less than 10% is spent on research into the health problems that account for 90% of the GBD [4], [23], [24]. This gap is also reflected in the low proportion of publications from research performed in low- and middle-income countries [25].\nContrary to previous research, we focused on 2 public health priorities (tobacco use, responsible for non-communicable chronic diseases, and HIV infection, a communicable chronic disease) relevant for high-, low- and middle-income countries. Our study is the first to focus on existing evidence (trials included in systematic reviews) and on awaiting evidence that will guide future evidence-based practices.\nWith the evidence of the effectiveness of an intervention, some might assume that the challenge is to make the intervention available in low- and middle-income countries [26]. However, the results of trials performed in high-income countries may not be easily applied to low- and middle-income countries. Many interventions shown to be efficacious in high-income countries are not similarly effective when carried out in other contexts [26], [27]. The populations can differ, with people consulting late, frequently using self-medication, unable or unwilling to adhere to treatment, and having several co-morbidities (malnourishment, anaemia, malaria, etc.), in addition to the behavioural and cultural differences [28], [29]. Further, the contextual factors differ greatly, particularly the social and cultural context, as well as the available facilities and infrastructures. The lack of studies performed in a relevant location is particularly problematic for nonpharmacologic interventions targeting behavioural changes (e.g., education, counselling), because the results of such trials could be strongly influenced by cultural conditions [20], [30]. The influence of the cultural context is important, even between different high-income countries [31].\nRecently, the Global Alliance for Chronic Disease (GACD) was created to address this imbalance issue [32]. The GACD brings together 6 major national health research councils representing 80% of all public research funding in the world to coordinate research activities that address the prevention and treatment of chronic disease on a global scale. One essential goal of the GACD is to focus on chronic disease in low- and middle-income countries and to coordinate research into low-cost interventions. This initiative should be important in limiting the worldwide imbalance in health research resources [7].\nSimilarly, the High Level Taskforce on Innovative International Financing for Health Systems was created in September 2008 to help fill national financing gaps to reach the health Millennium Development Goals in low-income countries. One key challenge is to obtain evidence for what works according to different settings [33].\nOur study has some limitations. First, we focused on only 2 medical areas, and these results should be confirmed in other medical areas. However, we chose tobacco use and HIV infection because they are among the first 5 causes of mortality in the world [9]. Second, we selected only RCTs included in Cochrane systematic reviews because these systematic reviews are known to be of high quality, and we did not consider non-Cochrane systematic reviews [34], [35], [36], [37], [38]. As well, the Cochrane collaboration seeks to actively encourage the participation of reviewers from developing countries and to prioritize reviews focusing on determinants of health that are particularly pertinent to low- and middle-income countries [1], [4]. Finally, our results may underestimate the number of ongoing RCTs performed in low- and middle-income countries because we focused on only available data (i.e., trials that were registered in a trial registry available on the WHO portal). We cannot exclude that some trials performed in low-income countries were not registered.\nIn conclusion, clinical research into tobacco use and HIV infection is not performed in the world locations most affected in terms of GBD and mortality because we found important underrepresentation of such research in low- and middle-income countries. The GBD measured by mortality or DALYs could be a useful aid in deciding where to conduct health research. The measure should help in conducting RCTs in an appropriate context to improve clinical research practice and decrease the GBD [39]." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Methods", "I. Classification of countries", "II. Where research is conducted", "a. Identification of RCTs included in Cochrane systematic reviews", "b. Identification of ongoing RCTs", "c. Data collection", "III. Where research is needed", "1. Prevalence of tobacco use and HIV infection", "2. Mortality and global burden of disease (GBD)", "IV. Geographic representation", "Statistical analysis", "Results", "Where research is conducted", "Geographic representation", "Tobacco use trials", "HIV infection trials", "Discussion" ]

[ "The main challenge for public health policies is to reduce mortality and the burden of disease. This aim implies having adequate evidence to guide health care providers and policymakers for implementing the most efficient and cost-effective interventions. More than two-thirds of the world's population live in low- and middle-income countries [1], [2], and 93% of the burden of preventable mortality occurs in these countries [3]. The shortage of resources in low- and middle-income countries paradoxically increases the need for reliable healthcare evidence to prioritize the use of these scarce resources [4], [5]. Such evidence is essential to determine the prevention and therapeutic strategies that work best but also under which circumstances they work and how best to deliver them [6].\nIn this context, it could be useful to tailor research to the needs of the particular population and to the context (socio-cultural, access to services, etc.). Addressing context-specific questions is fundamental to designing interventions that improve health [7], [8]. Results of research performed in high-income countries cannot be easily transposed to low- and middle-income countries. The extent to which questions addressed by researchers in high-income countries are relevant to patients and physicians in low-income countries is largely unknown.\nThis study aimed to compare where research is conducted to where research is needed according to the attributable burden of disease, mortality and prevalence. We compared high-income countries to low- or middle-income countries defined by the 2008 World Bank classification. We considered both current available evidence (i.e., randomized controlled trials [RCTs] included in systematic reviews) and awaiting evidence (i.e., RCTs registered on the World Health Organization's International Clinical Trials Registry Platform [WHO-ICTRP]) that provide data for current clinical decision making and future public health policies. We focused on 2 public-health priorities, tobacco use and HIV infection, because they are ranked among the top 5 leading causes of mortality in the world [9].", "[SUBTITLE] I. Classification of countries [SUBSECTION] We classified the countries according to the 2008 World Bank classification that divides the world into income categories (low, middle, and high) according to economies in terms of gross region national income per capita and in geographic regions for low- and middle-income economies only. We considered trials performed in high-income countries (country with a per capita income of ≥$11 906 in 2008 [e.g., USA, European Union [EU], Canada, Australia and other]) as one group and trials performed in low- and middle-income countries as another group [9], [10]. In addition, we grouped low- and middle-income countries into 6 geographic regions according to the 2008 World Bank classification (i.e., East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa).\nWe classified the countries according to the 2008 World Bank classification that divides the world into income categories (low, middle, and high) according to economies in terms of gross region national income per capita and in geographic regions for low- and middle-income economies only. We considered trials performed in high-income countries (country with a per capita income of ≥$11 906 in 2008 [e.g., USA, European Union [EU], Canada, Australia and other]) as one group and trials performed in low- and middle-income countries as another group [9], [10]. In addition, we grouped low- and middle-income countries into 6 geographic regions according to the 2008 World Bank classification (i.e., East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa).\n[SUBTITLE] II. Where research is conducted [SUBSECTION] To describe the context (i.e., geographic region and socioeconomic situation of the countries where the trial took place) of research performed in these fields, we focused on the context of currently available evidence and awaiting evidence that could be used for clinical decision making. Because RCTs are the gold standard to assess the effectiveness of interventions, we evaluated the context of all RCTs included in Cochrane systematic reviews (i.e., context of current evidence) and all ongoing RCTs in the registries of the WHO-ICTRP (i.e., context of awaiting evidence). We focused on systematic reviews performed by the Cochrane collaboration because this international, not-for-profit organisation is a source of high-quality, reliable health information providing up-to-date knowledge about the effects of health care. Further, an important goal of the Cochrane collaboration is the dissemination of information to low- and middle-income countries. The Cochrane Library is freely available to all residents of low-income countries.\n[SUBTITLE] a. Identification of RCTs included in Cochrane systematic reviews [SUBSECTION] We identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).\nWe identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).\n[SUBTITLE] b. Identification of ongoing RCTs [SUBSECTION] In January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.\nIn January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.\n[SUBTITLE] c. Data collection [SUBSECTION] One of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.\nOne of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.\nTo describe the context (i.e., geographic region and socioeconomic situation of the countries where the trial took place) of research performed in these fields, we focused on the context of currently available evidence and awaiting evidence that could be used for clinical decision making. Because RCTs are the gold standard to assess the effectiveness of interventions, we evaluated the context of all RCTs included in Cochrane systematic reviews (i.e., context of current evidence) and all ongoing RCTs in the registries of the WHO-ICTRP (i.e., context of awaiting evidence). We focused on systematic reviews performed by the Cochrane collaboration because this international, not-for-profit organisation is a source of high-quality, reliable health information providing up-to-date knowledge about the effects of health care. Further, an important goal of the Cochrane collaboration is the dissemination of information to low- and middle-income countries. The Cochrane Library is freely available to all residents of low-income countries.\n[SUBTITLE] a. Identification of RCTs included in Cochrane systematic reviews [SUBSECTION] We identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).\nWe identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).\n[SUBTITLE] b. Identification of ongoing RCTs [SUBSECTION] In January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.\nIn January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.\n[SUBTITLE] c. Data collection [SUBSECTION] One of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.\nOne of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.\n[SUBTITLE] III. Where research is needed [SUBSECTION] The following indicators were elaborated:\n[SUBTITLE] 1. Prevalence of tobacco use and HIV infection [SUBSECTION] To obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].\nTo obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].\n[SUBTITLE] 2. Mortality and global burden of disease (GBD) [SUBSECTION] The global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].\nThe global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].\nThe following indicators were elaborated:\n[SUBTITLE] 1. Prevalence of tobacco use and HIV infection [SUBSECTION] To obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].\nTo obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].\n[SUBTITLE] 2. Mortality and global burden of disease (GBD) [SUBSECTION] The global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].\nThe global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].\n[SUBTITLE] IV. Geographic representation [SUBSECTION] For an adequate representation of the contrast between the area where trials are needed and the area where trials are conducted, we constructed an area cartogram according to the method developed by Gastner and colleagues [17]. Area cartograms are maps in which the sizes of geographic regions such as countries are not proportional to the area on the ground but, rather, appear in proportion to their population or some other demographic feature such as disease prevalence. We used such maps to compare the area for number of people smoking and with HIV infection to where RCTs have been and are being conducted.\nFor an adequate representation of the contrast between the area where trials are needed and the area where trials are conducted, we constructed an area cartogram according to the method developed by Gastner and colleagues [17]. Area cartograms are maps in which the sizes of geographic regions such as countries are not proportional to the area on the ground but, rather, appear in proportion to their population or some other demographic feature such as disease prevalence. We used such maps to compare the area for number of people smoking and with HIV infection to where RCTs have been and are being conducted.\n[SUBTITLE] Statistical analysis [SUBSECTION] We used descriptive statistics; categorical variables are described with frequencies and percentages. All data analyses involved use of SAS for Windows, Release 9.1 (SAS Inst., Cary, NC).\nCartographic representation was developed with use of Migratio 8.0 (Stéphane Le Rouzic, Migratio.fr, Rennes, France) and ScapeToad cartography software (Chôros Laboratory, EPFL-ENAC-INTER, Lausanne, Switzerland).\nWe used descriptive statistics; categorical variables are described with frequencies and percentages. All data analyses involved use of SAS for Windows, Release 9.1 (SAS Inst., Cary, NC).\nCartographic representation was developed with use of Migratio 8.0 (Stéphane Le Rouzic, Migratio.fr, Rennes, France) and ScapeToad cartography software (Chôros Laboratory, EPFL-ENAC-INTER, Lausanne, Switzerland).", "We classified the countries according to the 2008 World Bank classification that divides the world into income categories (low, middle, and high) according to economies in terms of gross region national income per capita and in geographic regions for low- and middle-income economies only. We considered trials performed in high-income countries (country with a per capita income of ≥$11 906 in 2008 [e.g., USA, European Union [EU], Canada, Australia and other]) as one group and trials performed in low- and middle-income countries as another group [9], [10]. In addition, we grouped low- and middle-income countries into 6 geographic regions according to the 2008 World Bank classification (i.e., East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa).", "To describe the context (i.e., geographic region and socioeconomic situation of the countries where the trial took place) of research performed in these fields, we focused on the context of currently available evidence and awaiting evidence that could be used for clinical decision making. Because RCTs are the gold standard to assess the effectiveness of interventions, we evaluated the context of all RCTs included in Cochrane systematic reviews (i.e., context of current evidence) and all ongoing RCTs in the registries of the WHO-ICTRP (i.e., context of awaiting evidence). We focused on systematic reviews performed by the Cochrane collaboration because this international, not-for-profit organisation is a source of high-quality, reliable health information providing up-to-date knowledge about the effects of health care. Further, an important goal of the Cochrane collaboration is the dissemination of information to low- and middle-income countries. The Cochrane Library is freely available to all residents of low-income countries.\n[SUBTITLE] a. Identification of RCTs included in Cochrane systematic reviews [SUBSECTION] We identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).\nWe identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).\n[SUBTITLE] b. Identification of ongoing RCTs [SUBSECTION] In January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.\nIn January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.\n[SUBTITLE] c. Data collection [SUBSECTION] One of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.\nOne of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.", "We identified all systematic reviews of interventions aimed at reducing or stopping tobacco use and those aimed at treating or preventing HIV infection published in the Cochrane database of systematic reviews between January 1997 and December 2007. We searched for the following terms in the title, abstract or the MeSH terms: “smoking cessation” OR “tobacco use cessation” OR “smoking reduction” OR “tobacco reduction” OR “smoking abstinence” OR “tobacco abstinence” for tobacco use and “HIV” OR “human immunodeficiency virus” OR “AIDS” OR “acquired immunodeficiency syndrome” OR “sexually transmitted diseases” for HIV infection. Titles and abstracts were then screened by one of us (NA) to identify the relevant systematic reviews. The full texts for selected abstracts were retrieved and reviewed by one of us (NA) to determine the eligibility of systematic reviews for inclusion. Another reviewer (IB) checked the adequate selection of the abstracts and confirmed the adequate exclusion of full-text articles.\nReports were included if the study was identified as a systematic review or meta-analysis of interventions aimed at reducing tobacco consumption or preventing or treating HIV infection. We excluded systematic reviews focusing on a specific setting, except when this setting concerned developing countries (e.g., intervention for tobacco cessation in a dental setting) or a specific population (e.g., tobacco cessation for hospitalised patients). Excluding these systematic reviews allowed for a relatively homogeneous sample of systematic reviews aimed at providing conclusions on treatment effect whatever the context. We also excluded systematic reviews of prevention or treatment of complications of HIV infection (e.g., opportunistic infections, Kaposi's sarcoma) and those evaluating an intervention for another disease among subjects infected with HIV (e.g., treatment of anaemia or tuberculosis infection in people with HIV infection).", "In January 2009, we searched the WHO-ICTRP (http://www.who.int/trialsearch/) for all ongoing RCTs registered in the platform's 10 clinical trials registries: Australian New Zealand Clinical Trials Registry (ANZCTR), Chinese Clinical Trial Register (ChiCTR), ClinicalTrials.gov, Clinical Trials Registry - India (CTRI), German Clinical Trials Register (DRKS), Iranian Registry of Clinical Trials (IRCT), International Standard Randomised Controlled Trials Number (ISRCTN.org), Japan Primary Registries Network, The Netherlands National Trial Register (NTR), and Sri Lanka Clinical Trials Registry (SLCTR). We used this platform because it allows access to all primary registries meeting the WHO criteria and the requirements of the International Committee of Medical Journal Editors [11].\nWe searched for RCTs in the “conditions” menu of the platform using the topics “smoking” for tobacco use and “HIV infection” for HIV infection and in the “recruitment” menu of the “advanced search” feature, using “recruiting” to select only ongoing RCTs.", "One of us (NA) screened all records obtained by this search to select the relevant RCTs (i.e., all RCTs of interventions aimed at reducing or stopping tobacco consumption or preventing or treating HIV infection).\nFor RCTs included in systematic reviews, we collected data on the individual RCTs included in the review. To avoid collecting data several times from an RCT included in several systematic reviews, we recorded all RCTs in an EndNote data file (EndNote for Windows Version X2, Bld (3210)). We systematically searched this file for reports with the same authors and excluded duplicates.\nIf the country where the trial was performed was not mentioned in the systematic review, we retrieved the published report for the trial to obtain more information.\nAs a quality assurance exercise, another author (AD) collected data on a random sample of 102 RCTs included in systematic reviews.\nFor ongoing trials, we collected data from the trial records available in the WHO-ICTRP and in the primary registry's records.\nFor each RCT included in a systematic review or registered, we checked whether the trial was performed in a high-income or a low- or middle-income country. We considered trials performed in both high-income and low- and middle-income countries as trials of low- and middle-income countries.\nWe also collected data on the intervention (pharmacological or nonpharmacologic treatment) and the number of participating countries.", "The following indicators were elaborated:\n[SUBTITLE] 1. Prevalence of tobacco use and HIV infection [SUBSECTION] To obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].\nTo obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].\n[SUBTITLE] 2. Mortality and global burden of disease (GBD) [SUBSECTION] The global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].\nThe global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].", "To obtain the prevalence of tobacco use and HIV infection in the world, in high-income countries and in different regions of low- and middle-income countries, we used the World Bank reports for number of smokers [12] and WHO reports for number of people living with HIV infection [13], [14].", "The global burden of disease (GBD) reflects the burden of disease by disability-adjusted life years (DALYs). This time-based measure combines years of life lost due to premature mortality and that lost due to time lived in states of less than full health. The DALY measure was developed in the original GBD 1990 study to assess the burden of disease consistently across diseases, risk factors and regions [15].\nWe used the data (WHO assessment of the GBD for 2000–2002 compiled and updated by the World Bank) related to the GBD (DALYs) and mortality attributable to tobacco use and HIV infection in the world, in high-income countries and in low- and middle-income countries [16].", "For an adequate representation of the contrast between the area where trials are needed and the area where trials are conducted, we constructed an area cartogram according to the method developed by Gastner and colleagues [17]. Area cartograms are maps in which the sizes of geographic regions such as countries are not proportional to the area on the ground but, rather, appear in proportion to their population or some other demographic feature such as disease prevalence. We used such maps to compare the area for number of people smoking and with HIV infection to where RCTs have been and are being conducted.", "We used descriptive statistics; categorical variables are described with frequencies and percentages. All data analyses involved use of SAS for Windows, Release 9.1 (SAS Inst., Cary, NC).\nCartographic representation was developed with use of Migratio 8.0 (Stéphane Le Rouzic, Migratio.fr, Rennes, France) and ScapeToad cartography software (Chôros Laboratory, EPFL-ENAC-INTER, Lausanne, Switzerland).", "[SUBTITLE] Where research is conducted [SUBSECTION] The flow of RCTs included in the selected systematic reviews and ongoing RCTs through the study is in Figures 1 and 2, respectively. The electronic search yielded 118 systematic reviews, of which 57 (28 of tobacco use, 29 of HIV infection) were selected. From these systematic reviews, data for 1 012 trials were retrieved; 143 duplicate reports and reports of 129 nonrandomised trials were excluded. We included in the final analysis data for 740 RCTs (556 of tobacco use, 184 of HIV infection). Of the 940 ongoing RCTs (266 of tobacco use, 674 of HIV infection) identified in the international clinical trial registries, we included in the final analysis data for 346 (112 of tobacco use, 234 of HIV infection).\nThe characteristics of the RCTs and ongoing RCTs are in Table 1.\n* International clinical trial registries: Clinicaltrials.gov, Australian New Zealand Clinical Trials Registry, Chinese Clinical Trial Register, Clinical Trials Registry - India, German Clinical Trials Register, Iranian Registry of Clinical Trials, Japan Primary Registries Network, International Standard Randomised Controlled Trials Number, The Netherlands National Trial Register, Sri Lanka Clinical Trials Registry.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention.\nThe flow of RCTs included in the selected systematic reviews and ongoing RCTs through the study is in Figures 1 and 2, respectively. The electronic search yielded 118 systematic reviews, of which 57 (28 of tobacco use, 29 of HIV infection) were selected. From these systematic reviews, data for 1 012 trials were retrieved; 143 duplicate reports and reports of 129 nonrandomised trials were excluded. We included in the final analysis data for 740 RCTs (556 of tobacco use, 184 of HIV infection). Of the 940 ongoing RCTs (266 of tobacco use, 674 of HIV infection) identified in the international clinical trial registries, we included in the final analysis data for 346 (112 of tobacco use, 234 of HIV infection).\nThe characteristics of the RCTs and ongoing RCTs are in Table 1.\n* International clinical trial registries: Clinicaltrials.gov, Australian New Zealand Clinical Trials Registry, Chinese Clinical Trial Register, Clinical Trials Registry - India, German Clinical Trials Register, Iranian Registry of Clinical Trials, Japan Primary Registries Network, International Standard Randomised Controlled Trials Number, The Netherlands National Trial Register, Sri Lanka Clinical Trials Registry.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention.\n[SUBTITLE] Geographic representation [SUBSECTION] [SUBTITLE] Tobacco use trials [SUBSECTION] The geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nThe geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\n[SUBTITLE] HIV infection trials [SUBSECTION] The geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nThe geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\n[SUBTITLE] Tobacco use trials [SUBSECTION] The geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nThe geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\n[SUBTITLE] HIV infection trials [SUBSECTION] The geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nThe geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.", "The flow of RCTs included in the selected systematic reviews and ongoing RCTs through the study is in Figures 1 and 2, respectively. The electronic search yielded 118 systematic reviews, of which 57 (28 of tobacco use, 29 of HIV infection) were selected. From these systematic reviews, data for 1 012 trials were retrieved; 143 duplicate reports and reports of 129 nonrandomised trials were excluded. We included in the final analysis data for 740 RCTs (556 of tobacco use, 184 of HIV infection). Of the 940 ongoing RCTs (266 of tobacco use, 674 of HIV infection) identified in the international clinical trial registries, we included in the final analysis data for 346 (112 of tobacco use, 234 of HIV infection).\nThe characteristics of the RCTs and ongoing RCTs are in Table 1.\n* International clinical trial registries: Clinicaltrials.gov, Australian New Zealand Clinical Trials Registry, Chinese Clinical Trial Register, Clinical Trials Registry - India, German Clinical Trials Register, Iranian Registry of Clinical Trials, Japan Primary Registries Network, International Standard Randomised Controlled Trials Number, The Netherlands National Trial Register, Sri Lanka Clinical Trials Registry.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention.", "[SUBTITLE] Tobacco use trials [SUBSECTION] The geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nThe geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\n[SUBTITLE] HIV infection trials [SUBSECTION] The geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nThe geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.", "The geographic comparison of location of RCTs of tobacco use and mortality and GBD (in DALYs) attributable to tobacco use is in Table 2 and Figure 3A. In total, 96% of RCTs included in systematic reviews and 98% of ongoing RCTs were performed in high-income countries. Both kinds of trials were performed mainly in the USA (314 [57%] and 68 [61%], respectively) and the EU (144 [26%] and 27 [24%], respectively). Only 24 (4%) RCTs included in systematic reviews and 2 (2%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 74% of the GBD (in DALYs) and 70% of the mortality attributable to tobacco use and 82% of the smokers. Only 11 of 341 (3%) trials and 2 of 61 (3%) ongoing RCTs assessing nonpharmacologic treatment were performed in low- and middle-income countries. The highest prevalence of tobacco use concerns East Asia and Pacific countries (38%). This area represents 22% of the mortality attributable to tobacco use. In this area, only 9 RCTs were performed, and only 2 registered RCTs are ongoing. Similarly, no ongoing trial is registered in Eastern Europe and Central Asia, South Asia, Latin America and Caribbean, the Middle East and North Africa, and Sub-Saharan Africa, which overall represent nearly half of the mortality attributable to tobacco use. Figure 4 highlights the differences in geographic representation between regions with a high prevalence of smokers and where trials are conducted.\n* Source: World Bank's 2005 World Development Indicators, 2005. Table 2.18 Health: risk factors and future challenges. The World Bank sourced this in turn from the World Health Organization's 2004 World Health Report, Tobacco Control Country Profiles 2003. † The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.", "The geographical comparison of location of RCTs and mortality and GBD (in DALYs) attributable to HIV infection is in Table 3 and Figure 3B. In total, 70% of RCTs included in systematic reviews and 67% of ongoing RCTs were performed in high-income countries. Only 54 (31%) RCTs and 77 (33%) ongoing RCTs were carried out in low- or middle-income countries, although these countries represented 99% of the GBD, 99% of the mortality and 94% of the people with HIV infection. About 16% (16/99) of trials and 26% (20/78) of ongoing RCTs assessing nonpharmacologic treatments were performed in low- and middle-income countries.\nRCTs: randomized controlled trials. PT: pharmacologic intervention. NPT: nonpharmacologic intervention. GBD: global burden of disease.\nOnly 37 (21%) RCTs and 40 (17%) ongoing RCTs were performed in Sub-Saharan Africa, which represents 84% of the GBD and 84% of the mortality attributable to HIV infection and 68% of the people with HIV infection. Similarly, only 2 ongoing RCTs are being performed in South Asia, although this area represents 9% of the GBD, 9% of the mortality and 12% of the people with HIV infection. Figure 5 highlights the differences in geographic representation between the regions with a high prevalence of people with HIV infection and where trials are conducted.\n*Source: report on the global AIDS epidemic, UNAIDS/WHO, July 2008. †The number of RCTs (primary and ongoing) was classified by main regions according to the 2008 World Bank classification: high-income countries, East Asia and Pacific, Eastern Europe and Central Asia, Latin America and Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa. For the construction of the map we divided the number of trials performed in each region by the number of countries in each region.", "This study evaluated the context of clinical research (i.e., where research is performed) for 2 public health priorities: tobacco use and HIV infection. We described the location of 740 RCTs included in Cochrane systematic reviews and 346 ongoing RCTs identified through international clinical trial registries. Our results highlight a gap between where trials are conducted and where research is needed in terms of GBD, mortality and prevalence of tobacco use and HIV infection. Most ongoing and RCTs were performed in high-income countries, even though most of the mortality and GBD concerns low- and middle-income countries. Only 4% and 32% of the total and ongoing RCTs performed in the fields of tobacco use and HIV infection, respectively, were conducted in low- and middle-income countries. However, low- and middle-income countries represent 70% and 99% of the mortality attributable to tobacco use and HIV infection, respectively [13], [18].\nThe number of trials of tobacco use performed in low- and middle-income countries was particularly low (4%) as compared with trials of HIV infection performed in these countries (about one-third).\nPrevious studies have highlighted the underrepresentation of research addressing priority issues for low- and middle-income countries [19]. Some authors highlighted the poor association of GBD and reports of RCTs published in high-impact-factor journals [19], [20]. Swingler et al. [21] analysed nearly 3 000 systematic reviews and assessed the correlation between the number of systematic reviews undertaken and the burden of disease: only a few systematic reviews focused on some diseases affecting a large number of the world's population. Sheriff showed that only 3% of mental disease research is performed in low- and middle-income countries [22], but the religious and cultural individuality of these countries could greatly hamper implementation of any mental health interventions recommended by research performed in high-income countries. These low research percentages are probably related to the limited financial investment in research for low- and middle-income countries because of the phenomenon of the “10/90 gap”. The Global Forum has highlighted that of the US$73 billion invested annually in global health research, less than 10% is spent on research into the health problems that account for 90% of the GBD [4], [23], [24]. This gap is also reflected in the low proportion of publications from research performed in low- and middle-income countries [25].\nContrary to previous research, we focused on 2 public health priorities (tobacco use, responsible for non-communicable chronic diseases, and HIV infection, a communicable chronic disease) relevant for high-, low- and middle-income countries. Our study is the first to focus on existing evidence (trials included in systematic reviews) and on awaiting evidence that will guide future evidence-based practices.\nWith the evidence of the effectiveness of an intervention, some might assume that the challenge is to make the intervention available in low- and middle-income countries [26]. However, the results of trials performed in high-income countries may not be easily applied to low- and middle-income countries. Many interventions shown to be efficacious in high-income countries are not similarly effective when carried out in other contexts [26], [27]. The populations can differ, with people consulting late, frequently using self-medication, unable or unwilling to adhere to treatment, and having several co-morbidities (malnourishment, anaemia, malaria, etc.), in addition to the behavioural and cultural differences [28], [29]. Further, the contextual factors differ greatly, particularly the social and cultural context, as well as the available facilities and infrastructures. The lack of studies performed in a relevant location is particularly problematic for nonpharmacologic interventions targeting behavioural changes (e.g., education, counselling), because the results of such trials could be strongly influenced by cultural conditions [20], [30]. The influence of the cultural context is important, even between different high-income countries [31].\nRecently, the Global Alliance for Chronic Disease (GACD) was created to address this imbalance issue [32]. The GACD brings together 6 major national health research councils representing 80% of all public research funding in the world to coordinate research activities that address the prevention and treatment of chronic disease on a global scale. One essential goal of the GACD is to focus on chronic disease in low- and middle-income countries and to coordinate research into low-cost interventions. This initiative should be important in limiting the worldwide imbalance in health research resources [7].\nSimilarly, the High Level Taskforce on Innovative International Financing for Health Systems was created in September 2008 to help fill national financing gaps to reach the health Millennium Development Goals in low-income countries. One key challenge is to obtain evidence for what works according to different settings [33].\nOur study has some limitations. First, we focused on only 2 medical areas, and these results should be confirmed in other medical areas. However, we chose tobacco use and HIV infection because they are among the first 5 causes of mortality in the world [9]. Second, we selected only RCTs included in Cochrane systematic reviews because these systematic reviews are known to be of high quality, and we did not consider non-Cochrane systematic reviews [34], [35], [36], [37], [38]. As well, the Cochrane collaboration seeks to actively encourage the participation of reviewers from developing countries and to prioritize reviews focusing on determinants of health that are particularly pertinent to low- and middle-income countries [1], [4]. Finally, our results may underestimate the number of ongoing RCTs performed in low- and middle-income countries because we focused on only available data (i.e., trials that were registered in a trial registry available on the WHO portal). We cannot exclude that some trials performed in low-income countries were not registered.\nIn conclusion, clinical research into tobacco use and HIV infection is not performed in the world locations most affected in terms of GBD and mortality because we found important underrepresentation of such research in low- and middle-income countries. The GBD measured by mortality or DALYs could be a useful aid in deciding where to conduct health research. The measure should help in conducting RCTs in an appropriate context to improve clinical research practice and decrease the GBD [39]." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Animals", "Bromodeoxyuridine", "Capsaicin", "Cell Differentiation", "Cholera Toxin", "DNA", "Enteric Nervous System", "Gene Knockout Techniques", "In Vitro Techniques", "Intestinal Mucosa", "Male", "Mice", "Neurons, Afferent", "Rats", "Receptors, Muscarinic", "Receptors, Peptide", "Stem Cells", "TRPV Cation Channels", "Thymidine Kinase" ]

[ "Introduction", "Results", "Rapid cell renewal in the intestine is located to the crypts", "Cholera toxin does not increase rate of epithelial renewal", "Luminal capsaicin increases rate of epithelial renewal", "The capsaicin effect on epithelial renewal is mediated by nerves", "The capsaicin effect on epithelial renewal is mediated by muscarinic and peptidergic receptors", "Rate of epithelial renewal is decreased in capsaicin receptor knock-out mice", "Discussion", "Ethical approval", "Rat experiments", "Anesthesia and operative procedures", "Experimental procedures", "Lidocaine experiments", "Neurotransmitter receptor blockade", "Neurotransmitter i.a. infusions", "BrdU experiments", "Chronic denervation experiments", "Determination of thymidine kinase activity", "Thymidine incorporation into DNA", "Immunohistochemistry", "Autoradiography experiments", "Mouse experiments", "Chemicals", "Recordings of heart rate", "Statistics" ]

[ "The intestinal epithelium consists of a single layer of columnar cells about 25 µm high. It represents an outer surface of the organism exposed to luminal contents of widely varying composition. The epithelium is of great importance for the survival of the organism, since chemicals and/or microorganisms, if allowed to pass the epithelial barrier, may represent a deadly threat. Thus, the intestinal epithelium may be looked upon as being an important part of innate immunity. The maintenance of the epithelium is secured by a rapid renewal, the epithelium being replaced every 3–5 days in mammals [1]. The key cells in this event are the intestinal stem and progenitor cells located in the intestinal crypts. The present study describes animal experiments, which suggest that there is a nervous reflex control of the stem/progenitor cells and, hence, rate of epithelial cell renewal.\nThe nervous control of the gastrointestinal tract is exerted by two systems: the extrinsic, efferent sympathetic and parasympathetic nerves and the intrinsic enteric nervous system (ENS). The latter represents a nervous system in the gastrointestinal tract, controlling epithelial transport, motility and blood flow [2]. ENS is composed of two major nerve plexuses, the myenteric and the submucosal and of their interconnections. Most of the neurons of the ENS are confined to the gastrointestinal wall, but extrinsic nerves are also found in the ENS. Two major types of nervous reflexes can be identified in the ENS: intramural reflexes and axon reflexes. Intramural reflexes are confined to the intestinal wall. The peristaltic reflex is an example of an intramural reflex initiated by luminal distension [2]. Branching external afferent nerves constitute the morphological bases of axon reflexes.\nMost investigations of the nervous control of gastrointestinal functions have been concerned with its control of motility, blood flow or epithelial transport. Although the extrinsic control of epithelial cell proliferation usually is discussed in terms of growth factors, peptide hormones and inflammatory mediators (for review, see reference [3]), there are observations to suggest also a nervous control of intestinal cell renewal. In most of these studies the influence of the extrinsic nerves has been investigated, suggesting that both branches of the extrinsic innervation may increase mitotic rate in the crypt cells [4]–[6]. Chemical ablation of the myenteric plexus leads to an increased crypt cell proliferation both in the small intestine [6]–[8] and in the colon [9], [10], indicating that myenteric nerves may exert an inhibitory influence on intestinal cell renewal.\nWe have earlier demonstrated that ENS plays an important role in the control of fluid transport. In fact, in many, if not most, types of acute diarrhea, an activation of ENS is the major mechanism underlying the observed intestinal fluid loss [11]–[13]. Diarrhea can be looked upon as being an innate immunity response, since the secreted fluid dilutes the potentially noxious agent(s). The maintenance of the epithelial barrier is also important for innate immunity. Therefore, we decided to investigate if intestinal contents, via the ENS, also may influence epithelial cell renewal. Most experiments were performed on intestinal segments that had been acutely denervated by cutting the nerves accompanying the superior mesenteric artery, minimizing the influence of the external innervation. The ENS was activated in two different ways: In one type of experiments an intramural secretory reflex was stimulated by exposing the intestinal mucosa to cholera toxin [11]. In another series we utilized capsaicin, the “hot” agent of red pepper, to activate thin mucosal afferents [14]. Capsaicin is known to activate the TRPV1 receptor. Cell renewal was studied in three different ways: We determined deoxythymidine kinase (TK) activity in the intestinal wall. TK facilitates the phosphorylation of deoxythymidine, associated with the formation of DNA [15]. We also measured the rate of incorporation of radioactively labeled thymidine into intestinal DNA. Finally, epithelial cell renewal was estimated by means of bromodeoxyuridine (BrdU), a thymidine analogue possible to localize at the cellular level in tissue sections with immunohistochemistry.", "[SUBTITLE] Rapid cell renewal in the intestine is located to the crypts [SUBSECTION] TK activity and methyl-3H-thymidine incorporation into DNA was measured in the whole intestine. Hence, it was important to ensure that the recorded activity/incorporation mainly reflected events in the intestinal crypt stem/progenitor cells. This was tested in two ways. First, the left panel of Figure 1 shows an autoradiograph of the small intestine from a rat given methyl-3H-thymidine i.v. 6 h prior to the removal of the intestine. Dark field microscopy shows that the labeled thymidine was mainly located to the crypts, where intestinal epithelial stem/progenitor cells are found [16]–[18]. The right panel shows the same histological section in regular light microscopy. Figure 2 illustrates the localization of labeled thymidine at a higher magnification in intestinal crypts. The photographic grains are only seen above the nuclei of the cells (indicated by arrows). Second, in another type of experiments BrdU (a thymidine analogue) was given i.v. (50 mg kg−1). Two h after the i.v. injection, BrdU immunoreactivity was exclusively found in the crypt epithelial cells (Figure 3).\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine The left panel illustrates dark field microscopy of an autoradiogram showing radiolabeled cells (white spots) almost exclusively located in the epithelial layer of the crypts of Lieberkühn. To the right the same section is seen in light microscopy. Van Gieson.\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine. Only cell nuclei were found to be labeled (arrows).\nBrdU (50 mg/kg b.wt.), a thymidine analogue, was given 2 h prior to removing the intestine. The dark labeling by BrdU is exclusively located to the nuclei of intestinal crypt cells. Mayer's haematoxylin.\nTK activity and methyl-3H-thymidine incorporation into DNA was measured in the whole intestine. Hence, it was important to ensure that the recorded activity/incorporation mainly reflected events in the intestinal crypt stem/progenitor cells. This was tested in two ways. First, the left panel of Figure 1 shows an autoradiograph of the small intestine from a rat given methyl-3H-thymidine i.v. 6 h prior to the removal of the intestine. Dark field microscopy shows that the labeled thymidine was mainly located to the crypts, where intestinal epithelial stem/progenitor cells are found [16]–[18]. The right panel shows the same histological section in regular light microscopy. Figure 2 illustrates the localization of labeled thymidine at a higher magnification in intestinal crypts. The photographic grains are only seen above the nuclei of the cells (indicated by arrows). Second, in another type of experiments BrdU (a thymidine analogue) was given i.v. (50 mg kg−1). Two h after the i.v. injection, BrdU immunoreactivity was exclusively found in the crypt epithelial cells (Figure 3).\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine The left panel illustrates dark field microscopy of an autoradiogram showing radiolabeled cells (white spots) almost exclusively located in the epithelial layer of the crypts of Lieberkühn. To the right the same section is seen in light microscopy. Van Gieson.\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine. Only cell nuclei were found to be labeled (arrows).\nBrdU (50 mg/kg b.wt.), a thymidine analogue, was given 2 h prior to removing the intestine. The dark labeling by BrdU is exclusively located to the nuclei of intestinal crypt cells. Mayer's haematoxylin.\n[SUBTITLE] Cholera toxin does not increase rate of epithelial renewal [SUBSECTION] Cholera toxin evokes an intestinal fluid secretion mainly by stimulating nervous reflex(es) contained within the ENS [11], [13]. Exposing the intestinal mucosa to cholera toxin (20 µg per mL) for three hours did not influence TK activity (91.2±11.2% (mean ± s.e.m) of control; n = 5; p = 0.5), but decreased the incorporation of labeled thymidine into DNA (84.4±5.5% of control; n = 5; p<0.05).We conclude that in the present short term experiments nervous reflexes activated by cholera toxin do not increase rate of epithelial renewal.\nCholera toxin evokes an intestinal fluid secretion mainly by stimulating nervous reflex(es) contained within the ENS [11], [13]. Exposing the intestinal mucosa to cholera toxin (20 µg per mL) for three hours did not influence TK activity (91.2±11.2% (mean ± s.e.m) of control; n = 5; p = 0.5), but decreased the incorporation of labeled thymidine into DNA (84.4±5.5% of control; n = 5; p<0.05).We conclude that in the present short term experiments nervous reflexes activated by cholera toxin do not increase rate of epithelial renewal.\n[SUBTITLE] Luminal capsaicin increases rate of epithelial renewal [SUBSECTION] Two experimental protocols were used in the capsaicin experiments, “non-perfusion experiments” and “perfusion experiments” (see Methods). Since both techniques indicated that a 1.6 mM capsaicin solution evoked a consistent increase of TK activity (152.7±8.6% of TK activity in control segments (non-perfusion experiments; n = 7; p-value <0.05) and 149.8±14.4% of control segments (perfusion experiments; n = 6; p-value <0.05; Figure 4), this concentration was used in subsequent experiments, most of which were performed with the perfusion technique.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). The experiments were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”; n = 6). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”; n = 7). Asterisks indicate statistical significance. Mean ± s.e.m.\nThe incorporation of labeled thymidine into DNA was significantly increased after 2 h capsaicin perfusion (178.0±30.1% of incorporation in control segments; n = 6; p<0.05). Furthermore, the number of BrdU labeled crypt cells increased concomitantly in the capsaicin group as shown on left part of Figure 5, based on counting labeled cells in 255 crypts in control and 249 crypts in capsaicin segments (control segments: 15.0±0.3 labeled cells/crypt; capsaicin segments: 17.9±0.3 labeled cells/crypt; n = 5; p<0.01). Hence, exposing the intestinal mucosa to 1.6 mM capsaicin increased TK activity and 3H-thymidine incorporation into DNA reflecting events in crypt epithelial cells as confirmed by the BrdU experiments.\nBrdU (50 mg/kg bwt) was given i.v. 2 h prior to removal of the intestinal segments. The number of labeled cells in 255 crypts in control (n = 5) and in 249 crypts in capsaicin segments (n = 5) were counted. Asterisk indicates statistical significance. Mean ± s.e.m. Mouse: BrdU labeled cells per crypt in wild type or capsaicin receptor knock-out mice. Five capsaicin receptor knock-out mice (−/−; 225 crypts) and in five wild type mice (WT; 215 crypts) were investigated. BrdU (100 mg/kg b.wt.) was administered i.p. 2 h before removal of the small intestine. Asterisk indicates statistical significance. Mean ± s.e.m.\nTwo experimental protocols were used in the capsaicin experiments, “non-perfusion experiments” and “perfusion experiments” (see Methods). Since both techniques indicated that a 1.6 mM capsaicin solution evoked a consistent increase of TK activity (152.7±8.6% of TK activity in control segments (non-perfusion experiments; n = 7; p-value <0.05) and 149.8±14.4% of control segments (perfusion experiments; n = 6; p-value <0.05; Figure 4), this concentration was used in subsequent experiments, most of which were performed with the perfusion technique.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). The experiments were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”; n = 6). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”; n = 7). Asterisks indicate statistical significance. Mean ± s.e.m.\nThe incorporation of labeled thymidine into DNA was significantly increased after 2 h capsaicin perfusion (178.0±30.1% of incorporation in control segments; n = 6; p<0.05). Furthermore, the number of BrdU labeled crypt cells increased concomitantly in the capsaicin group as shown on left part of Figure 5, based on counting labeled cells in 255 crypts in control and 249 crypts in capsaicin segments (control segments: 15.0±0.3 labeled cells/crypt; capsaicin segments: 17.9±0.3 labeled cells/crypt; n = 5; p<0.01). Hence, exposing the intestinal mucosa to 1.6 mM capsaicin increased TK activity and 3H-thymidine incorporation into DNA reflecting events in crypt epithelial cells as confirmed by the BrdU experiments.\nBrdU (50 mg/kg bwt) was given i.v. 2 h prior to removal of the intestinal segments. The number of labeled cells in 255 crypts in control (n = 5) and in 249 crypts in capsaicin segments (n = 5) were counted. Asterisk indicates statistical significance. Mean ± s.e.m. Mouse: BrdU labeled cells per crypt in wild type or capsaicin receptor knock-out mice. Five capsaicin receptor knock-out mice (−/−; 225 crypts) and in five wild type mice (WT; 215 crypts) were investigated. BrdU (100 mg/kg b.wt.) was administered i.p. 2 h before removal of the small intestine. Asterisk indicates statistical significance. Mean ± s.e.m.\n[SUBTITLE] The capsaicin effect on epithelial renewal is mediated by nerves [SUBSECTION] Capsaicin was used to activate selectively thin afferent nerve fibers presumably via the vallinoid receptor 1 (TRPV1) [14]. We explored if capsaicin indeed activated afferent nerves by monitoring changes of heart rate on line after luminal administration of a 1.6 mM capsaicin. The periarterial nerves were not severed in these experiments. Within ten minutes 1.6 mM capsaicin heart rate increased by 46.2±8.2 min−1, whereas heart rate was not affected (−4.7±5.1 min−1) in control experiments (n = 6; p<0.05), suggesting that luminal capsaicin activates nervous afferents.\nIn one series of experiments the extrinsic intestinal nerves were allowed to degenerate for 2–3 weeks after cutting the nerves around the superior mesenteric artery. The success of the denervation procedure was tested by immunohistochemical determinations of tyrosine hydroxylase in the intestine, located to extrinsic adrenergic nerve fibers [19]. Absence of labeled nerve fibers was taken to indicate a successful denervation. Exposing such “chronically” denervated intestinal segments to capsaicin did not significantly increase TK activity (102.7±3.1% of control; n = 5). This is significantly lower than the capsaicin effect observed in “acutely” denervated segments (144.1±8.3%; n = 18; p<0.01).\nTo further investigate the possible involvement of nerves in the capsaicin response, the intestinal mucosa was anesthetized with lidocaine (see Methods). In these experiments (Figure 6) no significant increase of TK activity was recorded (perfusion experiments: TK activity in capsaicin segments 97.5±14.4% of control; n = 8; p = 0.58; non-perfusion experiments: TK activity in capsaicin segments 103.1±13.7% of control; n = 7; p = 0.74), indicating a nervous involvement in the capsaicin response.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). Asterisks indicate statistical significance between the two groups of experiments (perfusion experiments, n = 8; non-perfusion experiments, n = 7). Mean ± s.e.m.\nIndirect evidence for an involvement of nerves was also obtained in experiments with the TPRV1 receptor blocker capsazepine (3 mg/kg i.v.) given prior to the intraluminal administration of capsaicin. In these experiments the TK activity of the capsaicin segment was 97.2±7.3% (n = 7) of that measured in corresponding control segments (p<0.01 compared to a value of 144.1±8.3 per cent in experiments without capsazepine; n = 18). Hence, the capsaicin effect on TK activity was mediated by a TPRV1 receptor. Since the TPRV1 receptor is known to be located on many extrinsic afferent axons[14], the results support the proposal that the capsaicin effect involves nerves.\nFinally, hexamethonium (a cholinergic, nicotinic receptor blocker; 10 mg/kg b.wt.; n = 8) significantly attenuated the effect of capsaicin on TK activity measuring 96.5±10.7% of the activity in capsaicin free segments (p<0.01 compared to 144.1±8.3 per cent in experiments without hexamethonium; n = 18). These results suggest that the capsaicin effect was mediated by ENS reflex(es) containing at least one cholinergic synapse.\nCapsaicin was used to activate selectively thin afferent nerve fibers presumably via the vallinoid receptor 1 (TRPV1) [14]. We explored if capsaicin indeed activated afferent nerves by monitoring changes of heart rate on line after luminal administration of a 1.6 mM capsaicin. The periarterial nerves were not severed in these experiments. Within ten minutes 1.6 mM capsaicin heart rate increased by 46.2±8.2 min−1, whereas heart rate was not affected (−4.7±5.1 min−1) in control experiments (n = 6; p<0.05), suggesting that luminal capsaicin activates nervous afferents.\nIn one series of experiments the extrinsic intestinal nerves were allowed to degenerate for 2–3 weeks after cutting the nerves around the superior mesenteric artery. The success of the denervation procedure was tested by immunohistochemical determinations of tyrosine hydroxylase in the intestine, located to extrinsic adrenergic nerve fibers [19]. Absence of labeled nerve fibers was taken to indicate a successful denervation. Exposing such “chronically” denervated intestinal segments to capsaicin did not significantly increase TK activity (102.7±3.1% of control; n = 5). This is significantly lower than the capsaicin effect observed in “acutely” denervated segments (144.1±8.3%; n = 18; p<0.01).\nTo further investigate the possible involvement of nerves in the capsaicin response, the intestinal mucosa was anesthetized with lidocaine (see Methods). In these experiments (Figure 6) no significant increase of TK activity was recorded (perfusion experiments: TK activity in capsaicin segments 97.5±14.4% of control; n = 8; p = 0.58; non-perfusion experiments: TK activity in capsaicin segments 103.1±13.7% of control; n = 7; p = 0.74), indicating a nervous involvement in the capsaicin response.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). Asterisks indicate statistical significance between the two groups of experiments (perfusion experiments, n = 8; non-perfusion experiments, n = 7). Mean ± s.e.m.\nIndirect evidence for an involvement of nerves was also obtained in experiments with the TPRV1 receptor blocker capsazepine (3 mg/kg i.v.) given prior to the intraluminal administration of capsaicin. In these experiments the TK activity of the capsaicin segment was 97.2±7.3% (n = 7) of that measured in corresponding control segments (p<0.01 compared to a value of 144.1±8.3 per cent in experiments without capsazepine; n = 18). Hence, the capsaicin effect on TK activity was mediated by a TPRV1 receptor. Since the TPRV1 receptor is known to be located on many extrinsic afferent axons[14], the results support the proposal that the capsaicin effect involves nerves.\nFinally, hexamethonium (a cholinergic, nicotinic receptor blocker; 10 mg/kg b.wt.; n = 8) significantly attenuated the effect of capsaicin on TK activity measuring 96.5±10.7% of the activity in capsaicin free segments (p<0.01 compared to 144.1±8.3 per cent in experiments without hexamethonium; n = 18). These results suggest that the capsaicin effect was mediated by ENS reflex(es) containing at least one cholinergic synapse.\n[SUBTITLE] The capsaicin effect on epithelial renewal is mediated by muscarinic and peptidergic receptors [SUBSECTION] The experiments with chronically denervated intestines described above suggested that capsaicin stimulates an axon reflex. In many organs calcitonin gene related polypeptide (CGRP) or substance P (SP) have been shown to be involved in axon reflexes [see e.g. references 20]–[22]. To elucidate if this also was the case in the present experiments two types of studies were performed. First, blocking the CGRP receptors (human 8–37 α-CGRP; 1 mg kg−1 b.wt. i.v.) or the SP receptors (more specifically the NK1 receptors: Sendide; 1 mg kg−1 b.wt. i.v.) abolished the increase of TK activity evoked by capsaicin (Figure 7; CGRP receptors: TK activity in capsaicin segments 78.5±9.7% of control segments; n = 6; p<0.001 compared to control experiments; NK1 receptors: TK activity in capsaicin segments 82.6±6.8% of TK activity in control intestines; n = 6; p<0.001 compared to control experiments) and the augmented incorporation of 3H-thymidine into DNA (Figure 8; CGRP receptors: incorporation in capsaicin segments 68.9±14% of control intestines; n = 6; p<0.01; NK1 receptors: incorporation in capsaicin segments 90.6±9.8% of control; n = 6; p<0.001) induced by luminal capsaicin. The doses used have been previously shown to block effects of CGRP and SP [22]–[24]. Second, infusing the neurotransmitters retrogradely in the carotid artery at rates of 10−9 mol per min for 1 h increased significantly intestinal TK activity compared to activity at the start of infusion (Figure 9; CGRP: TK activity 141.0±7.6% of activity at time 0; n = 5; p<0.05; SP: TK activity 143.9±20.7% of activity at time 0; n = 5; p<0.05).\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). Thymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). N = 6 for each antagonist experiments. Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). The receptor blockers were given as described in Methods. 3H-thymidine incorporation is expressed in per cent of incorporation measured in control segments (no capsaicin). Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe peptides were infused for one hour at a rate of 10−9 mol per minute with or without prior administration of atropine (0.5 mg per kg b.wt. i.v.). Note that in these experiments thymidine kinase activity is expressed in per cent of activity measured in segments removed at time 0. Asterisks indicate statistical significance. Mean ± s.e.m.\nThere are two major types of cholinergic receptors, nicotinic and muscarinic. Above we described results from experiments with hexamethonium, a nicotinic receptor blocker. To block the muscarinic receptors, atropine (0.5 mg kg−1 b.wt.) was given i.v. Atropine significantly diminished the intestinal TK response to capsaicin (Figure 7; TK activity in capsaicin segments 91.8±5.6% of control; n = 6; p<0.001 compared to experiments without atropine) and the capsaicin evoked increase of 3H-thymidine incorporation into DNA (Figure 8; incorporation in capsaicin intestines 70.9±16.5% of control; n = 6; p = 0.002 compared to experiments without atropine).\nIn the atropine experiments it was possible to investigate the effect of atropine on rate of 3H-thymidine incorporation into DNA in control segments. After 2 h labeled nucleotide incorporation into DNA was 1017±81 DPM per mg DNA in animals not receiving atropine and 380±65 DPM per mg DNA in experiments with atropine (n = 6 in both groups; p<0.01 comparing the two groups). Hence, atropine lowered incorporation to about 40% of control suggesting an ongoing cholinergic, nervous influence in the absence of luminal capsaicin.\nA possible interaction between muscarinic and peptidergic transmission was investigated by intravascular infusion of CGRP or SP after atropine administration. No CGRP or SP effect on TK activity could be demonstrated after giving atropine. (Figure 9; CGRP: 90.8±7.1% of TK activity at start of peptide infusion; n = 7; p<0.05 compared to CGRP experiments without atropine; SP: 73.8±5.7% of TK activity at time 0; n = 7; p<0.05). The simplest explanation for these observations is that CGRP and SP influence cholinergic neuron(s), which, in turn, control the intestinal stem cells.\nTo elucidate if muscarinic receptors were present on the intestinal stem/progenitor cells, we performed immunofluorescence studies using antibodies against the five different muscarinic receptors (M1–M5) described in the literature [25]. As on many other effector cells innervated by cholinergic neurons, immunoreactivity for more than one type of muscarinic receptors (M3 and M5) was identified in the intestinal crypts [25]. The left panel of figure 10 illustrates the distribution of the immunoreactivity for the M5 receptor in the crypts as revealed by confocal microscopy in a 4 µm thin section of a crypt wall. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The disclosed fluorescence pattern is consistent with the immunoreactivy being located to cell membranes. The round structure in the middle of the figure is in all probability a cross-sectioned goblet cell surrounded by immunoreactivity. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (Figure 10, right panel).\nThe picture is photographed by confocal microscopy corresponding to a 4 µm thick section of the wall of the crypts. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The left panel of the figure suggests that the M5 receptor is located to the plasma membrane of the crypt cells. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (right panel).\nCheng and Leblond [26] proposed that the so called base columnar cells were the intestinal stem cells generating all types of epithelial cells, a hypothesis recently supported by Barker et al. [17], [18]. Thus, the intestinal stem cells seem to be the slender cells situated between the Paneth cells. We studied the possible presence of muscarinic receptors on these cells. Two types of muscarinic receptors M3 and M5, were investigated. Immunoreactivity to M3 and M5 receptors was identified on the presumed stem cell membrane, as illustrated for the M3 receptor on the left panel of Figure 11. The right panel shows the appearance of the tissue when exposed only to the secondary antibody.\nThe stem cells, indicated by arrows, are slender cells located between the Paneth cells (PC) at the very base of the crypts [17], [26]. The left part of the figure indicates that the muscarinic receptor M3 is located to the plasma membrane of the presumed intestinal stem cells. The tissue section in the right panel was only exposed to the secondary antibody. Mayer's haematoxylin. CL  =  crypt lumen.\nThe experiments with chronically denervated intestines described above suggested that capsaicin stimulates an axon reflex. In many organs calcitonin gene related polypeptide (CGRP) or substance P (SP) have been shown to be involved in axon reflexes [see e.g. references 20]–[22]. To elucidate if this also was the case in the present experiments two types of studies were performed. First, blocking the CGRP receptors (human 8–37 α-CGRP; 1 mg kg−1 b.wt. i.v.) or the SP receptors (more specifically the NK1 receptors: Sendide; 1 mg kg−1 b.wt. i.v.) abolished the increase of TK activity evoked by capsaicin (Figure 7; CGRP receptors: TK activity in capsaicin segments 78.5±9.7% of control segments; n = 6; p<0.001 compared to control experiments; NK1 receptors: TK activity in capsaicin segments 82.6±6.8% of TK activity in control intestines; n = 6; p<0.001 compared to control experiments) and the augmented incorporation of 3H-thymidine into DNA (Figure 8; CGRP receptors: incorporation in capsaicin segments 68.9±14% of control intestines; n = 6; p<0.01; NK1 receptors: incorporation in capsaicin segments 90.6±9.8% of control; n = 6; p<0.001) induced by luminal capsaicin. The doses used have been previously shown to block effects of CGRP and SP [22]–[24]. Second, infusing the neurotransmitters retrogradely in the carotid artery at rates of 10−9 mol per min for 1 h increased significantly intestinal TK activity compared to activity at the start of infusion (Figure 9; CGRP: TK activity 141.0±7.6% of activity at time 0; n = 5; p<0.05; SP: TK activity 143.9±20.7% of activity at time 0; n = 5; p<0.05).\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). Thymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). N = 6 for each antagonist experiments. Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). The receptor blockers were given as described in Methods. 3H-thymidine incorporation is expressed in per cent of incorporation measured in control segments (no capsaicin). Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe peptides were infused for one hour at a rate of 10−9 mol per minute with or without prior administration of atropine (0.5 mg per kg b.wt. i.v.). Note that in these experiments thymidine kinase activity is expressed in per cent of activity measured in segments removed at time 0. Asterisks indicate statistical significance. Mean ± s.e.m.\nThere are two major types of cholinergic receptors, nicotinic and muscarinic. Above we described results from experiments with hexamethonium, a nicotinic receptor blocker. To block the muscarinic receptors, atropine (0.5 mg kg−1 b.wt.) was given i.v. Atropine significantly diminished the intestinal TK response to capsaicin (Figure 7; TK activity in capsaicin segments 91.8±5.6% of control; n = 6; p<0.001 compared to experiments without atropine) and the capsaicin evoked increase of 3H-thymidine incorporation into DNA (Figure 8; incorporation in capsaicin intestines 70.9±16.5% of control; n = 6; p = 0.002 compared to experiments without atropine).\nIn the atropine experiments it was possible to investigate the effect of atropine on rate of 3H-thymidine incorporation into DNA in control segments. After 2 h labeled nucleotide incorporation into DNA was 1017±81 DPM per mg DNA in animals not receiving atropine and 380±65 DPM per mg DNA in experiments with atropine (n = 6 in both groups; p<0.01 comparing the two groups). Hence, atropine lowered incorporation to about 40% of control suggesting an ongoing cholinergic, nervous influence in the absence of luminal capsaicin.\nA possible interaction between muscarinic and peptidergic transmission was investigated by intravascular infusion of CGRP or SP after atropine administration. No CGRP or SP effect on TK activity could be demonstrated after giving atropine. (Figure 9; CGRP: 90.8±7.1% of TK activity at start of peptide infusion; n = 7; p<0.05 compared to CGRP experiments without atropine; SP: 73.8±5.7% of TK activity at time 0; n = 7; p<0.05). The simplest explanation for these observations is that CGRP and SP influence cholinergic neuron(s), which, in turn, control the intestinal stem cells.\nTo elucidate if muscarinic receptors were present on the intestinal stem/progenitor cells, we performed immunofluorescence studies using antibodies against the five different muscarinic receptors (M1–M5) described in the literature [25]. As on many other effector cells innervated by cholinergic neurons, immunoreactivity for more than one type of muscarinic receptors (M3 and M5) was identified in the intestinal crypts [25]. The left panel of figure 10 illustrates the distribution of the immunoreactivity for the M5 receptor in the crypts as revealed by confocal microscopy in a 4 µm thin section of a crypt wall. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The disclosed fluorescence pattern is consistent with the immunoreactivy being located to cell membranes. The round structure in the middle of the figure is in all probability a cross-sectioned goblet cell surrounded by immunoreactivity. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (Figure 10, right panel).\nThe picture is photographed by confocal microscopy corresponding to a 4 µm thick section of the wall of the crypts. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The left panel of the figure suggests that the M5 receptor is located to the plasma membrane of the crypt cells. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (right panel).\nCheng and Leblond [26] proposed that the so called base columnar cells were the intestinal stem cells generating all types of epithelial cells, a hypothesis recently supported by Barker et al. [17], [18]. Thus, the intestinal stem cells seem to be the slender cells situated between the Paneth cells. We studied the possible presence of muscarinic receptors on these cells. Two types of muscarinic receptors M3 and M5, were investigated. Immunoreactivity to M3 and M5 receptors was identified on the presumed stem cell membrane, as illustrated for the M3 receptor on the left panel of Figure 11. The right panel shows the appearance of the tissue when exposed only to the secondary antibody.\nThe stem cells, indicated by arrows, are slender cells located between the Paneth cells (PC) at the very base of the crypts [17], [26]. The left part of the figure indicates that the muscarinic receptor M3 is located to the plasma membrane of the presumed intestinal stem cells. The tissue section in the right panel was only exposed to the secondary antibody. Mayer's haematoxylin. CL  =  crypt lumen.\n[SUBTITLE] Rate of epithelial renewal is decreased in capsaicin receptor knock-out mice [SUBSECTION] To investigate if the capsaicin receptor was of importance for cell renewal during “normal” conditions, experiments were performed on wild type mice and on mice lacking the TRPV1 gene [27]. 3H-thymidine (2 µCi) or BrdU (100 mg/kg b.wt.) were injected i.p. 2 h prior to extirpating the intestinal segments. 3H-thymidine incorporation into DNA was 529±46 DPM per mg DNA (n = 5) in control mice and 340±72 DPM per mg DNA (n = 5) in TRPV1 knock out mice (p<0.05).\nThe BrdU labeled cells were exclusively found in the intestinal crypts (cf.\nFigure 3). BrdU labeled cells in 215 crypts in wild type mice and 225 crypts in TRPV1−/− mice were counted. The number of labeled cells per crypt was significantly lower in TRPV1−/− than in wild type mice (right part of Figure 5; wild type: 11.4±0.2 BrdU labeled cells/crypt; knock-out: 9.2±0.2 BrdU labeled cells/crypt; n = 5 in both groups; p<0.01). Thus, in mice devoid of capsaicin receptors rate of 3H-thymidine incorporation into DNA and number of BrdU labeled crypt cells were significantly lower than in wild type mice.\nTo investigate if the capsaicin receptor was of importance for cell renewal during “normal” conditions, experiments were performed on wild type mice and on mice lacking the TRPV1 gene [27]. 3H-thymidine (2 µCi) or BrdU (100 mg/kg b.wt.) were injected i.p. 2 h prior to extirpating the intestinal segments. 3H-thymidine incorporation into DNA was 529±46 DPM per mg DNA (n = 5) in control mice and 340±72 DPM per mg DNA (n = 5) in TRPV1 knock out mice (p<0.05).\nThe BrdU labeled cells were exclusively found in the intestinal crypts (cf.\nFigure 3). BrdU labeled cells in 215 crypts in wild type mice and 225 crypts in TRPV1−/− mice were counted. The number of labeled cells per crypt was significantly lower in TRPV1−/− than in wild type mice (right part of Figure 5; wild type: 11.4±0.2 BrdU labeled cells/crypt; knock-out: 9.2±0.2 BrdU labeled cells/crypt; n = 5 in both groups; p<0.01). Thus, in mice devoid of capsaicin receptors rate of 3H-thymidine incorporation into DNA and number of BrdU labeled crypt cells were significantly lower than in wild type mice.", "TK activity and methyl-3H-thymidine incorporation into DNA was measured in the whole intestine. Hence, it was important to ensure that the recorded activity/incorporation mainly reflected events in the intestinal crypt stem/progenitor cells. This was tested in two ways. First, the left panel of Figure 1 shows an autoradiograph of the small intestine from a rat given methyl-3H-thymidine i.v. 6 h prior to the removal of the intestine. Dark field microscopy shows that the labeled thymidine was mainly located to the crypts, where intestinal epithelial stem/progenitor cells are found [16]–[18]. The right panel shows the same histological section in regular light microscopy. Figure 2 illustrates the localization of labeled thymidine at a higher magnification in intestinal crypts. The photographic grains are only seen above the nuclei of the cells (indicated by arrows). Second, in another type of experiments BrdU (a thymidine analogue) was given i.v. (50 mg kg−1). Two h after the i.v. injection, BrdU immunoreactivity was exclusively found in the crypt epithelial cells (Figure 3).\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine The left panel illustrates dark field microscopy of an autoradiogram showing radiolabeled cells (white spots) almost exclusively located in the epithelial layer of the crypts of Lieberkühn. To the right the same section is seen in light microscopy. Van Gieson.\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine. Only cell nuclei were found to be labeled (arrows).\nBrdU (50 mg/kg b.wt.), a thymidine analogue, was given 2 h prior to removing the intestine. The dark labeling by BrdU is exclusively located to the nuclei of intestinal crypt cells. Mayer's haematoxylin.", "Cholera toxin evokes an intestinal fluid secretion mainly by stimulating nervous reflex(es) contained within the ENS [11], [13]. Exposing the intestinal mucosa to cholera toxin (20 µg per mL) for three hours did not influence TK activity (91.2±11.2% (mean ± s.e.m) of control; n = 5; p = 0.5), but decreased the incorporation of labeled thymidine into DNA (84.4±5.5% of control; n = 5; p<0.05).We conclude that in the present short term experiments nervous reflexes activated by cholera toxin do not increase rate of epithelial renewal.", "Two experimental protocols were used in the capsaicin experiments, “non-perfusion experiments” and “perfusion experiments” (see Methods). Since both techniques indicated that a 1.6 mM capsaicin solution evoked a consistent increase of TK activity (152.7±8.6% of TK activity in control segments (non-perfusion experiments; n = 7; p-value <0.05) and 149.8±14.4% of control segments (perfusion experiments; n = 6; p-value <0.05; Figure 4), this concentration was used in subsequent experiments, most of which were performed with the perfusion technique.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). The experiments were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”; n = 6). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”; n = 7). Asterisks indicate statistical significance. Mean ± s.e.m.\nThe incorporation of labeled thymidine into DNA was significantly increased after 2 h capsaicin perfusion (178.0±30.1% of incorporation in control segments; n = 6; p<0.05). Furthermore, the number of BrdU labeled crypt cells increased concomitantly in the capsaicin group as shown on left part of Figure 5, based on counting labeled cells in 255 crypts in control and 249 crypts in capsaicin segments (control segments: 15.0±0.3 labeled cells/crypt; capsaicin segments: 17.9±0.3 labeled cells/crypt; n = 5; p<0.01). Hence, exposing the intestinal mucosa to 1.6 mM capsaicin increased TK activity and 3H-thymidine incorporation into DNA reflecting events in crypt epithelial cells as confirmed by the BrdU experiments.\nBrdU (50 mg/kg bwt) was given i.v. 2 h prior to removal of the intestinal segments. The number of labeled cells in 255 crypts in control (n = 5) and in 249 crypts in capsaicin segments (n = 5) were counted. Asterisk indicates statistical significance. Mean ± s.e.m. Mouse: BrdU labeled cells per crypt in wild type or capsaicin receptor knock-out mice. Five capsaicin receptor knock-out mice (−/−; 225 crypts) and in five wild type mice (WT; 215 crypts) were investigated. BrdU (100 mg/kg b.wt.) was administered i.p. 2 h before removal of the small intestine. Asterisk indicates statistical significance. Mean ± s.e.m.", "Capsaicin was used to activate selectively thin afferent nerve fibers presumably via the vallinoid receptor 1 (TRPV1) [14]. We explored if capsaicin indeed activated afferent nerves by monitoring changes of heart rate on line after luminal administration of a 1.6 mM capsaicin. The periarterial nerves were not severed in these experiments. Within ten minutes 1.6 mM capsaicin heart rate increased by 46.2±8.2 min−1, whereas heart rate was not affected (−4.7±5.1 min−1) in control experiments (n = 6; p<0.05), suggesting that luminal capsaicin activates nervous afferents.\nIn one series of experiments the extrinsic intestinal nerves were allowed to degenerate for 2–3 weeks after cutting the nerves around the superior mesenteric artery. The success of the denervation procedure was tested by immunohistochemical determinations of tyrosine hydroxylase in the intestine, located to extrinsic adrenergic nerve fibers [19]. Absence of labeled nerve fibers was taken to indicate a successful denervation. Exposing such “chronically” denervated intestinal segments to capsaicin did not significantly increase TK activity (102.7±3.1% of control; n = 5). This is significantly lower than the capsaicin effect observed in “acutely” denervated segments (144.1±8.3%; n = 18; p<0.01).\nTo further investigate the possible involvement of nerves in the capsaicin response, the intestinal mucosa was anesthetized with lidocaine (see Methods). In these experiments (Figure 6) no significant increase of TK activity was recorded (perfusion experiments: TK activity in capsaicin segments 97.5±14.4% of control; n = 8; p = 0.58; non-perfusion experiments: TK activity in capsaicin segments 103.1±13.7% of control; n = 7; p = 0.74), indicating a nervous involvement in the capsaicin response.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). Asterisks indicate statistical significance between the two groups of experiments (perfusion experiments, n = 8; non-perfusion experiments, n = 7). Mean ± s.e.m.\nIndirect evidence for an involvement of nerves was also obtained in experiments with the TPRV1 receptor blocker capsazepine (3 mg/kg i.v.) given prior to the intraluminal administration of capsaicin. In these experiments the TK activity of the capsaicin segment was 97.2±7.3% (n = 7) of that measured in corresponding control segments (p<0.01 compared to a value of 144.1±8.3 per cent in experiments without capsazepine; n = 18). Hence, the capsaicin effect on TK activity was mediated by a TPRV1 receptor. Since the TPRV1 receptor is known to be located on many extrinsic afferent axons[14], the results support the proposal that the capsaicin effect involves nerves.\nFinally, hexamethonium (a cholinergic, nicotinic receptor blocker; 10 mg/kg b.wt.; n = 8) significantly attenuated the effect of capsaicin on TK activity measuring 96.5±10.7% of the activity in capsaicin free segments (p<0.01 compared to 144.1±8.3 per cent in experiments without hexamethonium; n = 18). These results suggest that the capsaicin effect was mediated by ENS reflex(es) containing at least one cholinergic synapse.", "The experiments with chronically denervated intestines described above suggested that capsaicin stimulates an axon reflex. In many organs calcitonin gene related polypeptide (CGRP) or substance P (SP) have been shown to be involved in axon reflexes [see e.g. references 20]–[22]. To elucidate if this also was the case in the present experiments two types of studies were performed. First, blocking the CGRP receptors (human 8–37 α-CGRP; 1 mg kg−1 b.wt. i.v.) or the SP receptors (more specifically the NK1 receptors: Sendide; 1 mg kg−1 b.wt. i.v.) abolished the increase of TK activity evoked by capsaicin (Figure 7; CGRP receptors: TK activity in capsaicin segments 78.5±9.7% of control segments; n = 6; p<0.001 compared to control experiments; NK1 receptors: TK activity in capsaicin segments 82.6±6.8% of TK activity in control intestines; n = 6; p<0.001 compared to control experiments) and the augmented incorporation of 3H-thymidine into DNA (Figure 8; CGRP receptors: incorporation in capsaicin segments 68.9±14% of control intestines; n = 6; p<0.01; NK1 receptors: incorporation in capsaicin segments 90.6±9.8% of control; n = 6; p<0.001) induced by luminal capsaicin. The doses used have been previously shown to block effects of CGRP and SP [22]–[24]. Second, infusing the neurotransmitters retrogradely in the carotid artery at rates of 10−9 mol per min for 1 h increased significantly intestinal TK activity compared to activity at the start of infusion (Figure 9; CGRP: TK activity 141.0±7.6% of activity at time 0; n = 5; p<0.05; SP: TK activity 143.9±20.7% of activity at time 0; n = 5; p<0.05).\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). Thymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). N = 6 for each antagonist experiments. Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). The receptor blockers were given as described in Methods. 3H-thymidine incorporation is expressed in per cent of incorporation measured in control segments (no capsaicin). Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe peptides were infused for one hour at a rate of 10−9 mol per minute with or without prior administration of atropine (0.5 mg per kg b.wt. i.v.). Note that in these experiments thymidine kinase activity is expressed in per cent of activity measured in segments removed at time 0. Asterisks indicate statistical significance. Mean ± s.e.m.\nThere are two major types of cholinergic receptors, nicotinic and muscarinic. Above we described results from experiments with hexamethonium, a nicotinic receptor blocker. To block the muscarinic receptors, atropine (0.5 mg kg−1 b.wt.) was given i.v. Atropine significantly diminished the intestinal TK response to capsaicin (Figure 7; TK activity in capsaicin segments 91.8±5.6% of control; n = 6; p<0.001 compared to experiments without atropine) and the capsaicin evoked increase of 3H-thymidine incorporation into DNA (Figure 8; incorporation in capsaicin intestines 70.9±16.5% of control; n = 6; p = 0.002 compared to experiments without atropine).\nIn the atropine experiments it was possible to investigate the effect of atropine on rate of 3H-thymidine incorporation into DNA in control segments. After 2 h labeled nucleotide incorporation into DNA was 1017±81 DPM per mg DNA in animals not receiving atropine and 380±65 DPM per mg DNA in experiments with atropine (n = 6 in both groups; p<0.01 comparing the two groups). Hence, atropine lowered incorporation to about 40% of control suggesting an ongoing cholinergic, nervous influence in the absence of luminal capsaicin.\nA possible interaction between muscarinic and peptidergic transmission was investigated by intravascular infusion of CGRP or SP after atropine administration. No CGRP or SP effect on TK activity could be demonstrated after giving atropine. (Figure 9; CGRP: 90.8±7.1% of TK activity at start of peptide infusion; n = 7; p<0.05 compared to CGRP experiments without atropine; SP: 73.8±5.7% of TK activity at time 0; n = 7; p<0.05). The simplest explanation for these observations is that CGRP and SP influence cholinergic neuron(s), which, in turn, control the intestinal stem cells.\nTo elucidate if muscarinic receptors were present on the intestinal stem/progenitor cells, we performed immunofluorescence studies using antibodies against the five different muscarinic receptors (M1–M5) described in the literature [25]. As on many other effector cells innervated by cholinergic neurons, immunoreactivity for more than one type of muscarinic receptors (M3 and M5) was identified in the intestinal crypts [25]. The left panel of figure 10 illustrates the distribution of the immunoreactivity for the M5 receptor in the crypts as revealed by confocal microscopy in a 4 µm thin section of a crypt wall. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The disclosed fluorescence pattern is consistent with the immunoreactivy being located to cell membranes. The round structure in the middle of the figure is in all probability a cross-sectioned goblet cell surrounded by immunoreactivity. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (Figure 10, right panel).\nThe picture is photographed by confocal microscopy corresponding to a 4 µm thick section of the wall of the crypts. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The left panel of the figure suggests that the M5 receptor is located to the plasma membrane of the crypt cells. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (right panel).\nCheng and Leblond [26] proposed that the so called base columnar cells were the intestinal stem cells generating all types of epithelial cells, a hypothesis recently supported by Barker et al. [17], [18]. Thus, the intestinal stem cells seem to be the slender cells situated between the Paneth cells. We studied the possible presence of muscarinic receptors on these cells. Two types of muscarinic receptors M3 and M5, were investigated. Immunoreactivity to M3 and M5 receptors was identified on the presumed stem cell membrane, as illustrated for the M3 receptor on the left panel of Figure 11. The right panel shows the appearance of the tissue when exposed only to the secondary antibody.\nThe stem cells, indicated by arrows, are slender cells located between the Paneth cells (PC) at the very base of the crypts [17], [26]. The left part of the figure indicates that the muscarinic receptor M3 is located to the plasma membrane of the presumed intestinal stem cells. The tissue section in the right panel was only exposed to the secondary antibody. Mayer's haematoxylin. CL  =  crypt lumen.", "To investigate if the capsaicin receptor was of importance for cell renewal during “normal” conditions, experiments were performed on wild type mice and on mice lacking the TRPV1 gene [27]. 3H-thymidine (2 µCi) or BrdU (100 mg/kg b.wt.) were injected i.p. 2 h prior to extirpating the intestinal segments. 3H-thymidine incorporation into DNA was 529±46 DPM per mg DNA (n = 5) in control mice and 340±72 DPM per mg DNA (n = 5) in TRPV1 knock out mice (p<0.05).\nThe BrdU labeled cells were exclusively found in the intestinal crypts (cf.\nFigure 3). BrdU labeled cells in 215 crypts in wild type mice and 225 crypts in TRPV1−/− mice were counted. The number of labeled cells per crypt was significantly lower in TRPV1−/− than in wild type mice (right part of Figure 5; wild type: 11.4±0.2 BrdU labeled cells/crypt; knock-out: 9.2±0.2 BrdU labeled cells/crypt; n = 5 in both groups; p<0.01). Thus, in mice devoid of capsaicin receptors rate of 3H-thymidine incorporation into DNA and number of BrdU labeled crypt cells were significantly lower than in wild type mice.", "The intestinal epithelium is the most rapidly self-renewing epithelium in adult mammals, being replaced every 3–5 day. Each crypt contains 4–6 long-lived stem cells that are constantly dividing. The daughter cells of the stem cells (here called progenitor cells) divide every 12–16 h as they move “upwards” along the crypts generating some 300 cells per crypt every day [28]. The present study provides evidence for a nervous control of intestinal stem/progenitor cells in vivo. We have demonstrated that luminal capsaicin (1.6 mM), stimulating thin nervous afferents, increases intestinal TK activity, 3H-thymidine incorporation into DNA and number of BrdU labeled crypt cells. No such effects were evoked by cholera toxin. Several observations strongly indicate that the observed capsaicin effects were nervously mediated. Hence, lidocaine, a local anesthetic, and antagonists to four different neurotransmitter receptors (muscarinic, nicotinic, NK1 and CGRP receptors) significantly attenuated the capsaicin response. A nervous involvement was also indicated by the failure of capsaicin to influence TK activity after allowing a degeneration of the extrinsic nerves (“chronic denervation”). Finally, the capsaicin effect was mediated by TRPV1 receptors, known to be located on extrinsic afferent nerves [20], the capsaicin effect being abolished by the TRPV1 receptor blocker capsazepine.\nIn the present study the measured TK activity has been taken as indicator of rate of cell renewal of the intestinal epithelium. The reasons for this are the following: TK is an important enzyme in DNA replication, phosphorylating the nucleoside before being incorporated into DNA. The enzyme exists in a cytosolic and mitochondrial form [15]. Cytosolic TK is present only in proliferating cells. Its importance for cell renewal is suggested by the observations that TK activity in the crypts is about 20 times higher than in the villi when expressed per unit tissue weight or DNA weight [29], [30]. This conclusion is also supported by our observations that 3H-thymidine or BrdU injected intravenously was localized almost exclusively to the nuclei of crypt cells (Figures 1, 2 and 3), as also shown by Lipkin [31] for labeled thymidine. Furthermore, the increased incorporation of thymidine into DNA and the increased number of BrdU labeled crypt cells in response to luminal capsaicin strongly suggest that the augmented TK activity reflected a change of cytosolic TK activity. Finally, nerve transmitter receptor antagonists concomitantly abolished the increase of TK activity and thymidine incorporation into DNA (compare Figures 7 and 8).\nIt might be argued that the increase of total intestinal TK activity reflects a proliferation of other cells than the stem/progenitor cells of intestinal crypts. Observations reported in the literature indicate that this is unlikely. Smooth muscle cells show almost no renewal in the mouse stomach [32]. Fibroblasts in the intestinal submucosa are renewed with turn over times varying between 100 and 130 days [33]. Hence, the two most dominating tissues of the intestinal wall, apart from the epithelium, show no proliferation within the short time of the present experiments.\nThe “chronic denervation” experiments are key experiments in the analysis of the anatomical arrangement of the nerves influencing intestinal stem/progenitor cells, since such experiments make it possible to elucidate if a studied reflex is confined to the intestinal wall (“intramural reflex”) or mediated via an axon reflex. An axon reflex may be influenced by severing an afferent neuron, whereas an intramural reflex will not. Thus, the periarterial nerves were cut and acute experiments were performed 2–3 weeks later, when the extrinsic innervation (afferent and efferent) had degenerated. The absence of extrinsic nerves was confirmed by immunohistochemical investigations of tyrosine hydroxylase, a rate limiting enzyme in catecholamine synthesis [19]. In chronically denervated rats luminal capsaicin failed to cause an increase of TK activity, indicating that capsaicin exerted its effect via an axon reflex. These observations are consistent with the TRPV1 receptors being present only on extrinsic afferent nerves of the gut [20].\nThere exist a number of criteria for neurotransmitter candidates to be accepted as true transmitters. Observations reported in this and other studies indicate that most of these criteria have been fulfilled with regard to SP and CGRP and the afferent nervous control of intestinal cell renewal. First, immunohistochemical investigations have revealed intestinal, extrinsic, afferent, nerves containing SP and CGRP [14], [20]. Second, subgroups of the capsaicin sensitive neurons show SP and/or CGRP immunoreactivity [20]. Third, blocking the receptors for NK1 and CGRP significantly attenuated the effect of luminal capsaicin on TK activity and 3H-thymidine incorporation (Figures 7 and 8). Fourth, intra-arterial administration of the two peptides significantly increased TK activity (Figure 9). Fifth, a release of SP or CGRP from stimulated afferent neurons has been reported in the literature [22], [34]–[36]. Thus, taken together the observations clearly suggest that SP and CGRP are involved in the capsaicin effect on epithelial cell renewal. The current results indicate, however, that the NK1 or CGRP receptors are not localized to the stem/progenitor cells, since atropine abolished the TK increase induced by intra-arterially infused SP or CGRP (Figure 9).\nCapsaicin sensitivity is a trait of a subgroup of the unmyelinated C fibres but includes also some of the thinly myelinated Aδ fibres. Capsaicin stimulated receptors (TRPV1) are localized to extrinsic, afferent neurons usually considered to belong to the nociceptor system, a set of nerves that can be activated by stimuli (mechanical, thermal, chemical) capable of causing tissue damage. The TRPV1 receptor is located on free nerve endings of at least on two different types of neurons, one being a heat sensitive neuron and the other being a peptidergic, polymodal nociceptor neuron [37].\nCombining observations reported in the literature with the findings of the current investigation it is possible to propose a model to explain the findings of the present study. The model is shown on Figure 12 and the observations that constitute the basis for the model are briefly summarized in Table 1. It should be underlined that Figure 12 represents the simplest model that can be constructed from our findings and the observations reported in the literature. It is a working hypothesis, which can be used when designing further experiments to prove or disprove the proposed model. The synapse between the afferent neuron and the cholinergic neuron has been located to the submucosal plexus in the figure for two reasons: first, most neurons controlling mucosal functions have their cell somas in the submucosal plexus. Second, luminal capsaicin transiently reduces CGRP-like immunoreactivity in the submucosal but not in the myenteric plexus [39].\nThe model is in part based on the findings of the present study (cf.\nTable 1) and is consistent with known types of neurons in the intestinal wall. It implies that the direct control of the stem/progenitor cells is exerted by a cholinergic neuron, which, in turn, is under the influence of an axon reflex releasing CGRP/SP/acetylcholine. Ach: acetylcholine; CGRP: calcitonin gene related peptide; SP: substance P; CNS: central nervous system.\nTS: this study.\nThe proposed axon reflex control of intestinal cell renewal, may, however, not be the only nervous control of intestinal stem cells. We observed that atropine administration markedly attenuated nucleotide incorporation into DNA also in control segments. A rough calculation based on the averages of nucleotide incorporation (DPM per mg DNA) indicates that atropine decreased incorporation to about 40% of control. Hence, other enteric nerves than the thin extrinsic mucosal afferents may influence stem cell/progenitor cell proliferation, the cholinergic neuron being the final common pathway for more than one reflex. In fact, the results indicate that nervously released acetylcholine may play a very important role in the stem cell niche of the small intestine. With regard to the type of epithelial cells controlled by enteric nerves, observations reported by Bjerknes and Cheng [40], [41] indicate that the enteric nerves control the proliferation of the columnar daughter cells.\nThere is a fairly extensive literature dealing with muscarinic receptors and cell proliferation, in normal or cancerous cells (for reviews see [42], [43]). The muscarinic influence on proliferation is, at least in part, mediated by the “classical” signalling pathways for muscarinic receptors, the inositol-phospholipid signalling cascade [42] involving two second messengers, diacylglycerol and inositol 1,4,5-trisphosphate (IP3). This pathway is, however, not the only one mediating the muscarinic effect. Several studies report that the muscarinic receptor also is coupled to the mitogen-activated protein kinase (MAPK) pathway probably via a transactivation of the epidermal growth factor (EGF) [43]–[47] receptor located on the basolateral membrane of the crypt cell epithelium [48].\nBesides luminal capsaicin we investigated the effect of cholera toxin on TK activity and thymidine incorporation into DNA. It was observed that the toxin, if anything, decreased cell renewal to judge by attenuated nucleotide incorporation into DNA. This is consistent with earlier observations that the final neurotransmitter in the secretory reflex activated by cholera toxin is not acetylcholine but vasoactive intestinal polypeptide (VIP) [49]–[52]. Atropine does not attenuate cholera secretion [13] excluding a muscarinic mechanism. Furthermore, in contrast to the present findings chronic denervation does not influence the secretory response to cholera toxin [53]. We conclude that cholera toxin and capsaicin stimulate two different reflexes with different final neurotransmitters, one being confined to the intestinal wall (cholera toxin reflex), releasing VIP, and the other being an axon reflex (capsaicin reflex) releasing acetylcholine at the effector cell.\nWe have earlier reported that ENS is of great importance in explaining fluid secretion in acute diarrhoea [11]–[13]. As pointed out in the introduction diarrhoea can be looked upon as being an innate immunity response, since the secreted fluid dilutes the noxious agent(s). Furthermore, fluid secretion is often accompanied by a nervously induced motility response propelling the intestinal contents aborally [54]. Capsaicin activated intestinal nerves evoke an intestinal vasodilatation [22] and may be involved in the control of mucus secretion [55] and intestinal homing of leukocytes [56]. Taken together with the present observations it suggests that ENS coordinates several innate immunity mechanisms in the gut.\nAs discussed above the capsaicin receptor TPRV1 is usually considered to be a nociceptor sensing potentially noxious stimuli. It may, however, also participate in physiological mechanisms. The contents of the intestinal lumen influences mucosal morphology, parenteral feeding leading to a decreased mucosal thickness [57]. This is usually explained in terms of trophic factors, such as hormones. However, the studied reflex may be involved. There are at least three physiological ways by which the intestinal contents may influence mucosal afferent nerves. First, the TRPV1 receptor seems to be an osmoreceptor activated by hyperosmolality [58]. A tissue hyperosmolality has been demonstrated in the upper parts of intestinal villi in several species with different techniques [59], measuring about 800 mOsm in humans and cats provided the intestinal contents contain sodium and glucose. We propose that TRPV1 receptors are activated by a villous tissue hyperosmolality. Such a mechanism may explain the present findings in the knock-out mice implying that the capsaicin receptor is important for cell renewal even in the absence of luminal capsaicin.\nSecond, intake of food leads to the presence of pancreatic proteases in the intestinal lumen. It has been shown that proteinase-activated receptors 1 and 2 (PAR-1, PAR-2) are present on the apical border of enterocytes and activated by trypsin [60], [61]. In the case of PAR-2 receptors indirect evidence supports the proposal that a stimulation of this receptor may influence firing in afferent nerves from the gut. Thus, Cenac et al. [62], [63] and Nguyen et al. [64] reported that exposing the colonic mucosa to a PAR-2 receptor agonist causes a neurogenic inflammation via capsaicin sensitive nerves. Furthermore, the PAR-2 induced inflammation was attenuated by NK1 and CGRP antagonists.\nThird, one may speculate that the intestinal contents via the discharge of hormones from enteroendocrine cells activate mucosal afferents to increase epithelial cell renewal. A similar chain of events has been proposed for the activation of intramural secretory nervous reflexes by e.g. cholera toxin [11], [13] and rotavirus [12]. The finding reported by Bjerknes and Cheng [65] that glucagon-like-peptide-2 exerts its stimulatory effect on epithelial proliferation via nerves may be explained by such a mechanism.\nMicroorganisms in the intestinal lumen are obvious examples of potential noxious agents. Their presence may be recorded by the intestinal epithelium, functioning as a mucosal defence early warning system. Using membrane-associated or intracellular pattern-recognition molecules, such as toll-like receptors and Nod proteins, the epithelial cells sense the microorganisms [66]. This, in turn, may evoke a release of cytokines, prostaglandins and/or nitrous oxide from the enterocytes [67], biologically potent compounds with receptors on capsaicin sensitive nervous dendrites. Invading microorganisms may evoke an inflammatory response. Many of the substances participating in an inflammation, such as cytokines, prostaglandins, adenosine, bradykinin, histamine and 5-hydroxytryptamine, may directly or indirectly via activation of immunocytes, stimulate the studied axon reflex [38], [68], [69]. A morphological characteristic of ulcerative colitis is the “reactive“ hyperplasia of the epithelium [70]. One may speculate this hyperplasia may, at least in part, reflect an inflammatory activation of the axon reflex controlling intestinal stem/progenitor cells. Similar mechanisms may exist in connection with inflammation of other mucosae, as in the air ways and the urinary tract.\nIt is well established that there is a relationship between tissue repair and cancer, implying that chronic tissue injury may lead to malignancy (see e.g. [71]). A well-known example is colon cancer developing in longstanding ulcerative colitis. It has been suggested that overall risk of cancer development depends on the number of activated stem cells. Any repeated stimulus leading to a chronic increase in stem cell numbers could result in a higher overall frequency of cancer formation [28], [72]. Furthermore, it has been pointed out that the efficiency of carcinogenesis in tumour models is increased when combining mutagenic agents with agents that induce cell proliferation [73]. In this context it is interesting to note that Frucht and coworkers [74] found that 6 out of 10 human colon cancer cell lines were provided with functional muscarinic cholinergic receptors. Therefore, in view of the findings of the present study, one may pose the question if long standing activation of mucosal afferent nerve fibres by e.g. inflammation, may, via a cholinergic stimulation of muscarinic receptors and transactivation of EGF receptors, contribute to the development of cancer by their influence on epithelial proliferation.", "The protocol used in the present experiments was approved by the Ethical committee for animal experimentation at Gothenburg University, Gothenburg, Sweden (approval 292-2009). The experiments were performed in accordance with the recommendations issued by the Swedish Department of Agriculture.", "Experiments were performed on male Sprague-Dawley rats, weighing 240–450 g (Möllegaards Breeding Centre Ltd, Ejby, Denmark or B&K Universal AB, Sollentuna, Sweden). The rats were housed for at least 7 days prior to experiments in animal quarters (22°C, 60% relative humidity, artificial lighting between 06.00 and 18.00 h). Before the experiments the rats were deprived of food for at least 12 h. All experiments were carried out between 9.30 and 15.00.\n[SUBTITLE] Anesthesia and operative procedures [SUBSECTION] Anesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.\nAnesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.\n[SUBTITLE] Experimental procedures [SUBSECTION] Two to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.\nTwo to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.\n[SUBTITLE] Lidocaine experiments [SUBSECTION] In one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.\nIn one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.\n[SUBTITLE] Neurotransmitter receptor blockade [SUBSECTION] Several neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.\nSeveral neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.\n[SUBTITLE] Neurotransmitter i.a. infusions [SUBSECTION] SP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.\nSP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.\n[SUBTITLE] BrdU experiments [SUBSECTION] In 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.\nIn 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.\n[SUBTITLE] Chronic denervation experiments [SUBSECTION] In 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).\nIn 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).\n[SUBTITLE] Determination of thymidine kinase activity [SUBSECTION] Thymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\nThymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[SUBTITLE] Thymidine incorporation into DNA [SUBSECTION] [Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[SUBTITLE] Immunohistochemistry [SUBSECTION] Biopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.\nBiopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.\n[SUBTITLE] Autoradiography experiments [SUBSECTION] Methyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.\nMethyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.", "Anesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.", "Two to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.", "In one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.", "Several neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.", "SP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.", "In 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.", "In 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).", "Thymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.", "[Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.", "Biopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.", "Methyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.", "Experiments were performed on mice devoid of the capsaicin receptor (TRPV1 [27]; strain B6.129X1-Trpv1) purchased from The Jackson Laboratory (Bar Harbor, Maine). Strain B6129SF2/J was used as controls as recommended by the distributor. The mice were kept in animal quarters under standardized environmental conditions (see above). At the time of the experiments the animals were 10 weeks old.\n\n3H-thymidin (2 µCi) or BrdU (100 mg/kg body weight) dissolved in physiological saline was administered i.p.. Two h later the mice were killed with an overdose of sodium pentobarbital i.p. 3H-thymidine incorporation into DNA was determined and BrdU labeled cells were stained and counted as described for rat experiments.", "Peptides were bought from Bachem AG, (Bubendorf, Switzerland). All other chemicals, for which any supplier has not been indicated in text, were purchased from Sigma-Aldrich Sweden, Stockholm, Sweden.", "Heart rate was monitored on line with a rate meter coupled to the blood pressure recorder sensing the arterial pressure oscillations and recorded on a Grass polygraph.", "Non-parametric statistics were used throughout. Significance between paired observations was determined using Wilcoxon matched-pairs test. The Mann-Whitney U test was utilized when comparing independent groups. All calculations were performed with SPSS for Windows. Level of statistical significance was set at p<0.05." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Results", "Rapid cell renewal in the intestine is located to the crypts", "Cholera toxin does not increase rate of epithelial renewal", "Luminal capsaicin increases rate of epithelial renewal", "The capsaicin effect on epithelial renewal is mediated by nerves", "The capsaicin effect on epithelial renewal is mediated by muscarinic and peptidergic receptors", "Rate of epithelial renewal is decreased in capsaicin receptor knock-out mice", "Discussion", "Methods", "Ethical approval", "Rat experiments", "Anesthesia and operative procedures", "Experimental procedures", "Lidocaine experiments", "Neurotransmitter receptor blockade", "Neurotransmitter i.a. infusions", "BrdU experiments", "Chronic denervation experiments", "Determination of thymidine kinase activity", "Thymidine incorporation into DNA", "Immunohistochemistry", "Autoradiography experiments", "Mouse experiments", "Chemicals", "Recordings of heart rate", "Statistics" ]

[ "The intestinal epithelium consists of a single layer of columnar cells about 25 µm high. It represents an outer surface of the organism exposed to luminal contents of widely varying composition. The epithelium is of great importance for the survival of the organism, since chemicals and/or microorganisms, if allowed to pass the epithelial barrier, may represent a deadly threat. Thus, the intestinal epithelium may be looked upon as being an important part of innate immunity. The maintenance of the epithelium is secured by a rapid renewal, the epithelium being replaced every 3–5 days in mammals [1]. The key cells in this event are the intestinal stem and progenitor cells located in the intestinal crypts. The present study describes animal experiments, which suggest that there is a nervous reflex control of the stem/progenitor cells and, hence, rate of epithelial cell renewal.\nThe nervous control of the gastrointestinal tract is exerted by two systems: the extrinsic, efferent sympathetic and parasympathetic nerves and the intrinsic enteric nervous system (ENS). The latter represents a nervous system in the gastrointestinal tract, controlling epithelial transport, motility and blood flow [2]. ENS is composed of two major nerve plexuses, the myenteric and the submucosal and of their interconnections. Most of the neurons of the ENS are confined to the gastrointestinal wall, but extrinsic nerves are also found in the ENS. Two major types of nervous reflexes can be identified in the ENS: intramural reflexes and axon reflexes. Intramural reflexes are confined to the intestinal wall. The peristaltic reflex is an example of an intramural reflex initiated by luminal distension [2]. Branching external afferent nerves constitute the morphological bases of axon reflexes.\nMost investigations of the nervous control of gastrointestinal functions have been concerned with its control of motility, blood flow or epithelial transport. Although the extrinsic control of epithelial cell proliferation usually is discussed in terms of growth factors, peptide hormones and inflammatory mediators (for review, see reference [3]), there are observations to suggest also a nervous control of intestinal cell renewal. In most of these studies the influence of the extrinsic nerves has been investigated, suggesting that both branches of the extrinsic innervation may increase mitotic rate in the crypt cells [4]–[6]. Chemical ablation of the myenteric plexus leads to an increased crypt cell proliferation both in the small intestine [6]–[8] and in the colon [9], [10], indicating that myenteric nerves may exert an inhibitory influence on intestinal cell renewal.\nWe have earlier demonstrated that ENS plays an important role in the control of fluid transport. In fact, in many, if not most, types of acute diarrhea, an activation of ENS is the major mechanism underlying the observed intestinal fluid loss [11]–[13]. Diarrhea can be looked upon as being an innate immunity response, since the secreted fluid dilutes the potentially noxious agent(s). The maintenance of the epithelial barrier is also important for innate immunity. Therefore, we decided to investigate if intestinal contents, via the ENS, also may influence epithelial cell renewal. Most experiments were performed on intestinal segments that had been acutely denervated by cutting the nerves accompanying the superior mesenteric artery, minimizing the influence of the external innervation. The ENS was activated in two different ways: In one type of experiments an intramural secretory reflex was stimulated by exposing the intestinal mucosa to cholera toxin [11]. In another series we utilized capsaicin, the “hot” agent of red pepper, to activate thin mucosal afferents [14]. Capsaicin is known to activate the TRPV1 receptor. Cell renewal was studied in three different ways: We determined deoxythymidine kinase (TK) activity in the intestinal wall. TK facilitates the phosphorylation of deoxythymidine, associated with the formation of DNA [15]. We also measured the rate of incorporation of radioactively labeled thymidine into intestinal DNA. Finally, epithelial cell renewal was estimated by means of bromodeoxyuridine (BrdU), a thymidine analogue possible to localize at the cellular level in tissue sections with immunohistochemistry.", "[SUBTITLE] Rapid cell renewal in the intestine is located to the crypts [SUBSECTION] TK activity and methyl-3H-thymidine incorporation into DNA was measured in the whole intestine. Hence, it was important to ensure that the recorded activity/incorporation mainly reflected events in the intestinal crypt stem/progenitor cells. This was tested in two ways. First, the left panel of Figure 1 shows an autoradiograph of the small intestine from a rat given methyl-3H-thymidine i.v. 6 h prior to the removal of the intestine. Dark field microscopy shows that the labeled thymidine was mainly located to the crypts, where intestinal epithelial stem/progenitor cells are found [16]–[18]. The right panel shows the same histological section in regular light microscopy. Figure 2 illustrates the localization of labeled thymidine at a higher magnification in intestinal crypts. The photographic grains are only seen above the nuclei of the cells (indicated by arrows). Second, in another type of experiments BrdU (a thymidine analogue) was given i.v. (50 mg kg−1). Two h after the i.v. injection, BrdU immunoreactivity was exclusively found in the crypt epithelial cells (Figure 3).\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine The left panel illustrates dark field microscopy of an autoradiogram showing radiolabeled cells (white spots) almost exclusively located in the epithelial layer of the crypts of Lieberkühn. To the right the same section is seen in light microscopy. Van Gieson.\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine. Only cell nuclei were found to be labeled (arrows).\nBrdU (50 mg/kg b.wt.), a thymidine analogue, was given 2 h prior to removing the intestine. The dark labeling by BrdU is exclusively located to the nuclei of intestinal crypt cells. Mayer's haematoxylin.\nTK activity and methyl-3H-thymidine incorporation into DNA was measured in the whole intestine. Hence, it was important to ensure that the recorded activity/incorporation mainly reflected events in the intestinal crypt stem/progenitor cells. This was tested in two ways. First, the left panel of Figure 1 shows an autoradiograph of the small intestine from a rat given methyl-3H-thymidine i.v. 6 h prior to the removal of the intestine. Dark field microscopy shows that the labeled thymidine was mainly located to the crypts, where intestinal epithelial stem/progenitor cells are found [16]–[18]. The right panel shows the same histological section in regular light microscopy. Figure 2 illustrates the localization of labeled thymidine at a higher magnification in intestinal crypts. The photographic grains are only seen above the nuclei of the cells (indicated by arrows). Second, in another type of experiments BrdU (a thymidine analogue) was given i.v. (50 mg kg−1). Two h after the i.v. injection, BrdU immunoreactivity was exclusively found in the crypt epithelial cells (Figure 3).\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine The left panel illustrates dark field microscopy of an autoradiogram showing radiolabeled cells (white spots) almost exclusively located in the epithelial layer of the crypts of Lieberkühn. To the right the same section is seen in light microscopy. Van Gieson.\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine. Only cell nuclei were found to be labeled (arrows).\nBrdU (50 mg/kg b.wt.), a thymidine analogue, was given 2 h prior to removing the intestine. The dark labeling by BrdU is exclusively located to the nuclei of intestinal crypt cells. Mayer's haematoxylin.\n[SUBTITLE] Cholera toxin does not increase rate of epithelial renewal [SUBSECTION] Cholera toxin evokes an intestinal fluid secretion mainly by stimulating nervous reflex(es) contained within the ENS [11], [13]. Exposing the intestinal mucosa to cholera toxin (20 µg per mL) for three hours did not influence TK activity (91.2±11.2% (mean ± s.e.m) of control; n = 5; p = 0.5), but decreased the incorporation of labeled thymidine into DNA (84.4±5.5% of control; n = 5; p<0.05).We conclude that in the present short term experiments nervous reflexes activated by cholera toxin do not increase rate of epithelial renewal.\nCholera toxin evokes an intestinal fluid secretion mainly by stimulating nervous reflex(es) contained within the ENS [11], [13]. Exposing the intestinal mucosa to cholera toxin (20 µg per mL) for three hours did not influence TK activity (91.2±11.2% (mean ± s.e.m) of control; n = 5; p = 0.5), but decreased the incorporation of labeled thymidine into DNA (84.4±5.5% of control; n = 5; p<0.05).We conclude that in the present short term experiments nervous reflexes activated by cholera toxin do not increase rate of epithelial renewal.\n[SUBTITLE] Luminal capsaicin increases rate of epithelial renewal [SUBSECTION] Two experimental protocols were used in the capsaicin experiments, “non-perfusion experiments” and “perfusion experiments” (see Methods). Since both techniques indicated that a 1.6 mM capsaicin solution evoked a consistent increase of TK activity (152.7±8.6% of TK activity in control segments (non-perfusion experiments; n = 7; p-value <0.05) and 149.8±14.4% of control segments (perfusion experiments; n = 6; p-value <0.05; Figure 4), this concentration was used in subsequent experiments, most of which were performed with the perfusion technique.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). The experiments were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”; n = 6). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”; n = 7). Asterisks indicate statistical significance. Mean ± s.e.m.\nThe incorporation of labeled thymidine into DNA was significantly increased after 2 h capsaicin perfusion (178.0±30.1% of incorporation in control segments; n = 6; p<0.05). Furthermore, the number of BrdU labeled crypt cells increased concomitantly in the capsaicin group as shown on left part of Figure 5, based on counting labeled cells in 255 crypts in control and 249 crypts in capsaicin segments (control segments: 15.0±0.3 labeled cells/crypt; capsaicin segments: 17.9±0.3 labeled cells/crypt; n = 5; p<0.01). Hence, exposing the intestinal mucosa to 1.6 mM capsaicin increased TK activity and 3H-thymidine incorporation into DNA reflecting events in crypt epithelial cells as confirmed by the BrdU experiments.\nBrdU (50 mg/kg bwt) was given i.v. 2 h prior to removal of the intestinal segments. The number of labeled cells in 255 crypts in control (n = 5) and in 249 crypts in capsaicin segments (n = 5) were counted. Asterisk indicates statistical significance. Mean ± s.e.m. Mouse: BrdU labeled cells per crypt in wild type or capsaicin receptor knock-out mice. Five capsaicin receptor knock-out mice (−/−; 225 crypts) and in five wild type mice (WT; 215 crypts) were investigated. BrdU (100 mg/kg b.wt.) was administered i.p. 2 h before removal of the small intestine. Asterisk indicates statistical significance. Mean ± s.e.m.\nTwo experimental protocols were used in the capsaicin experiments, “non-perfusion experiments” and “perfusion experiments” (see Methods). Since both techniques indicated that a 1.6 mM capsaicin solution evoked a consistent increase of TK activity (152.7±8.6% of TK activity in control segments (non-perfusion experiments; n = 7; p-value <0.05) and 149.8±14.4% of control segments (perfusion experiments; n = 6; p-value <0.05; Figure 4), this concentration was used in subsequent experiments, most of which were performed with the perfusion technique.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). The experiments were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”; n = 6). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”; n = 7). Asterisks indicate statistical significance. Mean ± s.e.m.\nThe incorporation of labeled thymidine into DNA was significantly increased after 2 h capsaicin perfusion (178.0±30.1% of incorporation in control segments; n = 6; p<0.05). Furthermore, the number of BrdU labeled crypt cells increased concomitantly in the capsaicin group as shown on left part of Figure 5, based on counting labeled cells in 255 crypts in control and 249 crypts in capsaicin segments (control segments: 15.0±0.3 labeled cells/crypt; capsaicin segments: 17.9±0.3 labeled cells/crypt; n = 5; p<0.01). Hence, exposing the intestinal mucosa to 1.6 mM capsaicin increased TK activity and 3H-thymidine incorporation into DNA reflecting events in crypt epithelial cells as confirmed by the BrdU experiments.\nBrdU (50 mg/kg bwt) was given i.v. 2 h prior to removal of the intestinal segments. The number of labeled cells in 255 crypts in control (n = 5) and in 249 crypts in capsaicin segments (n = 5) were counted. Asterisk indicates statistical significance. Mean ± s.e.m. Mouse: BrdU labeled cells per crypt in wild type or capsaicin receptor knock-out mice. Five capsaicin receptor knock-out mice (−/−; 225 crypts) and in five wild type mice (WT; 215 crypts) were investigated. BrdU (100 mg/kg b.wt.) was administered i.p. 2 h before removal of the small intestine. Asterisk indicates statistical significance. Mean ± s.e.m.\n[SUBTITLE] The capsaicin effect on epithelial renewal is mediated by nerves [SUBSECTION] Capsaicin was used to activate selectively thin afferent nerve fibers presumably via the vallinoid receptor 1 (TRPV1) [14]. We explored if capsaicin indeed activated afferent nerves by monitoring changes of heart rate on line after luminal administration of a 1.6 mM capsaicin. The periarterial nerves were not severed in these experiments. Within ten minutes 1.6 mM capsaicin heart rate increased by 46.2±8.2 min−1, whereas heart rate was not affected (−4.7±5.1 min−1) in control experiments (n = 6; p<0.05), suggesting that luminal capsaicin activates nervous afferents.\nIn one series of experiments the extrinsic intestinal nerves were allowed to degenerate for 2–3 weeks after cutting the nerves around the superior mesenteric artery. The success of the denervation procedure was tested by immunohistochemical determinations of tyrosine hydroxylase in the intestine, located to extrinsic adrenergic nerve fibers [19]. Absence of labeled nerve fibers was taken to indicate a successful denervation. Exposing such “chronically” denervated intestinal segments to capsaicin did not significantly increase TK activity (102.7±3.1% of control; n = 5). This is significantly lower than the capsaicin effect observed in “acutely” denervated segments (144.1±8.3%; n = 18; p<0.01).\nTo further investigate the possible involvement of nerves in the capsaicin response, the intestinal mucosa was anesthetized with lidocaine (see Methods). In these experiments (Figure 6) no significant increase of TK activity was recorded (perfusion experiments: TK activity in capsaicin segments 97.5±14.4% of control; n = 8; p = 0.58; non-perfusion experiments: TK activity in capsaicin segments 103.1±13.7% of control; n = 7; p = 0.74), indicating a nervous involvement in the capsaicin response.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). Asterisks indicate statistical significance between the two groups of experiments (perfusion experiments, n = 8; non-perfusion experiments, n = 7). Mean ± s.e.m.\nIndirect evidence for an involvement of nerves was also obtained in experiments with the TPRV1 receptor blocker capsazepine (3 mg/kg i.v.) given prior to the intraluminal administration of capsaicin. In these experiments the TK activity of the capsaicin segment was 97.2±7.3% (n = 7) of that measured in corresponding control segments (p<0.01 compared to a value of 144.1±8.3 per cent in experiments without capsazepine; n = 18). Hence, the capsaicin effect on TK activity was mediated by a TPRV1 receptor. Since the TPRV1 receptor is known to be located on many extrinsic afferent axons[14], the results support the proposal that the capsaicin effect involves nerves.\nFinally, hexamethonium (a cholinergic, nicotinic receptor blocker; 10 mg/kg b.wt.; n = 8) significantly attenuated the effect of capsaicin on TK activity measuring 96.5±10.7% of the activity in capsaicin free segments (p<0.01 compared to 144.1±8.3 per cent in experiments without hexamethonium; n = 18). These results suggest that the capsaicin effect was mediated by ENS reflex(es) containing at least one cholinergic synapse.\nCapsaicin was used to activate selectively thin afferent nerve fibers presumably via the vallinoid receptor 1 (TRPV1) [14]. We explored if capsaicin indeed activated afferent nerves by monitoring changes of heart rate on line after luminal administration of a 1.6 mM capsaicin. The periarterial nerves were not severed in these experiments. Within ten minutes 1.6 mM capsaicin heart rate increased by 46.2±8.2 min−1, whereas heart rate was not affected (−4.7±5.1 min−1) in control experiments (n = 6; p<0.05), suggesting that luminal capsaicin activates nervous afferents.\nIn one series of experiments the extrinsic intestinal nerves were allowed to degenerate for 2–3 weeks after cutting the nerves around the superior mesenteric artery. The success of the denervation procedure was tested by immunohistochemical determinations of tyrosine hydroxylase in the intestine, located to extrinsic adrenergic nerve fibers [19]. Absence of labeled nerve fibers was taken to indicate a successful denervation. Exposing such “chronically” denervated intestinal segments to capsaicin did not significantly increase TK activity (102.7±3.1% of control; n = 5). This is significantly lower than the capsaicin effect observed in “acutely” denervated segments (144.1±8.3%; n = 18; p<0.01).\nTo further investigate the possible involvement of nerves in the capsaicin response, the intestinal mucosa was anesthetized with lidocaine (see Methods). In these experiments (Figure 6) no significant increase of TK activity was recorded (perfusion experiments: TK activity in capsaicin segments 97.5±14.4% of control; n = 8; p = 0.58; non-perfusion experiments: TK activity in capsaicin segments 103.1±13.7% of control; n = 7; p = 0.74), indicating a nervous involvement in the capsaicin response.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). Asterisks indicate statistical significance between the two groups of experiments (perfusion experiments, n = 8; non-perfusion experiments, n = 7). Mean ± s.e.m.\nIndirect evidence for an involvement of nerves was also obtained in experiments with the TPRV1 receptor blocker capsazepine (3 mg/kg i.v.) given prior to the intraluminal administration of capsaicin. In these experiments the TK activity of the capsaicin segment was 97.2±7.3% (n = 7) of that measured in corresponding control segments (p<0.01 compared to a value of 144.1±8.3 per cent in experiments without capsazepine; n = 18). Hence, the capsaicin effect on TK activity was mediated by a TPRV1 receptor. Since the TPRV1 receptor is known to be located on many extrinsic afferent axons[14], the results support the proposal that the capsaicin effect involves nerves.\nFinally, hexamethonium (a cholinergic, nicotinic receptor blocker; 10 mg/kg b.wt.; n = 8) significantly attenuated the effect of capsaicin on TK activity measuring 96.5±10.7% of the activity in capsaicin free segments (p<0.01 compared to 144.1±8.3 per cent in experiments without hexamethonium; n = 18). These results suggest that the capsaicin effect was mediated by ENS reflex(es) containing at least one cholinergic synapse.\n[SUBTITLE] The capsaicin effect on epithelial renewal is mediated by muscarinic and peptidergic receptors [SUBSECTION] The experiments with chronically denervated intestines described above suggested that capsaicin stimulates an axon reflex. In many organs calcitonin gene related polypeptide (CGRP) or substance P (SP) have been shown to be involved in axon reflexes [see e.g. references 20]–[22]. To elucidate if this also was the case in the present experiments two types of studies were performed. First, blocking the CGRP receptors (human 8–37 α-CGRP; 1 mg kg−1 b.wt. i.v.) or the SP receptors (more specifically the NK1 receptors: Sendide; 1 mg kg−1 b.wt. i.v.) abolished the increase of TK activity evoked by capsaicin (Figure 7; CGRP receptors: TK activity in capsaicin segments 78.5±9.7% of control segments; n = 6; p<0.001 compared to control experiments; NK1 receptors: TK activity in capsaicin segments 82.6±6.8% of TK activity in control intestines; n = 6; p<0.001 compared to control experiments) and the augmented incorporation of 3H-thymidine into DNA (Figure 8; CGRP receptors: incorporation in capsaicin segments 68.9±14% of control intestines; n = 6; p<0.01; NK1 receptors: incorporation in capsaicin segments 90.6±9.8% of control; n = 6; p<0.001) induced by luminal capsaicin. The doses used have been previously shown to block effects of CGRP and SP [22]–[24]. Second, infusing the neurotransmitters retrogradely in the carotid artery at rates of 10−9 mol per min for 1 h increased significantly intestinal TK activity compared to activity at the start of infusion (Figure 9; CGRP: TK activity 141.0±7.6% of activity at time 0; n = 5; p<0.05; SP: TK activity 143.9±20.7% of activity at time 0; n = 5; p<0.05).\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). Thymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). N = 6 for each antagonist experiments. Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). The receptor blockers were given as described in Methods. 3H-thymidine incorporation is expressed in per cent of incorporation measured in control segments (no capsaicin). Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe peptides were infused for one hour at a rate of 10−9 mol per minute with or without prior administration of atropine (0.5 mg per kg b.wt. i.v.). Note that in these experiments thymidine kinase activity is expressed in per cent of activity measured in segments removed at time 0. Asterisks indicate statistical significance. Mean ± s.e.m.\nThere are two major types of cholinergic receptors, nicotinic and muscarinic. Above we described results from experiments with hexamethonium, a nicotinic receptor blocker. To block the muscarinic receptors, atropine (0.5 mg kg−1 b.wt.) was given i.v. Atropine significantly diminished the intestinal TK response to capsaicin (Figure 7; TK activity in capsaicin segments 91.8±5.6% of control; n = 6; p<0.001 compared to experiments without atropine) and the capsaicin evoked increase of 3H-thymidine incorporation into DNA (Figure 8; incorporation in capsaicin intestines 70.9±16.5% of control; n = 6; p = 0.002 compared to experiments without atropine).\nIn the atropine experiments it was possible to investigate the effect of atropine on rate of 3H-thymidine incorporation into DNA in control segments. After 2 h labeled nucleotide incorporation into DNA was 1017±81 DPM per mg DNA in animals not receiving atropine and 380±65 DPM per mg DNA in experiments with atropine (n = 6 in both groups; p<0.01 comparing the two groups). Hence, atropine lowered incorporation to about 40% of control suggesting an ongoing cholinergic, nervous influence in the absence of luminal capsaicin.\nA possible interaction between muscarinic and peptidergic transmission was investigated by intravascular infusion of CGRP or SP after atropine administration. No CGRP or SP effect on TK activity could be demonstrated after giving atropine. (Figure 9; CGRP: 90.8±7.1% of TK activity at start of peptide infusion; n = 7; p<0.05 compared to CGRP experiments without atropine; SP: 73.8±5.7% of TK activity at time 0; n = 7; p<0.05). The simplest explanation for these observations is that CGRP and SP influence cholinergic neuron(s), which, in turn, control the intestinal stem cells.\nTo elucidate if muscarinic receptors were present on the intestinal stem/progenitor cells, we performed immunofluorescence studies using antibodies against the five different muscarinic receptors (M1–M5) described in the literature [25]. As on many other effector cells innervated by cholinergic neurons, immunoreactivity for more than one type of muscarinic receptors (M3 and M5) was identified in the intestinal crypts [25]. The left panel of figure 10 illustrates the distribution of the immunoreactivity for the M5 receptor in the crypts as revealed by confocal microscopy in a 4 µm thin section of a crypt wall. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The disclosed fluorescence pattern is consistent with the immunoreactivy being located to cell membranes. The round structure in the middle of the figure is in all probability a cross-sectioned goblet cell surrounded by immunoreactivity. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (Figure 10, right panel).\nThe picture is photographed by confocal microscopy corresponding to a 4 µm thick section of the wall of the crypts. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The left panel of the figure suggests that the M5 receptor is located to the plasma membrane of the crypt cells. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (right panel).\nCheng and Leblond [26] proposed that the so called base columnar cells were the intestinal stem cells generating all types of epithelial cells, a hypothesis recently supported by Barker et al. [17], [18]. Thus, the intestinal stem cells seem to be the slender cells situated between the Paneth cells. We studied the possible presence of muscarinic receptors on these cells. Two types of muscarinic receptors M3 and M5, were investigated. Immunoreactivity to M3 and M5 receptors was identified on the presumed stem cell membrane, as illustrated for the M3 receptor on the left panel of Figure 11. The right panel shows the appearance of the tissue when exposed only to the secondary antibody.\nThe stem cells, indicated by arrows, are slender cells located between the Paneth cells (PC) at the very base of the crypts [17], [26]. The left part of the figure indicates that the muscarinic receptor M3 is located to the plasma membrane of the presumed intestinal stem cells. The tissue section in the right panel was only exposed to the secondary antibody. Mayer's haematoxylin. CL  =  crypt lumen.\nThe experiments with chronically denervated intestines described above suggested that capsaicin stimulates an axon reflex. In many organs calcitonin gene related polypeptide (CGRP) or substance P (SP) have been shown to be involved in axon reflexes [see e.g. references 20]–[22]. To elucidate if this also was the case in the present experiments two types of studies were performed. First, blocking the CGRP receptors (human 8–37 α-CGRP; 1 mg kg−1 b.wt. i.v.) or the SP receptors (more specifically the NK1 receptors: Sendide; 1 mg kg−1 b.wt. i.v.) abolished the increase of TK activity evoked by capsaicin (Figure 7; CGRP receptors: TK activity in capsaicin segments 78.5±9.7% of control segments; n = 6; p<0.001 compared to control experiments; NK1 receptors: TK activity in capsaicin segments 82.6±6.8% of TK activity in control intestines; n = 6; p<0.001 compared to control experiments) and the augmented incorporation of 3H-thymidine into DNA (Figure 8; CGRP receptors: incorporation in capsaicin segments 68.9±14% of control intestines; n = 6; p<0.01; NK1 receptors: incorporation in capsaicin segments 90.6±9.8% of control; n = 6; p<0.001) induced by luminal capsaicin. The doses used have been previously shown to block effects of CGRP and SP [22]–[24]. Second, infusing the neurotransmitters retrogradely in the carotid artery at rates of 10−9 mol per min for 1 h increased significantly intestinal TK activity compared to activity at the start of infusion (Figure 9; CGRP: TK activity 141.0±7.6% of activity at time 0; n = 5; p<0.05; SP: TK activity 143.9±20.7% of activity at time 0; n = 5; p<0.05).\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). Thymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). N = 6 for each antagonist experiments. Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). The receptor blockers were given as described in Methods. 3H-thymidine incorporation is expressed in per cent of incorporation measured in control segments (no capsaicin). Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe peptides were infused for one hour at a rate of 10−9 mol per minute with or without prior administration of atropine (0.5 mg per kg b.wt. i.v.). Note that in these experiments thymidine kinase activity is expressed in per cent of activity measured in segments removed at time 0. Asterisks indicate statistical significance. Mean ± s.e.m.\nThere are two major types of cholinergic receptors, nicotinic and muscarinic. Above we described results from experiments with hexamethonium, a nicotinic receptor blocker. To block the muscarinic receptors, atropine (0.5 mg kg−1 b.wt.) was given i.v. Atropine significantly diminished the intestinal TK response to capsaicin (Figure 7; TK activity in capsaicin segments 91.8±5.6% of control; n = 6; p<0.001 compared to experiments without atropine) and the capsaicin evoked increase of 3H-thymidine incorporation into DNA (Figure 8; incorporation in capsaicin intestines 70.9±16.5% of control; n = 6; p = 0.002 compared to experiments without atropine).\nIn the atropine experiments it was possible to investigate the effect of atropine on rate of 3H-thymidine incorporation into DNA in control segments. After 2 h labeled nucleotide incorporation into DNA was 1017±81 DPM per mg DNA in animals not receiving atropine and 380±65 DPM per mg DNA in experiments with atropine (n = 6 in both groups; p<0.01 comparing the two groups). Hence, atropine lowered incorporation to about 40% of control suggesting an ongoing cholinergic, nervous influence in the absence of luminal capsaicin.\nA possible interaction between muscarinic and peptidergic transmission was investigated by intravascular infusion of CGRP or SP after atropine administration. No CGRP or SP effect on TK activity could be demonstrated after giving atropine. (Figure 9; CGRP: 90.8±7.1% of TK activity at start of peptide infusion; n = 7; p<0.05 compared to CGRP experiments without atropine; SP: 73.8±5.7% of TK activity at time 0; n = 7; p<0.05). The simplest explanation for these observations is that CGRP and SP influence cholinergic neuron(s), which, in turn, control the intestinal stem cells.\nTo elucidate if muscarinic receptors were present on the intestinal stem/progenitor cells, we performed immunofluorescence studies using antibodies against the five different muscarinic receptors (M1–M5) described in the literature [25]. As on many other effector cells innervated by cholinergic neurons, immunoreactivity for more than one type of muscarinic receptors (M3 and M5) was identified in the intestinal crypts [25]. The left panel of figure 10 illustrates the distribution of the immunoreactivity for the M5 receptor in the crypts as revealed by confocal microscopy in a 4 µm thin section of a crypt wall. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The disclosed fluorescence pattern is consistent with the immunoreactivy being located to cell membranes. The round structure in the middle of the figure is in all probability a cross-sectioned goblet cell surrounded by immunoreactivity. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (Figure 10, right panel).\nThe picture is photographed by confocal microscopy corresponding to a 4 µm thick section of the wall of the crypts. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The left panel of the figure suggests that the M5 receptor is located to the plasma membrane of the crypt cells. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (right panel).\nCheng and Leblond [26] proposed that the so called base columnar cells were the intestinal stem cells generating all types of epithelial cells, a hypothesis recently supported by Barker et al. [17], [18]. Thus, the intestinal stem cells seem to be the slender cells situated between the Paneth cells. We studied the possible presence of muscarinic receptors on these cells. Two types of muscarinic receptors M3 and M5, were investigated. Immunoreactivity to M3 and M5 receptors was identified on the presumed stem cell membrane, as illustrated for the M3 receptor on the left panel of Figure 11. The right panel shows the appearance of the tissue when exposed only to the secondary antibody.\nThe stem cells, indicated by arrows, are slender cells located between the Paneth cells (PC) at the very base of the crypts [17], [26]. The left part of the figure indicates that the muscarinic receptor M3 is located to the plasma membrane of the presumed intestinal stem cells. The tissue section in the right panel was only exposed to the secondary antibody. Mayer's haematoxylin. CL  =  crypt lumen.\n[SUBTITLE] Rate of epithelial renewal is decreased in capsaicin receptor knock-out mice [SUBSECTION] To investigate if the capsaicin receptor was of importance for cell renewal during “normal” conditions, experiments were performed on wild type mice and on mice lacking the TRPV1 gene [27]. 3H-thymidine (2 µCi) or BrdU (100 mg/kg b.wt.) were injected i.p. 2 h prior to extirpating the intestinal segments. 3H-thymidine incorporation into DNA was 529±46 DPM per mg DNA (n = 5) in control mice and 340±72 DPM per mg DNA (n = 5) in TRPV1 knock out mice (p<0.05).\nThe BrdU labeled cells were exclusively found in the intestinal crypts (cf.\nFigure 3). BrdU labeled cells in 215 crypts in wild type mice and 225 crypts in TRPV1−/− mice were counted. The number of labeled cells per crypt was significantly lower in TRPV1−/− than in wild type mice (right part of Figure 5; wild type: 11.4±0.2 BrdU labeled cells/crypt; knock-out: 9.2±0.2 BrdU labeled cells/crypt; n = 5 in both groups; p<0.01). Thus, in mice devoid of capsaicin receptors rate of 3H-thymidine incorporation into DNA and number of BrdU labeled crypt cells were significantly lower than in wild type mice.\nTo investigate if the capsaicin receptor was of importance for cell renewal during “normal” conditions, experiments were performed on wild type mice and on mice lacking the TRPV1 gene [27]. 3H-thymidine (2 µCi) or BrdU (100 mg/kg b.wt.) were injected i.p. 2 h prior to extirpating the intestinal segments. 3H-thymidine incorporation into DNA was 529±46 DPM per mg DNA (n = 5) in control mice and 340±72 DPM per mg DNA (n = 5) in TRPV1 knock out mice (p<0.05).\nThe BrdU labeled cells were exclusively found in the intestinal crypts (cf.\nFigure 3). BrdU labeled cells in 215 crypts in wild type mice and 225 crypts in TRPV1−/− mice were counted. The number of labeled cells per crypt was significantly lower in TRPV1−/− than in wild type mice (right part of Figure 5; wild type: 11.4±0.2 BrdU labeled cells/crypt; knock-out: 9.2±0.2 BrdU labeled cells/crypt; n = 5 in both groups; p<0.01). Thus, in mice devoid of capsaicin receptors rate of 3H-thymidine incorporation into DNA and number of BrdU labeled crypt cells were significantly lower than in wild type mice.", "TK activity and methyl-3H-thymidine incorporation into DNA was measured in the whole intestine. Hence, it was important to ensure that the recorded activity/incorporation mainly reflected events in the intestinal crypt stem/progenitor cells. This was tested in two ways. First, the left panel of Figure 1 shows an autoradiograph of the small intestine from a rat given methyl-3H-thymidine i.v. 6 h prior to the removal of the intestine. Dark field microscopy shows that the labeled thymidine was mainly located to the crypts, where intestinal epithelial stem/progenitor cells are found [16]–[18]. The right panel shows the same histological section in regular light microscopy. Figure 2 illustrates the localization of labeled thymidine at a higher magnification in intestinal crypts. The photographic grains are only seen above the nuclei of the cells (indicated by arrows). Second, in another type of experiments BrdU (a thymidine analogue) was given i.v. (50 mg kg−1). Two h after the i.v. injection, BrdU immunoreactivity was exclusively found in the crypt epithelial cells (Figure 3).\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine The left panel illustrates dark field microscopy of an autoradiogram showing radiolabeled cells (white spots) almost exclusively located in the epithelial layer of the crypts of Lieberkühn. To the right the same section is seen in light microscopy. Van Gieson.\nThe labeled compound (50 µCi 3H-methyl-thymidine) was administred 6 h prior to removing the intestine. Only cell nuclei were found to be labeled (arrows).\nBrdU (50 mg/kg b.wt.), a thymidine analogue, was given 2 h prior to removing the intestine. The dark labeling by BrdU is exclusively located to the nuclei of intestinal crypt cells. Mayer's haematoxylin.", "Cholera toxin evokes an intestinal fluid secretion mainly by stimulating nervous reflex(es) contained within the ENS [11], [13]. Exposing the intestinal mucosa to cholera toxin (20 µg per mL) for three hours did not influence TK activity (91.2±11.2% (mean ± s.e.m) of control; n = 5; p = 0.5), but decreased the incorporation of labeled thymidine into DNA (84.4±5.5% of control; n = 5; p<0.05).We conclude that in the present short term experiments nervous reflexes activated by cholera toxin do not increase rate of epithelial renewal.", "Two experimental protocols were used in the capsaicin experiments, “non-perfusion experiments” and “perfusion experiments” (see Methods). Since both techniques indicated that a 1.6 mM capsaicin solution evoked a consistent increase of TK activity (152.7±8.6% of TK activity in control segments (non-perfusion experiments; n = 7; p-value <0.05) and 149.8±14.4% of control segments (perfusion experiments; n = 6; p-value <0.05; Figure 4), this concentration was used in subsequent experiments, most of which were performed with the perfusion technique.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). The experiments were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”; n = 6). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”; n = 7). Asterisks indicate statistical significance. Mean ± s.e.m.\nThe incorporation of labeled thymidine into DNA was significantly increased after 2 h capsaicin perfusion (178.0±30.1% of incorporation in control segments; n = 6; p<0.05). Furthermore, the number of BrdU labeled crypt cells increased concomitantly in the capsaicin group as shown on left part of Figure 5, based on counting labeled cells in 255 crypts in control and 249 crypts in capsaicin segments (control segments: 15.0±0.3 labeled cells/crypt; capsaicin segments: 17.9±0.3 labeled cells/crypt; n = 5; p<0.01). Hence, exposing the intestinal mucosa to 1.6 mM capsaicin increased TK activity and 3H-thymidine incorporation into DNA reflecting events in crypt epithelial cells as confirmed by the BrdU experiments.\nBrdU (50 mg/kg bwt) was given i.v. 2 h prior to removal of the intestinal segments. The number of labeled cells in 255 crypts in control (n = 5) and in 249 crypts in capsaicin segments (n = 5) were counted. Asterisk indicates statistical significance. Mean ± s.e.m. Mouse: BrdU labeled cells per crypt in wild type or capsaicin receptor knock-out mice. Five capsaicin receptor knock-out mice (−/−; 225 crypts) and in five wild type mice (WT; 215 crypts) were investigated. BrdU (100 mg/kg b.wt.) was administered i.p. 2 h before removal of the small intestine. Asterisk indicates statistical significance. Mean ± s.e.m.", "Capsaicin was used to activate selectively thin afferent nerve fibers presumably via the vallinoid receptor 1 (TRPV1) [14]. We explored if capsaicin indeed activated afferent nerves by monitoring changes of heart rate on line after luminal administration of a 1.6 mM capsaicin. The periarterial nerves were not severed in these experiments. Within ten minutes 1.6 mM capsaicin heart rate increased by 46.2±8.2 min−1, whereas heart rate was not affected (−4.7±5.1 min−1) in control experiments (n = 6; p<0.05), suggesting that luminal capsaicin activates nervous afferents.\nIn one series of experiments the extrinsic intestinal nerves were allowed to degenerate for 2–3 weeks after cutting the nerves around the superior mesenteric artery. The success of the denervation procedure was tested by immunohistochemical determinations of tyrosine hydroxylase in the intestine, located to extrinsic adrenergic nerve fibers [19]. Absence of labeled nerve fibers was taken to indicate a successful denervation. Exposing such “chronically” denervated intestinal segments to capsaicin did not significantly increase TK activity (102.7±3.1% of control; n = 5). This is significantly lower than the capsaicin effect observed in “acutely” denervated segments (144.1±8.3%; n = 18; p<0.01).\nTo further investigate the possible involvement of nerves in the capsaicin response, the intestinal mucosa was anesthetized with lidocaine (see Methods). In these experiments (Figure 6) no significant increase of TK activity was recorded (perfusion experiments: TK activity in capsaicin segments 97.5±14.4% of control; n = 8; p = 0.58; non-perfusion experiments: TK activity in capsaicin segments 103.1±13.7% of control; n = 7; p = 0.74), indicating a nervous involvement in the capsaicin response.\nThymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). Asterisks indicate statistical significance between the two groups of experiments (perfusion experiments, n = 8; non-perfusion experiments, n = 7). Mean ± s.e.m.\nIndirect evidence for an involvement of nerves was also obtained in experiments with the TPRV1 receptor blocker capsazepine (3 mg/kg i.v.) given prior to the intraluminal administration of capsaicin. In these experiments the TK activity of the capsaicin segment was 97.2±7.3% (n = 7) of that measured in corresponding control segments (p<0.01 compared to a value of 144.1±8.3 per cent in experiments without capsazepine; n = 18). Hence, the capsaicin effect on TK activity was mediated by a TPRV1 receptor. Since the TPRV1 receptor is known to be located on many extrinsic afferent axons[14], the results support the proposal that the capsaicin effect involves nerves.\nFinally, hexamethonium (a cholinergic, nicotinic receptor blocker; 10 mg/kg b.wt.; n = 8) significantly attenuated the effect of capsaicin on TK activity measuring 96.5±10.7% of the activity in capsaicin free segments (p<0.01 compared to 144.1±8.3 per cent in experiments without hexamethonium; n = 18). These results suggest that the capsaicin effect was mediated by ENS reflex(es) containing at least one cholinergic synapse.", "The experiments with chronically denervated intestines described above suggested that capsaicin stimulates an axon reflex. In many organs calcitonin gene related polypeptide (CGRP) or substance P (SP) have been shown to be involved in axon reflexes [see e.g. references 20]–[22]. To elucidate if this also was the case in the present experiments two types of studies were performed. First, blocking the CGRP receptors (human 8–37 α-CGRP; 1 mg kg−1 b.wt. i.v.) or the SP receptors (more specifically the NK1 receptors: Sendide; 1 mg kg−1 b.wt. i.v.) abolished the increase of TK activity evoked by capsaicin (Figure 7; CGRP receptors: TK activity in capsaicin segments 78.5±9.7% of control segments; n = 6; p<0.001 compared to control experiments; NK1 receptors: TK activity in capsaicin segments 82.6±6.8% of TK activity in control intestines; n = 6; p<0.001 compared to control experiments) and the augmented incorporation of 3H-thymidine into DNA (Figure 8; CGRP receptors: incorporation in capsaicin segments 68.9±14% of control intestines; n = 6; p<0.01; NK1 receptors: incorporation in capsaicin segments 90.6±9.8% of control; n = 6; p<0.001) induced by luminal capsaicin. The doses used have been previously shown to block effects of CGRP and SP [22]–[24]. Second, infusing the neurotransmitters retrogradely in the carotid artery at rates of 10−9 mol per min for 1 h increased significantly intestinal TK activity compared to activity at the start of infusion (Figure 9; CGRP: TK activity 141.0±7.6% of activity at time 0; n = 5; p<0.05; SP: TK activity 143.9±20.7% of activity at time 0; n = 5; p<0.05).\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). Thymidine kinase activity is expressed in per cent of activity measured in control segments (no capsaicin). N = 6 for each antagonist experiments. Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe receptor antagonists tested were: Sendide (NK1 receptor antagonist); 8–37 α-GGRP (CGRP receptor antagonist); atropine (muscarinic receptor antagonist). The receptor blockers were given as described in Methods. 3H-thymidine incorporation is expressed in per cent of incorporation measured in control segments (no capsaicin). Asterisks indicate statistical significance compared to control. Mean ± s.e.m.\nThe peptides were infused for one hour at a rate of 10−9 mol per minute with or without prior administration of atropine (0.5 mg per kg b.wt. i.v.). Note that in these experiments thymidine kinase activity is expressed in per cent of activity measured in segments removed at time 0. Asterisks indicate statistical significance. Mean ± s.e.m.\nThere are two major types of cholinergic receptors, nicotinic and muscarinic. Above we described results from experiments with hexamethonium, a nicotinic receptor blocker. To block the muscarinic receptors, atropine (0.5 mg kg−1 b.wt.) was given i.v. Atropine significantly diminished the intestinal TK response to capsaicin (Figure 7; TK activity in capsaicin segments 91.8±5.6% of control; n = 6; p<0.001 compared to experiments without atropine) and the capsaicin evoked increase of 3H-thymidine incorporation into DNA (Figure 8; incorporation in capsaicin intestines 70.9±16.5% of control; n = 6; p = 0.002 compared to experiments without atropine).\nIn the atropine experiments it was possible to investigate the effect of atropine on rate of 3H-thymidine incorporation into DNA in control segments. After 2 h labeled nucleotide incorporation into DNA was 1017±81 DPM per mg DNA in animals not receiving atropine and 380±65 DPM per mg DNA in experiments with atropine (n = 6 in both groups; p<0.01 comparing the two groups). Hence, atropine lowered incorporation to about 40% of control suggesting an ongoing cholinergic, nervous influence in the absence of luminal capsaicin.\nA possible interaction between muscarinic and peptidergic transmission was investigated by intravascular infusion of CGRP or SP after atropine administration. No CGRP or SP effect on TK activity could be demonstrated after giving atropine. (Figure 9; CGRP: 90.8±7.1% of TK activity at start of peptide infusion; n = 7; p<0.05 compared to CGRP experiments without atropine; SP: 73.8±5.7% of TK activity at time 0; n = 7; p<0.05). The simplest explanation for these observations is that CGRP and SP influence cholinergic neuron(s), which, in turn, control the intestinal stem cells.\nTo elucidate if muscarinic receptors were present on the intestinal stem/progenitor cells, we performed immunofluorescence studies using antibodies against the five different muscarinic receptors (M1–M5) described in the literature [25]. As on many other effector cells innervated by cholinergic neurons, immunoreactivity for more than one type of muscarinic receptors (M3 and M5) was identified in the intestinal crypts [25]. The left panel of figure 10 illustrates the distribution of the immunoreactivity for the M5 receptor in the crypts as revealed by confocal microscopy in a 4 µm thin section of a crypt wall. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The disclosed fluorescence pattern is consistent with the immunoreactivy being located to cell membranes. The round structure in the middle of the figure is in all probability a cross-sectioned goblet cell surrounded by immunoreactivity. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (Figure 10, right panel).\nThe picture is photographed by confocal microscopy corresponding to a 4 µm thick section of the wall of the crypts. Hence, no crypt lumen is seen and the crypt cells are cross-sectioned. The left panel of the figure suggests that the M5 receptor is located to the plasma membrane of the crypt cells. Tissue sections exposed only to the secondary antibody did not exhibit any fluorescence (right panel).\nCheng and Leblond [26] proposed that the so called base columnar cells were the intestinal stem cells generating all types of epithelial cells, a hypothesis recently supported by Barker et al. [17], [18]. Thus, the intestinal stem cells seem to be the slender cells situated between the Paneth cells. We studied the possible presence of muscarinic receptors on these cells. Two types of muscarinic receptors M3 and M5, were investigated. Immunoreactivity to M3 and M5 receptors was identified on the presumed stem cell membrane, as illustrated for the M3 receptor on the left panel of Figure 11. The right panel shows the appearance of the tissue when exposed only to the secondary antibody.\nThe stem cells, indicated by arrows, are slender cells located between the Paneth cells (PC) at the very base of the crypts [17], [26]. The left part of the figure indicates that the muscarinic receptor M3 is located to the plasma membrane of the presumed intestinal stem cells. The tissue section in the right panel was only exposed to the secondary antibody. Mayer's haematoxylin. CL  =  crypt lumen.", "To investigate if the capsaicin receptor was of importance for cell renewal during “normal” conditions, experiments were performed on wild type mice and on mice lacking the TRPV1 gene [27]. 3H-thymidine (2 µCi) or BrdU (100 mg/kg b.wt.) were injected i.p. 2 h prior to extirpating the intestinal segments. 3H-thymidine incorporation into DNA was 529±46 DPM per mg DNA (n = 5) in control mice and 340±72 DPM per mg DNA (n = 5) in TRPV1 knock out mice (p<0.05).\nThe BrdU labeled cells were exclusively found in the intestinal crypts (cf.\nFigure 3). BrdU labeled cells in 215 crypts in wild type mice and 225 crypts in TRPV1−/− mice were counted. The number of labeled cells per crypt was significantly lower in TRPV1−/− than in wild type mice (right part of Figure 5; wild type: 11.4±0.2 BrdU labeled cells/crypt; knock-out: 9.2±0.2 BrdU labeled cells/crypt; n = 5 in both groups; p<0.01). Thus, in mice devoid of capsaicin receptors rate of 3H-thymidine incorporation into DNA and number of BrdU labeled crypt cells were significantly lower than in wild type mice.", "The intestinal epithelium is the most rapidly self-renewing epithelium in adult mammals, being replaced every 3–5 day. Each crypt contains 4–6 long-lived stem cells that are constantly dividing. The daughter cells of the stem cells (here called progenitor cells) divide every 12–16 h as they move “upwards” along the crypts generating some 300 cells per crypt every day [28]. The present study provides evidence for a nervous control of intestinal stem/progenitor cells in vivo. We have demonstrated that luminal capsaicin (1.6 mM), stimulating thin nervous afferents, increases intestinal TK activity, 3H-thymidine incorporation into DNA and number of BrdU labeled crypt cells. No such effects were evoked by cholera toxin. Several observations strongly indicate that the observed capsaicin effects were nervously mediated. Hence, lidocaine, a local anesthetic, and antagonists to four different neurotransmitter receptors (muscarinic, nicotinic, NK1 and CGRP receptors) significantly attenuated the capsaicin response. A nervous involvement was also indicated by the failure of capsaicin to influence TK activity after allowing a degeneration of the extrinsic nerves (“chronic denervation”). Finally, the capsaicin effect was mediated by TRPV1 receptors, known to be located on extrinsic afferent nerves [20], the capsaicin effect being abolished by the TRPV1 receptor blocker capsazepine.\nIn the present study the measured TK activity has been taken as indicator of rate of cell renewal of the intestinal epithelium. The reasons for this are the following: TK is an important enzyme in DNA replication, phosphorylating the nucleoside before being incorporated into DNA. The enzyme exists in a cytosolic and mitochondrial form [15]. Cytosolic TK is present only in proliferating cells. Its importance for cell renewal is suggested by the observations that TK activity in the crypts is about 20 times higher than in the villi when expressed per unit tissue weight or DNA weight [29], [30]. This conclusion is also supported by our observations that 3H-thymidine or BrdU injected intravenously was localized almost exclusively to the nuclei of crypt cells (Figures 1, 2 and 3), as also shown by Lipkin [31] for labeled thymidine. Furthermore, the increased incorporation of thymidine into DNA and the increased number of BrdU labeled crypt cells in response to luminal capsaicin strongly suggest that the augmented TK activity reflected a change of cytosolic TK activity. Finally, nerve transmitter receptor antagonists concomitantly abolished the increase of TK activity and thymidine incorporation into DNA (compare Figures 7 and 8).\nIt might be argued that the increase of total intestinal TK activity reflects a proliferation of other cells than the stem/progenitor cells of intestinal crypts. Observations reported in the literature indicate that this is unlikely. Smooth muscle cells show almost no renewal in the mouse stomach [32]. Fibroblasts in the intestinal submucosa are renewed with turn over times varying between 100 and 130 days [33]. Hence, the two most dominating tissues of the intestinal wall, apart from the epithelium, show no proliferation within the short time of the present experiments.\nThe “chronic denervation” experiments are key experiments in the analysis of the anatomical arrangement of the nerves influencing intestinal stem/progenitor cells, since such experiments make it possible to elucidate if a studied reflex is confined to the intestinal wall (“intramural reflex”) or mediated via an axon reflex. An axon reflex may be influenced by severing an afferent neuron, whereas an intramural reflex will not. Thus, the periarterial nerves were cut and acute experiments were performed 2–3 weeks later, when the extrinsic innervation (afferent and efferent) had degenerated. The absence of extrinsic nerves was confirmed by immunohistochemical investigations of tyrosine hydroxylase, a rate limiting enzyme in catecholamine synthesis [19]. In chronically denervated rats luminal capsaicin failed to cause an increase of TK activity, indicating that capsaicin exerted its effect via an axon reflex. These observations are consistent with the TRPV1 receptors being present only on extrinsic afferent nerves of the gut [20].\nThere exist a number of criteria for neurotransmitter candidates to be accepted as true transmitters. Observations reported in this and other studies indicate that most of these criteria have been fulfilled with regard to SP and CGRP and the afferent nervous control of intestinal cell renewal. First, immunohistochemical investigations have revealed intestinal, extrinsic, afferent, nerves containing SP and CGRP [14], [20]. Second, subgroups of the capsaicin sensitive neurons show SP and/or CGRP immunoreactivity [20]. Third, blocking the receptors for NK1 and CGRP significantly attenuated the effect of luminal capsaicin on TK activity and 3H-thymidine incorporation (Figures 7 and 8). Fourth, intra-arterial administration of the two peptides significantly increased TK activity (Figure 9). Fifth, a release of SP or CGRP from stimulated afferent neurons has been reported in the literature [22], [34]–[36]. Thus, taken together the observations clearly suggest that SP and CGRP are involved in the capsaicin effect on epithelial cell renewal. The current results indicate, however, that the NK1 or CGRP receptors are not localized to the stem/progenitor cells, since atropine abolished the TK increase induced by intra-arterially infused SP or CGRP (Figure 9).\nCapsaicin sensitivity is a trait of a subgroup of the unmyelinated C fibres but includes also some of the thinly myelinated Aδ fibres. Capsaicin stimulated receptors (TRPV1) are localized to extrinsic, afferent neurons usually considered to belong to the nociceptor system, a set of nerves that can be activated by stimuli (mechanical, thermal, chemical) capable of causing tissue damage. The TRPV1 receptor is located on free nerve endings of at least on two different types of neurons, one being a heat sensitive neuron and the other being a peptidergic, polymodal nociceptor neuron [37].\nCombining observations reported in the literature with the findings of the current investigation it is possible to propose a model to explain the findings of the present study. The model is shown on Figure 12 and the observations that constitute the basis for the model are briefly summarized in Table 1. It should be underlined that Figure 12 represents the simplest model that can be constructed from our findings and the observations reported in the literature. It is a working hypothesis, which can be used when designing further experiments to prove or disprove the proposed model. The synapse between the afferent neuron and the cholinergic neuron has been located to the submucosal plexus in the figure for two reasons: first, most neurons controlling mucosal functions have their cell somas in the submucosal plexus. Second, luminal capsaicin transiently reduces CGRP-like immunoreactivity in the submucosal but not in the myenteric plexus [39].\nThe model is in part based on the findings of the present study (cf.\nTable 1) and is consistent with known types of neurons in the intestinal wall. It implies that the direct control of the stem/progenitor cells is exerted by a cholinergic neuron, which, in turn, is under the influence of an axon reflex releasing CGRP/SP/acetylcholine. Ach: acetylcholine; CGRP: calcitonin gene related peptide; SP: substance P; CNS: central nervous system.\nTS: this study.\nThe proposed axon reflex control of intestinal cell renewal, may, however, not be the only nervous control of intestinal stem cells. We observed that atropine administration markedly attenuated nucleotide incorporation into DNA also in control segments. A rough calculation based on the averages of nucleotide incorporation (DPM per mg DNA) indicates that atropine decreased incorporation to about 40% of control. Hence, other enteric nerves than the thin extrinsic mucosal afferents may influence stem cell/progenitor cell proliferation, the cholinergic neuron being the final common pathway for more than one reflex. In fact, the results indicate that nervously released acetylcholine may play a very important role in the stem cell niche of the small intestine. With regard to the type of epithelial cells controlled by enteric nerves, observations reported by Bjerknes and Cheng [40], [41] indicate that the enteric nerves control the proliferation of the columnar daughter cells.\nThere is a fairly extensive literature dealing with muscarinic receptors and cell proliferation, in normal or cancerous cells (for reviews see [42], [43]). The muscarinic influence on proliferation is, at least in part, mediated by the “classical” signalling pathways for muscarinic receptors, the inositol-phospholipid signalling cascade [42] involving two second messengers, diacylglycerol and inositol 1,4,5-trisphosphate (IP3). This pathway is, however, not the only one mediating the muscarinic effect. Several studies report that the muscarinic receptor also is coupled to the mitogen-activated protein kinase (MAPK) pathway probably via a transactivation of the epidermal growth factor (EGF) [43]–[47] receptor located on the basolateral membrane of the crypt cell epithelium [48].\nBesides luminal capsaicin we investigated the effect of cholera toxin on TK activity and thymidine incorporation into DNA. It was observed that the toxin, if anything, decreased cell renewal to judge by attenuated nucleotide incorporation into DNA. This is consistent with earlier observations that the final neurotransmitter in the secretory reflex activated by cholera toxin is not acetylcholine but vasoactive intestinal polypeptide (VIP) [49]–[52]. Atropine does not attenuate cholera secretion [13] excluding a muscarinic mechanism. Furthermore, in contrast to the present findings chronic denervation does not influence the secretory response to cholera toxin [53]. We conclude that cholera toxin and capsaicin stimulate two different reflexes with different final neurotransmitters, one being confined to the intestinal wall (cholera toxin reflex), releasing VIP, and the other being an axon reflex (capsaicin reflex) releasing acetylcholine at the effector cell.\nWe have earlier reported that ENS is of great importance in explaining fluid secretion in acute diarrhoea [11]–[13]. As pointed out in the introduction diarrhoea can be looked upon as being an innate immunity response, since the secreted fluid dilutes the noxious agent(s). Furthermore, fluid secretion is often accompanied by a nervously induced motility response propelling the intestinal contents aborally [54]. Capsaicin activated intestinal nerves evoke an intestinal vasodilatation [22] and may be involved in the control of mucus secretion [55] and intestinal homing of leukocytes [56]. Taken together with the present observations it suggests that ENS coordinates several innate immunity mechanisms in the gut.\nAs discussed above the capsaicin receptor TPRV1 is usually considered to be a nociceptor sensing potentially noxious stimuli. It may, however, also participate in physiological mechanisms. The contents of the intestinal lumen influences mucosal morphology, parenteral feeding leading to a decreased mucosal thickness [57]. This is usually explained in terms of trophic factors, such as hormones. However, the studied reflex may be involved. There are at least three physiological ways by which the intestinal contents may influence mucosal afferent nerves. First, the TRPV1 receptor seems to be an osmoreceptor activated by hyperosmolality [58]. A tissue hyperosmolality has been demonstrated in the upper parts of intestinal villi in several species with different techniques [59], measuring about 800 mOsm in humans and cats provided the intestinal contents contain sodium and glucose. We propose that TRPV1 receptors are activated by a villous tissue hyperosmolality. Such a mechanism may explain the present findings in the knock-out mice implying that the capsaicin receptor is important for cell renewal even in the absence of luminal capsaicin.\nSecond, intake of food leads to the presence of pancreatic proteases in the intestinal lumen. It has been shown that proteinase-activated receptors 1 and 2 (PAR-1, PAR-2) are present on the apical border of enterocytes and activated by trypsin [60], [61]. In the case of PAR-2 receptors indirect evidence supports the proposal that a stimulation of this receptor may influence firing in afferent nerves from the gut. Thus, Cenac et al. [62], [63] and Nguyen et al. [64] reported that exposing the colonic mucosa to a PAR-2 receptor agonist causes a neurogenic inflammation via capsaicin sensitive nerves. Furthermore, the PAR-2 induced inflammation was attenuated by NK1 and CGRP antagonists.\nThird, one may speculate that the intestinal contents via the discharge of hormones from enteroendocrine cells activate mucosal afferents to increase epithelial cell renewal. A similar chain of events has been proposed for the activation of intramural secretory nervous reflexes by e.g. cholera toxin [11], [13] and rotavirus [12]. The finding reported by Bjerknes and Cheng [65] that glucagon-like-peptide-2 exerts its stimulatory effect on epithelial proliferation via nerves may be explained by such a mechanism.\nMicroorganisms in the intestinal lumen are obvious examples of potential noxious agents. Their presence may be recorded by the intestinal epithelium, functioning as a mucosal defence early warning system. Using membrane-associated or intracellular pattern-recognition molecules, such as toll-like receptors and Nod proteins, the epithelial cells sense the microorganisms [66]. This, in turn, may evoke a release of cytokines, prostaglandins and/or nitrous oxide from the enterocytes [67], biologically potent compounds with receptors on capsaicin sensitive nervous dendrites. Invading microorganisms may evoke an inflammatory response. Many of the substances participating in an inflammation, such as cytokines, prostaglandins, adenosine, bradykinin, histamine and 5-hydroxytryptamine, may directly or indirectly via activation of immunocytes, stimulate the studied axon reflex [38], [68], [69]. A morphological characteristic of ulcerative colitis is the “reactive“ hyperplasia of the epithelium [70]. One may speculate this hyperplasia may, at least in part, reflect an inflammatory activation of the axon reflex controlling intestinal stem/progenitor cells. Similar mechanisms may exist in connection with inflammation of other mucosae, as in the air ways and the urinary tract.\nIt is well established that there is a relationship between tissue repair and cancer, implying that chronic tissue injury may lead to malignancy (see e.g. [71]). A well-known example is colon cancer developing in longstanding ulcerative colitis. It has been suggested that overall risk of cancer development depends on the number of activated stem cells. Any repeated stimulus leading to a chronic increase in stem cell numbers could result in a higher overall frequency of cancer formation [28], [72]. Furthermore, it has been pointed out that the efficiency of carcinogenesis in tumour models is increased when combining mutagenic agents with agents that induce cell proliferation [73]. In this context it is interesting to note that Frucht and coworkers [74] found that 6 out of 10 human colon cancer cell lines were provided with functional muscarinic cholinergic receptors. Therefore, in view of the findings of the present study, one may pose the question if long standing activation of mucosal afferent nerve fibres by e.g. inflammation, may, via a cholinergic stimulation of muscarinic receptors and transactivation of EGF receptors, contribute to the development of cancer by their influence on epithelial proliferation.", "[SUBTITLE] Ethical approval [SUBSECTION] The protocol used in the present experiments was approved by the Ethical committee for animal experimentation at Gothenburg University, Gothenburg, Sweden (approval 292-2009). The experiments were performed in accordance with the recommendations issued by the Swedish Department of Agriculture.\nThe protocol used in the present experiments was approved by the Ethical committee for animal experimentation at Gothenburg University, Gothenburg, Sweden (approval 292-2009). The experiments were performed in accordance with the recommendations issued by the Swedish Department of Agriculture.\n[SUBTITLE] Rat experiments [SUBSECTION] Experiments were performed on male Sprague-Dawley rats, weighing 240–450 g (Möllegaards Breeding Centre Ltd, Ejby, Denmark or B&K Universal AB, Sollentuna, Sweden). The rats were housed for at least 7 days prior to experiments in animal quarters (22°C, 60% relative humidity, artificial lighting between 06.00 and 18.00 h). Before the experiments the rats were deprived of food for at least 12 h. All experiments were carried out between 9.30 and 15.00.\n[SUBTITLE] Anesthesia and operative procedures [SUBSECTION] Anesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.\nAnesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.\n[SUBTITLE] Experimental procedures [SUBSECTION] Two to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.\nTwo to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.\n[SUBTITLE] Lidocaine experiments [SUBSECTION] In one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.\nIn one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.\n[SUBTITLE] Neurotransmitter receptor blockade [SUBSECTION] Several neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.\nSeveral neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.\n[SUBTITLE] Neurotransmitter i.a. infusions [SUBSECTION] SP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.\nSP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.\n[SUBTITLE] BrdU experiments [SUBSECTION] In 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.\nIn 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.\n[SUBTITLE] Chronic denervation experiments [SUBSECTION] In 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).\nIn 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).\n[SUBTITLE] Determination of thymidine kinase activity [SUBSECTION] Thymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\nThymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[SUBTITLE] Thymidine incorporation into DNA [SUBSECTION] [Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[SUBTITLE] Immunohistochemistry [SUBSECTION] Biopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.\nBiopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.\n[SUBTITLE] Autoradiography experiments [SUBSECTION] Methyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.\nMethyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.\nExperiments were performed on male Sprague-Dawley rats, weighing 240–450 g (Möllegaards Breeding Centre Ltd, Ejby, Denmark or B&K Universal AB, Sollentuna, Sweden). The rats were housed for at least 7 days prior to experiments in animal quarters (22°C, 60% relative humidity, artificial lighting between 06.00 and 18.00 h). Before the experiments the rats were deprived of food for at least 12 h. All experiments were carried out between 9.30 and 15.00.\n[SUBTITLE] Anesthesia and operative procedures [SUBSECTION] Anesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.\nAnesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.\n[SUBTITLE] Experimental procedures [SUBSECTION] Two to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.\nTwo to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.\n[SUBTITLE] Lidocaine experiments [SUBSECTION] In one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.\nIn one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.\n[SUBTITLE] Neurotransmitter receptor blockade [SUBSECTION] Several neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.\nSeveral neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.\n[SUBTITLE] Neurotransmitter i.a. infusions [SUBSECTION] SP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.\nSP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.\n[SUBTITLE] BrdU experiments [SUBSECTION] In 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.\nIn 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.\n[SUBTITLE] Chronic denervation experiments [SUBSECTION] In 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).\nIn 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).\n[SUBTITLE] Determination of thymidine kinase activity [SUBSECTION] Thymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\nThymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[SUBTITLE] Thymidine incorporation into DNA [SUBSECTION] [Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[SUBTITLE] Immunohistochemistry [SUBSECTION] Biopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.\nBiopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.\n[SUBTITLE] Autoradiography experiments [SUBSECTION] Methyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.\nMethyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.\n[SUBTITLE] Mouse experiments [SUBSECTION] Experiments were performed on mice devoid of the capsaicin receptor (TRPV1 [27]; strain B6.129X1-Trpv1) purchased from The Jackson Laboratory (Bar Harbor, Maine). Strain B6129SF2/J was used as controls as recommended by the distributor. The mice were kept in animal quarters under standardized environmental conditions (see above). At the time of the experiments the animals were 10 weeks old.\n\n3H-thymidin (2 µCi) or BrdU (100 mg/kg body weight) dissolved in physiological saline was administered i.p.. Two h later the mice were killed with an overdose of sodium pentobarbital i.p. 3H-thymidine incorporation into DNA was determined and BrdU labeled cells were stained and counted as described for rat experiments.\nExperiments were performed on mice devoid of the capsaicin receptor (TRPV1 [27]; strain B6.129X1-Trpv1) purchased from The Jackson Laboratory (Bar Harbor, Maine). Strain B6129SF2/J was used as controls as recommended by the distributor. The mice were kept in animal quarters under standardized environmental conditions (see above). At the time of the experiments the animals were 10 weeks old.\n\n3H-thymidin (2 µCi) or BrdU (100 mg/kg body weight) dissolved in physiological saline was administered i.p.. Two h later the mice were killed with an overdose of sodium pentobarbital i.p. 3H-thymidine incorporation into DNA was determined and BrdU labeled cells were stained and counted as described for rat experiments.\n[SUBTITLE] Chemicals [SUBSECTION] Peptides were bought from Bachem AG, (Bubendorf, Switzerland). All other chemicals, for which any supplier has not been indicated in text, were purchased from Sigma-Aldrich Sweden, Stockholm, Sweden.\nPeptides were bought from Bachem AG, (Bubendorf, Switzerland). All other chemicals, for which any supplier has not been indicated in text, were purchased from Sigma-Aldrich Sweden, Stockholm, Sweden.\n[SUBTITLE] Recordings of heart rate [SUBSECTION] Heart rate was monitored on line with a rate meter coupled to the blood pressure recorder sensing the arterial pressure oscillations and recorded on a Grass polygraph.\nHeart rate was monitored on line with a rate meter coupled to the blood pressure recorder sensing the arterial pressure oscillations and recorded on a Grass polygraph.\n[SUBTITLE] Statistics [SUBSECTION] Non-parametric statistics were used throughout. Significance between paired observations was determined using Wilcoxon matched-pairs test. The Mann-Whitney U test was utilized when comparing independent groups. All calculations were performed with SPSS for Windows. Level of statistical significance was set at p<0.05.\nNon-parametric statistics were used throughout. Significance between paired observations was determined using Wilcoxon matched-pairs test. The Mann-Whitney U test was utilized when comparing independent groups. All calculations were performed with SPSS for Windows. Level of statistical significance was set at p<0.05.", "The protocol used in the present experiments was approved by the Ethical committee for animal experimentation at Gothenburg University, Gothenburg, Sweden (approval 292-2009). The experiments were performed in accordance with the recommendations issued by the Swedish Department of Agriculture.", "Experiments were performed on male Sprague-Dawley rats, weighing 240–450 g (Möllegaards Breeding Centre Ltd, Ejby, Denmark or B&K Universal AB, Sollentuna, Sweden). The rats were housed for at least 7 days prior to experiments in animal quarters (22°C, 60% relative humidity, artificial lighting between 06.00 and 18.00 h). Before the experiments the rats were deprived of food for at least 12 h. All experiments were carried out between 9.30 and 15.00.\n[SUBTITLE] Anesthesia and operative procedures [SUBSECTION] Anesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.\nAnesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.\n[SUBTITLE] Experimental procedures [SUBSECTION] Two to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.\nTwo to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.\n[SUBTITLE] Lidocaine experiments [SUBSECTION] In one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.\nIn one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.\n[SUBTITLE] Neurotransmitter receptor blockade [SUBSECTION] Several neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.\nSeveral neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.\n[SUBTITLE] Neurotransmitter i.a. infusions [SUBSECTION] SP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.\nSP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.\n[SUBTITLE] BrdU experiments [SUBSECTION] In 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.\nIn 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.\n[SUBTITLE] Chronic denervation experiments [SUBSECTION] In 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).\nIn 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).\n[SUBTITLE] Determination of thymidine kinase activity [SUBSECTION] Thymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\nThymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[SUBTITLE] Thymidine incorporation into DNA [SUBSECTION] [Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.\n[SUBTITLE] Immunohistochemistry [SUBSECTION] Biopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.\nBiopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.\n[SUBTITLE] Autoradiography experiments [SUBSECTION] Methyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.\nMethyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.", "Anesthesia was induced with pentobarbital (60 mg/kg body wt.) i.p. in all experiments except those in which the intestinal segments were chronically denervated (see below). A tracheal cannula was inserted and a femoral vein was cannulated. Arterial pressure was continuously monitored in a femoral artery on a Grass polygraph. The arterial catheter was also used to maintain anesthesia by a continuous infusion of α-chloralose (Alfa Aesar GmbH & Co, Karlsruhe, Germany; 3 mg/ml; 20 µL/min) in a solution containing 138 mM glucose and 33 mM NaHCO3. Body temperature was kept around 38°C with a heated operating table. At the end of the experiments the animals were killed by an i.v. injection of a saturated KCl solution while still anesthetized.\nA midline abdominal incision was performed. The influence of the extrinsic autonomic nervous supply to the intestinal segments was minimized by cutting the nerves along the superior mesenteric artery.", "Two to six 2–6 cm long jejunal segments were isolated 5–10 cm below the ligament of Treitz. The experiments with capsaicin were performed using two types of preparations. In most experiments only the proximal ends of the segments were cannulated with plastic tubing and the intestinal segment was perfused by means of a pump at a rate of 20 µL per min (“perfusion experiments”). In other experiments the distal ends were also provided with plastic tubing closed by plugs (“non-perfusion experiments”). The original capsaicin solution contained 16 mM capsaicin dissolved in a solution containing Tween 80 (1 part), ethanol (1 part) and NaCl (0.9%; 8 parts). The concentration of capsaicin most often used (1.6 mM) was obtained by a 10 times dilution of the 16 mM solution with physiological saline. The control solution did not contain capsaicin. In some experiments the intestinal mucosa was exposed to cholera toxin (List Biological Laboratories, Campbell, California; 20 µg per mL physiological saline) or to saline. The TK experiments (see below) lasted for 60 min, whereas the [methyl-3H]thymidine and BrdU experiments were prolonged to 120 min.\nAt predetermined times intestinal segments were removed. Macroscopically apparent Peyer's patches were removed. Intestinal segments to be utilized for immunohistochemistry or autoradiography were placed in buffered a 4% paraformaldehyde (pH 7.0) solution. Intestinal segments to be used for estimation of thymidine kinase or DNA were immediately frozen in liquid nitrogen and kept at −20°C or −85°C until determination. In all types of experiments the choice of segments was randomized.", "In one type of experiments the intestinal mucosa was exposed to lidocaine, a local anaesthetic, together with capsaicin. About 10 min before exposing the intestinal segment to capsaicin the segment was flushed with a 1% lidocaine (w/v) in physiological saline. Furthermore, the capsaicin solution contained 0.2% lidocaine (perfusion experiments) or 1% lidocaine (non-perfusion experiments). Control solutions lacked capsaicin. One may argue that tetrodotoxin, a more specific nerve blocking agent than lidocaine, would have been more appropriate. However, the use of tetrodotoxin in in vivo experiments is greatly complicated by its high general toxicity for the experimental animal.", "Several neurotransmitter antagonists were tested. Atropine (a muscarinic receptor blocker; 0.5 mg kg−1 b.wt.) or 8-37 α-GGRP (a CGRP receptor blocker; 1 mg kg−1 b.wt;) was administered i.v. at times 0 and 60 min. Hexamethonium (a nicotinic receptor blocker; 10 mg kg−1 b.wt) or Sendide (1 mg kg−1 b.wt; a NK1 receptor blocker) was administered i.v. every 30. min.", "SP or CGRP were infused i.a. retrogradely via a catheter in the left carotid artery at a rate of 10−9 mol per min. The peptides were dissolved in a 1% albumin-saline solution. In the cholera toxin and neurotransmitter experiments the intestines were not perfused.", "In 5 experiments bromodeoxyuridine (BrdU; a thymidine analogue; Roche Diagnostics Scandinavia, Bromma, Sweden; dissolved in physiological saline; 50 mg/kg b.wt.) was administered i.v.", "In 5 experiments the intestinal segments were denervated periarterially 2–3 weeks prior to the acute experiments. The rats were anesthetized with ketamin (Ketalar, Pfizer Animal Health; 75 mg/kg b.wt. i.p.) and medetomidine (Domitor vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor agonist; 0.5 mg/kg b.wt. i.p.). The superior mesenteric artery was dissected free about 10 mm from its origin from the aorta. A wide portion of the tissue surrounding the artery and containing the periarterial nerves was divided between ligatures. To ablate any nerve fiber bundles that were not severed by this procedure the arterial vessel was painted with a 5% phenol solution. The abdomen was closed by ligatures and the animals were placed in a heated cage. The animals were aroused by giving atipamezole (Antisedan vet., Orion Oyj, Espoo, Finland; α2-adrenoreceptor antagonist; 0.5 mg/kg b.wt. i.p.). The acute experiments carried out later were performed as described above (“perfusion experiments”). The success of denervation was tested by an immunohistochemical investigation of the presence of thyrosine hydroxylase (marker of adrenergic nerve fibers; see below).", "Thymidine kinase (TK) catalyses the phosphorylation of deoxythymidine to deoxythymidine-5-phosphate. TK activity in the intestine was measured in duplicate samples as described [30], [75], [76]. Intestinal biopsies were homogenized (Polytron PT 1200 homogeniser) in a 2 ml Eppendorf tube with a Tris-phosphate buffer (pH 7.3) at 0–4°C and centrifuged at 16 000× g for 20 min, a g-force above that which enriches the pellet with mitochondria [30]. Supernatant (100 µl) was added to a reaction mixture (volume 1.275 mL) containing ATP (12 mM), phosphoglycerate (10 mM), MgCl2 (4 mM), NaF (3 mM), thymidine (0.4 mM) and [methyl-3H]-thymidine (12.5 or 25 µCi corresponding to 0.2 or 0.4 µM; Amersham Biosciences, Buckinghamshire, U.K.) in a 0.05 M TrisHCl buffer (pH 8). After a 20 min incubation period at 37°C in a metabolic shaker, the reaction was stopped by filtering the solution with a small volume of NaCOOH through an ion exchange filter (Whatman DE 81) This filter retained labelled dTMP, whereas free thymidine was washed away. After repeated washings with NaCOOH and water, the filter was placed in a scintillation counting vial together with 1 ml of a solution containing HCl (2 M) and NaCl (1 M) in a 1∶1 relationship. After 10 min scintillation fluid (Quickszint 361, Zinsser Analytical, Frankfurt, Germany) was added. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.", "[Methyl-3H]thymidine (25 or 50 µCi; Amersham Biosciences) was given i.v. half an hour after exposing an intestinal segments to cholera toxin or just prior to perfusing intestinal segments with 1.6 mM capsaicin or a control solution. DNA was isolated as described [77]. The amount of DNA was estimated from measurements of absorbance at 260 nm [78]. The radioactivity was measured in a Packard 1900 TR Liquid scintillation analyser.", "Biopsies from the rat small intestine immersed in buffered 4% paraformaldehyde (pH 7.0) were subsequently embedded in paraffin.\nBromodeoxyuridine (BrdU) is a thymidine analogue, which can be localized in tissue with histochemical techniques. It is incorporated into the DNA during the S phase making it possible to quantify the number of newly formed cells. In the present experiments intestinal segments were extirpated 2 h after giving BrdU (50 mg/kg b.wt.) i.v. The tissue was immersed in buffered 4% paraformaldehyde (pH 7.0), embedded in paraffin and cut at 4 µm. Endogenous peroxidase was inactivated with 3% H2O2. Non-specific protein binding was blocked with 4% normal donkey or 1.5% normal rabbit serum in Tris-buffered saline (TBS) for 30 min. The sections were exposed to monoclonal rat BrdU antibodies (0.5 µg/ml; Adb Serotec, Oxford, UK) over night at 4°C followed by incubation with avidin-biotin-peroxidase complex (ABC Elite, Vector laboratories, Inc., Burlingame, California) for 60 min at room temperature. The immunoreactivity was visualized with 3.3′-diaminobenzidine tetrahydrochloride (DAB; 0.50 mg/ml). Most biopsies were counterstained with Mayer's hematoxyline. As a negative tissue sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue. The number of BrdU labelled crypt cells was counted in coded tissue sections at 200 times magnification using an Olympus BX60 microscope. The crypts had to fulfil two criteria to be included when estimating number of BrdU labelled cells. First, a crypt lumen was seen along the length of the crypt. Second, the bottom of the crypt was located close to the submucous layer.\nImmunohistochemical investigation of muscarinic receptor (mAChR) expression was in some experiments undertaken on 20 µm thick sections, which were deparaffined and re-hydrated by heating the slides to 60°C for one hour followed by subjection to xylene, decreasing concentrations of ethanol and TBS. The sections were then immersed in a 10 mM citrate buffer (pH 6.0) and microwaved for four cycles of 6 min. Endogenous peroxidase was inactivated with 0.03% H2O2. Non-specific protein binding was minimized with 5% bovine serum albumin (BSA) in TBS for 30 min. The sections were then incubated overnight at room temperature in a humidified chamber with polyclonal rabbit anti-mAChR subtype specific antibodies (Research and Diagnostic Antibodies, Berkley, California) diluted 100 times in TBS containing 1% BSA. The presence of the muscarinic receptor immunoreactivity was revealed using Alexa Fluor 488 goat anti-rabbit Ig G (Molecular Probes, Eugene, Oregon) and analyzed using a Radiance 2000 Confocal Imaging System (Bio-Rad, Hercules, California) and the LaserSharp 2000 software (Bio-Rad).\nCertain other sections (4 µm thick) were treated with an immunoenzymatic method. Generally speaking, the method was similar to that described above with two modifications: endogenous peroxidase was inactivated with 0.6% H2O2 and the sections were stained with a secondary biotinylated goat antirabbit antibody (Dako, Sweden AB, Solna, Sweden) followed by a horseradish peroxidase-labeled avidin-biotin complex (Dako). The colour was then developed with AEC substrate chromogen (Dako). The sections were analyzed with an Olympus BX60 microscope. Control sections were immunostained without exposure to the primary antibody. It has previously been shown that the binding of the muscarinic receptor antibodies used in the present study are blocked by preincubation with its specific antigen [79].\nWe have used the adrenergic innervation to test presence or absence of external innervation after severing the intestinal periarterial nerves. This method is based on the absence of adrenergic nerves in the ENS that are confined to the intestinal wall. Therefore, adrenergic nerves are extrinsic and their absence in intestinal segments indicates a successful surgical denervation. This method has been used extensively [for experiments on rats], [ see references 80]–[82]. Tissue from every experimental segment was immersed in buffered 4% paraformaldehyde (pH 7.0) and embedded in paraffin. Four µm thick tissue sections were collected at a regular distance of 40 µm between biopsies. The sections were deparaffined in xylene and decreasing concentrations of ethanol and then immersed in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidise activity was blocked by 0.3% H2O2 The sections were exposed to a mouse, monoclonal, tyrosine hydroxylase antibody (Affinity BioReagents, AH Diagnostics, Skärholmen, Sweden) for 45 min at room temperature. A biotin rat anti-mouse IgG1 (BD Biosciences Sweden, Stockholm, Sweden) was used as a secondary antibody. After exposure to a peroxidase-conjugated avidine-biotin complex (ABC-complex, Dako) the sections were stained with an AEC-substrate (Dako) and counterstained with Mayer's hematoxyline. Control sections were immunostained without exposure to the primary antibody, which did not result in any staining of the tissue.", "Methyl-3H-thymidine (50 or 100 µCi; specific activity 70–75 Ci/mmol; Amersham Biosciences, UK) in physiological saline, was given i.v. four or six hours prior to extirpation of the intestinal segment. After embedding in paraffin, the biopsy was cut at 4 µm. For autoradiography the histological sections were deparaffinated with xylene and dehydrated with ascending concentrations of ethanol. The sections were dipped in a photographic solution (Ilford K2) for 10 s, exposed usually for about 4 weeks, developed and fixed. The sections were then stained (Van Gieson). The autoradiographs were analyzed by conventional bright field microscopy as well as by dark field microscopy.", "Experiments were performed on mice devoid of the capsaicin receptor (TRPV1 [27]; strain B6.129X1-Trpv1) purchased from The Jackson Laboratory (Bar Harbor, Maine). Strain B6129SF2/J was used as controls as recommended by the distributor. The mice were kept in animal quarters under standardized environmental conditions (see above). At the time of the experiments the animals were 10 weeks old.\n\n3H-thymidin (2 µCi) or BrdU (100 mg/kg body weight) dissolved in physiological saline was administered i.p.. Two h later the mice were killed with an overdose of sodium pentobarbital i.p. 3H-thymidine incorporation into DNA was determined and BrdU labeled cells were stained and counted as described for rat experiments.", "Peptides were bought from Bachem AG, (Bubendorf, Switzerland). All other chemicals, for which any supplier has not been indicated in text, were purchased from Sigma-Aldrich Sweden, Stockholm, Sweden.", "Heart rate was monitored on line with a rate meter coupled to the blood pressure recorder sensing the arterial pressure oscillations and recorded on a Grass polygraph.", "Non-parametric statistics were used throughout. Significance between paired observations was determined using Wilcoxon matched-pairs test. The Mann-Whitney U test was utilized when comparing independent groups. All calculations were performed with SPSS for Windows. Level of statistical significance was set at p<0.05." ]

[ null, null, null, null, null, null, null, null, null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Biomedical Research", "Europe", "Journal Impact Factor", "Societies, Scientific", "Time Factors", "United States" ]

[ "Introduction", "Results", "Discussion" ]

[ "The most commonly used indicator of the quality or at least the popularity of scientific journals is the journal impact factor, which is not however exempt of some limitations [1], [2]. Over the past years, the impact factor of biomedical journals has generally shown a tendency to rise, which can in part be attributed to the expansion in the size of the relevant literature [3], [4]. It is well recognized also that the leading positions in productivity in biomedical research are held by developed countries, including those from North America and Western Europe [5], [6].\nIn this context, we sought to evaluate the trends in the impact factor of journals published on behalf of United States (US) and European biomedical scientific societies during the past 10 years.", "The 4 subject categories of our interest (Biology, Cell Biology, Critical Care Medicine, and Infectious Diseases) in the Journal Citation Reports® database, included in total 252, 280, 288, and 303 journals for each of the years 1999, 2002, 2005, and 2008, respectively. Among these, we identified 167 different journals that were published on behalf of US or EU biomedical scientific societies and had an impact factor for at least one of the above years. In Table 1, we present the distribution of the journals included in our analysis with regard to region and subject category.\nAbbreviations: EU: European Union; US: United States.\nIn Figure 1, we present graphically the temporal trends in the median impact factors of the journals included in each of the above 4 subject categories and published on behalf of US compared with EU scientific societies, in the 4 selected years between 1999 and 2008. As can be inferred from the graphs included in Figure 1, in 1998 US journals had clearly higher median impact factor compared with the European ones in the all examined subject categories, except for Biology. However, there appears to be a trend towards a greater increase in the median impact factor of journals published on behalf of European scientific societies compared with the US ones, in the subject categories of Critical Care Medicine and Infectious Diseases, over the past decade. The increase in the median impact factor of journals published on behalf of European scientific societies parallels that of the US ones, in the subject category of Cell Biology, whereas there is a trend for a greater increase in the median impact factor of journals published on behalf of the US societies, compared with the European ones, in the subject category of Biology.\nThe fitted line represents the temporal trend in the median impact factors.\nIn Table 2, we present the findings of our analysis regarding the percentage change in the impact factor of journals published on behalf of US and European scientific societies, cumulatively for all the 4 selected subject categories, between 1999 and 2008, as well as for the interval periods. Specifically, between 1999 and 2008, the impact factor of journals published on behalf of European scientific societies rose more than those published on behalf of the US ones (by 73.7±110.0%, as compared with 39.7±70.0%, p = 0.049). No relevant statistically significant difference was observed in any of the interval periods. With regard to each selected subject category, the impact factor of journals published on behalf of European scientific societies compared with US ones, rose, between 1999 and 2008, by 212.4±291.5% compared with 65.4±45.5% (p = 0.107) for Critical Care Medicine, 88.3±99.5% compared with 48.7±61.3% (p = 0.39) for Infectious Diseases, 61.7±82.0% compared with 16.3±68.7% (p = 0.029) for Cell Biology, and by 41.0±47.2% compared with 62.5±81.7% (p = 0. 43) for Biology (data not in table).\nAbbreviations: IF: impact factor; EU: European Union; SD: standard deviation; US: United States.", "The main finding of our study is that the impact factor of journals published by European scientific societies rose more than that of journals published by US ones, in cumulatively 4 select subject categories, in the fields of Medicine or Biology, over the past decade. However, considerable variability was observed in this regard between journals in specific subject categories; the difference in the rise of the impact factor between European and US journals was particularly seen in the medical subject categories (Critical Care Medicine and Infectious Diseases) and Cell Biology, whereas the opposite was observed for Biology.\nThe difference in favor of the European journals regarding the degree of change in the impact factor appeared less pronounced when the median impact factor of each subject category was analyzed than when the percentage change in the impact factor of the journals included in each category was analyzed. In the first analysis, all journals that had an impact factor for at least one of the studied years were included, in contrast with the second, in which only journals that had an impact factor for both the initial and the final year compared were included. The latter analysis plausibly refers to journals with substantial tradition and influence; those with a low and declining impact factor could instead have been left out of the Journal Citation Reports® database.\nA potential explanation for the relatively high increase in the impact factor of European Journals could lie in the role of the funding for research. Over the past decade, the European Union has given greater value than before on research funding, by allocating more financial resources, organizing better the process of the allocation of resources, and favoring scientific collaboration within the European Union [7]. On the other hand, the rate of research funding in the US appears to have slowed down during the past decade compared with the previous one [8]. The above differences could have resulted in a greater rise in the research productivity of Europe compared with the US [9], and this could be reflected in our study findings.\nThere are various factors, however, that can influence the impact factor of scientific journals [1], [10]. One of these, is the language of publication of the journals; the dominance of English in biomedical publications has well been consolidated [11]. It is possible, therefore, that editorial committees of European journals published in local languages have been tempted over the last years to change the language of publication into English, so as to increase the penetration of the journals in the global scientific community [12]. The latter could have led into a rise of their impact factor. European journals could also have achieved greater visibility among scientists through the development or advancement of electronic scientific databases, such as Scopus, that cover a greater proportion of European journals than PubMed [13].\nOur study findings may be of interest to the individual researchers for deciding to what journal they should submit their work for publication. There is an important time lag between article submission, publication, and assignment of the journal impact factors for the specific year [14]. Thus, knowing the temporal trends of the impact factors in specific categories of journals, may be useful in this regard.\nOur study has several potential limitations. First, it is limited to journals of 4 subject categories, two in Biology and two in Medicine. Although the biological subject categories evaluated can be considered as potentially representative of the whole field, it might not be so for the medical ones. Moreover, there was difference in our study findings between the subject categories evaluated. Other studies have also shown that the scientific contribution of researchers from different countries can differ by the discipline examined [5], [9].\nWe also limited our analysis to journals published on behalf of scientific societies. This has inevitably decreased the sample size we analyzed. Yet, we elected to include only the above-mentioned type of journals, as the identity of other types of journals with regard to geographical origin can be more difficult to establish with accuracy. Some bibliographic databases include such information, which is easy to retrieve, but it mostly reflects the country that the publisher of the journal is legally based in, rather than the origin of the scientific board of the journal. The journals published on behalf of scientific societies can, to our view, reflect more accurately the research productivity in each region.\nFinally, our findings should be interpreted in the context of the well-discussed limitations of the journal impact factor as an indicator of the quality of scientific journals. Several scientists consider that this indicator could primarily reflect journal popularity rather than quality.[15] Moreover, there are certain ways through which journal editors can ‘manipulate’ the impact factor.[10] Still, although other relevant indicators have been developed, none has taken at least thus far the place of the journal impact factor.[2], [15]\n\nIn conclusion, our study indicates that the journals that were published on behalf of European scientific societies in 4 select biomedical subject categories tended, over the past decade, to increase their impact factor more than the respective US journals. This finding varied considerably between the 4 subject categories examined. Our analysis cannot however reflect the entire scientific fields of Biology and Medicine. Our findings could be interpreted as potentially indicative of an effort of the European Union to close the ‘gap’ in research productivity with the US." ]

[ null, null, null ]

[ "Introduction", "Methods", "Results", "Discussion" ]

[ "The most commonly used indicator of the quality or at least the popularity of scientific journals is the journal impact factor, which is not however exempt of some limitations [1], [2]. Over the past years, the impact factor of biomedical journals has generally shown a tendency to rise, which can in part be attributed to the expansion in the size of the relevant literature [3], [4]. It is well recognized also that the leading positions in productivity in biomedical research are held by developed countries, including those from North America and Western Europe [5], [6].\nIn this context, we sought to evaluate the trends in the impact factor of journals published on behalf of United States (US) and European biomedical scientific societies during the past 10 years.", "We did a retrospective analysis of journal impact factors provided by Thomson Reuters Journal Citation Reports® for the last 10-year period (1999–2008). We focused on 2 biomedical scientific fields, Biology and Medicine and selected two subject categories (as classified in the above-mentioned database) from each field. Specifically, we selected Biology, Cell Biology, Critical Care Medicine, and Infectious Diseases. These subject categories were chosen among those with a high median category impact factor for each field.\nWe retrieved all journals by name and International Standard Serial Number (ISSN) that were indexed in the above 4 subject categories of Journal Citation Reports,® for the years 1999, 2002, 2005, and 2008. To identify the journals that were published on behalf of scientific societies or professional organizations we searched in the official website of each of the retrieved journals for relevant information. We selected for inclusion the journals representing US or European scientific societies; the European countries of interest were specifically the first 15 ones to participate in the European Union (EU-15). Journals representing international scientific societies were excluded.\nWe extracted the impact factors of each of the included journals for the years 1999, 2002, 2005, and 2008. We grouped these data into separate variables for the US and European journals, respectively. We calculated the median value for the journal impact factor in each category and plotted graphically the temporal trends for the studied period. Journals that had an impact factor for any one of the above-mentioned years were included in this analysis.\nWe additionally calculated for each of the included journals the percentage change in the impact factor between 2002 and 1999, 2005 and 2002, 2008 and 2005, and finally between 2008 and 1999. Only journals that had an impact factor for both years in regard could be included in this analysis. We grouped these data into separate variables for the US and European journals and the years in regard. We assessed for statistical differences between the above variables. Our primary comparison was the difference in the percentage change in the impact factor between the US and European journals from 1999 to 2009. We used the independent samples t-test statistic for this comparison; a p-value less than 0.05 was considered as statistically significant.", "The 4 subject categories of our interest (Biology, Cell Biology, Critical Care Medicine, and Infectious Diseases) in the Journal Citation Reports® database, included in total 252, 280, 288, and 303 journals for each of the years 1999, 2002, 2005, and 2008, respectively. Among these, we identified 167 different journals that were published on behalf of US or EU biomedical scientific societies and had an impact factor for at least one of the above years. In Table 1, we present the distribution of the journals included in our analysis with regard to region and subject category.\nAbbreviations: EU: European Union; US: United States.\nIn Figure 1, we present graphically the temporal trends in the median impact factors of the journals included in each of the above 4 subject categories and published on behalf of US compared with EU scientific societies, in the 4 selected years between 1999 and 2008. As can be inferred from the graphs included in Figure 1, in 1998 US journals had clearly higher median impact factor compared with the European ones in the all examined subject categories, except for Biology. However, there appears to be a trend towards a greater increase in the median impact factor of journals published on behalf of European scientific societies compared with the US ones, in the subject categories of Critical Care Medicine and Infectious Diseases, over the past decade. The increase in the median impact factor of journals published on behalf of European scientific societies parallels that of the US ones, in the subject category of Cell Biology, whereas there is a trend for a greater increase in the median impact factor of journals published on behalf of the US societies, compared with the European ones, in the subject category of Biology.\nThe fitted line represents the temporal trend in the median impact factors.\nIn Table 2, we present the findings of our analysis regarding the percentage change in the impact factor of journals published on behalf of US and European scientific societies, cumulatively for all the 4 selected subject categories, between 1999 and 2008, as well as for the interval periods. Specifically, between 1999 and 2008, the impact factor of journals published on behalf of European scientific societies rose more than those published on behalf of the US ones (by 73.7±110.0%, as compared with 39.7±70.0%, p = 0.049). No relevant statistically significant difference was observed in any of the interval periods. With regard to each selected subject category, the impact factor of journals published on behalf of European scientific societies compared with US ones, rose, between 1999 and 2008, by 212.4±291.5% compared with 65.4±45.5% (p = 0.107) for Critical Care Medicine, 88.3±99.5% compared with 48.7±61.3% (p = 0.39) for Infectious Diseases, 61.7±82.0% compared with 16.3±68.7% (p = 0.029) for Cell Biology, and by 41.0±47.2% compared with 62.5±81.7% (p = 0. 43) for Biology (data not in table).\nAbbreviations: IF: impact factor; EU: European Union; SD: standard deviation; US: United States.", "The main finding of our study is that the impact factor of journals published by European scientific societies rose more than that of journals published by US ones, in cumulatively 4 select subject categories, in the fields of Medicine or Biology, over the past decade. However, considerable variability was observed in this regard between journals in specific subject categories; the difference in the rise of the impact factor between European and US journals was particularly seen in the medical subject categories (Critical Care Medicine and Infectious Diseases) and Cell Biology, whereas the opposite was observed for Biology.\nThe difference in favor of the European journals regarding the degree of change in the impact factor appeared less pronounced when the median impact factor of each subject category was analyzed than when the percentage change in the impact factor of the journals included in each category was analyzed. In the first analysis, all journals that had an impact factor for at least one of the studied years were included, in contrast with the second, in which only journals that had an impact factor for both the initial and the final year compared were included. The latter analysis plausibly refers to journals with substantial tradition and influence; those with a low and declining impact factor could instead have been left out of the Journal Citation Reports® database.\nA potential explanation for the relatively high increase in the impact factor of European Journals could lie in the role of the funding for research. Over the past decade, the European Union has given greater value than before on research funding, by allocating more financial resources, organizing better the process of the allocation of resources, and favoring scientific collaboration within the European Union [7]. On the other hand, the rate of research funding in the US appears to have slowed down during the past decade compared with the previous one [8]. The above differences could have resulted in a greater rise in the research productivity of Europe compared with the US [9], and this could be reflected in our study findings.\nThere are various factors, however, that can influence the impact factor of scientific journals [1], [10]. One of these, is the language of publication of the journals; the dominance of English in biomedical publications has well been consolidated [11]. It is possible, therefore, that editorial committees of European journals published in local languages have been tempted over the last years to change the language of publication into English, so as to increase the penetration of the journals in the global scientific community [12]. The latter could have led into a rise of their impact factor. European journals could also have achieved greater visibility among scientists through the development or advancement of electronic scientific databases, such as Scopus, that cover a greater proportion of European journals than PubMed [13].\nOur study findings may be of interest to the individual researchers for deciding to what journal they should submit their work for publication. There is an important time lag between article submission, publication, and assignment of the journal impact factors for the specific year [14]. Thus, knowing the temporal trends of the impact factors in specific categories of journals, may be useful in this regard.\nOur study has several potential limitations. First, it is limited to journals of 4 subject categories, two in Biology and two in Medicine. Although the biological subject categories evaluated can be considered as potentially representative of the whole field, it might not be so for the medical ones. Moreover, there was difference in our study findings between the subject categories evaluated. Other studies have also shown that the scientific contribution of researchers from different countries can differ by the discipline examined [5], [9].\nWe also limited our analysis to journals published on behalf of scientific societies. This has inevitably decreased the sample size we analyzed. Yet, we elected to include only the above-mentioned type of journals, as the identity of other types of journals with regard to geographical origin can be more difficult to establish with accuracy. Some bibliographic databases include such information, which is easy to retrieve, but it mostly reflects the country that the publisher of the journal is legally based in, rather than the origin of the scientific board of the journal. The journals published on behalf of scientific societies can, to our view, reflect more accurately the research productivity in each region.\nFinally, our findings should be interpreted in the context of the well-discussed limitations of the journal impact factor as an indicator of the quality of scientific journals. Several scientists consider that this indicator could primarily reflect journal popularity rather than quality.[15] Moreover, there are certain ways through which journal editors can ‘manipulate’ the impact factor.[10] Still, although other relevant indicators have been developed, none has taken at least thus far the place of the journal impact factor.[2], [15]\n\nIn conclusion, our study indicates that the journals that were published on behalf of European scientific societies in 4 select biomedical subject categories tended, over the past decade, to increase their impact factor more than the respective US journals. This finding varied considerably between the 4 subject categories examined. Our analysis cannot however reflect the entire scientific fields of Biology and Medicine. Our findings could be interpreted as potentially indicative of an effort of the European Union to close the ‘gap’ in research productivity with the US." ]

[ null, "methods", null, null ]

[]

[ "Adult", "Antibodies, Viral", "China", "Disease Outbreaks", "Female", "Humans", "Influenza A Virus, H1N1 Subtype", "Influenza Vaccines", "Influenza, Human", "Male", "Middle Aged", "Recurrence", "Risk", "Vaccination", "Young Adult" ]

[ "Introduction", "Study Design", "Ethics Statement", "Laboratory Methods", "Real-time RT-PCR", "Hemagglutination Inhibition (HI) Assay", "Statistical Analysis", "Results", "Antibody seroconversion rate", "Geometric mean titers (GMT)", "Discussion" ]

[ "A pandemic influenza A (H1N1) virus spread worldwide since April 2009, resulting in more than 16,000 deaths until March 2010. On 10 August 2010, WHO Director-General Dr Margaret Chan announced that the H1N1 influenza virus has moved into the post-pandemic period [1]. Although 2009 pandemic influenza A (H1N1) has been controlled, its recurrence cannot be excluded yet [2]. Guangzhou, the capital city of Guangdong province in south China, is one of the earliest attacked areas by 2009 pandemic influenza A (H1N1) virus. An inactivated vaccine against 2009 pandemic influenza A (H1N1) virus had been urgently manufactured to be used as an economical and effective weapon for the prophylaxis. To evaluate the risk of the recurrence and the efficiency of the vaccination, we conducted a follow-up study by detecting serum specimens collected from virus infected cases in an outbreak of a boarding school and vaccinated people in Guangzhou. The antibody dynamics characteristics would provide useful information for evaluating risk of the potential recurrence and efficacy of the vaccination.", "We investigated antibody responses in 129 patients with the pandemic influenza HIN1 and 86 persons who received the pandemic H1N1 vaccine in Guangzhou China. Patients who showed influenza symptoms, temperature ≥37.5° and viral RNA and/or antibody seroconversion for the pandemic virus were recruited in an outbreak of 2009 pandemic influenza H1N1 in a boarding school from August 21st to October 15th, while vaccinated study subjects were recruited from healthy persons who received the vaccine provided by Ministry of Health of China on October 30th 2009. These patients or vaccinated persons showed antibody negative to the pandemic virus (HI titer <1∶20) at the onset day of the disease or when they received the vaccination (day 0). Serum samples were collected from these patients and vaccinated persons at day 0, 15, 30, 180 after the onset of the disease or the vaccination, respectively. The ages of the study subjects in patient group were from 14 to 20 years, and that in vaccinated people were from 19 to 57 years. There are 86 males and 43 females in the patient group and 46 males and 40 females in vaccinated group.\nThe influenza A/H1N1 monovalent, split-virus, non-adjuvanted vaccines were manufactured by Tianyuan Bio-Pharmaceutical Co., Ltd. (batch number 20090902) through the nationwide vaccination program. Each dose of 0.5 ml product contained 15 µg hemagglutinin as prescribed by national guidelines. The vaccine was administered through intramuscular injection in the deltoid muscle.", "This study was approved by the ethics committee of the Guangzhou Center for Disease Control and Prevention and written informed consent was obtained from the study subjects.", "[SUBTITLE] Real-time RT-PCR [SUBSECTION] The existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.\nThe existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.\n[SUBTITLE] Hemagglutination Inhibition (HI) Assay [SUBSECTION] Antibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.\nAntibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.", "The existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.", "Antibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.", "Data were analyzed using EpiInfo version 3.3.2 (CDC, USA) and SPSS version 13.0 (SPSS Inc., USA). The statistically significant criterion was P-value <0.05. We compared the seroconversion rate (HI ≥40) by χ2-test between day 15 and day 30 or day 30 and day 180 or between natural infection and vaccinated group. We compared geometric mean titers (GMT) of those with positive results by Independent-Samples T test between day 15 and day 30 or day 30 and day 180 or between natural infection and vaccinated group. For GMT calculations, antibody levels below the detection limit (<1∶20) were assigned the value of 1∶10.", "[SUBTITLE] Antibody seroconversion rate [SUBSECTION] HI antibody positive rate of the patients increased significantly from 0% to 60% (95%CI: 26–88%) at day 15 (χ2 = 78, P<0.001) and 100% (95%CI: 93–100%) at day 30 (χ2 = 23, P<0.001), but decreased significantly to 52% (95%CI: 42–62%) at day 180 (χ2 = 38, P<0.001), while that of vaccinated subjects increased from 0% to 78% (95%CI: 68–86%) at day 15 (χ2 = 110, P<0.001) and 81% (95%CI: 72–89%) at day 30 (χ2 = 0.32, P = 0.57), but decreased significantly to 34% (95%CI: 24–45%) at day 180 (χ2 = 39, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed lower seroconversion rate (χ2 = 11, P<0.001; χ2 = 5.9, P = 0.015) at day 30 and day 180, respectively (Fig. 1a). In the vaccinated group, we did not find a significant difference in antibody seroconversion rate in different age groups (Table 2).\nNote: Detection limitation (HI titer <20) is indicated by the dotted line. Error bar indicates ± standard deviation (SD) from different individual study subjects. * indicates significant differences (P<0.01) between results of day 30 and day 180.\n*p<0.05,\n**p<0.01.\nHI antibody positive rate of the patients increased significantly from 0% to 60% (95%CI: 26–88%) at day 15 (χ2 = 78, P<0.001) and 100% (95%CI: 93–100%) at day 30 (χ2 = 23, P<0.001), but decreased significantly to 52% (95%CI: 42–62%) at day 180 (χ2 = 38, P<0.001), while that of vaccinated subjects increased from 0% to 78% (95%CI: 68–86%) at day 15 (χ2 = 110, P<0.001) and 81% (95%CI: 72–89%) at day 30 (χ2 = 0.32, P = 0.57), but decreased significantly to 34% (95%CI: 24–45%) at day 180 (χ2 = 39, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed lower seroconversion rate (χ2 = 11, P<0.001; χ2 = 5.9, P = 0.015) at day 30 and day 180, respectively (Fig. 1a). In the vaccinated group, we did not find a significant difference in antibody seroconversion rate in different age groups (Table 2).\nNote: Detection limitation (HI titer <20) is indicated by the dotted line. Error bar indicates ± standard deviation (SD) from different individual study subjects. * indicates significant differences (P<0.01) between results of day 30 and day 180.\n*p<0.05,\n**p<0.01.\n[SUBTITLE] Geometric mean titers (GMT) [SUBSECTION] No significant difference was found in GMT of HI antibodies in positive samples collected from the patients between 63 (95%CI: 30–135) at day 15 and 80 (95%CI: 68–93) at day 30 (T = 0.92, P = 0.36), but it decreased significantly to 52 (95%CI: 47–58) at day 180 (T = 4.5, P<0.001). GMT of vaccinated persons increased significantly from 100 (95%CI: 84–120) at day 15 to 193 (95%CI: 154–242) at day 30 (T = 4.5, P<0.001), but deceased significantly to 74 (95%CI: 63–89) at day 180 (T = 5.1, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed higher GMT at day 30 (T = 6.0, P<0.001) and day 180 (T = 3.6, P = 0.001), respectively (Fig. 1b). In the vaccinated group, we did not find a significant difference in GMT between different age groups (Table 2).\nNo significant difference was found in GMT of HI antibodies in positive samples collected from the patients between 63 (95%CI: 30–135) at day 15 and 80 (95%CI: 68–93) at day 30 (T = 0.92, P = 0.36), but it decreased significantly to 52 (95%CI: 47–58) at day 180 (T = 4.5, P<0.001). GMT of vaccinated persons increased significantly from 100 (95%CI: 84–120) at day 15 to 193 (95%CI: 154–242) at day 30 (T = 4.5, P<0.001), but deceased significantly to 74 (95%CI: 63–89) at day 180 (T = 5.1, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed higher GMT at day 30 (T = 6.0, P<0.001) and day 180 (T = 3.6, P = 0.001), respectively (Fig. 1b). In the vaccinated group, we did not find a significant difference in GMT between different age groups (Table 2).", "HI antibody positive rate of the patients increased significantly from 0% to 60% (95%CI: 26–88%) at day 15 (χ2 = 78, P<0.001) and 100% (95%CI: 93–100%) at day 30 (χ2 = 23, P<0.001), but decreased significantly to 52% (95%CI: 42–62%) at day 180 (χ2 = 38, P<0.001), while that of vaccinated subjects increased from 0% to 78% (95%CI: 68–86%) at day 15 (χ2 = 110, P<0.001) and 81% (95%CI: 72–89%) at day 30 (χ2 = 0.32, P = 0.57), but decreased significantly to 34% (95%CI: 24–45%) at day 180 (χ2 = 39, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed lower seroconversion rate (χ2 = 11, P<0.001; χ2 = 5.9, P = 0.015) at day 30 and day 180, respectively (Fig. 1a). In the vaccinated group, we did not find a significant difference in antibody seroconversion rate in different age groups (Table 2).\nNote: Detection limitation (HI titer <20) is indicated by the dotted line. Error bar indicates ± standard deviation (SD) from different individual study subjects. * indicates significant differences (P<0.01) between results of day 30 and day 180.\n*p<0.05,\n**p<0.01.", "No significant difference was found in GMT of HI antibodies in positive samples collected from the patients between 63 (95%CI: 30–135) at day 15 and 80 (95%CI: 68–93) at day 30 (T = 0.92, P = 0.36), but it decreased significantly to 52 (95%CI: 47–58) at day 180 (T = 4.5, P<0.001). GMT of vaccinated persons increased significantly from 100 (95%CI: 84–120) at day 15 to 193 (95%CI: 154–242) at day 30 (T = 4.5, P<0.001), but deceased significantly to 74 (95%CI: 63–89) at day 180 (T = 5.1, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed higher GMT at day 30 (T = 6.0, P<0.001) and day 180 (T = 3.6, P = 0.001), respectively (Fig. 1b). In the vaccinated group, we did not find a significant difference in GMT between different age groups (Table 2).", "The 2009 pandemic influenza A (H1N1) virus is a completely new infectious agent to human being. To investigate antibody dynamics which induced by natural infection or vaccination of the virus, we conducted this prospective study by following-up the infected patients and vaccinated people in the same city for six months.\nConsistent with previous reports [5]–[7], our results demonstrated that the vaccination was effective. The antibody seroconversion rate in vaccinated people achieved to a stable level more rapidly than naturally infected patients (15 vs 30 days), which might be due to the virus needs time to replicate after the infection. However, only about 80% in vaccinated people acquired protective antibody, whereas all subjects (100%) in the group of infected patients yield protective antibody at 30 days after the natural infection. It has been reported that seroconversion rates of vaccinations with live virus vaccines reached 86–97% but that of vaccination with inactivated vaccines achieved only 50–80% [8]–[10]. Our results supported that live virus vaccines may be more effective than inactivated vaccines. Furthermore, we found that higher GMT of antibody was achieved in the vaccinated subjects than in the infected patients. Previous studies also reported that GMT in recipients of live attenuated influenza vaccines was lower than that in recipients of inactivated vaccines [11], [12]. In this regard, a boost dose may be help to improve protective antibody response for live virus vaccination.\nIn contrast with previous studies which reported that the positive rate and GMT in children or elder persons were usually lower than the other age groups [13], we did not find a significant difference in antibody seroconversion rate and GMT levels in different age groups of vaccinated people. This might be due to that the age of study subjects recruited in this study were ranged from 19 to 60 years, so that children and elder persons did not include.\nIn this study, we first reported that both positive rate and GMT of antibodies to 2009 pandemic influenza A (H1N1) virus were decreased quickly in the patients and vaccinated persons. Antibody positive rates had been dropped down from 100% and 81% at day 30 to 52% and 34% at day 180, while GMTs were decreased from 80 and 193 at day 30 to 52 and 74 at day 180 in patients and vaccinated people, respective. The observation of low level antibody responses to the pandemic influenza viral infection and quick decrease of protective antibody levels may be able to explain why some pandemic influenza patients acquired a re-infection shortly as reported by Perez et al [14]. Considering the antibody levels tend to further decrease subsequently, these recovered patients and vaccinated persons may probably have no sufficient immunity against the recurrence of the viral infection. Thus, vaccination or re-vaccination may be necessary for prevention of the recurrence." ]

[ null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Methods", "Study Design", "Ethics Statement", "Laboratory Methods", "Real-time RT-PCR", "Hemagglutination Inhibition (HI) Assay", "Statistical Analysis", "Results", "Antibody seroconversion rate", "Geometric mean titers (GMT)", "Discussion" ]

[ "A pandemic influenza A (H1N1) virus spread worldwide since April 2009, resulting in more than 16,000 deaths until March 2010. On 10 August 2010, WHO Director-General Dr Margaret Chan announced that the H1N1 influenza virus has moved into the post-pandemic period [1]. Although 2009 pandemic influenza A (H1N1) has been controlled, its recurrence cannot be excluded yet [2]. Guangzhou, the capital city of Guangdong province in south China, is one of the earliest attacked areas by 2009 pandemic influenza A (H1N1) virus. An inactivated vaccine against 2009 pandemic influenza A (H1N1) virus had been urgently manufactured to be used as an economical and effective weapon for the prophylaxis. To evaluate the risk of the recurrence and the efficiency of the vaccination, we conducted a follow-up study by detecting serum specimens collected from virus infected cases in an outbreak of a boarding school and vaccinated people in Guangzhou. The antibody dynamics characteristics would provide useful information for evaluating risk of the potential recurrence and efficacy of the vaccination.", "[SUBTITLE] Study Design [SUBSECTION] We investigated antibody responses in 129 patients with the pandemic influenza HIN1 and 86 persons who received the pandemic H1N1 vaccine in Guangzhou China. Patients who showed influenza symptoms, temperature ≥37.5° and viral RNA and/or antibody seroconversion for the pandemic virus were recruited in an outbreak of 2009 pandemic influenza H1N1 in a boarding school from August 21st to October 15th, while vaccinated study subjects were recruited from healthy persons who received the vaccine provided by Ministry of Health of China on October 30th 2009. These patients or vaccinated persons showed antibody negative to the pandemic virus (HI titer <1∶20) at the onset day of the disease or when they received the vaccination (day 0). Serum samples were collected from these patients and vaccinated persons at day 0, 15, 30, 180 after the onset of the disease or the vaccination, respectively. The ages of the study subjects in patient group were from 14 to 20 years, and that in vaccinated people were from 19 to 57 years. There are 86 males and 43 females in the patient group and 46 males and 40 females in vaccinated group.\nThe influenza A/H1N1 monovalent, split-virus, non-adjuvanted vaccines were manufactured by Tianyuan Bio-Pharmaceutical Co., Ltd. (batch number 20090902) through the nationwide vaccination program. Each dose of 0.5 ml product contained 15 µg hemagglutinin as prescribed by national guidelines. The vaccine was administered through intramuscular injection in the deltoid muscle.\nWe investigated antibody responses in 129 patients with the pandemic influenza HIN1 and 86 persons who received the pandemic H1N1 vaccine in Guangzhou China. Patients who showed influenza symptoms, temperature ≥37.5° and viral RNA and/or antibody seroconversion for the pandemic virus were recruited in an outbreak of 2009 pandemic influenza H1N1 in a boarding school from August 21st to October 15th, while vaccinated study subjects were recruited from healthy persons who received the vaccine provided by Ministry of Health of China on October 30th 2009. These patients or vaccinated persons showed antibody negative to the pandemic virus (HI titer <1∶20) at the onset day of the disease or when they received the vaccination (day 0). Serum samples were collected from these patients and vaccinated persons at day 0, 15, 30, 180 after the onset of the disease or the vaccination, respectively. The ages of the study subjects in patient group were from 14 to 20 years, and that in vaccinated people were from 19 to 57 years. There are 86 males and 43 females in the patient group and 46 males and 40 females in vaccinated group.\nThe influenza A/H1N1 monovalent, split-virus, non-adjuvanted vaccines were manufactured by Tianyuan Bio-Pharmaceutical Co., Ltd. (batch number 20090902) through the nationwide vaccination program. Each dose of 0.5 ml product contained 15 µg hemagglutinin as prescribed by national guidelines. The vaccine was administered through intramuscular injection in the deltoid muscle.\n[SUBTITLE] Ethics Statement [SUBSECTION] This study was approved by the ethics committee of the Guangzhou Center for Disease Control and Prevention and written informed consent was obtained from the study subjects.\nThis study was approved by the ethics committee of the Guangzhou Center for Disease Control and Prevention and written informed consent was obtained from the study subjects.\n[SUBTITLE] Laboratory Methods [SUBSECTION] [SUBTITLE] Real-time RT-PCR [SUBSECTION] The existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.\nThe existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.\n[SUBTITLE] Hemagglutination Inhibition (HI) Assay [SUBSECTION] Antibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.\nAntibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.\n[SUBTITLE] Real-time RT-PCR [SUBSECTION] The existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.\nThe existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.\n[SUBTITLE] Hemagglutination Inhibition (HI) Assay [SUBSECTION] Antibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.\nAntibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.\n[SUBTITLE] Statistical Analysis [SUBSECTION] Data were analyzed using EpiInfo version 3.3.2 (CDC, USA) and SPSS version 13.0 (SPSS Inc., USA). The statistically significant criterion was P-value <0.05. We compared the seroconversion rate (HI ≥40) by χ2-test between day 15 and day 30 or day 30 and day 180 or between natural infection and vaccinated group. We compared geometric mean titers (GMT) of those with positive results by Independent-Samples T test between day 15 and day 30 or day 30 and day 180 or between natural infection and vaccinated group. For GMT calculations, antibody levels below the detection limit (<1∶20) were assigned the value of 1∶10.\nData were analyzed using EpiInfo version 3.3.2 (CDC, USA) and SPSS version 13.0 (SPSS Inc., USA). The statistically significant criterion was P-value <0.05. We compared the seroconversion rate (HI ≥40) by χ2-test between day 15 and day 30 or day 30 and day 180 or between natural infection and vaccinated group. We compared geometric mean titers (GMT) of those with positive results by Independent-Samples T test between day 15 and day 30 or day 30 and day 180 or between natural infection and vaccinated group. For GMT calculations, antibody levels below the detection limit (<1∶20) were assigned the value of 1∶10.", "We investigated antibody responses in 129 patients with the pandemic influenza HIN1 and 86 persons who received the pandemic H1N1 vaccine in Guangzhou China. Patients who showed influenza symptoms, temperature ≥37.5° and viral RNA and/or antibody seroconversion for the pandemic virus were recruited in an outbreak of 2009 pandemic influenza H1N1 in a boarding school from August 21st to October 15th, while vaccinated study subjects were recruited from healthy persons who received the vaccine provided by Ministry of Health of China on October 30th 2009. These patients or vaccinated persons showed antibody negative to the pandemic virus (HI titer <1∶20) at the onset day of the disease or when they received the vaccination (day 0). Serum samples were collected from these patients and vaccinated persons at day 0, 15, 30, 180 after the onset of the disease or the vaccination, respectively. The ages of the study subjects in patient group were from 14 to 20 years, and that in vaccinated people were from 19 to 57 years. There are 86 males and 43 females in the patient group and 46 males and 40 females in vaccinated group.\nThe influenza A/H1N1 monovalent, split-virus, non-adjuvanted vaccines were manufactured by Tianyuan Bio-Pharmaceutical Co., Ltd. (batch number 20090902) through the nationwide vaccination program. Each dose of 0.5 ml product contained 15 µg hemagglutinin as prescribed by national guidelines. The vaccine was administered through intramuscular injection in the deltoid muscle.", "This study was approved by the ethics committee of the Guangzhou Center for Disease Control and Prevention and written informed consent was obtained from the study subjects.", "[SUBTITLE] Real-time RT-PCR [SUBSECTION] The existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.\nThe existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.\n[SUBTITLE] Hemagglutination Inhibition (HI) Assay [SUBSECTION] Antibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.\nAntibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.", "The existence of pandemic influenza H1N1/2009 virus was detected by real-time RT-PCR as described previously [3]. Briefly, viral RNA was extracted from 140 µL nasopharyngeal specimen using the QIAamp® Viral RNA Mini Kit (Qiagen, Cat# 52906) following the manufacturer instructions. Viral RNA copies were determined by real-time one-step RT-PCR assay using Invitrogen SuperScript™III Platinum® One-Step Quantitative Kit (Invitrogen, Cat# 11732-088) and primers/probe which sequences provided by WHO, i.e. forward primer 5′-GTG CTA TAA ACA CCA GCC TYC CA-3′, reverse primer 5′-CGG GAT ATT CCT TAA TCC TGT RGC-3′, and probe 5′-(FAM)CA GAA TAT ACA TCC RGT CAC AAT TGG ARA A (TAMRA)-3′. Reactions were first incubated at 50°C for 30 min, denatured at 95°C for 2 min, and then were thermal-cycled for 40 cycles (95°C for 15 sec, 55°C for 30 sec). Serially diluted positive viral RNA controls were used as calibrators in each run.", "Antibody titers in these serum samples were determined by haemagglutination inhibition (HI) assay as described previously [3], [4]. The serum samples were treated with receptor destroying enzyme (RDE) from Vibrio cholerae (Denka Seiken, Cat#370013) for 18 h at 37°C and then were heat-inactivated at 56°C for 30 min according to WHO's standard procedure. Serum samples were diluted in serial two-fold dilutions from 1∶20 to 1∶640 and then mixed with 1% suspension of chicken red blood cells and 4 hemagglutinating units of a local isolated virus strain A/Guangdong Liwan/SWL1538/2009(H1N1). Specific positive antiserum and negative serum controls were included in the assay.\nAll the experiments were manipulated in the same laboratory of Guangzhou center for disease control and prevention.", "Data were analyzed using EpiInfo version 3.3.2 (CDC, USA) and SPSS version 13.0 (SPSS Inc., USA). The statistically significant criterion was P-value <0.05. We compared the seroconversion rate (HI ≥40) by χ2-test between day 15 and day 30 or day 30 and day 180 or between natural infection and vaccinated group. We compared geometric mean titers (GMT) of those with positive results by Independent-Samples T test between day 15 and day 30 or day 30 and day 180 or between natural infection and vaccinated group. For GMT calculations, antibody levels below the detection limit (<1∶20) were assigned the value of 1∶10.", "[SUBTITLE] Antibody seroconversion rate [SUBSECTION] HI antibody positive rate of the patients increased significantly from 0% to 60% (95%CI: 26–88%) at day 15 (χ2 = 78, P<0.001) and 100% (95%CI: 93–100%) at day 30 (χ2 = 23, P<0.001), but decreased significantly to 52% (95%CI: 42–62%) at day 180 (χ2 = 38, P<0.001), while that of vaccinated subjects increased from 0% to 78% (95%CI: 68–86%) at day 15 (χ2 = 110, P<0.001) and 81% (95%CI: 72–89%) at day 30 (χ2 = 0.32, P = 0.57), but decreased significantly to 34% (95%CI: 24–45%) at day 180 (χ2 = 39, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed lower seroconversion rate (χ2 = 11, P<0.001; χ2 = 5.9, P = 0.015) at day 30 and day 180, respectively (Fig. 1a). In the vaccinated group, we did not find a significant difference in antibody seroconversion rate in different age groups (Table 2).\nNote: Detection limitation (HI titer <20) is indicated by the dotted line. Error bar indicates ± standard deviation (SD) from different individual study subjects. * indicates significant differences (P<0.01) between results of day 30 and day 180.\n*p<0.05,\n**p<0.01.\nHI antibody positive rate of the patients increased significantly from 0% to 60% (95%CI: 26–88%) at day 15 (χ2 = 78, P<0.001) and 100% (95%CI: 93–100%) at day 30 (χ2 = 23, P<0.001), but decreased significantly to 52% (95%CI: 42–62%) at day 180 (χ2 = 38, P<0.001), while that of vaccinated subjects increased from 0% to 78% (95%CI: 68–86%) at day 15 (χ2 = 110, P<0.001) and 81% (95%CI: 72–89%) at day 30 (χ2 = 0.32, P = 0.57), but decreased significantly to 34% (95%CI: 24–45%) at day 180 (χ2 = 39, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed lower seroconversion rate (χ2 = 11, P<0.001; χ2 = 5.9, P = 0.015) at day 30 and day 180, respectively (Fig. 1a). In the vaccinated group, we did not find a significant difference in antibody seroconversion rate in different age groups (Table 2).\nNote: Detection limitation (HI titer <20) is indicated by the dotted line. Error bar indicates ± standard deviation (SD) from different individual study subjects. * indicates significant differences (P<0.01) between results of day 30 and day 180.\n*p<0.05,\n**p<0.01.\n[SUBTITLE] Geometric mean titers (GMT) [SUBSECTION] No significant difference was found in GMT of HI antibodies in positive samples collected from the patients between 63 (95%CI: 30–135) at day 15 and 80 (95%CI: 68–93) at day 30 (T = 0.92, P = 0.36), but it decreased significantly to 52 (95%CI: 47–58) at day 180 (T = 4.5, P<0.001). GMT of vaccinated persons increased significantly from 100 (95%CI: 84–120) at day 15 to 193 (95%CI: 154–242) at day 30 (T = 4.5, P<0.001), but deceased significantly to 74 (95%CI: 63–89) at day 180 (T = 5.1, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed higher GMT at day 30 (T = 6.0, P<0.001) and day 180 (T = 3.6, P = 0.001), respectively (Fig. 1b). In the vaccinated group, we did not find a significant difference in GMT between different age groups (Table 2).\nNo significant difference was found in GMT of HI antibodies in positive samples collected from the patients between 63 (95%CI: 30–135) at day 15 and 80 (95%CI: 68–93) at day 30 (T = 0.92, P = 0.36), but it decreased significantly to 52 (95%CI: 47–58) at day 180 (T = 4.5, P<0.001). GMT of vaccinated persons increased significantly from 100 (95%CI: 84–120) at day 15 to 193 (95%CI: 154–242) at day 30 (T = 4.5, P<0.001), but deceased significantly to 74 (95%CI: 63–89) at day 180 (T = 5.1, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed higher GMT at day 30 (T = 6.0, P<0.001) and day 180 (T = 3.6, P = 0.001), respectively (Fig. 1b). In the vaccinated group, we did not find a significant difference in GMT between different age groups (Table 2).", "HI antibody positive rate of the patients increased significantly from 0% to 60% (95%CI: 26–88%) at day 15 (χ2 = 78, P<0.001) and 100% (95%CI: 93–100%) at day 30 (χ2 = 23, P<0.001), but decreased significantly to 52% (95%CI: 42–62%) at day 180 (χ2 = 38, P<0.001), while that of vaccinated subjects increased from 0% to 78% (95%CI: 68–86%) at day 15 (χ2 = 110, P<0.001) and 81% (95%CI: 72–89%) at day 30 (χ2 = 0.32, P = 0.57), but decreased significantly to 34% (95%CI: 24–45%) at day 180 (χ2 = 39, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed lower seroconversion rate (χ2 = 11, P<0.001; χ2 = 5.9, P = 0.015) at day 30 and day 180, respectively (Fig. 1a). In the vaccinated group, we did not find a significant difference in antibody seroconversion rate in different age groups (Table 2).\nNote: Detection limitation (HI titer <20) is indicated by the dotted line. Error bar indicates ± standard deviation (SD) from different individual study subjects. * indicates significant differences (P<0.01) between results of day 30 and day 180.\n*p<0.05,\n**p<0.01.", "No significant difference was found in GMT of HI antibodies in positive samples collected from the patients between 63 (95%CI: 30–135) at day 15 and 80 (95%CI: 68–93) at day 30 (T = 0.92, P = 0.36), but it decreased significantly to 52 (95%CI: 47–58) at day 180 (T = 4.5, P<0.001). GMT of vaccinated persons increased significantly from 100 (95%CI: 84–120) at day 15 to 193 (95%CI: 154–242) at day 30 (T = 4.5, P<0.001), but deceased significantly to 74 (95%CI: 63–89) at day 180 (T = 5.1, P<0.001) (Table 1). Compared to the patients, the vaccinated subjects showed higher GMT at day 30 (T = 6.0, P<0.001) and day 180 (T = 3.6, P = 0.001), respectively (Fig. 1b). In the vaccinated group, we did not find a significant difference in GMT between different age groups (Table 2).", "The 2009 pandemic influenza A (H1N1) virus is a completely new infectious agent to human being. To investigate antibody dynamics which induced by natural infection or vaccination of the virus, we conducted this prospective study by following-up the infected patients and vaccinated people in the same city for six months.\nConsistent with previous reports [5]–[7], our results demonstrated that the vaccination was effective. The antibody seroconversion rate in vaccinated people achieved to a stable level more rapidly than naturally infected patients (15 vs 30 days), which might be due to the virus needs time to replicate after the infection. However, only about 80% in vaccinated people acquired protective antibody, whereas all subjects (100%) in the group of infected patients yield protective antibody at 30 days after the natural infection. It has been reported that seroconversion rates of vaccinations with live virus vaccines reached 86–97% but that of vaccination with inactivated vaccines achieved only 50–80% [8]–[10]. Our results supported that live virus vaccines may be more effective than inactivated vaccines. Furthermore, we found that higher GMT of antibody was achieved in the vaccinated subjects than in the infected patients. Previous studies also reported that GMT in recipients of live attenuated influenza vaccines was lower than that in recipients of inactivated vaccines [11], [12]. In this regard, a boost dose may be help to improve protective antibody response for live virus vaccination.\nIn contrast with previous studies which reported that the positive rate and GMT in children or elder persons were usually lower than the other age groups [13], we did not find a significant difference in antibody seroconversion rate and GMT levels in different age groups of vaccinated people. This might be due to that the age of study subjects recruited in this study were ranged from 19 to 60 years, so that children and elder persons did not include.\nIn this study, we first reported that both positive rate and GMT of antibodies to 2009 pandemic influenza A (H1N1) virus were decreased quickly in the patients and vaccinated persons. Antibody positive rates had been dropped down from 100% and 81% at day 30 to 52% and 34% at day 180, while GMTs were decreased from 80 and 193 at day 30 to 52 and 74 at day 180 in patients and vaccinated people, respective. The observation of low level antibody responses to the pandemic influenza viral infection and quick decrease of protective antibody levels may be able to explain why some pandemic influenza patients acquired a re-infection shortly as reported by Perez et al [14]. Considering the antibody levels tend to further decrease subsequently, these recovered patients and vaccinated persons may probably have no sufficient immunity against the recurrence of the viral infection. Thus, vaccination or re-vaccination may be necessary for prevention of the recurrence." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null ]

[]

[ "Anti-Inflammatory Agents, Non-Steroidal", "Humans", "Myocardial Infarction", "Randomized Controlled Trials as Topic", "Risk" ]

[ "Introduction", "Results", "Discussion" ]

[ "Non-steroidal antiinflammatory drugs (NSAIDs) are cyclooxygenase inhibitors (COX-1 and −2) used commonly for the treatment of acute and chronic pain. With recent randomised studies demonstrating increased cardiovascular adverse events associated with selective COX-2 inhibitors (coxibs)[1]–[4], there is growing concern and evidence that NSAIDs predispose to myocardial infarction (MI), particularly in those patients at highest cardiovascular risk. The first study to show an increased risk of MI with a coxib was the VIGOR trial [1], which was designed to compare the gastrointestinal safety of rofecoxib and naproxen in patients with rheumatoid arthritis. Although rofecoxib demonstrated a lower risk of gastrointestinal events, there was a four- to five-fold increased risk of MI among users of rofecoxib, but restricted to non fatal events, with apparently no impact on coronary heart disease (CHD) deaths. Due to the low event rate, the clinical significance of this observation was uncertain. With many more studies published since then, there is an opportunity to assess more precisely the relationship between NSAIDs and MI according to its severity. No previous meta-analysis has addressed separately the risk of non-fatal and fatal MI risk associated with NSAID using both RCTs and observational studies.\nWe performed an analysis of all observational studies and randomized controlled trials published between January 1990 and March 2010 to summarize the impact of NSAIDs on fatal and non-fatal MI in different population types.", "Among observational studies we were able to obtain separate estimates for fatal and non-fatal MI in six studies (Table 1) [5]–[10]. Studies not eligible were those that did not include fatal cases, did not provide separate estimates, or considered as fatal cases all deaths or all cardiovascular deaths. A total of 47 studies were excluded for these reasons. Eligible studies include four cohort studies [5], [7], [9], [10] and two nested case-control studies [6], [8]. The total number of events in these studies ranged from 814 to 8852 non-fatal events and from 277 to 3119 fatal events. The summary OR estimate for fatal and non-fatal MI risk among the 6 eligible studies was 1.21 (95%CI, 1.07–1.37). Using a random effects model, the summary OR estimate for the association between current NSAID use and non-fatal MI was 1.30 (95% CI, 1.20–1.41) compared with 1.02 (95% CI, 0.89–1.17) for fatal MI (Figure 2). Therefore, no association was found between use of NSAIDs and fatal MI. The estimates for non-fatal MI were on average 25% larger than those for fatal MI (OR, 1.25; 95% CI, 1.11–1.42).\n*Number of days since last use, except in the study by Chan et al in which the current exposure was defined as more than 22 days of use in the last month.\n†At baseline (Women only).\nFollowed up after CHD hospitalization.\n§Cohort of aspirin users.\nC-C: Case-Control; CHD: Coronary Heart Disease; CI: Confidence Interval; GPRD: General Practitioner Research Database; MI:Myocardial Infarction; NA:Not Available; NSAIDs: non steroidal anti-inflammatory drugs; RR: Relative Risk; THIN: The Health Network; tNSAIDs: traditional NSAIDs.\nTwo observational studies included only individuals with prior CHD history [9] or prior cardiovascular events (defined as ischaemic heart disease or ischaemic cerebrovascular event)[10]. In this particular subgroup of studies, estimates for non-fatal MI were on average 58% greater than those for fatal MI (OR, 1.58; 95% CI, 1.26–1.98). The remaining four studies included either individuals with no prior history of CHD or a percentage of individuals with prior history below 25%. Among these studies the estimates for non-fatal MIs compared to fatal MIs were of smaller magnitude (OR, 1.19; 95% CI 1.08–1.31).\nNine randomised controlled trials were eligible for the analysis [1], [2], [11]–[17]. A total of 127 trials were excluded due to small sample size, short follow-up, or both. Sample sizes ranged from 1561 to 34,701 patients whereas average follow-up ranged from 6 months to about 3 years (Table 2). The number of events in individual studies ranged from 4 to 210 for non-fatal MIs and 6 to 23 for fatal events. NSAID treatment arms always included coxibs but the comparator group varied between placebo and tNSAIDs. Overall, the summary estimates for the combined fatal and non-fatal MI risk among the 5 placebo controlled trials and the 4 non placebo-controlled trials were 1.76 (95%CI, 1.09–2.83) and 1.31 (95%CI, 0.78–2.18), respectively. The summary RR estimate for non-fatal MI associated with current coxib use was 1.61 (95%CI, 1.04–2.50) using the random effects model while the corresponding estimate for fatal MI was 0.86 (95%CI, 0.51–1.47) (Figure 3). The estimate for non-fatal MI relative to fatal MI was of borderline statistical significance (OR, 1.86; 95%CI, 0.99–3.50) (Figure 4), although it included both placebo-controlled and tNSAIDs-controlled trials. A secondary analysis restricted to the 5 placebo-controlled trials showed an increased risk for non-fatal MIs (OR, 1.90; 95%CI, 1.05–3.41) but not for fatal events (OR, 1.37; 95%CI, 0.54–3.46). In this subgroup analysis, the OR of non-fatal vs fatal MIs was 1.30 (95% CI, 0.41–4.12). The 4 trials using tNSAIDs as comparators (naproxen, ibuprofen or diclofenac) did not show a significant increased risk of either non-fatal MI (OR, 1.48; 95%CI: 0.80–2.75) or of fatal MI (OR, 0.68; 95%CI, 0.36–1.31) associated with coxib treatment.\n*Crude Rate Ratios and exact 95% CI are calculated for each study dividing the rate of events in treatment arms over the rate of events in control arms. Crude rates are estimated from the reported number of events and the patient-years of exposure in each arm.\n†Prior History of Coronary Heart Disease.\nASA: Acetyl Salicylic Acid (aka aspirin); CHD: Coronary Heart Disease; CI: Confidence Interval; MI:Myocardial Infarction; RR: Relative Risk; Tx:Treatment; Crtl: Control.\nIncidence rates of non-fatal MI and CHD death were relatively similar between observational studies and randomised controlled trials. Except one outlier [9], observational studies reported incidence rates between 1 and 8 per 1,000 person-years for non-fatal MI and between 0.3 and 4 for CHD deaths. The corresponding estimates of incidence in randomised controlled trials were between 2 and 5 per 1,000 person-years and 0.4 and 3, respectively.", "This pooled analysis reviewing multiple and diverse studies supports the conclusion that NSAID treatment, including both tNSAIDs and coxibs, predisposes to non-fatal myocardial infarction, but only a marginal effect on coronary mortality risk was found. This is important because the elderly population, which has an increased cardiovascular risk, is the major consumer of NSAIDs, so the consistency of our finding of an increased risk of non-fatal MI is of great concern. However, the finding of a lack of coronary-related mortality increase in NSAID users may be considered somewhat reassuring for those who need to be treated with NSAIDs. The hazard of non-fatal cardiovascular events should be taken into consideration when prescribing these drugs in high-risk patients [18].\nBased on current knowledge, the underlying pathophysiology of NSAID-associated coronary risk can only be speculated. As patient characteristics are well balanced between patients assigned to the NSAID or control (placebo or active) arms in randomised trials, more advanced coronary disease or other comorbidities (e.g. heart failure, diabetes) are unlikely to be the major explanation for the selective increase in nonfatal cardiovascular events. The differential effects of COX-1 and COX-2 inhibition, regulating thrombus formation [4] is the most plausible mechanism. Selective COX-1 inhibition such as the one produced by low-dose aspirin has been shown to have an antithrombotic effect mediated by a relatively specific, potent, and irreversible inhibition of platelet COX-1, which reduces thromboxane A2 biosynthesis, a powerful trigger of platelet aggregation [19]. Randomised controlled trials with low dose aspirin have consistently shown a clear reduction in risk of non fatal MI and to a lesser extent fatal MI in high risk patients [21]–[25]. In contrast COX-2 inhibition with the resulting suppression of prostacyclin and other hemostatic mediators increase the risk of coronary thrombosis [20]. Thus, an incomplete antithrombotic effect of COX-1 inhibition may counteract the prothrombotic effects of COX-2 inhibition afforded by NSAIDs leading to non-fatal rather than fatal events. A recent study showed that NSAIDs only increase the risk of non-ST-segment elevation acute coronary syndromes but not of ST-segment elevation MI [26]. As the former are related to incomplete occlusion of major coronary arteries or occlusion of smaller coronary arteries [27], it was hypothesized that NSAID-related thrombosis might be less severe than spontaneous coronary thrombosis. Whether coronary thrombi generated under NSAID treatment are different in stability or location from spontaneous thrombi remains to be explored. To directly address the hypothesis that the COX-1 inhibition component of NSAIDs reduces damage caused by MI (shifts the balance towards non fatal MI), the case-fatality effect of NSAIDs individually according to their degree of COX-1 and COX-2 inhibition needs to be assessed. Unfortunately, the current study numbers at the individual level are insufficient to reach any meaningful conclusion.\nOnly a small number of studies originally reported fatal and non-fatal events separately and met our eligibility criteria. These may not be representative of the majority of published studies in the field. Yet, when we analyzed the combined estimate (fatal and non-fatal MI) in our subset of studies, the results are similar to those reported in previous meta-analyses of observational studies [28], [29] and RCTs [30]. Our review of studies addressing the risk of non-fatal MI and CHD deaths include observational studies, which are susceptible to bias and confounding but based upon the consistency of results across the observational studies and the RCTs, which have well balanced baseline characteristics, it is unlikely that our finding of can be entirely explained by bias. We found some evidence of heterogeneity in our analyses that reached statistical significance for the pooled estimate of fatal events among observational studies (Figure 2). In this context, the use of random effects models is favoured over fixed effects models. Additionally we were able to identify sources of heterogeneity by fitting a metaregression model and exploring the effect of different study characteristics (such as the inclusion criteria in observational studies or the use of placebo in RCTs) on the pooled estimate. Overall, there were insufficient data to assess the risk of non-fatal and fatal MI among individual traditional NSAIDs, or whether there is any difference between traditional NSAIDs and coxibs. This should be the ultimate goal as it has been previously shown that individual NSAIDs at commonly used doses, present variability in their magnitude of risk of ischemic coronary disease [4], [8]. Also, despite our criteria of minimum total sample size and follow-up requirements the power of RCTs to study these rare events is quite limited due to the paucity of events. The inclusion of smaller RCTs with few or no events would provide very little additional information to our risk estimates.\nNSAIDs increase the risk of non-fatal MI but do not increase coronary heart disease mortality. Due to the widespread use of NSAIDs in the population, particularly amongst the elderly, further research is urgently needed to unravel the specific mechanisms involved in NSAID-associated coronary thrombosis (including genetic and epigenetic markers) and the identification of patients at highest risk of such complication. Additionally, studies are warranted to reduce the coronary risk associated with NSAID use." ]

[ null, null, null ]

[ "Introduction", "Methods", "Results", "Discussion" ]

[ "Non-steroidal antiinflammatory drugs (NSAIDs) are cyclooxygenase inhibitors (COX-1 and −2) used commonly for the treatment of acute and chronic pain. With recent randomised studies demonstrating increased cardiovascular adverse events associated with selective COX-2 inhibitors (coxibs)[1]–[4], there is growing concern and evidence that NSAIDs predispose to myocardial infarction (MI), particularly in those patients at highest cardiovascular risk. The first study to show an increased risk of MI with a coxib was the VIGOR trial [1], which was designed to compare the gastrointestinal safety of rofecoxib and naproxen in patients with rheumatoid arthritis. Although rofecoxib demonstrated a lower risk of gastrointestinal events, there was a four- to five-fold increased risk of MI among users of rofecoxib, but restricted to non fatal events, with apparently no impact on coronary heart disease (CHD) deaths. Due to the low event rate, the clinical significance of this observation was uncertain. With many more studies published since then, there is an opportunity to assess more precisely the relationship between NSAIDs and MI according to its severity. No previous meta-analysis has addressed separately the risk of non-fatal and fatal MI risk associated with NSAID using both RCTs and observational studies.\nWe performed an analysis of all observational studies and randomized controlled trials published between January 1990 and March 2010 to summarize the impact of NSAIDs on fatal and non-fatal MI in different population types.", "We identified all observational studies and randomized controlled trials exploring the association between NSAID use and the occurrence of non-fatal MI and fatal-MI (including CHD death), indexed in Pubmed between January 1990 and March 2010 without language restrictions. Additionally we also considered those studies included in previously published systematic reviews. This was particularly useful in capturing data from unpublished RCTs. Studies evaluating traditional NSAIDs (tNSAIDs) and coxibs were considered when they provided either separate estimates of risk for non-fatal MI and fatal-MI according to NSAID use, or sufficient data to compute crude estimates. Because the power of randomised controlled trials to study these rare events is quite limited, we selected only those RCTs with a total sample size greater than 1500, and an average accrued follow-up longer than 6 months. Thus, even though individual studies meeting these sample size and follow-up criteria would still be largely underpowered, assuming an incidence rate of fatal cases around 1.5 per 1000 person-years it would be expected that they found at least one or more of these fatal cases. Additionally, these inclusion criteria would improve homogeneity between studies because only long-term follow-up studies were considered. Data was extracted in duplicate and any discrepancies were resolved through consensus. Extracted data included: type of design, study period, sample size, mean follow-up (in prospective studies), exposure definition, prior history of coronary heart disease (observational studies), use of aspirin (randomized controlled trials), fatal events, non fatal events, and specific estimates for the risk of fatal and non-fatal events. When endpoint specific adjusted estimates were not available and the number of fatal and non-fatal events, as well as the number of exposed and unexposed person-time, were available we estimated the crude incidence rate and relative risk (RR). The detailed results of the study selection process are shown in Figure 1.\nIn order to obtain weighted summary estimates we used random effects models (Stata 10.0). Among observational studies, we obtained separate pooled estimates according to whether prior history of cardiovascular disease was an eligibility criterion to enter the study. We also obtained separate pooled estimates for randomised controlled trials according to the nature of the comparison group (placebo/NSAID). Finally we explored heterogeneity of effects between non-fatal and fatal endpoints and formally tested this hypothesis by fitting a meta-regression model (stata metareg command). This model provides summary estimates of the relative increase in non-fatal estimates compared to fatal estimates in these studies associated with NSAIDs (odds ratio [OR] and 95% confidence interval [CI]) and the corresponding p values.", "Among observational studies we were able to obtain separate estimates for fatal and non-fatal MI in six studies (Table 1) [5]–[10]. Studies not eligible were those that did not include fatal cases, did not provide separate estimates, or considered as fatal cases all deaths or all cardiovascular deaths. A total of 47 studies were excluded for these reasons. Eligible studies include four cohort studies [5], [7], [9], [10] and two nested case-control studies [6], [8]. The total number of events in these studies ranged from 814 to 8852 non-fatal events and from 277 to 3119 fatal events. The summary OR estimate for fatal and non-fatal MI risk among the 6 eligible studies was 1.21 (95%CI, 1.07–1.37). Using a random effects model, the summary OR estimate for the association between current NSAID use and non-fatal MI was 1.30 (95% CI, 1.20–1.41) compared with 1.02 (95% CI, 0.89–1.17) for fatal MI (Figure 2). Therefore, no association was found between use of NSAIDs and fatal MI. The estimates for non-fatal MI were on average 25% larger than those for fatal MI (OR, 1.25; 95% CI, 1.11–1.42).\n*Number of days since last use, except in the study by Chan et al in which the current exposure was defined as more than 22 days of use in the last month.\n†At baseline (Women only).\nFollowed up after CHD hospitalization.\n§Cohort of aspirin users.\nC-C: Case-Control; CHD: Coronary Heart Disease; CI: Confidence Interval; GPRD: General Practitioner Research Database; MI:Myocardial Infarction; NA:Not Available; NSAIDs: non steroidal anti-inflammatory drugs; RR: Relative Risk; THIN: The Health Network; tNSAIDs: traditional NSAIDs.\nTwo observational studies included only individuals with prior CHD history [9] or prior cardiovascular events (defined as ischaemic heart disease or ischaemic cerebrovascular event)[10]. In this particular subgroup of studies, estimates for non-fatal MI were on average 58% greater than those for fatal MI (OR, 1.58; 95% CI, 1.26–1.98). The remaining four studies included either individuals with no prior history of CHD or a percentage of individuals with prior history below 25%. Among these studies the estimates for non-fatal MIs compared to fatal MIs were of smaller magnitude (OR, 1.19; 95% CI 1.08–1.31).\nNine randomised controlled trials were eligible for the analysis [1], [2], [11]–[17]. A total of 127 trials were excluded due to small sample size, short follow-up, or both. Sample sizes ranged from 1561 to 34,701 patients whereas average follow-up ranged from 6 months to about 3 years (Table 2). The number of events in individual studies ranged from 4 to 210 for non-fatal MIs and 6 to 23 for fatal events. NSAID treatment arms always included coxibs but the comparator group varied between placebo and tNSAIDs. Overall, the summary estimates for the combined fatal and non-fatal MI risk among the 5 placebo controlled trials and the 4 non placebo-controlled trials were 1.76 (95%CI, 1.09–2.83) and 1.31 (95%CI, 0.78–2.18), respectively. The summary RR estimate for non-fatal MI associated with current coxib use was 1.61 (95%CI, 1.04–2.50) using the random effects model while the corresponding estimate for fatal MI was 0.86 (95%CI, 0.51–1.47) (Figure 3). The estimate for non-fatal MI relative to fatal MI was of borderline statistical significance (OR, 1.86; 95%CI, 0.99–3.50) (Figure 4), although it included both placebo-controlled and tNSAIDs-controlled trials. A secondary analysis restricted to the 5 placebo-controlled trials showed an increased risk for non-fatal MIs (OR, 1.90; 95%CI, 1.05–3.41) but not for fatal events (OR, 1.37; 95%CI, 0.54–3.46). In this subgroup analysis, the OR of non-fatal vs fatal MIs was 1.30 (95% CI, 0.41–4.12). The 4 trials using tNSAIDs as comparators (naproxen, ibuprofen or diclofenac) did not show a significant increased risk of either non-fatal MI (OR, 1.48; 95%CI: 0.80–2.75) or of fatal MI (OR, 0.68; 95%CI, 0.36–1.31) associated with coxib treatment.\n*Crude Rate Ratios and exact 95% CI are calculated for each study dividing the rate of events in treatment arms over the rate of events in control arms. Crude rates are estimated from the reported number of events and the patient-years of exposure in each arm.\n†Prior History of Coronary Heart Disease.\nASA: Acetyl Salicylic Acid (aka aspirin); CHD: Coronary Heart Disease; CI: Confidence Interval; MI:Myocardial Infarction; RR: Relative Risk; Tx:Treatment; Crtl: Control.\nIncidence rates of non-fatal MI and CHD death were relatively similar between observational studies and randomised controlled trials. Except one outlier [9], observational studies reported incidence rates between 1 and 8 per 1,000 person-years for non-fatal MI and between 0.3 and 4 for CHD deaths. The corresponding estimates of incidence in randomised controlled trials were between 2 and 5 per 1,000 person-years and 0.4 and 3, respectively.", "This pooled analysis reviewing multiple and diverse studies supports the conclusion that NSAID treatment, including both tNSAIDs and coxibs, predisposes to non-fatal myocardial infarction, but only a marginal effect on coronary mortality risk was found. This is important because the elderly population, which has an increased cardiovascular risk, is the major consumer of NSAIDs, so the consistency of our finding of an increased risk of non-fatal MI is of great concern. However, the finding of a lack of coronary-related mortality increase in NSAID users may be considered somewhat reassuring for those who need to be treated with NSAIDs. The hazard of non-fatal cardiovascular events should be taken into consideration when prescribing these drugs in high-risk patients [18].\nBased on current knowledge, the underlying pathophysiology of NSAID-associated coronary risk can only be speculated. As patient characteristics are well balanced between patients assigned to the NSAID or control (placebo or active) arms in randomised trials, more advanced coronary disease or other comorbidities (e.g. heart failure, diabetes) are unlikely to be the major explanation for the selective increase in nonfatal cardiovascular events. The differential effects of COX-1 and COX-2 inhibition, regulating thrombus formation [4] is the most plausible mechanism. Selective COX-1 inhibition such as the one produced by low-dose aspirin has been shown to have an antithrombotic effect mediated by a relatively specific, potent, and irreversible inhibition of platelet COX-1, which reduces thromboxane A2 biosynthesis, a powerful trigger of platelet aggregation [19]. Randomised controlled trials with low dose aspirin have consistently shown a clear reduction in risk of non fatal MI and to a lesser extent fatal MI in high risk patients [21]–[25]. In contrast COX-2 inhibition with the resulting suppression of prostacyclin and other hemostatic mediators increase the risk of coronary thrombosis [20]. Thus, an incomplete antithrombotic effect of COX-1 inhibition may counteract the prothrombotic effects of COX-2 inhibition afforded by NSAIDs leading to non-fatal rather than fatal events. A recent study showed that NSAIDs only increase the risk of non-ST-segment elevation acute coronary syndromes but not of ST-segment elevation MI [26]. As the former are related to incomplete occlusion of major coronary arteries or occlusion of smaller coronary arteries [27], it was hypothesized that NSAID-related thrombosis might be less severe than spontaneous coronary thrombosis. Whether coronary thrombi generated under NSAID treatment are different in stability or location from spontaneous thrombi remains to be explored. To directly address the hypothesis that the COX-1 inhibition component of NSAIDs reduces damage caused by MI (shifts the balance towards non fatal MI), the case-fatality effect of NSAIDs individually according to their degree of COX-1 and COX-2 inhibition needs to be assessed. Unfortunately, the current study numbers at the individual level are insufficient to reach any meaningful conclusion.\nOnly a small number of studies originally reported fatal and non-fatal events separately and met our eligibility criteria. These may not be representative of the majority of published studies in the field. Yet, when we analyzed the combined estimate (fatal and non-fatal MI) in our subset of studies, the results are similar to those reported in previous meta-analyses of observational studies [28], [29] and RCTs [30]. Our review of studies addressing the risk of non-fatal MI and CHD deaths include observational studies, which are susceptible to bias and confounding but based upon the consistency of results across the observational studies and the RCTs, which have well balanced baseline characteristics, it is unlikely that our finding of can be entirely explained by bias. We found some evidence of heterogeneity in our analyses that reached statistical significance for the pooled estimate of fatal events among observational studies (Figure 2). In this context, the use of random effects models is favoured over fixed effects models. Additionally we were able to identify sources of heterogeneity by fitting a metaregression model and exploring the effect of different study characteristics (such as the inclusion criteria in observational studies or the use of placebo in RCTs) on the pooled estimate. Overall, there were insufficient data to assess the risk of non-fatal and fatal MI among individual traditional NSAIDs, or whether there is any difference between traditional NSAIDs and coxibs. This should be the ultimate goal as it has been previously shown that individual NSAIDs at commonly used doses, present variability in their magnitude of risk of ischemic coronary disease [4], [8]. Also, despite our criteria of minimum total sample size and follow-up requirements the power of RCTs to study these rare events is quite limited due to the paucity of events. The inclusion of smaller RCTs with few or no events would provide very little additional information to our risk estimates.\nNSAIDs increase the risk of non-fatal MI but do not increase coronary heart disease mortality. Due to the widespread use of NSAIDs in the population, particularly amongst the elderly, further research is urgently needed to unravel the specific mechanisms involved in NSAID-associated coronary thrombosis (including genetic and epigenetic markers) and the identification of patients at highest risk of such complication. Additionally, studies are warranted to reduce the coronary risk associated with NSAID use." ]

[ null, "methods", null, null ]

[]

[ "Adenosine", "Adult", "Aged", "Aged, 80 and over", "Ammonia", "Coronary Circulation", "Female", "Forearm", "Humans", "Male", "Middle Aged", "Myocardial Perfusion Imaging", "Positron-Emission Tomography", "Regional Blood Flow", "Vasodilation" ]

[ "Study Population", "Acquisition Protocol", "Measurement of Myocardial Perfusion", "Measurement of Peripheral Perfusion", "Perfusion Reserve Calculation", "Statistical Analysis", "Hemodynamics", "Myocardial Perfusion (Reserve)", "Peripheral Perfusion (Reserve)", "Comparison Between the Different Groups", "Correlation Between Myocardial and Peripheral Blood Flow During Adenosine Infusion", "Correlation Between MPR and PPR" ]

[ "The study population comprised 58 subjects retrospectively, submitted for cardiac [13N]ammonia PET analysis in our centre between 1995 and 2003, whose scans allowed for calculation of perfusion in an appropriate area of the proximal upper limb. These patients were divided into four groups: (I) 15 patients (9 male, 6 female) with documented CAD, defined as wall irregularities on coronary angiogram; (II) 14 patients (4 males, 10 females) with SX, defined as typical chest pain, positive exercise ECG-changes and a normal coronary angiogram; (III) 16 patients (12 males, 4 females) with idiopathic DCM without prior evidence of CAD or myocardial infarction; and (IV) 13 healthy control subjects (2 males, 11 females) without cardiac or pulmonary disease.\nPatients’ current medication was continued, except for dipyridamole and diuretics due to interaction with adenosine and to prevent urinary urge during acquisition, respectively. Characteristics of each group are shown in Table 1.Table 1Patients’ characteristicsCADSXDCMControlN15141613Mean age (range)64 (50–81)55 (34–76)78 (45–91)58 (48–73)Use of β-blocker (%)100217554Use of ACE-inhibitor (%)6736818Use of vasodilator (%)13575038Use of statin (%)5343015Use of diuretics (%)27145614Known with diabetes (%)0768\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\n\nPatients’ characteristics\n\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy", "PET scans were performed on an ECAT EXACT HR + PET camera (Siemens Medical Systems, Knoxville, TN, USA). Positioning of the subjects was done with the aid of a rectilinear scan. Photon attenuation was measured using an external ring source filled with 68Ge/68Ga and data were automatically corrected for accidental coincidence and dead time. Patients were continuously monitored with 12-lead electrocardiography and blood pressure was obtained every 10 minutes. During adenosine-induced stress, blood pressure was measured every minute. Myocardial perfusion was measured using [13N]ammonia as a perfusion tracer and adenosine as the pharmacological stress agent. After injection of 400 MBq of [13N]ammonia a dynamic rest imaging was performed for 15 minutes. Adenosine stress imaging was performed identically after 140 μg/kg/min of adenosine infusion 3 minutes prior to and 3 minutes following the injection of 400 MBq of [13N]ammonia.", "The quantification of myocardial perfusion has been described previously.6 In summary, [13N]ammonia was injected intravenously over 30 seconds while acquisition of a dynamic sequence of images to obtain time-activity curves from the blood pool and from the myocardial tissue was started. 17 regions of interest (ROIs) were placed within the left ventricular myocardium in the three territories of the major coronary arteries. These ROIs were subsequently copied to the dynamic image sequence. In this way, myocardial tissue time-activity curves for [13N]ammonia were obtained. The arterial input function was obtained from a small ROI in the left ventricular blood pool. Reproducibility has been assessed previously.7 Myocardial perfusion was calculated by fitting the corrected tissue and blood pool time-activity curves to a validated 2-tissue-compartment model for [13N]ammonia using the Hutchins model.8\n", "For each study transmission images were used to draw six ROIs around the upper limb (opposite to the injected arm) in six consecutive slices, excluding the areas of highest density correlating with the location of bone. These ROIs were then copied to the dynamic sequences to obtain time-activity curves for [13N]ammonia in the peripheral vascular bed. This is illustrated in Figure 1. A small ROI in the left ventricular blood pool was used for the arterial input function. Peripheral perfusion was calculated by fitting the tissue and blood pool time-activity curves to the aforementioned 2-tissue-compartment model for [13N]ammonia. Since the myocardial perfusion and peripheral perfusion were calculated in the same studies, all circumstances (blood pressure, heart rate) were equal for both measurements.Figure 1Measurement of peripheral perfusion. Regions of interest (ROI) are drawn on the transmission dataset in the upper limb, excluding the area of the humerus. These ROI are copied to the perfusion dataset (large arrows) and the measured time/activity curve is fitted to the compartment model for [13N] ammonia. In the opposite arm the influx of activity from the infusion site is just visible (small arrow)\n\nMeasurement of peripheral perfusion. Regions of interest (ROI) are drawn on the transmission dataset in the upper limb, excluding the area of the humerus. These ROI are copied to the perfusion dataset (large arrows) and the measured time/activity curve is fitted to the compartment model for [13N] ammonia. In the opposite arm the influx of activity from the infusion site is just visible (small arrow)", "Myocardial and peripheral perfusion flow reserves were calculated by dividing the perfusion results of the adenosine [13N]ammonia stress study by the results of the [13N]ammonia rest study.", "The mentioned values are the mean values ± SD. standard error of estimates (SEE) was added in graphs. A paired t test or the nonparametric Wilcoxon signed rank test was used to compare individual values. Correlations were sought by standard linear regression. A P value <.05 was considered significant.", "The hemodynamic values are listed in Table 2 for the four patient groups. During adenosine infusion, there were no statistically significant changes in blood pressure in any of the groups. The heart rate and accordingly the rate-pressure product increased in all groups, though the increase was not statistically significant in the SX-group. The highest increase was noted in the DCM-group, showing a mean rise in heart rate of 50%, as opposed to the other groups which all showed less than 25% increase in heart rate.Table 2Hemodynamic valuesCADSXDCMControlHeart rate at rest per minute64 ± 1082 ± 1776 ± 965 ± 7Heart rate at stress per minute72 ± 11*85 ± 23114 ± 10*81 ± 14*Systolic blood pressure at rest (mmHg)131 ± 16159 ± 3102 ± 5132 ± 9Systolic blood pressure at stress (mmHg)136 ± 15144 ± 895 ± 6132 ± 13Diastolic blood pressure at rest (mmHg)75 ± 1189 ± 957 ± 678 ± 7Diastolic blood pressure at stress (mmHg)79 ± 885 ± 866 ± 776 ± 7Rate-pressure product at rest83401193977528526Rate-pressure product at stress9757*1216610830*10704*\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathyAll values mean ± SD* Statistically significant increase compared to rest (P < 0.05)\n\nHemodynamic values\n\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\nAll values mean ± SD\n* Statistically significant increase compared to rest (P < 0.05)", "The mean resting flow in the myocardium in mL/min/100 g per group was: CAD 60.3 ± 16.2; SX 83.4 ± 24.6; DCM 50.4 ± 9.8; and NC 70.3 ± 15.8. During adenosine infusion, the measurements were: CAD 112.8 ± 26.4; SX 115.8 ± 40.6; DCM 80.1 ± 36.9; and NC 184.7 ± 47.6. All these changes were highly significant (CAD: P = .007, SX: P = .002, DCM: P = .005, NC: P = .005). The resulting calculated MPR per group was: CAD 1.99 ± 0.47; SX 1.39 ± 0.31; DCM 1.72 ± 0.69; NC 2.91 ± 0.78. The changes in myocardial perfusion between rest and pharmacological stress are shown in Figure 2.Figure 2Myocardial perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; MBF, myocardial blood flow\n\nMyocardial perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; MBF, myocardial blood flow", "The mean resting flow in the peripheral vasculature in mL/min/100 g per group was: CAD 3.0 ± 1.5; SX 6.5 ± 3.4; DCM 1.6 ± 0.8; and NC 3.9 ± 1.5. During adenosine infusion, the measurements were: CAD 3.1 ± 2.1; SX 8.8 ± 6.5; DCM 2.2 ± 1.0; and NC 5.3 ± 3.4. None of these changes were significant, nor were the differences between the groups. The resulting calculated PPR per group was: CAD 1.30 ± 0.79; SX 1.36 ± 0.71; DCM 1.60 ± 1.22; NC 1.27 ± 0.63. No correlation was found with the rate-pressure product. The changes in peripheral perfusion between rest and pharmacological stress are shown in Figure 3.Figure 3Peripheral perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PBF, peripheral blood flow\n\nPeripheral perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PBF, peripheral blood flow", "Myocardial blood flow during adenosine infusion was significantly higher in normal controls compared to all other groups (P < .001 vs all groups), as was the MPR (P ≤ .001 vs all groups).\nMyocardial blood flow was significantly higher in SX compared to DCM (P = .03) and in CAD compared to DCM (P = .02). The difference between CAD and SX was not significant.\nThe MPR was significantly higher in CAD compared to SX (P = .001) whereas all other groups comparisons showed no significant differences.\nPeripheral blood flow during adenosine infusion was significantly higher when comparing NC to DCM (P = .015), SX to CAD (P = .014), and SX to DCM (P = .005), other comparisons showed no significant differences.\nThere were no significant differences in PPR between any of the groups.", "No significant correlations were found between the two circulatory beds during adenosine infusion for the group as a whole (r = 0.13, SEE = 54.32, P = .36) or any of the subgroups (CAD: r = 0.03, SEE = 1.46; SX: r = 0.16, SEE = 43.57; DCM: r = −0.35, SEE = 37.15; NC: r = −0.31, SEE = 49.57).", "The correlations between the myocardial perfusion reserve and the peripheral perfusion reserve in the separate groups are shown in Figure 4. No correlations were found in the whole group (r = −0.07, SEE = 0.70, P = .47), or for any of the four separate groups (values provided in Figure 4). Figure 5 shows the Bland-Altman plot for the comparison in the whole group.Figure 4Correlations between MPR and PPR. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PPR, peripheral perfusion reserve; MPR, myocardial perfusion reserve\nFigure 5Bland-Altman plot for MPR/PPR comparison\n\nCorrelations between MPR and PPR. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PPR, peripheral perfusion reserve; MPR, myocardial perfusion reserve\nBland-Altman plot for MPR/PPR comparison" ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Methods", "Study Population", "Acquisition Protocol", "Measurement of Myocardial Perfusion", "Measurement of Peripheral Perfusion", "Perfusion Reserve Calculation", "Statistical Analysis", "Results", "Hemodynamics", "Myocardial Perfusion (Reserve)", "Peripheral Perfusion (Reserve)", "Comparison Between the Different Groups", "Correlation Between Myocardial and Peripheral Blood Flow During Adenosine Infusion", "Correlation Between MPR and PPR", "Discussion", "Conclusion" ]

[ "Evaluation of systemic arterial dilatory function, mostly by ultrasonography, is a widely available technique. Its findings have been shown to predict coronary artery disease.1 For the microvascular component of myocardial and peripheral perfusion, these correlations are less obvious and probably different.2 Patients with coronary artery disease (CAD) risk factors such as hypercholesterolemia and diabetes mellitus exhibit not only large-vessel dysfunction but also microvascular dysfunction, e.g., reduced responses to vasoactive substances such as adenosine and dipyridamole.3 PET is an accurate technique for the quantification of myocardial perfusion reserve (MPR). An earlier PET study showed no correlation between Doppler ultrasound flow measurement in the forearm and dipyridamole-induced myocardial hyperemia,2 whereas another study comparing acetylcholine-induced myocardial hyperemia to brachial artery dilation did find some weak correlation.4 A possible limitation of these previous studies is the differences in technique between the myocardial and the peripheral measurements. The microvascular component of myocardial and peripheral perfusion may however differ also. To evaluate this possible limitation we studied the peripheral and myocardial perfusion in rest [13N]ammonia PET and during adenosine stress [13N]ammonia PET and calculated the MPR and peripheral perfusion reserve (PPR) in patients with documented coronary artery disease (CAD), microcardiovascular disease (cardiac syndrome X, SX), idiopathic dilating cardiomyopathy (DCM), and normal controls (NC). By including three separate clinical entities we intended also to compare the differences between atherosclerotic macro- and microvascular disease (CAD) and microvascular disease without documented macrovascular disease as found in cardiac syndrome X and idiopathic DCM and ascertain whether peripheral perfusion responses were more strongly linked to any of the underlying pathologies. Perfusion reserve is an important parameter for prognosis in patients with cardiovascular disease5 and will be used in this study.\nThe aim of this study is to evaluate the correlation between the myocardial and the peripheral perfusion reserve when tested during the same PET perfusion session.", "[SUBTITLE] Study Population [SUBSECTION] The study population comprised 58 subjects retrospectively, submitted for cardiac [13N]ammonia PET analysis in our centre between 1995 and 2003, whose scans allowed for calculation of perfusion in an appropriate area of the proximal upper limb. These patients were divided into four groups: (I) 15 patients (9 male, 6 female) with documented CAD, defined as wall irregularities on coronary angiogram; (II) 14 patients (4 males, 10 females) with SX, defined as typical chest pain, positive exercise ECG-changes and a normal coronary angiogram; (III) 16 patients (12 males, 4 females) with idiopathic DCM without prior evidence of CAD or myocardial infarction; and (IV) 13 healthy control subjects (2 males, 11 females) without cardiac or pulmonary disease.\nPatients’ current medication was continued, except for dipyridamole and diuretics due to interaction with adenosine and to prevent urinary urge during acquisition, respectively. Characteristics of each group are shown in Table 1.Table 1Patients’ characteristicsCADSXDCMControlN15141613Mean age (range)64 (50–81)55 (34–76)78 (45–91)58 (48–73)Use of β-blocker (%)100217554Use of ACE-inhibitor (%)6736818Use of vasodilator (%)13575038Use of statin (%)5343015Use of diuretics (%)27145614Known with diabetes (%)0768\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\n\nPatients’ characteristics\n\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\nThe study population comprised 58 subjects retrospectively, submitted for cardiac [13N]ammonia PET analysis in our centre between 1995 and 2003, whose scans allowed for calculation of perfusion in an appropriate area of the proximal upper limb. These patients were divided into four groups: (I) 15 patients (9 male, 6 female) with documented CAD, defined as wall irregularities on coronary angiogram; (II) 14 patients (4 males, 10 females) with SX, defined as typical chest pain, positive exercise ECG-changes and a normal coronary angiogram; (III) 16 patients (12 males, 4 females) with idiopathic DCM without prior evidence of CAD or myocardial infarction; and (IV) 13 healthy control subjects (2 males, 11 females) without cardiac or pulmonary disease.\nPatients’ current medication was continued, except for dipyridamole and diuretics due to interaction with adenosine and to prevent urinary urge during acquisition, respectively. Characteristics of each group are shown in Table 1.Table 1Patients’ characteristicsCADSXDCMControlN15141613Mean age (range)64 (50–81)55 (34–76)78 (45–91)58 (48–73)Use of β-blocker (%)100217554Use of ACE-inhibitor (%)6736818Use of vasodilator (%)13575038Use of statin (%)5343015Use of diuretics (%)27145614Known with diabetes (%)0768\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\n\nPatients’ characteristics\n\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\n[SUBTITLE] Acquisition Protocol [SUBSECTION] PET scans were performed on an ECAT EXACT HR + PET camera (Siemens Medical Systems, Knoxville, TN, USA). Positioning of the subjects was done with the aid of a rectilinear scan. Photon attenuation was measured using an external ring source filled with 68Ge/68Ga and data were automatically corrected for accidental coincidence and dead time. Patients were continuously monitored with 12-lead electrocardiography and blood pressure was obtained every 10 minutes. During adenosine-induced stress, blood pressure was measured every minute. Myocardial perfusion was measured using [13N]ammonia as a perfusion tracer and adenosine as the pharmacological stress agent. After injection of 400 MBq of [13N]ammonia a dynamic rest imaging was performed for 15 minutes. Adenosine stress imaging was performed identically after 140 μg/kg/min of adenosine infusion 3 minutes prior to and 3 minutes following the injection of 400 MBq of [13N]ammonia.\nPET scans were performed on an ECAT EXACT HR + PET camera (Siemens Medical Systems, Knoxville, TN, USA). Positioning of the subjects was done with the aid of a rectilinear scan. Photon attenuation was measured using an external ring source filled with 68Ge/68Ga and data were automatically corrected for accidental coincidence and dead time. Patients were continuously monitored with 12-lead electrocardiography and blood pressure was obtained every 10 minutes. During adenosine-induced stress, blood pressure was measured every minute. Myocardial perfusion was measured using [13N]ammonia as a perfusion tracer and adenosine as the pharmacological stress agent. After injection of 400 MBq of [13N]ammonia a dynamic rest imaging was performed for 15 minutes. Adenosine stress imaging was performed identically after 140 μg/kg/min of adenosine infusion 3 minutes prior to and 3 minutes following the injection of 400 MBq of [13N]ammonia.\n[SUBTITLE] Measurement of Myocardial Perfusion [SUBSECTION] The quantification of myocardial perfusion has been described previously.6 In summary, [13N]ammonia was injected intravenously over 30 seconds while acquisition of a dynamic sequence of images to obtain time-activity curves from the blood pool and from the myocardial tissue was started. 17 regions of interest (ROIs) were placed within the left ventricular myocardium in the three territories of the major coronary arteries. These ROIs were subsequently copied to the dynamic image sequence. In this way, myocardial tissue time-activity curves for [13N]ammonia were obtained. The arterial input function was obtained from a small ROI in the left ventricular blood pool. Reproducibility has been assessed previously.7 Myocardial perfusion was calculated by fitting the corrected tissue and blood pool time-activity curves to a validated 2-tissue-compartment model for [13N]ammonia using the Hutchins model.8\n\nThe quantification of myocardial perfusion has been described previously.6 In summary, [13N]ammonia was injected intravenously over 30 seconds while acquisition of a dynamic sequence of images to obtain time-activity curves from the blood pool and from the myocardial tissue was started. 17 regions of interest (ROIs) were placed within the left ventricular myocardium in the three territories of the major coronary arteries. These ROIs were subsequently copied to the dynamic image sequence. In this way, myocardial tissue time-activity curves for [13N]ammonia were obtained. The arterial input function was obtained from a small ROI in the left ventricular blood pool. Reproducibility has been assessed previously.7 Myocardial perfusion was calculated by fitting the corrected tissue and blood pool time-activity curves to a validated 2-tissue-compartment model for [13N]ammonia using the Hutchins model.8\n\n[SUBTITLE] Measurement of Peripheral Perfusion [SUBSECTION] For each study transmission images were used to draw six ROIs around the upper limb (opposite to the injected arm) in six consecutive slices, excluding the areas of highest density correlating with the location of bone. These ROIs were then copied to the dynamic sequences to obtain time-activity curves for [13N]ammonia in the peripheral vascular bed. This is illustrated in Figure 1. A small ROI in the left ventricular blood pool was used for the arterial input function. Peripheral perfusion was calculated by fitting the tissue and blood pool time-activity curves to the aforementioned 2-tissue-compartment model for [13N]ammonia. Since the myocardial perfusion and peripheral perfusion were calculated in the same studies, all circumstances (blood pressure, heart rate) were equal for both measurements.Figure 1Measurement of peripheral perfusion. Regions of interest (ROI) are drawn on the transmission dataset in the upper limb, excluding the area of the humerus. These ROI are copied to the perfusion dataset (large arrows) and the measured time/activity curve is fitted to the compartment model for [13N] ammonia. In the opposite arm the influx of activity from the infusion site is just visible (small arrow)\n\nMeasurement of peripheral perfusion. Regions of interest (ROI) are drawn on the transmission dataset in the upper limb, excluding the area of the humerus. These ROI are copied to the perfusion dataset (large arrows) and the measured time/activity curve is fitted to the compartment model for [13N] ammonia. In the opposite arm the influx of activity from the infusion site is just visible (small arrow)\nFor each study transmission images were used to draw six ROIs around the upper limb (opposite to the injected arm) in six consecutive slices, excluding the areas of highest density correlating with the location of bone. These ROIs were then copied to the dynamic sequences to obtain time-activity curves for [13N]ammonia in the peripheral vascular bed. This is illustrated in Figure 1. A small ROI in the left ventricular blood pool was used for the arterial input function. Peripheral perfusion was calculated by fitting the tissue and blood pool time-activity curves to the aforementioned 2-tissue-compartment model for [13N]ammonia. Since the myocardial perfusion and peripheral perfusion were calculated in the same studies, all circumstances (blood pressure, heart rate) were equal for both measurements.Figure 1Measurement of peripheral perfusion. Regions of interest (ROI) are drawn on the transmission dataset in the upper limb, excluding the area of the humerus. These ROI are copied to the perfusion dataset (large arrows) and the measured time/activity curve is fitted to the compartment model for [13N] ammonia. In the opposite arm the influx of activity from the infusion site is just visible (small arrow)\n\nMeasurement of peripheral perfusion. Regions of interest (ROI) are drawn on the transmission dataset in the upper limb, excluding the area of the humerus. These ROI are copied to the perfusion dataset (large arrows) and the measured time/activity curve is fitted to the compartment model for [13N] ammonia. In the opposite arm the influx of activity from the infusion site is just visible (small arrow)\n[SUBTITLE] Perfusion Reserve Calculation [SUBSECTION] Myocardial and peripheral perfusion flow reserves were calculated by dividing the perfusion results of the adenosine [13N]ammonia stress study by the results of the [13N]ammonia rest study.\nMyocardial and peripheral perfusion flow reserves were calculated by dividing the perfusion results of the adenosine [13N]ammonia stress study by the results of the [13N]ammonia rest study.\n[SUBTITLE] Statistical Analysis [SUBSECTION] The mentioned values are the mean values ± SD. standard error of estimates (SEE) was added in graphs. A paired t test or the nonparametric Wilcoxon signed rank test was used to compare individual values. Correlations were sought by standard linear regression. A P value <.05 was considered significant.\nThe mentioned values are the mean values ± SD. standard error of estimates (SEE) was added in graphs. A paired t test or the nonparametric Wilcoxon signed rank test was used to compare individual values. Correlations were sought by standard linear regression. A P value <.05 was considered significant.", "The study population comprised 58 subjects retrospectively, submitted for cardiac [13N]ammonia PET analysis in our centre between 1995 and 2003, whose scans allowed for calculation of perfusion in an appropriate area of the proximal upper limb. These patients were divided into four groups: (I) 15 patients (9 male, 6 female) with documented CAD, defined as wall irregularities on coronary angiogram; (II) 14 patients (4 males, 10 females) with SX, defined as typical chest pain, positive exercise ECG-changes and a normal coronary angiogram; (III) 16 patients (12 males, 4 females) with idiopathic DCM without prior evidence of CAD or myocardial infarction; and (IV) 13 healthy control subjects (2 males, 11 females) without cardiac or pulmonary disease.\nPatients’ current medication was continued, except for dipyridamole and diuretics due to interaction with adenosine and to prevent urinary urge during acquisition, respectively. Characteristics of each group are shown in Table 1.Table 1Patients’ characteristicsCADSXDCMControlN15141613Mean age (range)64 (50–81)55 (34–76)78 (45–91)58 (48–73)Use of β-blocker (%)100217554Use of ACE-inhibitor (%)6736818Use of vasodilator (%)13575038Use of statin (%)5343015Use of diuretics (%)27145614Known with diabetes (%)0768\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\n\nPatients’ characteristics\n\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy", "PET scans were performed on an ECAT EXACT HR + PET camera (Siemens Medical Systems, Knoxville, TN, USA). Positioning of the subjects was done with the aid of a rectilinear scan. Photon attenuation was measured using an external ring source filled with 68Ge/68Ga and data were automatically corrected for accidental coincidence and dead time. Patients were continuously monitored with 12-lead electrocardiography and blood pressure was obtained every 10 minutes. During adenosine-induced stress, blood pressure was measured every minute. Myocardial perfusion was measured using [13N]ammonia as a perfusion tracer and adenosine as the pharmacological stress agent. After injection of 400 MBq of [13N]ammonia a dynamic rest imaging was performed for 15 minutes. Adenosine stress imaging was performed identically after 140 μg/kg/min of adenosine infusion 3 minutes prior to and 3 minutes following the injection of 400 MBq of [13N]ammonia.", "The quantification of myocardial perfusion has been described previously.6 In summary, [13N]ammonia was injected intravenously over 30 seconds while acquisition of a dynamic sequence of images to obtain time-activity curves from the blood pool and from the myocardial tissue was started. 17 regions of interest (ROIs) were placed within the left ventricular myocardium in the three territories of the major coronary arteries. These ROIs were subsequently copied to the dynamic image sequence. In this way, myocardial tissue time-activity curves for [13N]ammonia were obtained. The arterial input function was obtained from a small ROI in the left ventricular blood pool. Reproducibility has been assessed previously.7 Myocardial perfusion was calculated by fitting the corrected tissue and blood pool time-activity curves to a validated 2-tissue-compartment model for [13N]ammonia using the Hutchins model.8\n", "For each study transmission images were used to draw six ROIs around the upper limb (opposite to the injected arm) in six consecutive slices, excluding the areas of highest density correlating with the location of bone. These ROIs were then copied to the dynamic sequences to obtain time-activity curves for [13N]ammonia in the peripheral vascular bed. This is illustrated in Figure 1. A small ROI in the left ventricular blood pool was used for the arterial input function. Peripheral perfusion was calculated by fitting the tissue and blood pool time-activity curves to the aforementioned 2-tissue-compartment model for [13N]ammonia. Since the myocardial perfusion and peripheral perfusion were calculated in the same studies, all circumstances (blood pressure, heart rate) were equal for both measurements.Figure 1Measurement of peripheral perfusion. Regions of interest (ROI) are drawn on the transmission dataset in the upper limb, excluding the area of the humerus. These ROI are copied to the perfusion dataset (large arrows) and the measured time/activity curve is fitted to the compartment model for [13N] ammonia. In the opposite arm the influx of activity from the infusion site is just visible (small arrow)\n\nMeasurement of peripheral perfusion. Regions of interest (ROI) are drawn on the transmission dataset in the upper limb, excluding the area of the humerus. These ROI are copied to the perfusion dataset (large arrows) and the measured time/activity curve is fitted to the compartment model for [13N] ammonia. In the opposite arm the influx of activity from the infusion site is just visible (small arrow)", "Myocardial and peripheral perfusion flow reserves were calculated by dividing the perfusion results of the adenosine [13N]ammonia stress study by the results of the [13N]ammonia rest study.", "The mentioned values are the mean values ± SD. standard error of estimates (SEE) was added in graphs. A paired t test or the nonparametric Wilcoxon signed rank test was used to compare individual values. Correlations were sought by standard linear regression. A P value <.05 was considered significant.", "[SUBTITLE] Hemodynamics [SUBSECTION] The hemodynamic values are listed in Table 2 for the four patient groups. During adenosine infusion, there were no statistically significant changes in blood pressure in any of the groups. The heart rate and accordingly the rate-pressure product increased in all groups, though the increase was not statistically significant in the SX-group. The highest increase was noted in the DCM-group, showing a mean rise in heart rate of 50%, as opposed to the other groups which all showed less than 25% increase in heart rate.Table 2Hemodynamic valuesCADSXDCMControlHeart rate at rest per minute64 ± 1082 ± 1776 ± 965 ± 7Heart rate at stress per minute72 ± 11*85 ± 23114 ± 10*81 ± 14*Systolic blood pressure at rest (mmHg)131 ± 16159 ± 3102 ± 5132 ± 9Systolic blood pressure at stress (mmHg)136 ± 15144 ± 895 ± 6132 ± 13Diastolic blood pressure at rest (mmHg)75 ± 1189 ± 957 ± 678 ± 7Diastolic blood pressure at stress (mmHg)79 ± 885 ± 866 ± 776 ± 7Rate-pressure product at rest83401193977528526Rate-pressure product at stress9757*1216610830*10704*\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathyAll values mean ± SD* Statistically significant increase compared to rest (P < 0.05)\n\nHemodynamic values\n\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\nAll values mean ± SD\n* Statistically significant increase compared to rest (P < 0.05)\nThe hemodynamic values are listed in Table 2 for the four patient groups. During adenosine infusion, there were no statistically significant changes in blood pressure in any of the groups. The heart rate and accordingly the rate-pressure product increased in all groups, though the increase was not statistically significant in the SX-group. The highest increase was noted in the DCM-group, showing a mean rise in heart rate of 50%, as opposed to the other groups which all showed less than 25% increase in heart rate.Table 2Hemodynamic valuesCADSXDCMControlHeart rate at rest per minute64 ± 1082 ± 1776 ± 965 ± 7Heart rate at stress per minute72 ± 11*85 ± 23114 ± 10*81 ± 14*Systolic blood pressure at rest (mmHg)131 ± 16159 ± 3102 ± 5132 ± 9Systolic blood pressure at stress (mmHg)136 ± 15144 ± 895 ± 6132 ± 13Diastolic blood pressure at rest (mmHg)75 ± 1189 ± 957 ± 678 ± 7Diastolic blood pressure at stress (mmHg)79 ± 885 ± 866 ± 776 ± 7Rate-pressure product at rest83401193977528526Rate-pressure product at stress9757*1216610830*10704*\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathyAll values mean ± SD* Statistically significant increase compared to rest (P < 0.05)\n\nHemodynamic values\n\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\nAll values mean ± SD\n* Statistically significant increase compared to rest (P < 0.05)\n[SUBTITLE] Myocardial Perfusion (Reserve) [SUBSECTION] The mean resting flow in the myocardium in mL/min/100 g per group was: CAD 60.3 ± 16.2; SX 83.4 ± 24.6; DCM 50.4 ± 9.8; and NC 70.3 ± 15.8. During adenosine infusion, the measurements were: CAD 112.8 ± 26.4; SX 115.8 ± 40.6; DCM 80.1 ± 36.9; and NC 184.7 ± 47.6. All these changes were highly significant (CAD: P = .007, SX: P = .002, DCM: P = .005, NC: P = .005). The resulting calculated MPR per group was: CAD 1.99 ± 0.47; SX 1.39 ± 0.31; DCM 1.72 ± 0.69; NC 2.91 ± 0.78. The changes in myocardial perfusion between rest and pharmacological stress are shown in Figure 2.Figure 2Myocardial perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; MBF, myocardial blood flow\n\nMyocardial perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; MBF, myocardial blood flow\nThe mean resting flow in the myocardium in mL/min/100 g per group was: CAD 60.3 ± 16.2; SX 83.4 ± 24.6; DCM 50.4 ± 9.8; and NC 70.3 ± 15.8. During adenosine infusion, the measurements were: CAD 112.8 ± 26.4; SX 115.8 ± 40.6; DCM 80.1 ± 36.9; and NC 184.7 ± 47.6. All these changes were highly significant (CAD: P = .007, SX: P = .002, DCM: P = .005, NC: P = .005). The resulting calculated MPR per group was: CAD 1.99 ± 0.47; SX 1.39 ± 0.31; DCM 1.72 ± 0.69; NC 2.91 ± 0.78. The changes in myocardial perfusion between rest and pharmacological stress are shown in Figure 2.Figure 2Myocardial perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; MBF, myocardial blood flow\n\nMyocardial perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; MBF, myocardial blood flow\n[SUBTITLE] Peripheral Perfusion (Reserve) [SUBSECTION] The mean resting flow in the peripheral vasculature in mL/min/100 g per group was: CAD 3.0 ± 1.5; SX 6.5 ± 3.4; DCM 1.6 ± 0.8; and NC 3.9 ± 1.5. During adenosine infusion, the measurements were: CAD 3.1 ± 2.1; SX 8.8 ± 6.5; DCM 2.2 ± 1.0; and NC 5.3 ± 3.4. None of these changes were significant, nor were the differences between the groups. The resulting calculated PPR per group was: CAD 1.30 ± 0.79; SX 1.36 ± 0.71; DCM 1.60 ± 1.22; NC 1.27 ± 0.63. No correlation was found with the rate-pressure product. The changes in peripheral perfusion between rest and pharmacological stress are shown in Figure 3.Figure 3Peripheral perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PBF, peripheral blood flow\n\nPeripheral perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PBF, peripheral blood flow\nThe mean resting flow in the peripheral vasculature in mL/min/100 g per group was: CAD 3.0 ± 1.5; SX 6.5 ± 3.4; DCM 1.6 ± 0.8; and NC 3.9 ± 1.5. During adenosine infusion, the measurements were: CAD 3.1 ± 2.1; SX 8.8 ± 6.5; DCM 2.2 ± 1.0; and NC 5.3 ± 3.4. None of these changes were significant, nor were the differences between the groups. The resulting calculated PPR per group was: CAD 1.30 ± 0.79; SX 1.36 ± 0.71; DCM 1.60 ± 1.22; NC 1.27 ± 0.63. No correlation was found with the rate-pressure product. The changes in peripheral perfusion between rest and pharmacological stress are shown in Figure 3.Figure 3Peripheral perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PBF, peripheral blood flow\n\nPeripheral perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PBF, peripheral blood flow\n[SUBTITLE] Comparison Between the Different Groups [SUBSECTION] Myocardial blood flow during adenosine infusion was significantly higher in normal controls compared to all other groups (P < .001 vs all groups), as was the MPR (P ≤ .001 vs all groups).\nMyocardial blood flow was significantly higher in SX compared to DCM (P = .03) and in CAD compared to DCM (P = .02). The difference between CAD and SX was not significant.\nThe MPR was significantly higher in CAD compared to SX (P = .001) whereas all other groups comparisons showed no significant differences.\nPeripheral blood flow during adenosine infusion was significantly higher when comparing NC to DCM (P = .015), SX to CAD (P = .014), and SX to DCM (P = .005), other comparisons showed no significant differences.\nThere were no significant differences in PPR between any of the groups.\nMyocardial blood flow during adenosine infusion was significantly higher in normal controls compared to all other groups (P < .001 vs all groups), as was the MPR (P ≤ .001 vs all groups).\nMyocardial blood flow was significantly higher in SX compared to DCM (P = .03) and in CAD compared to DCM (P = .02). The difference between CAD and SX was not significant.\nThe MPR was significantly higher in CAD compared to SX (P = .001) whereas all other groups comparisons showed no significant differences.\nPeripheral blood flow during adenosine infusion was significantly higher when comparing NC to DCM (P = .015), SX to CAD (P = .014), and SX to DCM (P = .005), other comparisons showed no significant differences.\nThere were no significant differences in PPR between any of the groups.\n[SUBTITLE] Correlation Between Myocardial and Peripheral Blood Flow During Adenosine Infusion [SUBSECTION] No significant correlations were found between the two circulatory beds during adenosine infusion for the group as a whole (r = 0.13, SEE = 54.32, P = .36) or any of the subgroups (CAD: r = 0.03, SEE = 1.46; SX: r = 0.16, SEE = 43.57; DCM: r = −0.35, SEE = 37.15; NC: r = −0.31, SEE = 49.57).\nNo significant correlations were found between the two circulatory beds during adenosine infusion for the group as a whole (r = 0.13, SEE = 54.32, P = .36) or any of the subgroups (CAD: r = 0.03, SEE = 1.46; SX: r = 0.16, SEE = 43.57; DCM: r = −0.35, SEE = 37.15; NC: r = −0.31, SEE = 49.57).\n[SUBTITLE] Correlation Between MPR and PPR [SUBSECTION] The correlations between the myocardial perfusion reserve and the peripheral perfusion reserve in the separate groups are shown in Figure 4. No correlations were found in the whole group (r = −0.07, SEE = 0.70, P = .47), or for any of the four separate groups (values provided in Figure 4). Figure 5 shows the Bland-Altman plot for the comparison in the whole group.Figure 4Correlations between MPR and PPR. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PPR, peripheral perfusion reserve; MPR, myocardial perfusion reserve\nFigure 5Bland-Altman plot for MPR/PPR comparison\n\nCorrelations between MPR and PPR. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PPR, peripheral perfusion reserve; MPR, myocardial perfusion reserve\nBland-Altman plot for MPR/PPR comparison\nThe correlations between the myocardial perfusion reserve and the peripheral perfusion reserve in the separate groups are shown in Figure 4. No correlations were found in the whole group (r = −0.07, SEE = 0.70, P = .47), or for any of the four separate groups (values provided in Figure 4). Figure 5 shows the Bland-Altman plot for the comparison in the whole group.Figure 4Correlations between MPR and PPR. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PPR, peripheral perfusion reserve; MPR, myocardial perfusion reserve\nFigure 5Bland-Altman plot for MPR/PPR comparison\n\nCorrelations between MPR and PPR. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PPR, peripheral perfusion reserve; MPR, myocardial perfusion reserve\nBland-Altman plot for MPR/PPR comparison", "The hemodynamic values are listed in Table 2 for the four patient groups. During adenosine infusion, there were no statistically significant changes in blood pressure in any of the groups. The heart rate and accordingly the rate-pressure product increased in all groups, though the increase was not statistically significant in the SX-group. The highest increase was noted in the DCM-group, showing a mean rise in heart rate of 50%, as opposed to the other groups which all showed less than 25% increase in heart rate.Table 2Hemodynamic valuesCADSXDCMControlHeart rate at rest per minute64 ± 1082 ± 1776 ± 965 ± 7Heart rate at stress per minute72 ± 11*85 ± 23114 ± 10*81 ± 14*Systolic blood pressure at rest (mmHg)131 ± 16159 ± 3102 ± 5132 ± 9Systolic blood pressure at stress (mmHg)136 ± 15144 ± 895 ± 6132 ± 13Diastolic blood pressure at rest (mmHg)75 ± 1189 ± 957 ± 678 ± 7Diastolic blood pressure at stress (mmHg)79 ± 885 ± 866 ± 776 ± 7Rate-pressure product at rest83401193977528526Rate-pressure product at stress9757*1216610830*10704*\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathyAll values mean ± SD* Statistically significant increase compared to rest (P < 0.05)\n\nHemodynamic values\n\nCAD, Coronary artery disease; SX, syndrome X, DCM, dilating cardiomyopathy\nAll values mean ± SD\n* Statistically significant increase compared to rest (P < 0.05)", "The mean resting flow in the myocardium in mL/min/100 g per group was: CAD 60.3 ± 16.2; SX 83.4 ± 24.6; DCM 50.4 ± 9.8; and NC 70.3 ± 15.8. During adenosine infusion, the measurements were: CAD 112.8 ± 26.4; SX 115.8 ± 40.6; DCM 80.1 ± 36.9; and NC 184.7 ± 47.6. All these changes were highly significant (CAD: P = .007, SX: P = .002, DCM: P = .005, NC: P = .005). The resulting calculated MPR per group was: CAD 1.99 ± 0.47; SX 1.39 ± 0.31; DCM 1.72 ± 0.69; NC 2.91 ± 0.78. The changes in myocardial perfusion between rest and pharmacological stress are shown in Figure 2.Figure 2Myocardial perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; MBF, myocardial blood flow\n\nMyocardial perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; MBF, myocardial blood flow", "The mean resting flow in the peripheral vasculature in mL/min/100 g per group was: CAD 3.0 ± 1.5; SX 6.5 ± 3.4; DCM 1.6 ± 0.8; and NC 3.9 ± 1.5. During adenosine infusion, the measurements were: CAD 3.1 ± 2.1; SX 8.8 ± 6.5; DCM 2.2 ± 1.0; and NC 5.3 ± 3.4. None of these changes were significant, nor were the differences between the groups. The resulting calculated PPR per group was: CAD 1.30 ± 0.79; SX 1.36 ± 0.71; DCM 1.60 ± 1.22; NC 1.27 ± 0.63. No correlation was found with the rate-pressure product. The changes in peripheral perfusion between rest and pharmacological stress are shown in Figure 3.Figure 3Peripheral perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PBF, peripheral blood flow\n\nPeripheral perfusion at rest vs during adenosine stress. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PBF, peripheral blood flow", "Myocardial blood flow during adenosine infusion was significantly higher in normal controls compared to all other groups (P < .001 vs all groups), as was the MPR (P ≤ .001 vs all groups).\nMyocardial blood flow was significantly higher in SX compared to DCM (P = .03) and in CAD compared to DCM (P = .02). The difference between CAD and SX was not significant.\nThe MPR was significantly higher in CAD compared to SX (P = .001) whereas all other groups comparisons showed no significant differences.\nPeripheral blood flow during adenosine infusion was significantly higher when comparing NC to DCM (P = .015), SX to CAD (P = .014), and SX to DCM (P = .005), other comparisons showed no significant differences.\nThere were no significant differences in PPR between any of the groups.", "No significant correlations were found between the two circulatory beds during adenosine infusion for the group as a whole (r = 0.13, SEE = 54.32, P = .36) or any of the subgroups (CAD: r = 0.03, SEE = 1.46; SX: r = 0.16, SEE = 43.57; DCM: r = −0.35, SEE = 37.15; NC: r = −0.31, SEE = 49.57).", "The correlations between the myocardial perfusion reserve and the peripheral perfusion reserve in the separate groups are shown in Figure 4. No correlations were found in the whole group (r = −0.07, SEE = 0.70, P = .47), or for any of the four separate groups (values provided in Figure 4). Figure 5 shows the Bland-Altman plot for the comparison in the whole group.Figure 4Correlations between MPR and PPR. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PPR, peripheral perfusion reserve; MPR, myocardial perfusion reserve\nFigure 5Bland-Altman plot for MPR/PPR comparison\n\nCorrelations between MPR and PPR. CAD, Coronary artery disease; NC, normal controls; SX, syndrome X; DCM, dilating cardiomyopathy; PPR, peripheral perfusion reserve; MPR, myocardial perfusion reserve\nBland-Altman plot for MPR/PPR comparison", "In our study, we evaluated the effects of vasodilator-stress between the two vascular beds in various underlying cardiovascular pathologies.\nOur data showed that during the same dose of i.v. adenosine infusion the response in perfusion in the heart and the musculature of the upper limb is not related. Although we noted a minor increase in blood flow in the upper limb, this response was not statistically significant and showed no differences between normal controls and the three patient groups.\nThere has been an interest in the past in the correlation between abnormalities in the coronary and the peripheral vasculature. Endothelial dysfunction has been linked to atherosclerotic risk factors, paving the way for ultrasound-based assessment of endothelial dysfunction. It has been shown that endothelial dysfunction in the brachial artery has a predictive value for coronary artery disease.1\n\nThe resting myocardial perfusion values and the hemodynamic response to adenosine infusion in our patient groups were comparable to other findings,9-11 even though changes in the SX-group were not statistically significant. The myocardial blood flow response to vasodilator-stress in our four patient groups was also comparable to other studies.3,5,12,13 As expected, coronary perfusion increased more than 2-fold in normal controls, whereas in the three patient groups with vascular disease these diseases blocked an adequate increase in blood flow. In the group with CAD, inadequate blood flow response was more localized due to the relatively focal lesion of the vasculopathy when compared to the SX and DCM groups. As the blood flow response is normal in non-atherosclerotic vessels in patients with CAD, in contrast to generally diseased myocardium in SX and DCM, the global MPR in CAD will be less affected. The relatively low MPR found in the SX group is at least partially based on a higher blood flow in rest, presumably due to the inherently higher metabolic rate in these patients. A higher resting flow further decreases the ratio between rest and stress flow measurements, explaining why the MPR in this group was lower than in any of the other groups even though the blood flow during adenosine infusion was comparable to that of the CAD group.\nThe resting blood flow in the peripheral circulation, in our case mostly in the biceps and triceps muscles, was also comparable to findings in different peripheral muscles in other studies.14,15\n\nA major difference between our study and previous described literature about the relation between myocardial and peripheral blood flow is the used method of a single session adenosine infusion. Most other studies used ischemia-driven blood flow increase in the upper limb after a period of occlusion, usually by means of an inflatable cuff, coupled with hand exercises.\nThough the vasodilative effects of adenosine have been reported over 8 decades ago,16 the precise mechanism through which it induces vasodilation is still not completely understood. Biaggioni wrote a commentary17 where he listed a number of mechanisms that may play a role in the vasodilative effects of adenosine, pointing out that the findings of various studies on the subject are contradictory. In an overview article on the role of PET in the understanding of coronary physiology, Schindler et al18 describe the mechanism of dipyridamole and adenosine as being a mixture of predominantly endothelium-independent flow responses, with shear sensitive endothelial components adding to these responses. Adenosine-induced hyperemia is therefore partially endothelium-independent and partially endothelium-dependent.\nIt is therefore possible that the responses elicited with intravenous adenosine differ significantly from other vasodilative stimuli such as hypoxia (mostly myogenic with supposed added effects from prostaglandins and adenosine), acetylcholine (fully endothelium dependent), or papaverine (smooth muscle relaxation) due to the different pathways that govern each of these responses, and that these responses in turn differ between the coronary and the systemic circulation. Anderson et al4 did find some weak correlation between myocardial vessel response to acetylcholine and flow-mediated brachial artery diameter measurements, though with significant scatter in their findings. In this study, sublingual nitroglycerin, a non-endothelium dependent vasodilator, did not show the same effect as reactive hyperemia, showing the variability in response to differing vasodilators.\nEven dipyridamol elicits a different response to purely intravenous adenosine as it increases not only the intravenous but also the interstitial levels of adenosine. This affects the peripheral vasculature in a different way than purely intravenous administration of adenosine does, as adenosine concentrations are raised concomitantly on both sides of the vascular wall.\nIn conscious subjects, systemic infusion of adenosine has been shown to stimulate sympathetic tone, most probably due to stimulation of arterial chemoreceptors.19 Whether this affects flow in myocardial and peripheral vascular beds differently is not known.\nOur data do suggest a possible explanation for the differences between the two perfusion reserves, in that resting flow in the myocardial vasculature is about 20 times higher than in the periphery. As the total delivered dose of adenosine is flow-dependent, the dose delivered to the myocardial vessels will also be about 20 times higher than peripherally. This may account for the minimal effect of adenosine on the peripheral blood flow in all four patient groups, as a threshold dose may not be reached in the periphery due to the significantly lower flow in the resting phase.\nFew data exist on peripheral blood flow during intravenous adenosine stress only, especially as measured with PET. Heinonen et al14 in their study using [15O]-H2O as the PET perfusion tracer, reported an increase in peripheral muscle perfusion during adenosine infusion comparable to exercise that was absent in our study population. However, in their study adenosine was infused directly into the femoral artery which differs from our intravenous infusion. They also reported a decline in blood flow in the opposite limb during adenosine infusion, indicating that the hyperemic response to adenosine was localized rather than widespread throughout the body after passing into the systemic circulation.\nPedrinelli et al20 found that in patients with systemic hypertension the peripheral vascular reserve was lower than the myocardial reserve without correlation between the two values. This patient group is an interesting comparison with the patient groups with different included pathologies in our study. Again due to the retrospective nature of our study, we could not specify such a subgroup of patients with hypertension for inclusion. It is difficult to hypothesize about any relation between our current findings and those of Pedrinelli et al, because peripheral vasodilation was based on postischemic hyperemia as opposed to adenosine infusion.", "In this study, we confirmed that perfusion reserves of the myocardial and the peripheral microvasculature during the same PET session were different, indicating different levels of response for the two vascular beds during the same intravenous adenosine infusion session." ]

[ "introduction", "materials|methods", null, null, null, null, null, null, "results", null, null, null, null, null, null, "discussion", "conclusion" ]

[ "Myocardial perfusion imaging", "peripheral perfusion imaging", "coronary flow reserve", "vasodilator stress", "N-13 ammonia" ]

[ "Adult", "Bone Screws", "Follow-Up Studies", "Humans", "Joint Instability", "Male", "Middle Aged", "Minimally Invasive Surgical Procedures", "Models, Anatomic", "Pelvic Bones", "Postoperative Complications", "Radiography", "Retrospective Studies", "Sacrum", "Spinal Fractures", "Young Adult" ]

[ "Surgical technique" ]

[ "The patient is under general anaesthesia, in prone position, on a standard radiolucent orthopaedic table. The patient position is exactly symmetrical, with forward tilting of the pelvis achieved by insertion of a thoraco-pelvic support, the knees being flexed at about 30° to release the sciatic nerves and sacral roots. The C-arm fluoroscope is placed on the uninjured side of the patient, and adequate image rendering is verified before starting the operation. The patient’s body has to be placed as caudal as possible, to avoid impingement of the C-arm and table’s pedestal; this can be achieved by assembling two standard leg attachments. The posterior pelvis is then prepped and draped in usual fashion.\nThe posterior superior iliac spines (PSIS) are individuated bilaterally by palpation, and two incisions (about 1.2 inches each) are performed just lateral to them (Fig. 3); then a 3.2-mm drill is used to insert a cortical 4.5 screw in each side (Fig. 4). The drill is started on the posterior superior iliac spine, angling lateral approximately 40° in relation to the sagittal plane and slightly cranially to achieve placement in the direction of the iliac wing (Fig. 5a).Fig. 3Bony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the middle, the sacral spinous lineFig. 4Surgical access to the PSIS (right side) and 3.2-mm drilling to insert the 4.5-mm screw (left side)Fig. 5a Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. b Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction\nBony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the middle, the sacral spinous line\nSurgical access to the PSIS (right side) and 3.2-mm drilling to insert the 4.5-mm screw (left side)\na Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. b Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction\nNext, a large Jungbluth reduction clamp (Matta Pelvic System–Stryker Trauma AG, Bohnackerweg 1, CH-2545 Selzach, Switzerland) is connected to these screws (Fig. 5b). If an important rotational dislocation is detected, a standard 6.0-mm Shanz screw placed proximal to the 4.5 screw in the iliac wing can be used as a joystick. When the fracture displacement is satisfactory reduced, the clamp is tightened to maintain the reduction (Fig. 6). The reduction is checked by fluoroscopy in AP, inlet and outlet views on the posterior ring, and when obtained, it is common to see also the anterior part of the ring in a reduced position.Fig. 6When the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views\nWhen the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views\nFixation is usually achieved with ilio-sacral stainless-steel 6.5- or 8.0-mm cannulated screws (Asnis III; Stryker Orthopaedics, 325 Corporate Drive, Mahwah, NJ 07430, USA) under fluoroscopic control [1–5]. Most times, two screws are placed in the first sacral segment, even if it is possible to place one screw in S1 and one in S2, or a single screw may be used if the area of safe placement is limited [14]. At the end, screw position has to be confirmed on antero-posterior, inlet and outlet views [15–19]. In very unstable disruptions, and whenever possible and safe, use of a trans-sacral screw is suggested, as this type of fixation seems to provide greater stability [20].\nAfter removal of the Jungbluth clamp, it is possible to implant a slide-insertion posterior plate through the same incisions [21, 22] to improve fracture stabilization. This is particularly suggested in case of large displacement, long trauma to intervention time (15 days or more) or poor bone quality.\nAfter fixation of the posterior pelvic ring and closure of wounds, the patient is placed in supine position for anterior pelvic ring fixation, if necessary, by internal or external devices [23–25]." ]

[ null ]

[ "Introduction", "Materials and methods", "Surgical technique", "Results", "Discussion" ]

[ "A lot has been written about minimally invasive stabilization of sacral fractures with percutaneous ilio-sacral screws [1–5] or posterior sacral plates [6], but few articles give sufficient details on the reduction methods. The achievement of satisfactory reduction is the first hot point in the treatment of pelvic ring disruption [7]. Despite its importance, this step may be a real challenge even for the expert pelvic surgeon, because of several issues including the frequent finding of multi-planar and rotational displacement components even in so-called vertically displaced fracture. The classic traction method is often insufficient to deal with these difficulties, because of its exclusively axial effect, the large force required and the need to fix the intact hemi-pelvis in a strong and safe manner. Only some of these problems seem to be solved by use of special pelvic frames [8, 9], and moreover, radiolucent traction tables are expensive devices, available only in few hospitals.\nOtherwise, open reduction techniques can provide good results, but are expensive for these patients in terms of blood loss [7, 10]. Furthermore, surgical access has to be achieved via a skin area often damaged by trauma and in zones that are subject to bedsores.\nThe aim of this study is radiological and functional evaluation of a minimally invasive reduction technique for treatment of sacral fracture, which could be effective, tissue sparing and economic in terms of equipment needed.", "From November 2002 to March 2009, 82 patients suffering from sacral fracture were surgically treated at our institution. Among these, 51 patients presented 61-C1.3 fracture according to the Orthopaedic Trauma Association (OTA) [11].\nThe method of reduction was predominantly closed with traction (28 cases), then open with posterior surgical access. Starting from April 2007, we began to perform the minimally invasive technique described herein.\nFor this retrospective study, inclusion criteria were: 61-C1.3 fracture pattern according to the OTA [11] (type C pelvic ring disruption with monolateral sacral fracture), availability of complete clinical and radiological documentation and a minimum 12-month follow-up time. Exclusion criteria were presence of cognitive deficits, major head trauma, neurologic deficits related to extrapelvic lesions, major injuries to the upper and lower limbs, open fractures and pathological fractures. Moreover, patients with significantly impaired mobility or pain during gait already present before the trauma were excluded.\nWe finally included in this study 11 patients with average age of 40.2 years (range 24–59 years), who were referred to our institution from April 2007 to March 2008 and surgically treated by the same surgeons (A.M., A.N.) using the technique indicated in Table 1.Table 1Patients’ data, procedures and outcomesCaseType oftraumaAge (years)Time from trauma to surgery (days)DenisFollow-up time (months)Majeed scoreDisplacement (mm)PosteriorfixationAnteriorfixationAssociated lesionsPre-surgicalPost-surgicalFollow-up1Industrial accident247118462166Two IS screws, tension bandEx-fixUrethral tear2Industrial accident591011988734One IS screwFour-hole plateL4-L5 fracture(internal fixation)3Sports injury357117991444Two IS screwsEx-fix4Car accident571312581844Two IS screwsSix-hole plate, ex-fix5Sports injury37221598965One IS screwNoneTwo-column acetabular fracture6Car accident28222281855One IS screwEx-fix7Motorbike accident39622482734One IS screwFour-hole plateBladder tear8Motorbike accident43221871866One IS screwFour-hole plate, ex-fix9Industrial accident334214941278One IS screwEx-fixSplenic rupture10Sports injury481021884844One IS screwFour-hole plateL1 fracture (conservative)11Sports injury371631866171213Two IS screwsFour-hole plateTibial plateau fracture, tibial shaft fractureMean40.07.18–18.9180.9110.825.455.73–––SD11.024.68–3.4515.164.662.542.72–––SD standard deviation, IS ilio-sacral\nPatients’ data, procedures and outcomes\nL4-L5 fracture\n(internal fixation)\nSD standard deviation, IS ilio-sacral\nThe trauma was caused by road accident in 36% of cases, by sports injury in 36% and by industrial accident in 28%.\nSeven patients had associated injuries, six of which required surgery: two urological lesions, one spleen rupture, one two-column acetabular fracture, one L4-L5 vertebral fracture and one bilateral lower limb fracture.\nEvery patient was submitted to accurate clinical examination and pre-operative imaging planning, including at least antero-posterior (AP) radiogram of the pelvis, inlet and outlet views (Fig. 1) and a computed tomography scan. Every fracture was classified according to Denis classification [12]; the most common were type II (54%) and type I (36%). Pre-operative displacements were measured to the nearest millimetre as the maximum point-to-point distance between the fragments of the sacral fracture on each of the three views of the pelvis; all displacements were recorded.Fig. 1Illustrative case: pre-operative radiograms (antero-posterior, outlet, inlet) showing a multi-planar displacement with maximum value of 21 mm\nIllustrative case: pre-operative radiograms (antero-posterior, outlet, inlet) showing a multi-planar displacement with maximum value of 21 mm\nThe patients were brought to the operating room as soon as permitted by general health conditions, since it is well known that reduction becomes progressively more difficult with time; mean trauma to surgery time was 7.18 days [range 2–16 days, standard deviation (SD) 4.68 days]. Each fracture was reduced by the minimally invasive approach described below. Fixation of the posterior pelvic ring was achieved by one cannulated ilio-sacral screw in 64%, by two ilio-sacral screws in 28% and by two ilio-sacral screws plus tension-band plating in one case; all screws were placed in the body of the first sacral segment. Fixation of the anterior pelvic ring was achieved with a symphyseal plate in 36% and with an external anterior fixator in 36%; in two cases they were used together, while in one case no anterior fixation was performed.\nAfter surgery, AP, inlet and outlet view radiograms of the pelvis were taken for every patient and post-operative displacements were measured on all three views (Fig. 2); we considered the highest value as an index of quality of reduction according to Matta’s criteria [7].Fig. 2Illustrative case: post-operative radiograms. Fixation is achieved by two ilio-sacral screws, a posterior tension-band plate and anterior external fixation. Maximum residual displacement was 6 mm\nIllustrative case: post-operative radiograms. Fixation is achieved by two ilio-sacral screws, a posterior tension-band plate and anterior external fixation. Maximum residual displacement was 6 mm\nPatients were not allowed weight-bearing for 60 days. The external fixator, when used, was removed 2 months after its placement. They were then directed towards walking rehabilitation, with complete weight-bearing allowed 90 days after injury. Clinical and radiographic examination was carried out at least at 1, 2, 3, 6 and 12 months after surgery.\nFinally, all patients were evaluated functionally at least 1 year after surgery using Majeed’s grading scale for pelvic fracture [13], with mean follow-up time of 18.9 months (range 14–25 months, SD 3.45 months). The study conforms to the 1964 Declaration of Helsinki as revised in 2000 and was approved by our institutional ethical committee. All enrolled patients provided informed consent.\n[SUBTITLE] Surgical technique [SUBSECTION] The patient is under general anaesthesia, in prone position, on a standard radiolucent orthopaedic table. The patient position is exactly symmetrical, with forward tilting of the pelvis achieved by insertion of a thoraco-pelvic support, the knees being flexed at about 30° to release the sciatic nerves and sacral roots. The C-arm fluoroscope is placed on the uninjured side of the patient, and adequate image rendering is verified before starting the operation. The patient’s body has to be placed as caudal as possible, to avoid impingement of the C-arm and table’s pedestal; this can be achieved by assembling two standard leg attachments. The posterior pelvis is then prepped and draped in usual fashion.\nThe posterior superior iliac spines (PSIS) are individuated bilaterally by palpation, and two incisions (about 1.2 inches each) are performed just lateral to them (Fig. 3); then a 3.2-mm drill is used to insert a cortical 4.5 screw in each side (Fig. 4). The drill is started on the posterior superior iliac spine, angling lateral approximately 40° in relation to the sagittal plane and slightly cranially to achieve placement in the direction of the iliac wing (Fig. 5a).Fig. 3Bony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the middle, the sacral spinous lineFig. 4Surgical access to the PSIS (right side) and 3.2-mm drilling to insert the 4.5-mm screw (left side)Fig. 5a Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. b Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction\nBony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the middle, the sacral spinous line\nSurgical access to the PSIS (right side) and 3.2-mm drilling to insert the 4.5-mm screw (left side)\na Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. b Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction\nNext, a large Jungbluth reduction clamp (Matta Pelvic System–Stryker Trauma AG, Bohnackerweg 1, CH-2545 Selzach, Switzerland) is connected to these screws (Fig. 5b). If an important rotational dislocation is detected, a standard 6.0-mm Shanz screw placed proximal to the 4.5 screw in the iliac wing can be used as a joystick. When the fracture displacement is satisfactory reduced, the clamp is tightened to maintain the reduction (Fig. 6). The reduction is checked by fluoroscopy in AP, inlet and outlet views on the posterior ring, and when obtained, it is common to see also the anterior part of the ring in a reduced position.Fig. 6When the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views\nWhen the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views\nFixation is usually achieved with ilio-sacral stainless-steel 6.5- or 8.0-mm cannulated screws (Asnis III; Stryker Orthopaedics, 325 Corporate Drive, Mahwah, NJ 07430, USA) under fluoroscopic control [1–5]. Most times, two screws are placed in the first sacral segment, even if it is possible to place one screw in S1 and one in S2, or a single screw may be used if the area of safe placement is limited [14]. At the end, screw position has to be confirmed on antero-posterior, inlet and outlet views [15–19]. In very unstable disruptions, and whenever possible and safe, use of a trans-sacral screw is suggested, as this type of fixation seems to provide greater stability [20].\nAfter removal of the Jungbluth clamp, it is possible to implant a slide-insertion posterior plate through the same incisions [21, 22] to improve fracture stabilization. This is particularly suggested in case of large displacement, long trauma to intervention time (15 days or more) or poor bone quality.\nAfter fixation of the posterior pelvic ring and closure of wounds, the patient is placed in supine position for anterior pelvic ring fixation, if necessary, by internal or external devices [23–25].\nThe patient is under general anaesthesia, in prone position, on a standard radiolucent orthopaedic table. The patient position is exactly symmetrical, with forward tilting of the pelvis achieved by insertion of a thoraco-pelvic support, the knees being flexed at about 30° to release the sciatic nerves and sacral roots. The C-arm fluoroscope is placed on the uninjured side of the patient, and adequate image rendering is verified before starting the operation. The patient’s body has to be placed as caudal as possible, to avoid impingement of the C-arm and table’s pedestal; this can be achieved by assembling two standard leg attachments. The posterior pelvis is then prepped and draped in usual fashion.\nThe posterior superior iliac spines (PSIS) are individuated bilaterally by palpation, and two incisions (about 1.2 inches each) are performed just lateral to them (Fig. 3); then a 3.2-mm drill is used to insert a cortical 4.5 screw in each side (Fig. 4). The drill is started on the posterior superior iliac spine, angling lateral approximately 40° in relation to the sagittal plane and slightly cranially to achieve placement in the direction of the iliac wing (Fig. 5a).Fig. 3Bony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the middle, the sacral spinous lineFig. 4Surgical access to the PSIS (right side) and 3.2-mm drilling to insert the 4.5-mm screw (left side)Fig. 5a Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. b Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction\nBony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the middle, the sacral spinous line\nSurgical access to the PSIS (right side) and 3.2-mm drilling to insert the 4.5-mm screw (left side)\na Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. b Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction\nNext, a large Jungbluth reduction clamp (Matta Pelvic System–Stryker Trauma AG, Bohnackerweg 1, CH-2545 Selzach, Switzerland) is connected to these screws (Fig. 5b). If an important rotational dislocation is detected, a standard 6.0-mm Shanz screw placed proximal to the 4.5 screw in the iliac wing can be used as a joystick. When the fracture displacement is satisfactory reduced, the clamp is tightened to maintain the reduction (Fig. 6). The reduction is checked by fluoroscopy in AP, inlet and outlet views on the posterior ring, and when obtained, it is common to see also the anterior part of the ring in a reduced position.Fig. 6When the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views\nWhen the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views\nFixation is usually achieved with ilio-sacral stainless-steel 6.5- or 8.0-mm cannulated screws (Asnis III; Stryker Orthopaedics, 325 Corporate Drive, Mahwah, NJ 07430, USA) under fluoroscopic control [1–5]. Most times, two screws are placed in the first sacral segment, even if it is possible to place one screw in S1 and one in S2, or a single screw may be used if the area of safe placement is limited [14]. At the end, screw position has to be confirmed on antero-posterior, inlet and outlet views [15–19]. In very unstable disruptions, and whenever possible and safe, use of a trans-sacral screw is suggested, as this type of fixation seems to provide greater stability [20].\nAfter removal of the Jungbluth clamp, it is possible to implant a slide-insertion posterior plate through the same incisions [21, 22] to improve fracture stabilization. This is particularly suggested in case of large displacement, long trauma to intervention time (15 days or more) or poor bone quality.\nAfter fixation of the posterior pelvic ring and closure of wounds, the patient is placed in supine position for anterior pelvic ring fixation, if necessary, by internal or external devices [23–25].", "The patient is under general anaesthesia, in prone position, on a standard radiolucent orthopaedic table. The patient position is exactly symmetrical, with forward tilting of the pelvis achieved by insertion of a thoraco-pelvic support, the knees being flexed at about 30° to release the sciatic nerves and sacral roots. The C-arm fluoroscope is placed on the uninjured side of the patient, and adequate image rendering is verified before starting the operation. The patient’s body has to be placed as caudal as possible, to avoid impingement of the C-arm and table’s pedestal; this can be achieved by assembling two standard leg attachments. The posterior pelvis is then prepped and draped in usual fashion.\nThe posterior superior iliac spines (PSIS) are individuated bilaterally by palpation, and two incisions (about 1.2 inches each) are performed just lateral to them (Fig. 3); then a 3.2-mm drill is used to insert a cortical 4.5 screw in each side (Fig. 4). The drill is started on the posterior superior iliac spine, angling lateral approximately 40° in relation to the sagittal plane and slightly cranially to achieve placement in the direction of the iliac wing (Fig. 5a).Fig. 3Bony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the middle, the sacral spinous lineFig. 4Surgical access to the PSIS (right side) and 3.2-mm drilling to insert the 4.5-mm screw (left side)Fig. 5a Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. b Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction\nBony landmarks drawn on the skin. The PSIS and the sciatic notch are individuated bilaterally by palpation. In the middle, the sacral spinous line\nSurgical access to the PSIS (right side) and 3.2-mm drilling to insert the 4.5-mm screw (left side)\na Anatomic model showing typical starting and direction of the PSIS drilling, angled approximately 40° laterally and 10° cranially. b Connection of the Jungbluth clamp to the screws, which are later tightened to perform the fracture reduction\nNext, a large Jungbluth reduction clamp (Matta Pelvic System–Stryker Trauma AG, Bohnackerweg 1, CH-2545 Selzach, Switzerland) is connected to these screws (Fig. 5b). If an important rotational dislocation is detected, a standard 6.0-mm Shanz screw placed proximal to the 4.5 screw in the iliac wing can be used as a joystick. When the fracture displacement is satisfactory reduced, the clamp is tightened to maintain the reduction (Fig. 6). The reduction is checked by fluoroscopy in AP, inlet and outlet views on the posterior ring, and when obtained, it is common to see also the anterior part of the ring in a reduced position.Fig. 6When the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views\nWhen the reduction seems satisfactory, the Jungbluth clamp is tightened and the result is checked on the three standard fluoroscopy views\nFixation is usually achieved with ilio-sacral stainless-steel 6.5- or 8.0-mm cannulated screws (Asnis III; Stryker Orthopaedics, 325 Corporate Drive, Mahwah, NJ 07430, USA) under fluoroscopic control [1–5]. Most times, two screws are placed in the first sacral segment, even if it is possible to place one screw in S1 and one in S2, or a single screw may be used if the area of safe placement is limited [14]. At the end, screw position has to be confirmed on antero-posterior, inlet and outlet views [15–19]. In very unstable disruptions, and whenever possible and safe, use of a trans-sacral screw is suggested, as this type of fixation seems to provide greater stability [20].\nAfter removal of the Jungbluth clamp, it is possible to implant a slide-insertion posterior plate through the same incisions [21, 22] to improve fracture stabilization. This is particularly suggested in case of large displacement, long trauma to intervention time (15 days or more) or poor bone quality.\nAfter fixation of the posterior pelvic ring and closure of wounds, the patient is placed in supine position for anterior pelvic ring fixation, if necessary, by internal or external devices [23–25].", "Pre-operative displacements averaged 10.8 mm (range 7–21 mm, SD 4.66 mm) (Table 1). The largest displacement was seen on the 40° caudal view in 63% of cases. Post-operative displacements averaged 5.4 mm (range 3–12 mm, SD 2.54 mm). Using the grading criteria described by Matta [7], there were five excellent (45.5%), five good (45.5%), one fair and no poor reductions. All patients healed, with average displacement at 1-year follow-up of 5.73 mm (range 4–13 mm, SD 2.72 mm). The improvement obtained with surgery was strongly significant (paired-sample t-test: P < 0.0009), while the difference between post-operative and follow-up displacement was nonsignificant (paired-sample t-test: P = 0.192).\nThere were no operative complications; regarding nonsurgical peri-operative complications, there was one urinary tract infection. In one case, aseptic loosening of the symphyseal plate was detected at 2-month follow-up; the hardware was removed 10 days later, without need for further fixation. In two cases, ilio-sacral screw removal, due to screw-related pain, was performed 9 and 11 months after surgery. In one further case, treated with two ilio-sacral screws and a sacral plate, the hardware was removed 9 months after the accident (Fig. 7) because it generated discomfort while the patient was using sports equipment (scuba diving gear with air cylinder). We do not consider this to be a device-related complication but rather a sign of surgical success.Fig. 7Illustrative case: after removal of external and internal devices, performed about 9 months after the accident, there was no sign of fracture displacement\nIllustrative case: after removal of external and internal devices, performed about 9 months after the accident, there was no sign of fracture displacement\nOn functional evaluation, performed using the Majeed score system, 82% of patients obtained good or excellent results, with only one fair and one poor result; the average score was 80.91 points (range 46–99 points, SD 15.16 points). All but one patients returned to work, but the majority of them complained of reduced performance.\nA 23-year-old male affected by urethral disruption, treated with anterograde and retrograde endoscopic repair by a specialized urologic team, complained of reduced sexual function; at 12-month follow-up he reported fair improvement but not complete remission yet. In this case it is difficult to state whether this reflects a real neurologic issue, and whether it could be related to damage to periurethral neural fibres or sacral roots. We registered slight erectile dysfunction in two further males without urologic tears, who had almost complete recovery at 12-month follow-up. There were no further perineal deficits.\nThere were no major lower limb neurologic impairments, but in two cases we observed transitory lateral femoral cutaneous nerve hypoestesy, probably related to anterior external fixation; in one case recovery was complete within 6 months, in another case a partial sensory deficit remained at 12-month follow-up.", "Reduction of vertically displaced sacral fracture can be a difficult challenge for the pelvic surgeon. Open techniques can obtain good reduction but are costly for the patient in terms of blood loss and soft tissue damage. Tornetta and Matta [7] stated that 10 mm is an acceptable reduction for injury to the posterior pelvic ring, as they suggested that greater anatomic reduction of posterior injuries did not result in less posterior pain; those authors, performing open reduction and internal fixation of the injured pelvis, reported excellent or good results in 95% of 107 patients with 38 Bucholz type II (type B) injuries and 69 Bucholz type III (type C) injuries, with 1 case of loss of reduction and 3 cases of deep infection. Van den Bosch et al. [26] evaluated 37 patients (16 type B, 21 type C) treated with internal fixation, obtaining a mean score of 78.6 of 100 on Majeed functional evaluation. Lindahl and Hirvensalo [27] obtained excellent-good radiographic results in 90% of 101 patients treated with open reduction; the overall functional results, measured with a modified version [28] of the original Majeed scoring system [13], were good or excellent in 83% of patients.\nPercutaneous technique is becoming increasingly popular because it can reduce wound-related problems and blood loss [1, 2, 4, 15]. On the other hand, closed reduction using vertical traction can sometimes be insufficient, even with the adjunct of dedicated pelvic frames [8, 9]; moreover, it needs a radiolucent traction table. Routt and Simonian [4] defined reductions that showed less than 1 cm residual displacement in any plane as acceptable. They obtained only 3 malreductions among 60 sacral fractures treated with closed reduction by manipulation and traction method, but reported 5 cases of failure of fixation and 2 cases of non-union.\nSchweitzer et al. [2], revising 71 pelvic ring fractures (10 type B, 61 type C) treated with closed reduction and percutaneous screw fixation, obtained 69 satisfactory reductions and 62 good or excellent functional results according to the Majeed scoring system [13]. Nevertheless, they reported a 9% rate of surgical-related complications, a rate similar to previous reports regarding this technique [5].\nMinimally invasive transiliac plate osteosynthesis [21, 22] has been recently described, and its early results appear encouraging. However, this technique has usually been associated with the closed reduction by traction method.\nThe method described herein can obtain good fracture reduction with a limited surgical approach and with a standard radiolucent table; it is particularly indicated in simple, monolateral vertical sacral fractures (61-C1.3). Otherwise, we suggest reduction by traditional techniques for bilateral or complex sacral fractures, because of the extreme posterior instability, as well as for those disruptions which require direct vision of the fracture site.\nThe position of the patient allows traditional open exposure if minimally invasive reduction fails and allows internal fixation of the posterior ring and optimal placement of percutaneous ilio-sacral screws.\nUsing this technique we achieved good to excellent reductions in 91% of patients, coherent with most of the studies regarding sacral fractures which can be found in literature [2, 7, 27]. In only one patient was a fair result (>10 mm) obtained; he presented wide pre-operative displacement, and because of the long trauma to surgery time, interposition of soft tissue and fibrous callus formation prevented a better result; nevertheless, we decided not to perform open reduction because the skin and subcutaneous tissues were compromised. Unfortunately, he reported one of the lower functional scores at 1-year follow-up. We suggest that open reduction should be preferred, whenever possible, where the trauma to surgery time is extended beyond 15 days.\nThe main limitations of this study are the small sample size and the short follow-up; furthermore, its strength is blunted by all the implications of its retrospective design and the lack of a control group.\nIn conclusion, we found a minimally invasive reduction technique to be a satisfactory procedure for management of vertically displaced sacral fracture. Axial traction remains a good method, and more cases need to be operated using this technique to confirm its effectiveness in terms of reduction and determine possible complications." ]

[ "introduction", "materials|methods", null, "results", "discussion" ]

[ "Sacral fractures", "Minimally invasive reduction", "Pelvic instability", "Ilio-sacral screw fixation", "Slide-insertion plate" ]

[ "Aged", "Evidence-Based Medicine", "Health Care Costs", "Humans", "Male", "Middle Aged", "Practice Patterns, Physicians'", "Prostatic Neoplasms", "Quality of Life", "Registries", "Risk Factors", "Treatment Outcome" ]

[ "Trends in prostate cancer presentation and risk", "Health services research and trends in disease management", "Oncologic outcomes", "Quality of life outcomes" ]

[ "Temporal trends in the PSA era in patient risk at diagnosis are consistent with downward stage migration. In CaPSURE, the proportion of patients presenting with low-risk disease (PSA < 10 ng/mL, Gleason score < 7 with no pattern 4 or 5 disease on biopsy, and clinical stage T1 or T2 disease) has increased from 31% between 1989 and 1990 to 47% of patients between 2001 and 2002 and has remained relatively stable [2]. However, within this low-risk group, there is a trend toward lower-risk characteristics (based on PSA, clinical stage, and percent positive biopsy). Conversely, during the same time, men with high-risk disease (PSA > 20 ng/mL, Gleason 8-10 on biopsy, or stage T3) have decreased from 41 to 29%. Although the rate of high-risk prostate cancer has fallen, it has remained stable since 2000 and represents approximately 24% of patients in recent studies [3]. There is no evidence in the CaPSURE cohort of meaningful downward risk migration among high-risk patients over the past 15 years.\nDespite presentation with lower-risk disease due to increased PSA screening, recent studies have observed a disparity in prostate cancer presentation across sociodemographic groups. Dall’era et al. [4] examined 5,939 men enrolled in CaPSURE from 1995 to 2007, and found that patients who were older, less educated, and had Medicare for insurance (as opposed to VA or private coverage) were more likely to have intermediate or high-risk disease. Non-white race was also associated with high-risk disease at presentation (OR 1.83, 1.47–2.29, P < 0.01). Clinically insignificant disease (PSA < 10 ng/mL, <33% of biopsy cores involved, no Gleason pattern 4/5, and stage T1a or T2a) was more common in younger men (<60 years old), higher income/education, and men with private insurance. Within this group, younger age and private insurance were again associated with immediate treatment in lieu of AS.", "The majority of patients followed in CaPSURE were diagnosed during the PSA era and treated in community-based settings. One strength of this registry is that participating physicians treat according to their usual practices and patient preferences. CaPSURE provides a mix of locales and practice types, reflecting current contemporary urological practice at a national level. The CaPSURE sites are not a random sample of the US population. However, CaPSURE includes far richer clinical detail than population-based sources such as the Surveillance Epidemiology and End Results (SEER) database and SEER-Medicare and therefore is an excellent data source for national studies of disease management.\nImaging studies performed in men with prostate cancer may serve to facilitate optimal treatment planning. However, staging studies are associated with low but definite risks, significant costs to the health care system, and have minimal benefit in patients with low-risk disease characteristics. One early study from CaPSURE examined the use of imaging tests for staging clinically localized prostate cancer between 1989 and 1997. Kindrick et al. [5] found widespread and consistent overuse in the rates of bone scan, computed tomography, and magnetic resonance imaging among patients with low likelihood of extraprostatic disease. Follow-up analysis through 2001 showed that rates of bone scan and computerized tomography use have decreased in recent years with the greatest decreases in patients with lower-risk cancer [6]. Whereas among early CaPSURE patients, disease risk exerted no influence on the likelihood of imaging, the more recent CaPSURE data have illustrated a trend toward appropriate and evidence-based use of imaging tests. This highlights the value of registries such as CaPSURE that allows evaluation of adherence to guidelines based on high-level evidence into common clinical practice, as well as the extent to which evidence-based medicine is practiced in urology.\nMultiple recent CaPSURE studies have examined patterns in treatment selection for patients diagnosed with prostate cancer. An early study examined the use of ADT in patients with localized disease and found a higher than expected use of ADT monotherapy [7]. Among low-, intermediate-, and high-risk patients, ADT monotherapy rose dramatically, from 5 to 14%, 9 to 20%, and 33 to 48%, respectively, from 1989–1990 to 2000–2001. PADT monotherapy is considered to be investigational based on the American Urological Association’s clinical practice guidelines [8], and no controlled trials have established efficacy of this approach.\nWith decreasing risk migration, it would be expected that the use of AS would increase since more men are presenting with favorable risk disease. An early study of temporal trends found that the use of AS reached a nadir of 5.5% in 2000–2001 from 9.5% in 1992–1994, with the largest declines in low-risk patients (6.2%), although recent improvement to 10.2% in 2004–2006 was observed [2]. Another study by Barocas et al. [9] found that between 1999 and 2004, 16.4% of men met strict surveillance criteria (PSA < 10 ng/mL, clinical T1 or T2a, prostate-specific antigen density PSAD < 0.15, <33% biopsy cores positive, and absence of Gleason pattern 4/5 on biopsy), but only 9% of men in this low-risk category chose surveillance, highlighting the underuse of this management strategy in the United States.\nIn addition to the initial decline in AS, a decrease was observed in EBRT from 13% in 1993–1995 to 7% in 1999–2001 and RP (55–52%) and increases in PADT (7–12%) and brachytherapy (4–22%) in low-risk patients [10]. This treatment trend remained present in men >75 years with the use of surveillance in only 24%, declining from 52%, PADT use increasing from 23 to 30%, and brachytherapy from 3 to 31%. A reversal of this trend was observed with a decrease in brachytherapy to 13% in 2005–2006, a decrease in PADT to 6.6%, and an increase in RP to 60% [2].\nOverall, it appears that prostate cancer risk drives treatment selection in conjunction with age, comorbidity, and socioeconomic status. However, a recent analysis in CaPSURE examining practice patterns in the primary treatment of localized prostate cancer confirmed prevalent overtreatment of low-risk disease but also identified what appeared to be a troubling undertreatment of high-risk disease [3, 11]. Interestingly, treatment patterns were found to vary across clinical sites, and much of the variation could be attributed to practice site itself and not solely patient or tumor factors. Clinical practice site alone explained, for example, 13% variation seen in ADT, 30% in RP, 36% in brachytherapy, 20% in EBRT, and 74% in cryoablation [11].\nAnother unique feature of CaPSURE is the access to resource use data that offer a means of studying health care cost implications of different prostate cancer management strategies. Penson et al. [12] examined adjusted first-year costs associated with various treatment options based on Medicare payment schedules and found a trend toward higher costs for higher-stage patients. The mean cost of prostate cancer treatment in the first year after diagnosis was $6,375 and was not different between patients with RP and EBRT, but was higher for those receiving NADT prior to either treatment. Wilson et al. [13] analyzed all health care utilization of various treatments over a period of 5 years and found that the average annual cost was $7,740. This varied widely with AS costing the least at $5,843 and androgen deprivation therapy the most expensive at $12,590.", "Recent evidence suggests that information beyond Gleason score obtained at diagnostic biopsy contributes significantly to accurate risk assessment among patients with newly diagnosed prostate cancer. The percent of positive biopsy cores was initially validated as a prognostic marker among patients who underwent RP in CaPSURE and was significant across all risk groups, confirming that biopsy data obtained in the community setting using non-standardized techniques and assessed by diverse pathologists offers consistent and useful prognostic information [14].\nCancer of the Prostate Risk Assessment score (CAPRA) was developed using data from the CaPSURE registry. The CAPRA score simply and accurately predicts pathologic status, disease recurrence, and mortality after surgery [15]. Points are assigned based on age, clinical stage, PSA, Gleason grade, and percent of cores positive on biopsy. Scores range from 0 to 10 with each 2-point increase roughly doubling the risk of recurrence and progression. A 9-point variation of the CAPRA scoring system can alternatively be used if data regarding percent positive biopsy cores are not available. The strength of the CAPRA score is not only in its ease of use but in its ability to better discriminate between categories of risk in all practice settings when compared with other nomograms [16, 17]. The CAPRA score has been extensively validated in other disease registries and academic cohorts, both in the United States and Europe [18].\nUncertainty regarding the optimal treatment for localized prostate cancer has led to large variation in practice patterns as reported previously in CaPSURE [11]. To date, no adequate randomized trials have compared active treatments for localized prostate cancer due to difficulties with accrual, randomization, high cost, and need for long follow-up. Disease registries can provide important insights into outcomes and can provide an important source for comparative effectiveness analysis between various treatments. This is highlighted in a recent study by Cooperberg et al. [19] which compared risk-adjusted disease-specific and all-cause mortality after treatment of localized prostate cancer with RP, EBRT, and primary ADT. After adjusting for age, disease risk, and comorbidity, mortality at 10 years was less likely in men who underwent RP than EBRT or primary ADT, especially in men with intermediate or high-risk disease.", "The preservation of health-related quality of Life (HRQOL) is a priority in any discussion of prostate cancer treatment and outcomes due to the long natural history of disease. Any negative impact on quality of life must be minimized, as patients may experience it for an extended period of time. CaPSURE had proved to be an invaluable resource for the prospective, longitudinal assessment of patient-reported outcomes and it has enabled investigators to successfully address a number of questions in this area of prostate cancer research.\nCaPSURE HRQOL data are reported by both patients and physicians and provide insight into the differences between outcomes. An early study by Litwin et al. [20] found that physician assessment of treatment impact on HRQOL underestimated the impact experienced by patients, especially within general health parameters. When impairment was noted, urologists reported on urinary and sexual functions more often than pain in men who underwent treatment for localized prostate cancer. Many aspects of patient-reported HRQOL (physical function and general health) have also been significantly associated with survival in a similar cohort of men, over the course of disease from diagnosis to after treatment [21]. These studies highlight the continued importance of multidimensional, patient-reported HRQOL assessment in current prostate cancer treatment.\nUsing longitudinal HRQOL data within CaPSURE, comparisons can be made between outcomes among treatment options. Men who underwent RP had lower scores on both disease-specific and general HRQOL instruments immediately postoperative, which improved significantly at 1 year after treatment, and continued to improve in the domain of sexual function in the second year. Men who were treated with EBRT, AS, or primary ADT had scores that were relatively stable, except for sexual function, which declined with time. Overall, patients who underwent RP had the greatest decline initially, but also the greatest degree of recovery. Most men experienced the greatest recovery of both urinary and sexual functions within 2 years of treatment, with little change in reported HRQOL scores after 3 years [22]. Those who received multimodal therapy appeared to have greater declines in urinary and sexual functions than those who were treated with monotherapy [23].\nAlthough prostate cancer presentation and treatment patterns vary based on ethnicity, little was known with regards to any differences in HRQOL. Disease registries like CaPSURE are ideal to study ethnic groups that are underrepresented in most clinical studies. Using CaPSURE data, African American men were found to have lower HRQOL scores across many domains of QOL at baseline and after treatment, but had higher sexual function scores [24]. Over time, African American patients never recovered as well as white patients in reported HRQOL scores. Ethnicity was also found to play a role in primary treatment choice in men with equivalent disease characteristics [25].\nAlthough this article focuses primarily on lessons learned from CaPSURE, there are a number of other US national registries and databases available for research. Table 2 highlights the strengths of many of the resources available to study the various aspects of prostate cancer. For example, the Prostate Outcomes Cancer Study (PCOS), which utilizes Surveillance Epidemiology and End Results (SEER)-based data, was initiated by the National Cancer Institute (NCI) and used by investigators to report on HRQOL outcomes in American men diagnosed with prostate cancer [26]. Likewise, the Department of Defense Center for Prostate Disease Research (CPDR) has been used to report on prostate cancer-specific mortality and not only collects data on men with prostate cancer treated within the military medical system but seeks to standardize clinical practice among military sites [27].Table 2National registries in the United States used commonly in prostate cancer researchDatabaseDescriptionCaPSURECommunity practice-based longitudinal data from 31 sites as reported by urologists and patients\nhttp://www.capsure.net\nCPDRLongitudinal data on prostate cancer patients treated in the military health care system as reported by medical practitioners (urologists, medical/radiation oncologists)\nhttp://www.cdpr.org/database.html\nMedicarePopulation-based data on insurance claims for covered health services for eligible US persons over 65 years oldNISA cross-sectional database that is part of HCUP which collects information on over 8 million hospital stays annually\nhttp://http://www.hucp-us.ahrq.gov\nPCOSCross-sectional, population-based (SEER) data collected from patient chart abstraction\nhttp://healthservices.cancer.gov/pcos\nPROST-QALongitudinal data collected from prostate cancer patients and spouses at 9 US academic medical centers on prostate cancer treatment outcomesSEARCHLongitudinal data on men who underwent radical prostatectomy at 4 VA hospitals and 1 military centerSEERPopulation-based, longitudinal data collected on approximately 28% of all US cancer patients, including cancer staging\nhttp://seer.cancer.gov\nSEER-MedicareLinkage of two of the largest US population-based data sources\nhttp://healthservices.cancer.gov/seermedicare\n\nCaPSURE Cancer of the Prostate Strategic Urologic Research Endeavor\nCPDR Center for Prostate Disease Research\nHCUP Healthcare Cost and Utilization Project\nNIS Nationwide Inpatient Sample\nPCOS Prostate Cancer Outcomes Study\nPROST-QA The Prostate Cancer Outcomes and Satisfaction with Treatment Quality Assessment\nSEARCH Shared Equal Access Regional Cancer Hospital\nSEER Surveillance Epidemiology and End Results\nVA Veterans Affairs\n\nNational registries in the United States used commonly in prostate cancer research\n\nCaPSURE Cancer of the Prostate Strategic Urologic Research Endeavor\n\nCPDR Center for Prostate Disease Research\n\nHCUP Healthcare Cost and Utilization Project\n\nNIS Nationwide Inpatient Sample\n\nPCOS Prostate Cancer Outcomes Study\n\nPROST-QA The Prostate Cancer Outcomes and Satisfaction with Treatment Quality Assessment\n\nSEARCH Shared Equal Access Regional Cancer Hospital\n\nSEER Surveillance Epidemiology and End Results\n\nVA Veterans Affairs\nLimitations to disease registries in general center on methodology of data collection. The quality of data collected and integrity in data collection and follow-up determine the quality of the disease registry. In addition to bias introduced by data collection, disease registries do not capture a random sample of the population. These biases can be minimized (as in CaPSURE) by including community centers to increase population sampling. Centralizing data collection and analysis with strict monitoring by a statistician ensure quality control. Although CaPSURE does capture a diverse group of men, patients are enrolled by their urologists, rather than medical or radiation oncologists, which may exclude a proportion of men with prostate cancer.\nCaPSURE is a research partnership with industry (initially funded by TAP pharmaceuticals) under IRB approval. Data integrity has been maintained free of conflict of interest by thoughtful and transparent methodology and reporting. Data analyses and decision for publication of CaPSURE results have always rested with academic investigators without the influence of industry. Currently, CaPSURE is supported by a combination of gifted funds from Abbott Laboratories along with a growing portfolio of federal grants to ensure registry continuity and maintenance." ]

[ null, null, null, null ]

[ "Introduction", "Trends in prostate cancer presentation and risk", "Health services research and trends in disease management", "Oncologic outcomes", "Quality of life outcomes", "Conclusion" ]

[ "The randomized controlled trial (RCT) remains the gold standard for informing evidence-based clinical practice, in urology as in other medical domains. Limitations include time, significant expense, stringent inclusion criteria, and resistance to randomization by clinicians and patients. Ultimately, if patients enrolled in a RCT differ from the larger population of patients with a given condition, the external validity of the findings may be questionable. Although these limitations can be mitigated by utilizing specialized RCT designs (Table 1), disease registries have emerged as an important complement to RCTs. Disease registries, which accrue prospectively identified cohorts and follow them regardless of sociodemographic characteristics, clinical variables, treatment details or intermediate outcomes, have emerged as an important complement to RCTs. This reflection of “real world” treatment is a tremendous asset especially in prostate cancer research as it provides a relatively cost-effective tool to shed light on a disease with a long natural history and rapidly changing management practices that are subject to many different clinical, scientific, demographic, and economic dynamics. This article will highlight some advantages of disease registries focusing on the example of prostate cancer and the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE™) registry.Table 1Specialized classifications of randomized controlled trialsTypeDescriptionStrengthsLimitationsClusterRandomization of subjects as a group rather than on individual basisCan study interventions that cannot be directed toward selected individualsCan control for contamination across individuals (i.e., when one individual’s behavior can influence another’s)Needs more subjects to reach statistical power than standard RCTExplanatoryIndividual randomization of very selective subjects in a highly controlled settingUseful to test efficacy (i.e., whether an intervention causes a specific biologic response)Patients blindedExcellent internal validityGood for acute disease processesExternal validity and applicability to clinical practice is limited due to subject selectionPragmaticIndividual randomization of non-selective group of patients in a regular clinical settingUseful to test effectiveness of an intervention in everyday practiceGood for chronic disease processes and complex interventionsExcellent external validity and directly applicable to clinical practicePatients unblindedInternal validity limited due to broad inclusion criteriaExpertise-basedIndividual randomization of subjects to an expertise in the intervention in questionUseful when intervention is non-pharmacologic (i.e., surgical procedures)External validity limited to only those patients receiving care from a physician with expertise skills\n\nSpecialized classifications of randomized controlled trials\nCan study interventions that cannot be directed toward selected individuals\nCan control for contamination across individuals (i.e., when one individual’s behavior can influence another’s)\nUseful to test efficacy (i.e., whether an intervention causes a specific biologic response)\nPatients blinded\nExcellent internal validity\nGood for acute disease processes\nUseful to test effectiveness of an intervention in everyday practice\nGood for chronic disease processes and complex interventions\nExcellent external validity and directly applicable to clinical practice\nPatients unblinded\nInternal validity limited due to broad inclusion criteria\nCaPSURE was initiated in 1995 to document national trends in prostate cancer epidemiology, disease management, oncologic outcomes, and health-related quality of life (HRQOL) outcomes. It is a longitudinal, observational database accruing data from a total of 40 urologic practice sites over the history of the registry. The majority of sites are community based, although four university-affiliated centers, and Veterans Affairs (VA) medical centers are included. Men with biopsy-proven cancer are invited to join CaPSURE regardless of disease stage or treatment history. Informed consent for participation is obtained under institutional review board supervision.\nCaPSURE collects approximately 1,000 clinical and patient-reported variables. Clinical information is collected by the treating urologist at baseline and with each follow-up visit and includes history of prostate cancer diagnosis, biopsies, pathology, staging tests, primary and subsequent treatments (radical prostatectomy [RP], external beam radiotherapy [EBRT], brachytherapy, primary and neoadjuvant androgen deprivation therapy [PADT and NADT], cryosurgery, and watchful waiting/active surveillance [AS]), Karnofsky performance status scores, clinic procedures, and medications. At enrollment and every 6 months thereafter, a questionnaire is completed documenting sociodemographic parameters, comorbidities, and HRQOL using validated instruments. Other sections of the patient questionnaires assess use of health services, with hospitalization data verified by discharge summary review.\nPatients are treated according to their physicians’ usual practices and patient preferences and are followed until time of death or withdrawal from the study. Periodic, random sample chart review ensures completeness and accuracy of data collected and entered. Additional details regarding project methodology have been reported previously [1]. CaPSURE is managed by the Department of Urology at the UCSF Helen Diller Family Comprehensive Cancer Center. It was funded from inception to 2007 through an unrestricted education grant from TAP Pharmaceutical Products, Inc., and currently is supported through Abbott Labs (Chicago, IL) and several collaborative Federal grants.\nThere are currently 13,821 men enrolled in CaPSURE. The median patient age at diagnosis is 67, and 75% of men are between 60 and 79 years of age. Most patients are white, with approximately 10% black representation, and 3.5% Latino, Asian, and other ethnicities. There is a fairly even distribution across socioeconomic strata, based on education and income level. CaPSURE has yielded over 130 peer-reviewed publications, with several others in press. A summary of some key research findings follows.", "Temporal trends in the PSA era in patient risk at diagnosis are consistent with downward stage migration. In CaPSURE, the proportion of patients presenting with low-risk disease (PSA < 10 ng/mL, Gleason score < 7 with no pattern 4 or 5 disease on biopsy, and clinical stage T1 or T2 disease) has increased from 31% between 1989 and 1990 to 47% of patients between 2001 and 2002 and has remained relatively stable [2]. However, within this low-risk group, there is a trend toward lower-risk characteristics (based on PSA, clinical stage, and percent positive biopsy). Conversely, during the same time, men with high-risk disease (PSA > 20 ng/mL, Gleason 8-10 on biopsy, or stage T3) have decreased from 41 to 29%. Although the rate of high-risk prostate cancer has fallen, it has remained stable since 2000 and represents approximately 24% of patients in recent studies [3]. There is no evidence in the CaPSURE cohort of meaningful downward risk migration among high-risk patients over the past 15 years.\nDespite presentation with lower-risk disease due to increased PSA screening, recent studies have observed a disparity in prostate cancer presentation across sociodemographic groups. Dall’era et al. [4] examined 5,939 men enrolled in CaPSURE from 1995 to 2007, and found that patients who were older, less educated, and had Medicare for insurance (as opposed to VA or private coverage) were more likely to have intermediate or high-risk disease. Non-white race was also associated with high-risk disease at presentation (OR 1.83, 1.47–2.29, P < 0.01). Clinically insignificant disease (PSA < 10 ng/mL, <33% of biopsy cores involved, no Gleason pattern 4/5, and stage T1a or T2a) was more common in younger men (<60 years old), higher income/education, and men with private insurance. Within this group, younger age and private insurance were again associated with immediate treatment in lieu of AS.", "The majority of patients followed in CaPSURE were diagnosed during the PSA era and treated in community-based settings. One strength of this registry is that participating physicians treat according to their usual practices and patient preferences. CaPSURE provides a mix of locales and practice types, reflecting current contemporary urological practice at a national level. The CaPSURE sites are not a random sample of the US population. However, CaPSURE includes far richer clinical detail than population-based sources such as the Surveillance Epidemiology and End Results (SEER) database and SEER-Medicare and therefore is an excellent data source for national studies of disease management.\nImaging studies performed in men with prostate cancer may serve to facilitate optimal treatment planning. However, staging studies are associated with low but definite risks, significant costs to the health care system, and have minimal benefit in patients with low-risk disease characteristics. One early study from CaPSURE examined the use of imaging tests for staging clinically localized prostate cancer between 1989 and 1997. Kindrick et al. [5] found widespread and consistent overuse in the rates of bone scan, computed tomography, and magnetic resonance imaging among patients with low likelihood of extraprostatic disease. Follow-up analysis through 2001 showed that rates of bone scan and computerized tomography use have decreased in recent years with the greatest decreases in patients with lower-risk cancer [6]. Whereas among early CaPSURE patients, disease risk exerted no influence on the likelihood of imaging, the more recent CaPSURE data have illustrated a trend toward appropriate and evidence-based use of imaging tests. This highlights the value of registries such as CaPSURE that allows evaluation of adherence to guidelines based on high-level evidence into common clinical practice, as well as the extent to which evidence-based medicine is practiced in urology.\nMultiple recent CaPSURE studies have examined patterns in treatment selection for patients diagnosed with prostate cancer. An early study examined the use of ADT in patients with localized disease and found a higher than expected use of ADT monotherapy [7]. Among low-, intermediate-, and high-risk patients, ADT monotherapy rose dramatically, from 5 to 14%, 9 to 20%, and 33 to 48%, respectively, from 1989–1990 to 2000–2001. PADT monotherapy is considered to be investigational based on the American Urological Association’s clinical practice guidelines [8], and no controlled trials have established efficacy of this approach.\nWith decreasing risk migration, it would be expected that the use of AS would increase since more men are presenting with favorable risk disease. An early study of temporal trends found that the use of AS reached a nadir of 5.5% in 2000–2001 from 9.5% in 1992–1994, with the largest declines in low-risk patients (6.2%), although recent improvement to 10.2% in 2004–2006 was observed [2]. Another study by Barocas et al. [9] found that between 1999 and 2004, 16.4% of men met strict surveillance criteria (PSA < 10 ng/mL, clinical T1 or T2a, prostate-specific antigen density PSAD < 0.15, <33% biopsy cores positive, and absence of Gleason pattern 4/5 on biopsy), but only 9% of men in this low-risk category chose surveillance, highlighting the underuse of this management strategy in the United States.\nIn addition to the initial decline in AS, a decrease was observed in EBRT from 13% in 1993–1995 to 7% in 1999–2001 and RP (55–52%) and increases in PADT (7–12%) and brachytherapy (4–22%) in low-risk patients [10]. This treatment trend remained present in men >75 years with the use of surveillance in only 24%, declining from 52%, PADT use increasing from 23 to 30%, and brachytherapy from 3 to 31%. A reversal of this trend was observed with a decrease in brachytherapy to 13% in 2005–2006, a decrease in PADT to 6.6%, and an increase in RP to 60% [2].\nOverall, it appears that prostate cancer risk drives treatment selection in conjunction with age, comorbidity, and socioeconomic status. However, a recent analysis in CaPSURE examining practice patterns in the primary treatment of localized prostate cancer confirmed prevalent overtreatment of low-risk disease but also identified what appeared to be a troubling undertreatment of high-risk disease [3, 11]. Interestingly, treatment patterns were found to vary across clinical sites, and much of the variation could be attributed to practice site itself and not solely patient or tumor factors. Clinical practice site alone explained, for example, 13% variation seen in ADT, 30% in RP, 36% in brachytherapy, 20% in EBRT, and 74% in cryoablation [11].\nAnother unique feature of CaPSURE is the access to resource use data that offer a means of studying health care cost implications of different prostate cancer management strategies. Penson et al. [12] examined adjusted first-year costs associated with various treatment options based on Medicare payment schedules and found a trend toward higher costs for higher-stage patients. The mean cost of prostate cancer treatment in the first year after diagnosis was $6,375 and was not different between patients with RP and EBRT, but was higher for those receiving NADT prior to either treatment. Wilson et al. [13] analyzed all health care utilization of various treatments over a period of 5 years and found that the average annual cost was $7,740. This varied widely with AS costing the least at $5,843 and androgen deprivation therapy the most expensive at $12,590.", "Recent evidence suggests that information beyond Gleason score obtained at diagnostic biopsy contributes significantly to accurate risk assessment among patients with newly diagnosed prostate cancer. The percent of positive biopsy cores was initially validated as a prognostic marker among patients who underwent RP in CaPSURE and was significant across all risk groups, confirming that biopsy data obtained in the community setting using non-standardized techniques and assessed by diverse pathologists offers consistent and useful prognostic information [14].\nCancer of the Prostate Risk Assessment score (CAPRA) was developed using data from the CaPSURE registry. The CAPRA score simply and accurately predicts pathologic status, disease recurrence, and mortality after surgery [15]. Points are assigned based on age, clinical stage, PSA, Gleason grade, and percent of cores positive on biopsy. Scores range from 0 to 10 with each 2-point increase roughly doubling the risk of recurrence and progression. A 9-point variation of the CAPRA scoring system can alternatively be used if data regarding percent positive biopsy cores are not available. The strength of the CAPRA score is not only in its ease of use but in its ability to better discriminate between categories of risk in all practice settings when compared with other nomograms [16, 17]. The CAPRA score has been extensively validated in other disease registries and academic cohorts, both in the United States and Europe [18].\nUncertainty regarding the optimal treatment for localized prostate cancer has led to large variation in practice patterns as reported previously in CaPSURE [11]. To date, no adequate randomized trials have compared active treatments for localized prostate cancer due to difficulties with accrual, randomization, high cost, and need for long follow-up. Disease registries can provide important insights into outcomes and can provide an important source for comparative effectiveness analysis between various treatments. This is highlighted in a recent study by Cooperberg et al. [19] which compared risk-adjusted disease-specific and all-cause mortality after treatment of localized prostate cancer with RP, EBRT, and primary ADT. After adjusting for age, disease risk, and comorbidity, mortality at 10 years was less likely in men who underwent RP than EBRT or primary ADT, especially in men with intermediate or high-risk disease.", "The preservation of health-related quality of Life (HRQOL) is a priority in any discussion of prostate cancer treatment and outcomes due to the long natural history of disease. Any negative impact on quality of life must be minimized, as patients may experience it for an extended period of time. CaPSURE had proved to be an invaluable resource for the prospective, longitudinal assessment of patient-reported outcomes and it has enabled investigators to successfully address a number of questions in this area of prostate cancer research.\nCaPSURE HRQOL data are reported by both patients and physicians and provide insight into the differences between outcomes. An early study by Litwin et al. [20] found that physician assessment of treatment impact on HRQOL underestimated the impact experienced by patients, especially within general health parameters. When impairment was noted, urologists reported on urinary and sexual functions more often than pain in men who underwent treatment for localized prostate cancer. Many aspects of patient-reported HRQOL (physical function and general health) have also been significantly associated with survival in a similar cohort of men, over the course of disease from diagnosis to after treatment [21]. These studies highlight the continued importance of multidimensional, patient-reported HRQOL assessment in current prostate cancer treatment.\nUsing longitudinal HRQOL data within CaPSURE, comparisons can be made between outcomes among treatment options. Men who underwent RP had lower scores on both disease-specific and general HRQOL instruments immediately postoperative, which improved significantly at 1 year after treatment, and continued to improve in the domain of sexual function in the second year. Men who were treated with EBRT, AS, or primary ADT had scores that were relatively stable, except for sexual function, which declined with time. Overall, patients who underwent RP had the greatest decline initially, but also the greatest degree of recovery. Most men experienced the greatest recovery of both urinary and sexual functions within 2 years of treatment, with little change in reported HRQOL scores after 3 years [22]. Those who received multimodal therapy appeared to have greater declines in urinary and sexual functions than those who were treated with monotherapy [23].\nAlthough prostate cancer presentation and treatment patterns vary based on ethnicity, little was known with regards to any differences in HRQOL. Disease registries like CaPSURE are ideal to study ethnic groups that are underrepresented in most clinical studies. Using CaPSURE data, African American men were found to have lower HRQOL scores across many domains of QOL at baseline and after treatment, but had higher sexual function scores [24]. Over time, African American patients never recovered as well as white patients in reported HRQOL scores. Ethnicity was also found to play a role in primary treatment choice in men with equivalent disease characteristics [25].\nAlthough this article focuses primarily on lessons learned from CaPSURE, there are a number of other US national registries and databases available for research. Table 2 highlights the strengths of many of the resources available to study the various aspects of prostate cancer. For example, the Prostate Outcomes Cancer Study (PCOS), which utilizes Surveillance Epidemiology and End Results (SEER)-based data, was initiated by the National Cancer Institute (NCI) and used by investigators to report on HRQOL outcomes in American men diagnosed with prostate cancer [26]. Likewise, the Department of Defense Center for Prostate Disease Research (CPDR) has been used to report on prostate cancer-specific mortality and not only collects data on men with prostate cancer treated within the military medical system but seeks to standardize clinical practice among military sites [27].Table 2National registries in the United States used commonly in prostate cancer researchDatabaseDescriptionCaPSURECommunity practice-based longitudinal data from 31 sites as reported by urologists and patients\nhttp://www.capsure.net\nCPDRLongitudinal data on prostate cancer patients treated in the military health care system as reported by medical practitioners (urologists, medical/radiation oncologists)\nhttp://www.cdpr.org/database.html\nMedicarePopulation-based data on insurance claims for covered health services for eligible US persons over 65 years oldNISA cross-sectional database that is part of HCUP which collects information on over 8 million hospital stays annually\nhttp://http://www.hucp-us.ahrq.gov\nPCOSCross-sectional, population-based (SEER) data collected from patient chart abstraction\nhttp://healthservices.cancer.gov/pcos\nPROST-QALongitudinal data collected from prostate cancer patients and spouses at 9 US academic medical centers on prostate cancer treatment outcomesSEARCHLongitudinal data on men who underwent radical prostatectomy at 4 VA hospitals and 1 military centerSEERPopulation-based, longitudinal data collected on approximately 28% of all US cancer patients, including cancer staging\nhttp://seer.cancer.gov\nSEER-MedicareLinkage of two of the largest US population-based data sources\nhttp://healthservices.cancer.gov/seermedicare\n\nCaPSURE Cancer of the Prostate Strategic Urologic Research Endeavor\nCPDR Center for Prostate Disease Research\nHCUP Healthcare Cost and Utilization Project\nNIS Nationwide Inpatient Sample\nPCOS Prostate Cancer Outcomes Study\nPROST-QA The Prostate Cancer Outcomes and Satisfaction with Treatment Quality Assessment\nSEARCH Shared Equal Access Regional Cancer Hospital\nSEER Surveillance Epidemiology and End Results\nVA Veterans Affairs\n\nNational registries in the United States used commonly in prostate cancer research\n\nCaPSURE Cancer of the Prostate Strategic Urologic Research Endeavor\n\nCPDR Center for Prostate Disease Research\n\nHCUP Healthcare Cost and Utilization Project\n\nNIS Nationwide Inpatient Sample\n\nPCOS Prostate Cancer Outcomes Study\n\nPROST-QA The Prostate Cancer Outcomes and Satisfaction with Treatment Quality Assessment\n\nSEARCH Shared Equal Access Regional Cancer Hospital\n\nSEER Surveillance Epidemiology and End Results\n\nVA Veterans Affairs\nLimitations to disease registries in general center on methodology of data collection. The quality of data collected and integrity in data collection and follow-up determine the quality of the disease registry. In addition to bias introduced by data collection, disease registries do not capture a random sample of the population. These biases can be minimized (as in CaPSURE) by including community centers to increase population sampling. Centralizing data collection and analysis with strict monitoring by a statistician ensure quality control. Although CaPSURE does capture a diverse group of men, patients are enrolled by their urologists, rather than medical or radiation oncologists, which may exclude a proportion of men with prostate cancer.\nCaPSURE is a research partnership with industry (initially funded by TAP pharmaceuticals) under IRB approval. Data integrity has been maintained free of conflict of interest by thoughtful and transparent methodology and reporting. Data analyses and decision for publication of CaPSURE results have always rested with academic investigators without the influence of industry. Currently, CaPSURE is supported by a combination of gifted funds from Abbott Laboratories along with a growing portfolio of federal grants to ensure registry continuity and maintenance.", "While RCTs remain the gold standard for advancing knowledge in medicine, they can be difficult to complete and expensive; moreover, in a disease with a prolonged natural history, treatment strategies may evolve quickly that even a well-executed trial may not be relevant by the time it is published. Disease registries, like CaPSURE, have provided a means to better understand many aspects of prostate cancer epidemiology, practice patterns, outcomes, and costs of care. It remains a robust source of information and provides a cost-effective way of driving evidence-based decisions regarding treatment and health policy. Specialized observational disease registries such as CaPSURE provide insight and have broad implications for disease management and policy." ]

[ "introduction", null, null, null, null, "conclusion" ]

[ "Prostatic neoplasms", "CaPSURE", "Disease registries" ]

[ "Animals", "Cattle", "Chromosomes, Mammalian", "Female", "Lactation", "Microsatellite Repeats", "Milk", "Pedigree", "Polymorphism, Single Nucleotide", "Quantitative Trait Loci" ]

[ "Background", "Animals and phenotype", "DNA preparation, microsatellite marker selection and genotyping", "SNP selection, genotyping and haplotyping", "Linkage map construction", "QTL fine mapping", "LDL mapping by microsatellite markers", "LDL mapping by SNPs", "Multiple-QTL analysis using linkage disequilibrium and linkage (LDL) analysis method", "Estimation of model parameters and test statistics", "Results", "Genotypes and linkage map construction", "Haplotype analysis in a complex pedigree", "Combined linkage disequilibrium and linkage analysis", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Recent developments in molecular biology and statistical methodologies for quantitative trait loci (QTL) mapping have made it possible to identify genetic factors affecting economically important traits. Such developments have the potential to significantly increase the rate of genetic improvement of livestock species, through marker-assisted selection of specific loci, genome-wide selection, gene introgression and positional cloning [1]. However, after an initial exaggerated enthusiasm animal geneticists, like their colleagues in human genetics e.g. [2] have faced somewhat unexpected challenges.\nThe first step in QTL mapping usually involves a complete or partial genome scan, where the mapping population is genotyped for markers covering the entire genome or only selected chromosomes, respectively. The QTL are then mapped using linkage analysis (LA) methods. The resolution of this mapping approach is low because relatively few new recombination events are generated in the single generation separating parents and progeny. Typically, the size of confidence intervals for the most likely QTL positions ranges between 20 and 40 cM.\nFine-mapping approaches have been developed to reduce these confidence intervals e.g. [3-5], leading in some instances to the identification of the underlying causal mutation [6-9]. These approaches are usually based on the addition of new families, new markers and the use of statistical methods combining linkage-disequilibrium and linkage (LDL) analysis. In general, the marker density is increased by adding a few tens of new markers (microsatellite markers or single nucleotide polymorphism (SNP)) identified within the QTL region or candidate gene.\nAt present, high-throughput SNP analysis provides the opportunity to genotype many animals for hundreds or even thousands of SNP per bovine chromosome [10-12]. Therefore, the limiting factors in QTL fine-mapping studies have now switched partly from marker density to the applied methods and designs. Use of linkage-disequilibrium (LD) information increases the precision of QTL mapping because it exploits the entire number of recombinations accumulated since the original mutation generating the new QTL allele occurred [13].\nThe degree of LD in livestock populations has attracted much attention because it provides useful information regarding the possibility of fine-mapping QTL and the potential to use marker-assisted selection. In cattle, previous reports using a low density microsatellite map (10 cM interval on average) and Hedrick's normalized measure of LD [14] D' have shown that LD extends over several tens of centimorgans [10,15,16]. However, an exceedingly low long-range and non-syntenic LD has been estimated [17] when evaluated by the standardized chi-square measure of LD, which is related to the predictive ability of LD. Nevertheless, the extent of LD in cattle [18] is greater than in humans [19] but smaller than in dog [20].\nCombined linkage disequilibrium and linkage (LDL) analysis [3] makes it possible to exploit recombinations occurring both within and outside the pedigree and genotyped population. It also gives a clearer signal for QTL positions compared with LA or LD mapping alone [3]. Additionally, the LDL approach reduces the risk of false-positive QTL identification caused by accidental marker-phenotype associations when LA and LD are used separately, and also increases the power and resolution of QTL mapping by combining all available information [21].\nIn dairy cattle, several studies have reported the presence of one or more QTL affecting milk production traits on BTA5 e.g. [22-25], but the results differ among studies with respect to the number of QTL detected, their positions, and the extent to which the milk traits are affected by the QTL.\nThe present study aimed at refining the previously detected QTL affecting milk yield (MY1), milk protein yield (PY1) and milk fat yield (FY1) during first lactation in the distal part of BTA5 in the Fleckvieh dual-purpose cattle breed [24], and to define the candidate region for high-throughput sequencing. To achieve this, we sampled additional families carrying the low frequency allele of the putative QTL (minor QTL allele) and genotyped additional markers covering the most likely QTL region on BTA5. These new families were identified by combining results from QTL-mapping based on microsatellites and haplotype analysis based on SNP in a complex pedigree. Single- and multiple-QTL analyses based on the LDL method were performed in different sample-sets, in order to allocate the minor QTL allele to specific families and to use the increased frequency of the minor QTL allele for refined mapping.", "In this study, we analysed the same nine granddaughter (GD) families used in our previous study [24], in which we identified three GD families (G01, G02 and G03) as heterozygous for a QTL located in the distal region of BTA5. The grandsires of these three GD families are designated as G01, G02 and G03, respectively. Grandsires G01 and G02 are half-sibs and have inherited the same haplotype in the distal region of BTA5 from their common ancestor A0 [24]. By target sampling (see haplotyping section, below), we introduced two additional GD families; family G10 with 85 sons, and family G11 with 47 sons. Grandsire G10 (grandsire of family G10), was connected through his dam to A0. Grandsire G11 (grandsire of family G11) is a son of grandsire G02. In addition, we identified all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11 to add more, possibly recombinant, A0 haplotypes into the mapping population. In this way, we created three maternal grandsire (MGS) families, M02 with 21 grandsons, M10 with 32 grandsons and M11 with 33 grandsons, descendants of grandsires G02, G10 and G11, respectively. Samples of maternal grandsons were not available for grandsire G01. Thus, the analysis included 11 GD families: G01 to G11 and three MGS families (M02, M10 and M11). Figure 1 shows the relationships of all families included in this study. In some cases, mapping analyses were carried out on 173 additional animals available from other projects that are not descended from ancestor A0. Estimated breeding values (EBV) of the Fleckvieh bulls for milk production traits MY1, PY1, and FY1, (along with their reliability values) were obtained from the 2009 joint Austria-Germany genetic evaluation of the Fleckvieh population [26].\nFamilial relationships considered in this study and segregation of most important haplotypes. A complex pedigree of 38 sires (squares) of GD families (G), ten sires of daughter design (DD) families, three maternal grandsire (M) families and 26 sampled and genotyped relevant ancestors; the pedigree has been simplified by showing only ancestors who made it possible to trace haplotypes from family-sires to the most important ancestors (A0, A1, A2); furthermore, to reduce the complexity of the figure, ancestor A1 is represented more than once; correspondingly, letters and numbers within squares of family-sires represent the internal family ID; non-genotyped individuals are represented by smaller circles (females) and squares (males) marked with a diagonal line; the estimated haplotype of 25 markers (A0H1) comprising a derived QTL allele affecting MY1 and PY1 with 97% CI between 117.962 Mb and 119.018 Mb is graphically presented by yellow bars above the individual's symbol; five other most frequent haplotypes are represented by five different coloured bars; introgression of Red-Holstein genes into the mapping populations is represented by ancestor A2 and the corresponding haplotype presented by a red bar; to reduce the complexity of the figure, 77 low frequency haplotypes are omitted; the allelic composition of the respective haplotypes is presented within the figure; the pedigree MSPED2089 is a subset of the total material which can be constructed by keeping the families marked by a grey circle around squares and associated ancestors; pedigrees MSPED1038 and SNPPED421 are subsets of MSPED2089 which can be constructed by removing appropriate families as described in material and methods; the pedigree SNPPED308 consists of GD family G36 and animals across the entire mapping population but not descending from A0; the pedigree SNPPED723 is a sum of pedigrees SNPPED308 and SNPPED421\n.", "Genomic DNA was prepared from semen using standard methods, and from whole blood samples with QIAamp Blood-Kits (Qiagen), according to the manufacturer's protocol.\nTwelve evenly distributed microsatellite markers were added to the 28 microsatellite markers used in the previous study [24]. Twenty-one of these 40 microsatellite markers covered the most likely region containing the QTL in the distal part of BTA5 (Table 1) and were used in most analyses of the present study. Previously analysed animals were genotyped only for the new markers, but the five new families (G10, G11, M02, M10 and M11) were genotyped for all marker sets [24]. For 11 of the 12 markers, relevant information was obtained from the MARC-ARS-USDA public database at http://www.ars.usda.gov/Main/docs.htm?docid=12539 \n[27]. The new marker LMU0505 was obtained by a targeted search for dinucleotide repeats in genomic regions with a low marker density. The unique sequences flanking the newly identified dinucleotide repeats were tested for informativity by genotyping a small set of animals first. Primers for the 12 new microsatellite markers were optimized using Primer3 (v.0.4.0) according to the bovine genome sequence data currently available (i.e. Baylor release Btau_4.0, http://genome.ucsc.edu/cgi-bin/hgGateway) and the appropriate fragment size in the currently designed marker set. New markers were divided into two PCR multiplex sets (Table 1) that were combined again after PCR for electrophoresis and fragment analysis. The fragment analysis of the PCR products was performed on ABI377 and ABI Prism 310 sequencers. Genotypes were assigned using GENESCAN and GENOTYPER (Applied Biosystems) software programs. We performed double genotyping of all families and ancestors using two independent runs. For ambiguous genotypes, the raw data were re-evaluated and animals were re-genotyped if necessary.\nMicrosatellite markers used for QTL mapping\nMarker name, relative position (cM), physical position (bp), forward and reverse primer sequences and marker set (set: Set0 & Set1 as in previous study; Set3 and Set4 comprise multiplex 1&2 in this study)", "SNP genotyping was carried out by Tierzuchtforschung e. V. München using the commercial Illumina Bovine SNP50 Bead chip featuring 54 001 SNP (http://www.illumina.com/; Illumina, San Diego) that span the bovine genome, excluding Y-chromosome. The genotype calling was performed with the GenCall application, as implemented in Illumina Bead chip Genotyping analysis software. This application computes a Gencall score for each locus, which evaluates the quality of genotypes. We included only animals with confirmed paternity and with a call rate above 0.98. Furthermore, we only used markers with a call rate above 0.90. We excluded all markers producing more than 1% paternity problems in pairs with confirmed paternity, and also excluded all markers that were non-informative in the Fleckvieh population or with an unknown chromosomal position. This yielded 43 806 informative SNP available for the whole-genome analysis in the Fleckvieh population, of which 1 976 are found on the BTA5. Two hundred and forty of these covered the region most likely containing the QTL in the distal part of BTA5 and were used in the present study.\nWe performed SNP genotyping in two stages. First, 75 animals i.e. the gransires of the nine initial GD families and their ancestors, and also a number of potential GD-family sires and their ancestors, were genotyped with the SNP chip and their haplotypes were reconstructed with the BEAGLE program [28]. These 75 animals constitute a complex pedigree (Figure 1) in which it is possible to trace the segregating haplotypes five generations back to some important ancestors of the Fleckvieh population, born in the 1960's and 1970's. This pedigree represents almost all of the important bull lines originating from a wide range of dams. Considering this, and the fact that a large proportion of the included bull dams are unrelated (no common grand-parents), these 75 animals provide a good representation of the haplotype diversity in the breeding Fleckvieh population. Second, the new families (G10, G11, M02, M10 and M11) containing the target haplotype segment of ancestor A0 were genotyped with microsatellite markers and with the genome-wide SNP chip. These animals and 173 additional Fleckvieh animals not closely related to ancestor A0 (but genotyped with the SNP chip in other projects running in our laboratory) were also haplotyped using the BEAGLE program.", "The relative positions of microsatellite markers were re-evaluated by the CRI-MAP program [29]. A physical map was constructed according to the sequence data of all the markers (Table 1) using the basic alignment search tool (BLAST) and the latest cattle genome sequence http://genome.ucsc.edu/cgi-bin/hgGateway. Our genetic data was used to resolve cases where more than one marker order was obtained from published linkage and physical maps. When our genetic data supported a marker order different from that of the public linkage map, but in accordance to the physical map, we modified the relative position (cM) of the markers along with the corresponding sequence. The linkage and physical maps were used as a framework to insert the newly designed marker (LMU0505) with the build option of the CRI-MAP program. The resulting final map (Table 1) was used for all the following analyses.", "[SUBTITLE] LDL mapping by microsatellite markers [SUBSECTION] Joint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).\nJoint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).\n[SUBTITLE] LDL mapping by SNPs [SUBSECTION] Here we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.\nHere we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.", "Joint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).", "Here we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.", "We used the analysis method of Olsen et al. [33], i.e. the same model as for single-QTL analysis, but including a random QTL effect of a specified marker bracket. That is, the bracket that showed the highest LRT in the single-QTL analysis was included as a random effect in the QTL model in turn, and the analysis was repeated. These analyses searched for an additional QTL, given that the QTL in the specified marker bracket is accounted for, and is similar to the fitting of cofactors [34].", "The variance components and the logarithm of the likelihood (L) of a model containing a QTL as well as residual polygenic effects at position p (logLp) were estimated by AIREML [32,35], which is an integral part of the LDLRAMS and LDL programs. The likelihood of a model without QTL effect (logL0) was calculated on the basis of a polygenic model. The log-likelihood ratio (LRT) was calculated as double difference in logL between models with and without a QTL, i.e. LRT = -2 (logL0-logLp). The LRT test statistic is distributed approximately as chi-square with 1 degree of freedom [36]. The confidence interval (CI) for the QTL position was determined as 1-LOD support interval, which was constructed as the interval surrounding the QTL peak where the LRT exceeds LRTmax - 2 × ln (10), where LRTmax is the maximum LRT-value for the tested QTL [37].", "[SUBTITLE] Genotypes and linkage map construction [SUBSECTION] Genotypes for 40 microsatellite markers were available to build the BTA5 genetic map. In most of the LDL analyses, only the 21 most distal markers (Table 1) covering the 97% confidence interval were considered. When we controlled if the genotype and haplotype data were plausible, the most distal marker (MNB71), which was genotyped in previous projects [24], showed extensive double recombinations with the 12 markers added in the present project. To reduce possible mapping errors, we excluded this marker from all subsequent analyses. Using the build option of the CRI-MAP program, we re-estimated the marker distances and order.\nThe following changes with respect to the public USDA linkage map were made: (i) according to the physical map (i.e. bp position of release Btau_4.0) and confirmed by applying the build option of the CRI-MAP program to our own data, the positions of markers BM49 and BM733 are inverted (Table 1); (ii) markers DIK2035 and DIK5277 are both at the same position (120.85 cM) on the USDA linkage map but, according to our genotypes and the physical map results, they are separated, placing DIK2035 (120.38 cM) upstream of DIK5277 (120.82 cM); (iii) the new marker developed in this study (LMU0505) is highly informative for linkage analysis and its relative position between DIK5106 and ETH152 was estimated by applying the build option of the CRI-MAP program. The positions of both flanking markers DIK5106 and ETH152 also changed (Table 1).\nGenotypes for 40 microsatellite markers were available to build the BTA5 genetic map. In most of the LDL analyses, only the 21 most distal markers (Table 1) covering the 97% confidence interval were considered. When we controlled if the genotype and haplotype data were plausible, the most distal marker (MNB71), which was genotyped in previous projects [24], showed extensive double recombinations with the 12 markers added in the present project. To reduce possible mapping errors, we excluded this marker from all subsequent analyses. Using the build option of the CRI-MAP program, we re-estimated the marker distances and order.\nThe following changes with respect to the public USDA linkage map were made: (i) according to the physical map (i.e. bp position of release Btau_4.0) and confirmed by applying the build option of the CRI-MAP program to our own data, the positions of markers BM49 and BM733 are inverted (Table 1); (ii) markers DIK2035 and DIK5277 are both at the same position (120.85 cM) on the USDA linkage map but, according to our genotypes and the physical map results, they are separated, placing DIK2035 (120.38 cM) upstream of DIK5277 (120.82 cM); (iii) the new marker developed in this study (LMU0505) is highly informative for linkage analysis and its relative position between DIK5106 and ETH152 was estimated by applying the build option of the CRI-MAP program. The positions of both flanking markers DIK5106 and ETH152 also changed (Table 1).\n[SUBTITLE] Haplotype analysis in a complex pedigree [SUBSECTION] Using the algorithm implemented into the program BEAGLE, we haplotyped the 75 animals of the complex pedigree in Figure 1 with 1 976 SNP on BTA5 that are informative in the Fleckvieh population. Thus reconstructed haplotypes were used to identify families segregating for the QTL detected in the initial study [24]. As already shown by the microsatellite analysis, the grandsires of families G01 and G02 which are heterozygous at the QTL, inherited the same haplotype in the distal region of BTA5 from their ancestor A0 (Figure 1). This was confirmed by the haplotype reconstruction using the 1 976 SNP. This A0 ancestral haplotype is named \"haplotype 1\" or (A0H1) and its A0 alternative haplotype \"haplotype 2\" or (A0H2). Family G03, previously declared as heterozygous for the target QTL [24] but not identified here, has inherited haplotypes not related to A0H1 (Figure 1). All animals with haplotype A0H1 (surrounding the putative QTL position) can be traced back to A0. Two of these, grandsires G10 and G11 are paternal and maternal grandsons of A0, and are very important Fleckvieh bull sires. We have collected samples of all the available progeny-tested sons of these two grandsires and all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11, to add more recombinant A0 haplotypes into the mapping population. In total, 485 animals were genotyped by the SNP chip and haplotyped for BTA5. By calculating the independent haplotypes in the complex pedigrees, and considering the traceability of all A0H1 haplotypes to A0, we estimated a very low frequency (<0.005) of A0H1 in the Fleckvieh population. Consequently, throughout the rest of this paper, the less frequent putative QTL allele embedded in this less frequent haplotype is referred to as the minor QTL allele.\nUsing the algorithm implemented into the program BEAGLE, we haplotyped the 75 animals of the complex pedigree in Figure 1 with 1 976 SNP on BTA5 that are informative in the Fleckvieh population. Thus reconstructed haplotypes were used to identify families segregating for the QTL detected in the initial study [24]. As already shown by the microsatellite analysis, the grandsires of families G01 and G02 which are heterozygous at the QTL, inherited the same haplotype in the distal region of BTA5 from their ancestor A0 (Figure 1). This was confirmed by the haplotype reconstruction using the 1 976 SNP. This A0 ancestral haplotype is named \"haplotype 1\" or (A0H1) and its A0 alternative haplotype \"haplotype 2\" or (A0H2). Family G03, previously declared as heterozygous for the target QTL [24] but not identified here, has inherited haplotypes not related to A0H1 (Figure 1). All animals with haplotype A0H1 (surrounding the putative QTL position) can be traced back to A0. Two of these, grandsires G10 and G11 are paternal and maternal grandsons of A0, and are very important Fleckvieh bull sires. We have collected samples of all the available progeny-tested sons of these two grandsires and all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11, to add more recombinant A0 haplotypes into the mapping population. In total, 485 animals were genotyped by the SNP chip and haplotyped for BTA5. By calculating the independent haplotypes in the complex pedigrees, and considering the traceability of all A0H1 haplotypes to A0, we estimated a very low frequency (<0.005) of A0H1 in the Fleckvieh population. Consequently, throughout the rest of this paper, the less frequent putative QTL allele embedded in this less frequent haplotype is referred to as the minor QTL allele.\n[SUBTITLE] Combined linkage disequilibrium and linkage analysis [SUBSECTION] Thirty-seven microsatellite markers (three markers BM6026, BMS610 and MNB71 showed extensive recombinations and were excluded) and the complex pedigree MSPED2089 were used for initial LDL mapping analyses. As shown in Figure 2, we observed a highly significant QTL effect (LRT = 20 to 22, i.e. P = 0.0000077 to 0.0000027), but were unable to improve the mapping accuracy because of the presence of two or three peaks.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 2089 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 37 microsatellites, a complex pedigree of 2 089 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nAccording to previous results [24], and to the results obtained in the first part of this study, we have assumed that haplotype A0H1 has only introduced one QTL into the mapping population. Therefore, we performed a second LDL analysis using the 21 most distal markers, and limited to GD and MGS families descending from A0 and known to carry A0H1, i.e. pedigree MSPED1038 (Figure 3). Unlike the analysis of pedigree MSPED2089, Figure 3 illustrates a single rather broad peak between positions 119.005 cM and 120.166 cM. However, this highly significant QTL (P = 0.000062 to 0.000021) is still mapped with a low accuracy, i.e. 1-LOD drop-off support intervals are 4.7 cM for FY1, 10.4 cM for PY1 and 11.5 cM for MY1.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 1 038 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 21 microsatellites covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 1 038 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nSince the confidence interval achieved by LDL analyses using pedigree MSPED1038 was still too large for a positional candidate gene approach, we analysed pedigree SNPPED723 using the LDL approach. The results were similar to those obtained with microsatellite markers and pedigree MSPED2089, namely, multiple peaks suggesting multiple QTL or no QTL (Figure 4).\nLDL analysis by variance component approach using SNP in a complex pedigree of 723 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 723 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nTo resolve this dilemma, we divided pedigree SNPPED723 into pedigree SNPPED421 consisting of all progeny-tested animals descending from ancestor A0, and pedigree SNPPED308 consisting of the remaining progeny-tested animals. The LDL analyses of SNPPED308 pedigree showed a moderately flat, non-significant test statistic along the investigated chromosomal segment (Figure 5). Only LRT values for FY1 reached an indicative level of 3.99 (P = 0.046). Conversely, it was possible to map a QTL with pedigree SNPPED421 whose minor allele is most probably originating from ancestor A0 (Figure 6). There were two distinct peaks; one with LRT values over 17 (P < 0.000037) for both MY1 and PY1 in a region of 0.5 Mb (from 118.107 to 118.606 Mb), and one with a very high LRT value for only PY1 (LRT = 20.72, P = 0.0000053) at position 122.115 Mb. Considering 1-LOD drop-off support intervals, the 97% confidence intervals were located between 117.962 Mb and 119.018 Mb (i.e. 1.056 Mb) for the QTL affecting MY1 and PY1, and between 121.800 Mb and 122.200 Mb (i.e. 0.400 Mb) for the QTL affecting only PY1. There were two additional peaks with LRT values over 15 in regions around the positions 115.650 and 116.300 Mb, but they were not included in the 97% confidence interval for PY1 and were not supported by the highly correlated MY1 trait.\nLDL analysis by variance component approach using SNP in a complex pedigree of 308 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 308 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nLDL analysis by variance component approach using SNP in a complex pedigree of 421 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covering the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 421 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. The long homozygous region (~5 Mb) in ancestor A0 was shown (A0 Homo).\nThe two identified peaks (located between 118.107 Mb and 118.606 Mb and at 122.115 Mb, respectively) may be due to either the presence of more than one QTL, or the presence of one QTL with carryover effects to another region. Thus, a multiple-QTL analysis was performed. Two-QTL analyses using pedigree SNPPED421 for MY1 and PY1 fitting a QTL at position 118.202 Mb revealed a single QTL affecting only MY1 at this location and an additional QTL affecting PY1 at position 122.115 Mb (P = 0.019). However, two-QTL analyses accounting for the QTL at position 122.115 Mb did not rule out a possible second QTL affecting PY1 at position 118.202 Mb (P = 0.019).\nThirty-seven microsatellite markers (three markers BM6026, BMS610 and MNB71 showed extensive recombinations and were excluded) and the complex pedigree MSPED2089 were used for initial LDL mapping analyses. As shown in Figure 2, we observed a highly significant QTL effect (LRT = 20 to 22, i.e. P = 0.0000077 to 0.0000027), but were unable to improve the mapping accuracy because of the presence of two or three peaks.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 2089 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 37 microsatellites, a complex pedigree of 2 089 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nAccording to previous results [24], and to the results obtained in the first part of this study, we have assumed that haplotype A0H1 has only introduced one QTL into the mapping population. Therefore, we performed a second LDL analysis using the 21 most distal markers, and limited to GD and MGS families descending from A0 and known to carry A0H1, i.e. pedigree MSPED1038 (Figure 3). Unlike the analysis of pedigree MSPED2089, Figure 3 illustrates a single rather broad peak between positions 119.005 cM and 120.166 cM. However, this highly significant QTL (P = 0.000062 to 0.000021) is still mapped with a low accuracy, i.e. 1-LOD drop-off support intervals are 4.7 cM for FY1, 10.4 cM for PY1 and 11.5 cM for MY1.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 1 038 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 21 microsatellites covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 1 038 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nSince the confidence interval achieved by LDL analyses using pedigree MSPED1038 was still too large for a positional candidate gene approach, we analysed pedigree SNPPED723 using the LDL approach. The results were similar to those obtained with microsatellite markers and pedigree MSPED2089, namely, multiple peaks suggesting multiple QTL or no QTL (Figure 4).\nLDL analysis by variance component approach using SNP in a complex pedigree of 723 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 723 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nTo resolve this dilemma, we divided pedigree SNPPED723 into pedigree SNPPED421 consisting of all progeny-tested animals descending from ancestor A0, and pedigree SNPPED308 consisting of the remaining progeny-tested animals. The LDL analyses of SNPPED308 pedigree showed a moderately flat, non-significant test statistic along the investigated chromosomal segment (Figure 5). Only LRT values for FY1 reached an indicative level of 3.99 (P = 0.046). Conversely, it was possible to map a QTL with pedigree SNPPED421 whose minor allele is most probably originating from ancestor A0 (Figure 6). There were two distinct peaks; one with LRT values over 17 (P < 0.000037) for both MY1 and PY1 in a region of 0.5 Mb (from 118.107 to 118.606 Mb), and one with a very high LRT value for only PY1 (LRT = 20.72, P = 0.0000053) at position 122.115 Mb. Considering 1-LOD drop-off support intervals, the 97% confidence intervals were located between 117.962 Mb and 119.018 Mb (i.e. 1.056 Mb) for the QTL affecting MY1 and PY1, and between 121.800 Mb and 122.200 Mb (i.e. 0.400 Mb) for the QTL affecting only PY1. There were two additional peaks with LRT values over 15 in regions around the positions 115.650 and 116.300 Mb, but they were not included in the 97% confidence interval for PY1 and were not supported by the highly correlated MY1 trait.\nLDL analysis by variance component approach using SNP in a complex pedigree of 308 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 308 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nLDL analysis by variance component approach using SNP in a complex pedigree of 421 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covering the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 421 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. The long homozygous region (~5 Mb) in ancestor A0 was shown (A0 Homo).\nThe two identified peaks (located between 118.107 Mb and 118.606 Mb and at 122.115 Mb, respectively) may be due to either the presence of more than one QTL, or the presence of one QTL with carryover effects to another region. Thus, a multiple-QTL analysis was performed. Two-QTL analyses using pedigree SNPPED421 for MY1 and PY1 fitting a QTL at position 118.202 Mb revealed a single QTL affecting only MY1 at this location and an additional QTL affecting PY1 at position 122.115 Mb (P = 0.019). However, two-QTL analyses accounting for the QTL at position 122.115 Mb did not rule out a possible second QTL affecting PY1 at position 118.202 Mb (P = 0.019).", "Genotypes for 40 microsatellite markers were available to build the BTA5 genetic map. In most of the LDL analyses, only the 21 most distal markers (Table 1) covering the 97% confidence interval were considered. When we controlled if the genotype and haplotype data were plausible, the most distal marker (MNB71), which was genotyped in previous projects [24], showed extensive double recombinations with the 12 markers added in the present project. To reduce possible mapping errors, we excluded this marker from all subsequent analyses. Using the build option of the CRI-MAP program, we re-estimated the marker distances and order.\nThe following changes with respect to the public USDA linkage map were made: (i) according to the physical map (i.e. bp position of release Btau_4.0) and confirmed by applying the build option of the CRI-MAP program to our own data, the positions of markers BM49 and BM733 are inverted (Table 1); (ii) markers DIK2035 and DIK5277 are both at the same position (120.85 cM) on the USDA linkage map but, according to our genotypes and the physical map results, they are separated, placing DIK2035 (120.38 cM) upstream of DIK5277 (120.82 cM); (iii) the new marker developed in this study (LMU0505) is highly informative for linkage analysis and its relative position between DIK5106 and ETH152 was estimated by applying the build option of the CRI-MAP program. The positions of both flanking markers DIK5106 and ETH152 also changed (Table 1).", "Using the algorithm implemented into the program BEAGLE, we haplotyped the 75 animals of the complex pedigree in Figure 1 with 1 976 SNP on BTA5 that are informative in the Fleckvieh population. Thus reconstructed haplotypes were used to identify families segregating for the QTL detected in the initial study [24]. As already shown by the microsatellite analysis, the grandsires of families G01 and G02 which are heterozygous at the QTL, inherited the same haplotype in the distal region of BTA5 from their ancestor A0 (Figure 1). This was confirmed by the haplotype reconstruction using the 1 976 SNP. This A0 ancestral haplotype is named \"haplotype 1\" or (A0H1) and its A0 alternative haplotype \"haplotype 2\" or (A0H2). Family G03, previously declared as heterozygous for the target QTL [24] but not identified here, has inherited haplotypes not related to A0H1 (Figure 1). All animals with haplotype A0H1 (surrounding the putative QTL position) can be traced back to A0. Two of these, grandsires G10 and G11 are paternal and maternal grandsons of A0, and are very important Fleckvieh bull sires. We have collected samples of all the available progeny-tested sons of these two grandsires and all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11, to add more recombinant A0 haplotypes into the mapping population. In total, 485 animals were genotyped by the SNP chip and haplotyped for BTA5. By calculating the independent haplotypes in the complex pedigrees, and considering the traceability of all A0H1 haplotypes to A0, we estimated a very low frequency (<0.005) of A0H1 in the Fleckvieh population. Consequently, throughout the rest of this paper, the less frequent putative QTL allele embedded in this less frequent haplotype is referred to as the minor QTL allele.", "Thirty-seven microsatellite markers (three markers BM6026, BMS610 and MNB71 showed extensive recombinations and were excluded) and the complex pedigree MSPED2089 were used for initial LDL mapping analyses. As shown in Figure 2, we observed a highly significant QTL effect (LRT = 20 to 22, i.e. P = 0.0000077 to 0.0000027), but were unable to improve the mapping accuracy because of the presence of two or three peaks.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 2089 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 37 microsatellites, a complex pedigree of 2 089 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nAccording to previous results [24], and to the results obtained in the first part of this study, we have assumed that haplotype A0H1 has only introduced one QTL into the mapping population. Therefore, we performed a second LDL analysis using the 21 most distal markers, and limited to GD and MGS families descending from A0 and known to carry A0H1, i.e. pedigree MSPED1038 (Figure 3). Unlike the analysis of pedigree MSPED2089, Figure 3 illustrates a single rather broad peak between positions 119.005 cM and 120.166 cM. However, this highly significant QTL (P = 0.000062 to 0.000021) is still mapped with a low accuracy, i.e. 1-LOD drop-off support intervals are 4.7 cM for FY1, 10.4 cM for PY1 and 11.5 cM for MY1.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 1 038 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 21 microsatellites covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 1 038 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nSince the confidence interval achieved by LDL analyses using pedigree MSPED1038 was still too large for a positional candidate gene approach, we analysed pedigree SNPPED723 using the LDL approach. The results were similar to those obtained with microsatellite markers and pedigree MSPED2089, namely, multiple peaks suggesting multiple QTL or no QTL (Figure 4).\nLDL analysis by variance component approach using SNP in a complex pedigree of 723 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 723 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nTo resolve this dilemma, we divided pedigree SNPPED723 into pedigree SNPPED421 consisting of all progeny-tested animals descending from ancestor A0, and pedigree SNPPED308 consisting of the remaining progeny-tested animals. The LDL analyses of SNPPED308 pedigree showed a moderately flat, non-significant test statistic along the investigated chromosomal segment (Figure 5). Only LRT values for FY1 reached an indicative level of 3.99 (P = 0.046). Conversely, it was possible to map a QTL with pedigree SNPPED421 whose minor allele is most probably originating from ancestor A0 (Figure 6). There were two distinct peaks; one with LRT values over 17 (P < 0.000037) for both MY1 and PY1 in a region of 0.5 Mb (from 118.107 to 118.606 Mb), and one with a very high LRT value for only PY1 (LRT = 20.72, P = 0.0000053) at position 122.115 Mb. Considering 1-LOD drop-off support intervals, the 97% confidence intervals were located between 117.962 Mb and 119.018 Mb (i.e. 1.056 Mb) for the QTL affecting MY1 and PY1, and between 121.800 Mb and 122.200 Mb (i.e. 0.400 Mb) for the QTL affecting only PY1. There were two additional peaks with LRT values over 15 in regions around the positions 115.650 and 116.300 Mb, but they were not included in the 97% confidence interval for PY1 and were not supported by the highly correlated MY1 trait.\nLDL analysis by variance component approach using SNP in a complex pedigree of 308 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 308 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nLDL analysis by variance component approach using SNP in a complex pedigree of 421 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covering the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 421 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. The long homozygous region (~5 Mb) in ancestor A0 was shown (A0 Homo).\nThe two identified peaks (located between 118.107 Mb and 118.606 Mb and at 122.115 Mb, respectively) may be due to either the presence of more than one QTL, or the presence of one QTL with carryover effects to another region. Thus, a multiple-QTL analysis was performed. Two-QTL analyses using pedigree SNPPED421 for MY1 and PY1 fitting a QTL at position 118.202 Mb revealed a single QTL affecting only MY1 at this location and an additional QTL affecting PY1 at position 122.115 Mb (P = 0.019). However, two-QTL analyses accounting for the QTL at position 122.115 Mb did not rule out a possible second QTL affecting PY1 at position 118.202 Mb (P = 0.019).", "The aim of this study was to refine the position of a previously mapped QTL by increasing the marker density in the region, target sampling of additional families and adapting fine mapping methods. According to our previous results [24] and to results from the initial part of this study, we hypothesized the presence of a minor QTL allele with a strong effect, but at a very low frequency, in the Fleckvieh dual-purpose cattle breed. In such a situation, random sampling of additional families for confirmation and fine-mapping purposes can result in an increased frequency of the common QTL allele in the mapping design Thus, the capacity to differentiate between genetic background noise and the initially targeted QTL will be decreased. The reduced accuracy of QTL position estimates when using all genotyped animals (pedigrees MSPED2089 or SNPPED723) compared to a subset of animals (pedigrees MSPED1038 or SNPPED421) is counterintuitive to the general notion that the use of more information should result in better estimates. To further explore this unexpected result, we have investigated several possible explanations, including the effects of the haplotype distribution and the possibility of additional QTL. To study the haplotype distribution in the Fleckvieh population, 485 animals were genotyped with the Illumina 50 K SNP chip. Of these, a subset of 144 animals were not progeny-tested and not relevant for QTL mapping, but were very informative for the study of haplotype distribution. In particular, considering the putative QTL affecting MY1 and PY1 located within the 97% CI (between 117.962 Mb and 119.018 Mb), a haplotype of 25 markers (A0H1) covering this region was detected in 89 of 485 animals. This haplotype A0H1, most probably carrying the minor QTL allele, could be traced back to the ancestor A0 in all 89 cases (Figure 1). The alternative haplotype A0H2, most probably carrying the common QTL allele, was found in 13 cases but was traced back to the ancestor A0 only in three. A perfect LD between the minor QTL allele and A0H1 (and only A0H1) would result in a relatively low allele frequency (0.137) of the minor QTL allele in phenotyped animals of pedigree SNPPED723, and in a frequency about double (0.254) in pedigree SNPPED421. The mapping results did reflect this difference too.\nIn contrast, consider the six markers located within the 97% CI (between 121.800 Mb and 122.200 Mb) of the putative QTL region affecting only PY1. Ancestor A0 is homozygous for a very long segment of this region i.e. from positions 118.266 Mb to 123.347 Mb (three SNP telomeric to the main peak of QTL affecting MY1 and PY1). This segment of 5.080 Mb includes 109 informative markers in the Fleckvieh population. Comparison of mapping results from pedigrees SNPPED723\n(Figures 4), SNPPED421 (Figure 6),and SNPPED308 (Figure 5) revealed a highly significant QTL allele affecting PY1 only when the pedigree included families segregating for haplotype A0H1 (see comparison between Figures 4 and 6). Excluding these families yielded LRT values below 3.99 (P > 0.045) for all three milk yield traits and for the complete investigated region (Figure 5, between 113.500 Mb and 123.700 Mb). We therefore mainly used the linkage information in the SNPPED421 pedigree (A0H1 always traceable to A0), to map a QTL affecting both MY1 and PY1 in a 97% CI of 1 Mb.\nHaplotype and LDL analyses by microsatellite markers (Figures 2 and 3) and SNP (Figures 4 and 6) clearly suggest that the minor QTL allele associated with the putative QTL around the physical position 118 Mb (97% CI between 117.962 Mb to 119.018 Mb) has been introduced by ancestor A0 into the mapping population. The explanation of the second possible QTL that maps to the physical position 122.115 Mb and affects only PY1 is different. First, this QTL should also be associated with ancestor A0 haplotypes, i.e. absence of effect in the smaller SNPPED308 pedigree (Figure 5). Second, both ancestor haplotypes at the physical position 122.115 Mb are most probably identical by descent (i.e. homozygous for a 5.080 Mb segment with 109 informative SNP). Therefore, ancestor A0 is most probably homozygous for the putative QTL at this position too. Third, this part of the haplotype is not unique to A0, but also segregates in other families, i.e. there is LD information for mapping, too. The relatively sharp LRT peak at position 122.115 Mb and homozygosity of A0 suggest an essential contribution of LD to this mapping result. Fourth, analyses with the two-QTL model did not rule out the possibility of a second QTL affecting PY1 within the candidate region on BTA5. And finally, despite the overall presence of haplotypes with a high IBD to ancestor haplotypes around position 122.115 Mb, the complete absence of this peak in SNPPED308 pedigree can be explained by either a novel mutation in ancestor A0 or by the incapacity of the method and design used here to map it in a relatively small pedigree like SNPPED308. More reasonable explanations may be the lower statistical power of the pedigree SNPPED308, possible local inconsistencies in the map order (which was based on map release Btau_4.0), the presence of a strong QTL at position 118.000 Mb with carryover effects to other regions, or a combination of all these explanations.\nThe LDL analysis using SNPs and pedigree SNPPED723 indicate several peaks affecting MY1 and PY1 in the region investigated here. In principle, these results (Figure 4) are comparable to the fine-mapping results reported on BTA3 by Druet et al. [38]. In this study, the authors have also first carried out mapping by linkage analysis and finally ended up with LDL analyses and multiple LRT peaks. We used larger overlapping marker windows (80 SNP) than Druet et al. [38]. By dividing the data set according to the results of linkage and haplotype analyses, most of the multiple peaks were explained as genetic background noise in a larger family set. The multiple peak profile could be explained by the heterogeneous LD structure within the QTL region or by the use of LD in the model when there is no LD information at all [38]. This might be increased by possible local inconsistencies in the map order, which was based on the draft assembly, or on comparative map information. Moreover, the method and the data structure may not make it possible to discard some regions even though they do not harbour the QTL [38].\nTo check for possible effects of the data structure on the reported mapping results, we tested regression of EBV on genetic distance from ancestor A0 for all carriers of haplotype 1 (A0H1). The apparent lack of this regression suggests that we are looking at a real QTL effect, and not an artifact of pedigree-tracking.\nSearching the region between 117.900 and 119.100 Mb for candidate genes revealed 27 genes, 13 of which had no known function. Based on current biological information, the genes with partly known function could only be indirectly related to milk yield traits.", "In the present study, we have performed a haplotype-assisted extension of the mapping design and thus increased the allele frequency of the minor QTL allele in mapping families. Alternative analyses with family subsets resulted in a substantial reduction of the genetic background noise and an increased frequency of the minor QTL allele. Using these subsets, we succeeded in refining the map position of the previously detected QTL for milk production traits on BTA5 to a 1 Mb interval. In spite of implementing a two-QTL analysis, the possibility of a second QTL affecting only PY1 could not be ruled out. All in all, the results of both this study and the previous study by Awad et al. [24] support the presence of a QTL affecting both MY1 and PY1 that is close to the centromeric part of the long homozygous region (~5 Mb) in ancestor A0. Therefore, positional cloning and high-throughput sequencing of the candidate region located between 117.900 Mb and 119.100 Mb should now be considered, but should also not neglect the second possible QTL around position 122.115 Mb.", "The authors declare that they have no competing interests.", "AA carried out DNA extraction, microsatellite genotyping; AA and IM performed all data analysis and wrote the paper; IM and MF designed the study; IR performed SNP genotyping and partly performed sampling. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Animals and phenotype", "DNA preparation, microsatellite marker selection and genotyping", "SNP selection, genotyping and haplotyping", "Linkage map construction", "QTL fine mapping", "LDL mapping by microsatellite markers", "LDL mapping by SNPs", "Multiple-QTL analysis using linkage disequilibrium and linkage (LDL) analysis method", "Estimation of model parameters and test statistics", "Results", "Genotypes and linkage map construction", "Haplotype analysis in a complex pedigree", "Combined linkage disequilibrium and linkage analysis", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Recent developments in molecular biology and statistical methodologies for quantitative trait loci (QTL) mapping have made it possible to identify genetic factors affecting economically important traits. Such developments have the potential to significantly increase the rate of genetic improvement of livestock species, through marker-assisted selection of specific loci, genome-wide selection, gene introgression and positional cloning [1]. However, after an initial exaggerated enthusiasm animal geneticists, like their colleagues in human genetics e.g. [2] have faced somewhat unexpected challenges.\nThe first step in QTL mapping usually involves a complete or partial genome scan, where the mapping population is genotyped for markers covering the entire genome or only selected chromosomes, respectively. The QTL are then mapped using linkage analysis (LA) methods. The resolution of this mapping approach is low because relatively few new recombination events are generated in the single generation separating parents and progeny. Typically, the size of confidence intervals for the most likely QTL positions ranges between 20 and 40 cM.\nFine-mapping approaches have been developed to reduce these confidence intervals e.g. [3-5], leading in some instances to the identification of the underlying causal mutation [6-9]. These approaches are usually based on the addition of new families, new markers and the use of statistical methods combining linkage-disequilibrium and linkage (LDL) analysis. In general, the marker density is increased by adding a few tens of new markers (microsatellite markers or single nucleotide polymorphism (SNP)) identified within the QTL region or candidate gene.\nAt present, high-throughput SNP analysis provides the opportunity to genotype many animals for hundreds or even thousands of SNP per bovine chromosome [10-12]. Therefore, the limiting factors in QTL fine-mapping studies have now switched partly from marker density to the applied methods and designs. Use of linkage-disequilibrium (LD) information increases the precision of QTL mapping because it exploits the entire number of recombinations accumulated since the original mutation generating the new QTL allele occurred [13].\nThe degree of LD in livestock populations has attracted much attention because it provides useful information regarding the possibility of fine-mapping QTL and the potential to use marker-assisted selection. In cattle, previous reports using a low density microsatellite map (10 cM interval on average) and Hedrick's normalized measure of LD [14] D' have shown that LD extends over several tens of centimorgans [10,15,16]. However, an exceedingly low long-range and non-syntenic LD has been estimated [17] when evaluated by the standardized chi-square measure of LD, which is related to the predictive ability of LD. Nevertheless, the extent of LD in cattle [18] is greater than in humans [19] but smaller than in dog [20].\nCombined linkage disequilibrium and linkage (LDL) analysis [3] makes it possible to exploit recombinations occurring both within and outside the pedigree and genotyped population. It also gives a clearer signal for QTL positions compared with LA or LD mapping alone [3]. Additionally, the LDL approach reduces the risk of false-positive QTL identification caused by accidental marker-phenotype associations when LA and LD are used separately, and also increases the power and resolution of QTL mapping by combining all available information [21].\nIn dairy cattle, several studies have reported the presence of one or more QTL affecting milk production traits on BTA5 e.g. [22-25], but the results differ among studies with respect to the number of QTL detected, their positions, and the extent to which the milk traits are affected by the QTL.\nThe present study aimed at refining the previously detected QTL affecting milk yield (MY1), milk protein yield (PY1) and milk fat yield (FY1) during first lactation in the distal part of BTA5 in the Fleckvieh dual-purpose cattle breed [24], and to define the candidate region for high-throughput sequencing. To achieve this, we sampled additional families carrying the low frequency allele of the putative QTL (minor QTL allele) and genotyped additional markers covering the most likely QTL region on BTA5. These new families were identified by combining results from QTL-mapping based on microsatellites and haplotype analysis based on SNP in a complex pedigree. Single- and multiple-QTL analyses based on the LDL method were performed in different sample-sets, in order to allocate the minor QTL allele to specific families and to use the increased frequency of the minor QTL allele for refined mapping.", "[SUBTITLE] Animals and phenotype [SUBSECTION] In this study, we analysed the same nine granddaughter (GD) families used in our previous study [24], in which we identified three GD families (G01, G02 and G03) as heterozygous for a QTL located in the distal region of BTA5. The grandsires of these three GD families are designated as G01, G02 and G03, respectively. Grandsires G01 and G02 are half-sibs and have inherited the same haplotype in the distal region of BTA5 from their common ancestor A0 [24]. By target sampling (see haplotyping section, below), we introduced two additional GD families; family G10 with 85 sons, and family G11 with 47 sons. Grandsire G10 (grandsire of family G10), was connected through his dam to A0. Grandsire G11 (grandsire of family G11) is a son of grandsire G02. In addition, we identified all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11 to add more, possibly recombinant, A0 haplotypes into the mapping population. In this way, we created three maternal grandsire (MGS) families, M02 with 21 grandsons, M10 with 32 grandsons and M11 with 33 grandsons, descendants of grandsires G02, G10 and G11, respectively. Samples of maternal grandsons were not available for grandsire G01. Thus, the analysis included 11 GD families: G01 to G11 and three MGS families (M02, M10 and M11). Figure 1 shows the relationships of all families included in this study. In some cases, mapping analyses were carried out on 173 additional animals available from other projects that are not descended from ancestor A0. Estimated breeding values (EBV) of the Fleckvieh bulls for milk production traits MY1, PY1, and FY1, (along with their reliability values) were obtained from the 2009 joint Austria-Germany genetic evaluation of the Fleckvieh population [26].\nFamilial relationships considered in this study and segregation of most important haplotypes. A complex pedigree of 38 sires (squares) of GD families (G), ten sires of daughter design (DD) families, three maternal grandsire (M) families and 26 sampled and genotyped relevant ancestors; the pedigree has been simplified by showing only ancestors who made it possible to trace haplotypes from family-sires to the most important ancestors (A0, A1, A2); furthermore, to reduce the complexity of the figure, ancestor A1 is represented more than once; correspondingly, letters and numbers within squares of family-sires represent the internal family ID; non-genotyped individuals are represented by smaller circles (females) and squares (males) marked with a diagonal line; the estimated haplotype of 25 markers (A0H1) comprising a derived QTL allele affecting MY1 and PY1 with 97% CI between 117.962 Mb and 119.018 Mb is graphically presented by yellow bars above the individual's symbol; five other most frequent haplotypes are represented by five different coloured bars; introgression of Red-Holstein genes into the mapping populations is represented by ancestor A2 and the corresponding haplotype presented by a red bar; to reduce the complexity of the figure, 77 low frequency haplotypes are omitted; the allelic composition of the respective haplotypes is presented within the figure; the pedigree MSPED2089 is a subset of the total material which can be constructed by keeping the families marked by a grey circle around squares and associated ancestors; pedigrees MSPED1038 and SNPPED421 are subsets of MSPED2089 which can be constructed by removing appropriate families as described in material and methods; the pedigree SNPPED308 consists of GD family G36 and animals across the entire mapping population but not descending from A0; the pedigree SNPPED723 is a sum of pedigrees SNPPED308 and SNPPED421\n.\nIn this study, we analysed the same nine granddaughter (GD) families used in our previous study [24], in which we identified three GD families (G01, G02 and G03) as heterozygous for a QTL located in the distal region of BTA5. The grandsires of these three GD families are designated as G01, G02 and G03, respectively. Grandsires G01 and G02 are half-sibs and have inherited the same haplotype in the distal region of BTA5 from their common ancestor A0 [24]. By target sampling (see haplotyping section, below), we introduced two additional GD families; family G10 with 85 sons, and family G11 with 47 sons. Grandsire G10 (grandsire of family G10), was connected through his dam to A0. Grandsire G11 (grandsire of family G11) is a son of grandsire G02. In addition, we identified all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11 to add more, possibly recombinant, A0 haplotypes into the mapping population. In this way, we created three maternal grandsire (MGS) families, M02 with 21 grandsons, M10 with 32 grandsons and M11 with 33 grandsons, descendants of grandsires G02, G10 and G11, respectively. Samples of maternal grandsons were not available for grandsire G01. Thus, the analysis included 11 GD families: G01 to G11 and three MGS families (M02, M10 and M11). Figure 1 shows the relationships of all families included in this study. In some cases, mapping analyses were carried out on 173 additional animals available from other projects that are not descended from ancestor A0. Estimated breeding values (EBV) of the Fleckvieh bulls for milk production traits MY1, PY1, and FY1, (along with their reliability values) were obtained from the 2009 joint Austria-Germany genetic evaluation of the Fleckvieh population [26].\nFamilial relationships considered in this study and segregation of most important haplotypes. A complex pedigree of 38 sires (squares) of GD families (G), ten sires of daughter design (DD) families, three maternal grandsire (M) families and 26 sampled and genotyped relevant ancestors; the pedigree has been simplified by showing only ancestors who made it possible to trace haplotypes from family-sires to the most important ancestors (A0, A1, A2); furthermore, to reduce the complexity of the figure, ancestor A1 is represented more than once; correspondingly, letters and numbers within squares of family-sires represent the internal family ID; non-genotyped individuals are represented by smaller circles (females) and squares (males) marked with a diagonal line; the estimated haplotype of 25 markers (A0H1) comprising a derived QTL allele affecting MY1 and PY1 with 97% CI between 117.962 Mb and 119.018 Mb is graphically presented by yellow bars above the individual's symbol; five other most frequent haplotypes are represented by five different coloured bars; introgression of Red-Holstein genes into the mapping populations is represented by ancestor A2 and the corresponding haplotype presented by a red bar; to reduce the complexity of the figure, 77 low frequency haplotypes are omitted; the allelic composition of the respective haplotypes is presented within the figure; the pedigree MSPED2089 is a subset of the total material which can be constructed by keeping the families marked by a grey circle around squares and associated ancestors; pedigrees MSPED1038 and SNPPED421 are subsets of MSPED2089 which can be constructed by removing appropriate families as described in material and methods; the pedigree SNPPED308 consists of GD family G36 and animals across the entire mapping population but not descending from A0; the pedigree SNPPED723 is a sum of pedigrees SNPPED308 and SNPPED421\n.\n[SUBTITLE] DNA preparation, microsatellite marker selection and genotyping [SUBSECTION] Genomic DNA was prepared from semen using standard methods, and from whole blood samples with QIAamp Blood-Kits (Qiagen), according to the manufacturer's protocol.\nTwelve evenly distributed microsatellite markers were added to the 28 microsatellite markers used in the previous study [24]. Twenty-one of these 40 microsatellite markers covered the most likely region containing the QTL in the distal part of BTA5 (Table 1) and were used in most analyses of the present study. Previously analysed animals were genotyped only for the new markers, but the five new families (G10, G11, M02, M10 and M11) were genotyped for all marker sets [24]. For 11 of the 12 markers, relevant information was obtained from the MARC-ARS-USDA public database at http://www.ars.usda.gov/Main/docs.htm?docid=12539 \n[27]. The new marker LMU0505 was obtained by a targeted search for dinucleotide repeats in genomic regions with a low marker density. The unique sequences flanking the newly identified dinucleotide repeats were tested for informativity by genotyping a small set of animals first. Primers for the 12 new microsatellite markers were optimized using Primer3 (v.0.4.0) according to the bovine genome sequence data currently available (i.e. Baylor release Btau_4.0, http://genome.ucsc.edu/cgi-bin/hgGateway) and the appropriate fragment size in the currently designed marker set. New markers were divided into two PCR multiplex sets (Table 1) that were combined again after PCR for electrophoresis and fragment analysis. The fragment analysis of the PCR products was performed on ABI377 and ABI Prism 310 sequencers. Genotypes were assigned using GENESCAN and GENOTYPER (Applied Biosystems) software programs. We performed double genotyping of all families and ancestors using two independent runs. For ambiguous genotypes, the raw data were re-evaluated and animals were re-genotyped if necessary.\nMicrosatellite markers used for QTL mapping\nMarker name, relative position (cM), physical position (bp), forward and reverse primer sequences and marker set (set: Set0 & Set1 as in previous study; Set3 and Set4 comprise multiplex 1&2 in this study)\nGenomic DNA was prepared from semen using standard methods, and from whole blood samples with QIAamp Blood-Kits (Qiagen), according to the manufacturer's protocol.\nTwelve evenly distributed microsatellite markers were added to the 28 microsatellite markers used in the previous study [24]. Twenty-one of these 40 microsatellite markers covered the most likely region containing the QTL in the distal part of BTA5 (Table 1) and were used in most analyses of the present study. Previously analysed animals were genotyped only for the new markers, but the five new families (G10, G11, M02, M10 and M11) were genotyped for all marker sets [24]. For 11 of the 12 markers, relevant information was obtained from the MARC-ARS-USDA public database at http://www.ars.usda.gov/Main/docs.htm?docid=12539 \n[27]. The new marker LMU0505 was obtained by a targeted search for dinucleotide repeats in genomic regions with a low marker density. The unique sequences flanking the newly identified dinucleotide repeats were tested for informativity by genotyping a small set of animals first. Primers for the 12 new microsatellite markers were optimized using Primer3 (v.0.4.0) according to the bovine genome sequence data currently available (i.e. Baylor release Btau_4.0, http://genome.ucsc.edu/cgi-bin/hgGateway) and the appropriate fragment size in the currently designed marker set. New markers were divided into two PCR multiplex sets (Table 1) that were combined again after PCR for electrophoresis and fragment analysis. The fragment analysis of the PCR products was performed on ABI377 and ABI Prism 310 sequencers. Genotypes were assigned using GENESCAN and GENOTYPER (Applied Biosystems) software programs. We performed double genotyping of all families and ancestors using two independent runs. For ambiguous genotypes, the raw data were re-evaluated and animals were re-genotyped if necessary.\nMicrosatellite markers used for QTL mapping\nMarker name, relative position (cM), physical position (bp), forward and reverse primer sequences and marker set (set: Set0 & Set1 as in previous study; Set3 and Set4 comprise multiplex 1&2 in this study)\n[SUBTITLE] SNP selection, genotyping and haplotyping [SUBSECTION] SNP genotyping was carried out by Tierzuchtforschung e. V. München using the commercial Illumina Bovine SNP50 Bead chip featuring 54 001 SNP (http://www.illumina.com/; Illumina, San Diego) that span the bovine genome, excluding Y-chromosome. The genotype calling was performed with the GenCall application, as implemented in Illumina Bead chip Genotyping analysis software. This application computes a Gencall score for each locus, which evaluates the quality of genotypes. We included only animals with confirmed paternity and with a call rate above 0.98. Furthermore, we only used markers with a call rate above 0.90. We excluded all markers producing more than 1% paternity problems in pairs with confirmed paternity, and also excluded all markers that were non-informative in the Fleckvieh population or with an unknown chromosomal position. This yielded 43 806 informative SNP available for the whole-genome analysis in the Fleckvieh population, of which 1 976 are found on the BTA5. Two hundred and forty of these covered the region most likely containing the QTL in the distal part of BTA5 and were used in the present study.\nWe performed SNP genotyping in two stages. First, 75 animals i.e. the gransires of the nine initial GD families and their ancestors, and also a number of potential GD-family sires and their ancestors, were genotyped with the SNP chip and their haplotypes were reconstructed with the BEAGLE program [28]. These 75 animals constitute a complex pedigree (Figure 1) in which it is possible to trace the segregating haplotypes five generations back to some important ancestors of the Fleckvieh population, born in the 1960's and 1970's. This pedigree represents almost all of the important bull lines originating from a wide range of dams. Considering this, and the fact that a large proportion of the included bull dams are unrelated (no common grand-parents), these 75 animals provide a good representation of the haplotype diversity in the breeding Fleckvieh population. Second, the new families (G10, G11, M02, M10 and M11) containing the target haplotype segment of ancestor A0 were genotyped with microsatellite markers and with the genome-wide SNP chip. These animals and 173 additional Fleckvieh animals not closely related to ancestor A0 (but genotyped with the SNP chip in other projects running in our laboratory) were also haplotyped using the BEAGLE program.\nSNP genotyping was carried out by Tierzuchtforschung e. V. München using the commercial Illumina Bovine SNP50 Bead chip featuring 54 001 SNP (http://www.illumina.com/; Illumina, San Diego) that span the bovine genome, excluding Y-chromosome. The genotype calling was performed with the GenCall application, as implemented in Illumina Bead chip Genotyping analysis software. This application computes a Gencall score for each locus, which evaluates the quality of genotypes. We included only animals with confirmed paternity and with a call rate above 0.98. Furthermore, we only used markers with a call rate above 0.90. We excluded all markers producing more than 1% paternity problems in pairs with confirmed paternity, and also excluded all markers that were non-informative in the Fleckvieh population or with an unknown chromosomal position. This yielded 43 806 informative SNP available for the whole-genome analysis in the Fleckvieh population, of which 1 976 are found on the BTA5. Two hundred and forty of these covered the region most likely containing the QTL in the distal part of BTA5 and were used in the present study.\nWe performed SNP genotyping in two stages. First, 75 animals i.e. the gransires of the nine initial GD families and their ancestors, and also a number of potential GD-family sires and their ancestors, were genotyped with the SNP chip and their haplotypes were reconstructed with the BEAGLE program [28]. These 75 animals constitute a complex pedigree (Figure 1) in which it is possible to trace the segregating haplotypes five generations back to some important ancestors of the Fleckvieh population, born in the 1960's and 1970's. This pedigree represents almost all of the important bull lines originating from a wide range of dams. Considering this, and the fact that a large proportion of the included bull dams are unrelated (no common grand-parents), these 75 animals provide a good representation of the haplotype diversity in the breeding Fleckvieh population. Second, the new families (G10, G11, M02, M10 and M11) containing the target haplotype segment of ancestor A0 were genotyped with microsatellite markers and with the genome-wide SNP chip. These animals and 173 additional Fleckvieh animals not closely related to ancestor A0 (but genotyped with the SNP chip in other projects running in our laboratory) were also haplotyped using the BEAGLE program.\n[SUBTITLE] Linkage map construction [SUBSECTION] The relative positions of microsatellite markers were re-evaluated by the CRI-MAP program [29]. A physical map was constructed according to the sequence data of all the markers (Table 1) using the basic alignment search tool (BLAST) and the latest cattle genome sequence http://genome.ucsc.edu/cgi-bin/hgGateway. Our genetic data was used to resolve cases where more than one marker order was obtained from published linkage and physical maps. When our genetic data supported a marker order different from that of the public linkage map, but in accordance to the physical map, we modified the relative position (cM) of the markers along with the corresponding sequence. The linkage and physical maps were used as a framework to insert the newly designed marker (LMU0505) with the build option of the CRI-MAP program. The resulting final map (Table 1) was used for all the following analyses.\nThe relative positions of microsatellite markers were re-evaluated by the CRI-MAP program [29]. A physical map was constructed according to the sequence data of all the markers (Table 1) using the basic alignment search tool (BLAST) and the latest cattle genome sequence http://genome.ucsc.edu/cgi-bin/hgGateway. Our genetic data was used to resolve cases where more than one marker order was obtained from published linkage and physical maps. When our genetic data supported a marker order different from that of the public linkage map, but in accordance to the physical map, we modified the relative position (cM) of the markers along with the corresponding sequence. The linkage and physical maps were used as a framework to insert the newly designed marker (LMU0505) with the build option of the CRI-MAP program. The resulting final map (Table 1) was used for all the following analyses.\n[SUBTITLE] QTL fine mapping [SUBSECTION] [SUBTITLE] LDL mapping by microsatellite markers [SUBSECTION] Joint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).\nJoint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).\n[SUBTITLE] LDL mapping by SNPs [SUBSECTION] Here we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.\nHere we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.\n[SUBTITLE] LDL mapping by microsatellite markers [SUBSECTION] Joint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).\nJoint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).\n[SUBTITLE] LDL mapping by SNPs [SUBSECTION] Here we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.\nHere we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.\n[SUBTITLE] Multiple-QTL analysis using linkage disequilibrium and linkage (LDL) analysis method [SUBSECTION] We used the analysis method of Olsen et al. [33], i.e. the same model as for single-QTL analysis, but including a random QTL effect of a specified marker bracket. That is, the bracket that showed the highest LRT in the single-QTL analysis was included as a random effect in the QTL model in turn, and the analysis was repeated. These analyses searched for an additional QTL, given that the QTL in the specified marker bracket is accounted for, and is similar to the fitting of cofactors [34].\nWe used the analysis method of Olsen et al. [33], i.e. the same model as for single-QTL analysis, but including a random QTL effect of a specified marker bracket. That is, the bracket that showed the highest LRT in the single-QTL analysis was included as a random effect in the QTL model in turn, and the analysis was repeated. These analyses searched for an additional QTL, given that the QTL in the specified marker bracket is accounted for, and is similar to the fitting of cofactors [34].\n[SUBTITLE] Estimation of model parameters and test statistics [SUBSECTION] The variance components and the logarithm of the likelihood (L) of a model containing a QTL as well as residual polygenic effects at position p (logLp) were estimated by AIREML [32,35], which is an integral part of the LDLRAMS and LDL programs. The likelihood of a model without QTL effect (logL0) was calculated on the basis of a polygenic model. The log-likelihood ratio (LRT) was calculated as double difference in logL between models with and without a QTL, i.e. LRT = -2 (logL0-logLp). The LRT test statistic is distributed approximately as chi-square with 1 degree of freedom [36]. The confidence interval (CI) for the QTL position was determined as 1-LOD support interval, which was constructed as the interval surrounding the QTL peak where the LRT exceeds LRTmax - 2 × ln (10), where LRTmax is the maximum LRT-value for the tested QTL [37].\nThe variance components and the logarithm of the likelihood (L) of a model containing a QTL as well as residual polygenic effects at position p (logLp) were estimated by AIREML [32,35], which is an integral part of the LDLRAMS and LDL programs. The likelihood of a model without QTL effect (logL0) was calculated on the basis of a polygenic model. The log-likelihood ratio (LRT) was calculated as double difference in logL between models with and without a QTL, i.e. LRT = -2 (logL0-logLp). The LRT test statistic is distributed approximately as chi-square with 1 degree of freedom [36]. The confidence interval (CI) for the QTL position was determined as 1-LOD support interval, which was constructed as the interval surrounding the QTL peak where the LRT exceeds LRTmax - 2 × ln (10), where LRTmax is the maximum LRT-value for the tested QTL [37].", "In this study, we analysed the same nine granddaughter (GD) families used in our previous study [24], in which we identified three GD families (G01, G02 and G03) as heterozygous for a QTL located in the distal region of BTA5. The grandsires of these three GD families are designated as G01, G02 and G03, respectively. Grandsires G01 and G02 are half-sibs and have inherited the same haplotype in the distal region of BTA5 from their common ancestor A0 [24]. By target sampling (see haplotyping section, below), we introduced two additional GD families; family G10 with 85 sons, and family G11 with 47 sons. Grandsire G10 (grandsire of family G10), was connected through his dam to A0. Grandsire G11 (grandsire of family G11) is a son of grandsire G02. In addition, we identified all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11 to add more, possibly recombinant, A0 haplotypes into the mapping population. In this way, we created three maternal grandsire (MGS) families, M02 with 21 grandsons, M10 with 32 grandsons and M11 with 33 grandsons, descendants of grandsires G02, G10 and G11, respectively. Samples of maternal grandsons were not available for grandsire G01. Thus, the analysis included 11 GD families: G01 to G11 and three MGS families (M02, M10 and M11). Figure 1 shows the relationships of all families included in this study. In some cases, mapping analyses were carried out on 173 additional animals available from other projects that are not descended from ancestor A0. Estimated breeding values (EBV) of the Fleckvieh bulls for milk production traits MY1, PY1, and FY1, (along with their reliability values) were obtained from the 2009 joint Austria-Germany genetic evaluation of the Fleckvieh population [26].\nFamilial relationships considered in this study and segregation of most important haplotypes. A complex pedigree of 38 sires (squares) of GD families (G), ten sires of daughter design (DD) families, three maternal grandsire (M) families and 26 sampled and genotyped relevant ancestors; the pedigree has been simplified by showing only ancestors who made it possible to trace haplotypes from family-sires to the most important ancestors (A0, A1, A2); furthermore, to reduce the complexity of the figure, ancestor A1 is represented more than once; correspondingly, letters and numbers within squares of family-sires represent the internal family ID; non-genotyped individuals are represented by smaller circles (females) and squares (males) marked with a diagonal line; the estimated haplotype of 25 markers (A0H1) comprising a derived QTL allele affecting MY1 and PY1 with 97% CI between 117.962 Mb and 119.018 Mb is graphically presented by yellow bars above the individual's symbol; five other most frequent haplotypes are represented by five different coloured bars; introgression of Red-Holstein genes into the mapping populations is represented by ancestor A2 and the corresponding haplotype presented by a red bar; to reduce the complexity of the figure, 77 low frequency haplotypes are omitted; the allelic composition of the respective haplotypes is presented within the figure; the pedigree MSPED2089 is a subset of the total material which can be constructed by keeping the families marked by a grey circle around squares and associated ancestors; pedigrees MSPED1038 and SNPPED421 are subsets of MSPED2089 which can be constructed by removing appropriate families as described in material and methods; the pedigree SNPPED308 consists of GD family G36 and animals across the entire mapping population but not descending from A0; the pedigree SNPPED723 is a sum of pedigrees SNPPED308 and SNPPED421\n.", "Genomic DNA was prepared from semen using standard methods, and from whole blood samples with QIAamp Blood-Kits (Qiagen), according to the manufacturer's protocol.\nTwelve evenly distributed microsatellite markers were added to the 28 microsatellite markers used in the previous study [24]. Twenty-one of these 40 microsatellite markers covered the most likely region containing the QTL in the distal part of BTA5 (Table 1) and were used in most analyses of the present study. Previously analysed animals were genotyped only for the new markers, but the five new families (G10, G11, M02, M10 and M11) were genotyped for all marker sets [24]. For 11 of the 12 markers, relevant information was obtained from the MARC-ARS-USDA public database at http://www.ars.usda.gov/Main/docs.htm?docid=12539 \n[27]. The new marker LMU0505 was obtained by a targeted search for dinucleotide repeats in genomic regions with a low marker density. The unique sequences flanking the newly identified dinucleotide repeats were tested for informativity by genotyping a small set of animals first. Primers for the 12 new microsatellite markers were optimized using Primer3 (v.0.4.0) according to the bovine genome sequence data currently available (i.e. Baylor release Btau_4.0, http://genome.ucsc.edu/cgi-bin/hgGateway) and the appropriate fragment size in the currently designed marker set. New markers were divided into two PCR multiplex sets (Table 1) that were combined again after PCR for electrophoresis and fragment analysis. The fragment analysis of the PCR products was performed on ABI377 and ABI Prism 310 sequencers. Genotypes were assigned using GENESCAN and GENOTYPER (Applied Biosystems) software programs. We performed double genotyping of all families and ancestors using two independent runs. For ambiguous genotypes, the raw data were re-evaluated and animals were re-genotyped if necessary.\nMicrosatellite markers used for QTL mapping\nMarker name, relative position (cM), physical position (bp), forward and reverse primer sequences and marker set (set: Set0 & Set1 as in previous study; Set3 and Set4 comprise multiplex 1&2 in this study)", "SNP genotyping was carried out by Tierzuchtforschung e. V. München using the commercial Illumina Bovine SNP50 Bead chip featuring 54 001 SNP (http://www.illumina.com/; Illumina, San Diego) that span the bovine genome, excluding Y-chromosome. The genotype calling was performed with the GenCall application, as implemented in Illumina Bead chip Genotyping analysis software. This application computes a Gencall score for each locus, which evaluates the quality of genotypes. We included only animals with confirmed paternity and with a call rate above 0.98. Furthermore, we only used markers with a call rate above 0.90. We excluded all markers producing more than 1% paternity problems in pairs with confirmed paternity, and also excluded all markers that were non-informative in the Fleckvieh population or with an unknown chromosomal position. This yielded 43 806 informative SNP available for the whole-genome analysis in the Fleckvieh population, of which 1 976 are found on the BTA5. Two hundred and forty of these covered the region most likely containing the QTL in the distal part of BTA5 and were used in the present study.\nWe performed SNP genotyping in two stages. First, 75 animals i.e. the gransires of the nine initial GD families and their ancestors, and also a number of potential GD-family sires and their ancestors, were genotyped with the SNP chip and their haplotypes were reconstructed with the BEAGLE program [28]. These 75 animals constitute a complex pedigree (Figure 1) in which it is possible to trace the segregating haplotypes five generations back to some important ancestors of the Fleckvieh population, born in the 1960's and 1970's. This pedigree represents almost all of the important bull lines originating from a wide range of dams. Considering this, and the fact that a large proportion of the included bull dams are unrelated (no common grand-parents), these 75 animals provide a good representation of the haplotype diversity in the breeding Fleckvieh population. Second, the new families (G10, G11, M02, M10 and M11) containing the target haplotype segment of ancestor A0 were genotyped with microsatellite markers and with the genome-wide SNP chip. These animals and 173 additional Fleckvieh animals not closely related to ancestor A0 (but genotyped with the SNP chip in other projects running in our laboratory) were also haplotyped using the BEAGLE program.", "The relative positions of microsatellite markers were re-evaluated by the CRI-MAP program [29]. A physical map was constructed according to the sequence data of all the markers (Table 1) using the basic alignment search tool (BLAST) and the latest cattle genome sequence http://genome.ucsc.edu/cgi-bin/hgGateway. Our genetic data was used to resolve cases where more than one marker order was obtained from published linkage and physical maps. When our genetic data supported a marker order different from that of the public linkage map, but in accordance to the physical map, we modified the relative position (cM) of the markers along with the corresponding sequence. The linkage and physical maps were used as a framework to insert the newly designed marker (LMU0505) with the build option of the CRI-MAP program. The resulting final map (Table 1) was used for all the following analyses.", "[SUBTITLE] LDL mapping by microsatellite markers [SUBSECTION] Joint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).\nJoint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).\n[SUBTITLE] LDL mapping by SNPs [SUBSECTION] Here we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.\nHere we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.", "Joint linkage disequilibrium and linkage (LDL) analysis is a variance component approach and we used linear mixed models to estimate variance components as described previously [24]. Thereby, we used the Markov chain Monte Carlo (MCMC) implemented in the program LDLRAMS [30-32] (version 1.76) to estimate IBD probabilities in general complex pedigrees [30-32]. To estimate LD-based IBD probabilities, we assumed the number of generations since the base population (mutation age) and the past effective population size to be 100, and the initial homozygosity at each microsatellite marker in the base population was set to 0.35. In addition, the program LDLRAMS exploits allele frequencies in the population. To calculate an unbiased estimation of allele frequencies in the Fleckvieh population, we performed allele counting within the complex pedigree. We counted both alleles of all genotyped founder individuals and only the maternal allele of descendents in the pedigree. Two complex pedigrees consisting of 2 089 (MSPED2089) and 1 038 (MSPED1038) animals, respectively, were analysed by LDLRAMS. The MSPED2089\npedigree included nine GD families from the previous study (G01 to G09), two additional GD families (G10 and G11), three maternal grandsire families (M02, M10 and M11), some highly related animals and some important ancestors (paternal and maternal grandsires of phenotyped sons and of family sires). The MSPED1038 pedigree included two GD families (G01 and G02) found to be segregating for QTL in the previous study, two additional GD (G10 and G11) families and three MGS families (M02, M10 and M11) sampled according to the results of the haplotype analysis. For both LDL analyses, as implemented in the MCMC approach of the program LDLRAMS, we used an initial burn-in of 500 iterations followed by 2 500 iterations, with parameter estimates collected for each iteration. To avoid entrapment in a local maximum, we performed two independent sampling procedures (i.e. two LDLRAMS runs with different random number seeds).", "Here we used three complex pedigrees for LDL mapping by SNPs. The first pedigree, SNPPED723, was based on all progeny-tested Fleckvieh animals genotyped with the SNP chip, and consisted of 325 genotyped and phenotyped sons, and 16 genotyped and 382 non genotyped ancestors. The second pedigree, SNPPED421, was based on progeny-tested animals that could be traced back to ancestor A0, and consisted of 175 genotyped and phenotyped sons, eight genotyped and 238 non genotyped ancestors. The third pedigree, SNPPED308, was based on animals not related to ancestor A0 according to the known pedigree, and consisted of 144 genotyped and phenotyped animals, 12 genotyped and 152 non genotyped ancestors. These pedigrees were analysed with LDLRAMS using a dense map of 240 SNPs covering the region from 112.650 to 124.780 Mb on BTA5, i.e. a region larger than the 97% confidence interval as determined by 1-LOD support interval [24]. Due to computing constraints, the total marker set was divided into five overlapping sets of 80 SNP each. Since IBD estimates are most accurate in the middle of an investigated marker set, we present log-likelihood ratio (LRT) values only for the internal 40 marker intervals within these windows (that is, excluding the most proximal and most distal 20 markers). We used the model described above, setting the initial homozygosity at each SNP in the base population to 0.75 and using an initial burn-in of 500 iterations followed by 2 500 iterations. The parameter estimates were collected after each iteration. Two independent MCMC sampling procedures (i.e. two LDLRAMS runs with different random number seeds) indicated convergence to a global maximum.", "We used the analysis method of Olsen et al. [33], i.e. the same model as for single-QTL analysis, but including a random QTL effect of a specified marker bracket. That is, the bracket that showed the highest LRT in the single-QTL analysis was included as a random effect in the QTL model in turn, and the analysis was repeated. These analyses searched for an additional QTL, given that the QTL in the specified marker bracket is accounted for, and is similar to the fitting of cofactors [34].", "The variance components and the logarithm of the likelihood (L) of a model containing a QTL as well as residual polygenic effects at position p (logLp) were estimated by AIREML [32,35], which is an integral part of the LDLRAMS and LDL programs. The likelihood of a model without QTL effect (logL0) was calculated on the basis of a polygenic model. The log-likelihood ratio (LRT) was calculated as double difference in logL between models with and without a QTL, i.e. LRT = -2 (logL0-logLp). The LRT test statistic is distributed approximately as chi-square with 1 degree of freedom [36]. The confidence interval (CI) for the QTL position was determined as 1-LOD support interval, which was constructed as the interval surrounding the QTL peak where the LRT exceeds LRTmax - 2 × ln (10), where LRTmax is the maximum LRT-value for the tested QTL [37].", "[SUBTITLE] Genotypes and linkage map construction [SUBSECTION] Genotypes for 40 microsatellite markers were available to build the BTA5 genetic map. In most of the LDL analyses, only the 21 most distal markers (Table 1) covering the 97% confidence interval were considered. When we controlled if the genotype and haplotype data were plausible, the most distal marker (MNB71), which was genotyped in previous projects [24], showed extensive double recombinations with the 12 markers added in the present project. To reduce possible mapping errors, we excluded this marker from all subsequent analyses. Using the build option of the CRI-MAP program, we re-estimated the marker distances and order.\nThe following changes with respect to the public USDA linkage map were made: (i) according to the physical map (i.e. bp position of release Btau_4.0) and confirmed by applying the build option of the CRI-MAP program to our own data, the positions of markers BM49 and BM733 are inverted (Table 1); (ii) markers DIK2035 and DIK5277 are both at the same position (120.85 cM) on the USDA linkage map but, according to our genotypes and the physical map results, they are separated, placing DIK2035 (120.38 cM) upstream of DIK5277 (120.82 cM); (iii) the new marker developed in this study (LMU0505) is highly informative for linkage analysis and its relative position between DIK5106 and ETH152 was estimated by applying the build option of the CRI-MAP program. The positions of both flanking markers DIK5106 and ETH152 also changed (Table 1).\nGenotypes for 40 microsatellite markers were available to build the BTA5 genetic map. In most of the LDL analyses, only the 21 most distal markers (Table 1) covering the 97% confidence interval were considered. When we controlled if the genotype and haplotype data were plausible, the most distal marker (MNB71), which was genotyped in previous projects [24], showed extensive double recombinations with the 12 markers added in the present project. To reduce possible mapping errors, we excluded this marker from all subsequent analyses. Using the build option of the CRI-MAP program, we re-estimated the marker distances and order.\nThe following changes with respect to the public USDA linkage map were made: (i) according to the physical map (i.e. bp position of release Btau_4.0) and confirmed by applying the build option of the CRI-MAP program to our own data, the positions of markers BM49 and BM733 are inverted (Table 1); (ii) markers DIK2035 and DIK5277 are both at the same position (120.85 cM) on the USDA linkage map but, according to our genotypes and the physical map results, they are separated, placing DIK2035 (120.38 cM) upstream of DIK5277 (120.82 cM); (iii) the new marker developed in this study (LMU0505) is highly informative for linkage analysis and its relative position between DIK5106 and ETH152 was estimated by applying the build option of the CRI-MAP program. The positions of both flanking markers DIK5106 and ETH152 also changed (Table 1).\n[SUBTITLE] Haplotype analysis in a complex pedigree [SUBSECTION] Using the algorithm implemented into the program BEAGLE, we haplotyped the 75 animals of the complex pedigree in Figure 1 with 1 976 SNP on BTA5 that are informative in the Fleckvieh population. Thus reconstructed haplotypes were used to identify families segregating for the QTL detected in the initial study [24]. As already shown by the microsatellite analysis, the grandsires of families G01 and G02 which are heterozygous at the QTL, inherited the same haplotype in the distal region of BTA5 from their ancestor A0 (Figure 1). This was confirmed by the haplotype reconstruction using the 1 976 SNP. This A0 ancestral haplotype is named \"haplotype 1\" or (A0H1) and its A0 alternative haplotype \"haplotype 2\" or (A0H2). Family G03, previously declared as heterozygous for the target QTL [24] but not identified here, has inherited haplotypes not related to A0H1 (Figure 1). All animals with haplotype A0H1 (surrounding the putative QTL position) can be traced back to A0. Two of these, grandsires G10 and G11 are paternal and maternal grandsons of A0, and are very important Fleckvieh bull sires. We have collected samples of all the available progeny-tested sons of these two grandsires and all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11, to add more recombinant A0 haplotypes into the mapping population. In total, 485 animals were genotyped by the SNP chip and haplotyped for BTA5. By calculating the independent haplotypes in the complex pedigrees, and considering the traceability of all A0H1 haplotypes to A0, we estimated a very low frequency (<0.005) of A0H1 in the Fleckvieh population. Consequently, throughout the rest of this paper, the less frequent putative QTL allele embedded in this less frequent haplotype is referred to as the minor QTL allele.\nUsing the algorithm implemented into the program BEAGLE, we haplotyped the 75 animals of the complex pedigree in Figure 1 with 1 976 SNP on BTA5 that are informative in the Fleckvieh population. Thus reconstructed haplotypes were used to identify families segregating for the QTL detected in the initial study [24]. As already shown by the microsatellite analysis, the grandsires of families G01 and G02 which are heterozygous at the QTL, inherited the same haplotype in the distal region of BTA5 from their ancestor A0 (Figure 1). This was confirmed by the haplotype reconstruction using the 1 976 SNP. This A0 ancestral haplotype is named \"haplotype 1\" or (A0H1) and its A0 alternative haplotype \"haplotype 2\" or (A0H2). Family G03, previously declared as heterozygous for the target QTL [24] but not identified here, has inherited haplotypes not related to A0H1 (Figure 1). All animals with haplotype A0H1 (surrounding the putative QTL position) can be traced back to A0. Two of these, grandsires G10 and G11 are paternal and maternal grandsons of A0, and are very important Fleckvieh bull sires. We have collected samples of all the available progeny-tested sons of these two grandsires and all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11, to add more recombinant A0 haplotypes into the mapping population. In total, 485 animals were genotyped by the SNP chip and haplotyped for BTA5. By calculating the independent haplotypes in the complex pedigrees, and considering the traceability of all A0H1 haplotypes to A0, we estimated a very low frequency (<0.005) of A0H1 in the Fleckvieh population. Consequently, throughout the rest of this paper, the less frequent putative QTL allele embedded in this less frequent haplotype is referred to as the minor QTL allele.\n[SUBTITLE] Combined linkage disequilibrium and linkage analysis [SUBSECTION] Thirty-seven microsatellite markers (three markers BM6026, BMS610 and MNB71 showed extensive recombinations and were excluded) and the complex pedigree MSPED2089 were used for initial LDL mapping analyses. As shown in Figure 2, we observed a highly significant QTL effect (LRT = 20 to 22, i.e. P = 0.0000077 to 0.0000027), but were unable to improve the mapping accuracy because of the presence of two or three peaks.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 2089 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 37 microsatellites, a complex pedigree of 2 089 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nAccording to previous results [24], and to the results obtained in the first part of this study, we have assumed that haplotype A0H1 has only introduced one QTL into the mapping population. Therefore, we performed a second LDL analysis using the 21 most distal markers, and limited to GD and MGS families descending from A0 and known to carry A0H1, i.e. pedigree MSPED1038 (Figure 3). Unlike the analysis of pedigree MSPED2089, Figure 3 illustrates a single rather broad peak between positions 119.005 cM and 120.166 cM. However, this highly significant QTL (P = 0.000062 to 0.000021) is still mapped with a low accuracy, i.e. 1-LOD drop-off support intervals are 4.7 cM for FY1, 10.4 cM for PY1 and 11.5 cM for MY1.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 1 038 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 21 microsatellites covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 1 038 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nSince the confidence interval achieved by LDL analyses using pedigree MSPED1038 was still too large for a positional candidate gene approach, we analysed pedigree SNPPED723 using the LDL approach. The results were similar to those obtained with microsatellite markers and pedigree MSPED2089, namely, multiple peaks suggesting multiple QTL or no QTL (Figure 4).\nLDL analysis by variance component approach using SNP in a complex pedigree of 723 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 723 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nTo resolve this dilemma, we divided pedigree SNPPED723 into pedigree SNPPED421 consisting of all progeny-tested animals descending from ancestor A0, and pedigree SNPPED308 consisting of the remaining progeny-tested animals. The LDL analyses of SNPPED308 pedigree showed a moderately flat, non-significant test statistic along the investigated chromosomal segment (Figure 5). Only LRT values for FY1 reached an indicative level of 3.99 (P = 0.046). Conversely, it was possible to map a QTL with pedigree SNPPED421 whose minor allele is most probably originating from ancestor A0 (Figure 6). There were two distinct peaks; one with LRT values over 17 (P < 0.000037) for both MY1 and PY1 in a region of 0.5 Mb (from 118.107 to 118.606 Mb), and one with a very high LRT value for only PY1 (LRT = 20.72, P = 0.0000053) at position 122.115 Mb. Considering 1-LOD drop-off support intervals, the 97% confidence intervals were located between 117.962 Mb and 119.018 Mb (i.e. 1.056 Mb) for the QTL affecting MY1 and PY1, and between 121.800 Mb and 122.200 Mb (i.e. 0.400 Mb) for the QTL affecting only PY1. There were two additional peaks with LRT values over 15 in regions around the positions 115.650 and 116.300 Mb, but they were not included in the 97% confidence interval for PY1 and were not supported by the highly correlated MY1 trait.\nLDL analysis by variance component approach using SNP in a complex pedigree of 308 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 308 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nLDL analysis by variance component approach using SNP in a complex pedigree of 421 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covering the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 421 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. The long homozygous region (~5 Mb) in ancestor A0 was shown (A0 Homo).\nThe two identified peaks (located between 118.107 Mb and 118.606 Mb and at 122.115 Mb, respectively) may be due to either the presence of more than one QTL, or the presence of one QTL with carryover effects to another region. Thus, a multiple-QTL analysis was performed. Two-QTL analyses using pedigree SNPPED421 for MY1 and PY1 fitting a QTL at position 118.202 Mb revealed a single QTL affecting only MY1 at this location and an additional QTL affecting PY1 at position 122.115 Mb (P = 0.019). However, two-QTL analyses accounting for the QTL at position 122.115 Mb did not rule out a possible second QTL affecting PY1 at position 118.202 Mb (P = 0.019).\nThirty-seven microsatellite markers (three markers BM6026, BMS610 and MNB71 showed extensive recombinations and were excluded) and the complex pedigree MSPED2089 were used for initial LDL mapping analyses. As shown in Figure 2, we observed a highly significant QTL effect (LRT = 20 to 22, i.e. P = 0.0000077 to 0.0000027), but were unable to improve the mapping accuracy because of the presence of two or three peaks.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 2089 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 37 microsatellites, a complex pedigree of 2 089 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nAccording to previous results [24], and to the results obtained in the first part of this study, we have assumed that haplotype A0H1 has only introduced one QTL into the mapping population. Therefore, we performed a second LDL analysis using the 21 most distal markers, and limited to GD and MGS families descending from A0 and known to carry A0H1, i.e. pedigree MSPED1038 (Figure 3). Unlike the analysis of pedigree MSPED2089, Figure 3 illustrates a single rather broad peak between positions 119.005 cM and 120.166 cM. However, this highly significant QTL (P = 0.000062 to 0.000021) is still mapped with a low accuracy, i.e. 1-LOD drop-off support intervals are 4.7 cM for FY1, 10.4 cM for PY1 and 11.5 cM for MY1.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 1 038 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 21 microsatellites covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 1 038 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nSince the confidence interval achieved by LDL analyses using pedigree MSPED1038 was still too large for a positional candidate gene approach, we analysed pedigree SNPPED723 using the LDL approach. The results were similar to those obtained with microsatellite markers and pedigree MSPED2089, namely, multiple peaks suggesting multiple QTL or no QTL (Figure 4).\nLDL analysis by variance component approach using SNP in a complex pedigree of 723 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 723 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nTo resolve this dilemma, we divided pedigree SNPPED723 into pedigree SNPPED421 consisting of all progeny-tested animals descending from ancestor A0, and pedigree SNPPED308 consisting of the remaining progeny-tested animals. The LDL analyses of SNPPED308 pedigree showed a moderately flat, non-significant test statistic along the investigated chromosomal segment (Figure 5). Only LRT values for FY1 reached an indicative level of 3.99 (P = 0.046). Conversely, it was possible to map a QTL with pedigree SNPPED421 whose minor allele is most probably originating from ancestor A0 (Figure 6). There were two distinct peaks; one with LRT values over 17 (P < 0.000037) for both MY1 and PY1 in a region of 0.5 Mb (from 118.107 to 118.606 Mb), and one with a very high LRT value for only PY1 (LRT = 20.72, P = 0.0000053) at position 122.115 Mb. Considering 1-LOD drop-off support intervals, the 97% confidence intervals were located between 117.962 Mb and 119.018 Mb (i.e. 1.056 Mb) for the QTL affecting MY1 and PY1, and between 121.800 Mb and 122.200 Mb (i.e. 0.400 Mb) for the QTL affecting only PY1. There were two additional peaks with LRT values over 15 in regions around the positions 115.650 and 116.300 Mb, but they were not included in the 97% confidence interval for PY1 and were not supported by the highly correlated MY1 trait.\nLDL analysis by variance component approach using SNP in a complex pedigree of 308 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 308 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nLDL analysis by variance component approach using SNP in a complex pedigree of 421 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covering the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 421 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. The long homozygous region (~5 Mb) in ancestor A0 was shown (A0 Homo).\nThe two identified peaks (located between 118.107 Mb and 118.606 Mb and at 122.115 Mb, respectively) may be due to either the presence of more than one QTL, or the presence of one QTL with carryover effects to another region. Thus, a multiple-QTL analysis was performed. Two-QTL analyses using pedigree SNPPED421 for MY1 and PY1 fitting a QTL at position 118.202 Mb revealed a single QTL affecting only MY1 at this location and an additional QTL affecting PY1 at position 122.115 Mb (P = 0.019). However, two-QTL analyses accounting for the QTL at position 122.115 Mb did not rule out a possible second QTL affecting PY1 at position 118.202 Mb (P = 0.019).", "Genotypes for 40 microsatellite markers were available to build the BTA5 genetic map. In most of the LDL analyses, only the 21 most distal markers (Table 1) covering the 97% confidence interval were considered. When we controlled if the genotype and haplotype data were plausible, the most distal marker (MNB71), which was genotyped in previous projects [24], showed extensive double recombinations with the 12 markers added in the present project. To reduce possible mapping errors, we excluded this marker from all subsequent analyses. Using the build option of the CRI-MAP program, we re-estimated the marker distances and order.\nThe following changes with respect to the public USDA linkage map were made: (i) according to the physical map (i.e. bp position of release Btau_4.0) and confirmed by applying the build option of the CRI-MAP program to our own data, the positions of markers BM49 and BM733 are inverted (Table 1); (ii) markers DIK2035 and DIK5277 are both at the same position (120.85 cM) on the USDA linkage map but, according to our genotypes and the physical map results, they are separated, placing DIK2035 (120.38 cM) upstream of DIK5277 (120.82 cM); (iii) the new marker developed in this study (LMU0505) is highly informative for linkage analysis and its relative position between DIK5106 and ETH152 was estimated by applying the build option of the CRI-MAP program. The positions of both flanking markers DIK5106 and ETH152 also changed (Table 1).", "Using the algorithm implemented into the program BEAGLE, we haplotyped the 75 animals of the complex pedigree in Figure 1 with 1 976 SNP on BTA5 that are informative in the Fleckvieh population. Thus reconstructed haplotypes were used to identify families segregating for the QTL detected in the initial study [24]. As already shown by the microsatellite analysis, the grandsires of families G01 and G02 which are heterozygous at the QTL, inherited the same haplotype in the distal region of BTA5 from their ancestor A0 (Figure 1). This was confirmed by the haplotype reconstruction using the 1 976 SNP. This A0 ancestral haplotype is named \"haplotype 1\" or (A0H1) and its A0 alternative haplotype \"haplotype 2\" or (A0H2). Family G03, previously declared as heterozygous for the target QTL [24] but not identified here, has inherited haplotypes not related to A0H1 (Figure 1). All animals with haplotype A0H1 (surrounding the putative QTL position) can be traced back to A0. Two of these, grandsires G10 and G11 are paternal and maternal grandsons of A0, and are very important Fleckvieh bull sires. We have collected samples of all the available progeny-tested sons of these two grandsires and all available progeny-tested maternal grandsons of grandsires G01, G02, G10 and G11, to add more recombinant A0 haplotypes into the mapping population. In total, 485 animals were genotyped by the SNP chip and haplotyped for BTA5. By calculating the independent haplotypes in the complex pedigrees, and considering the traceability of all A0H1 haplotypes to A0, we estimated a very low frequency (<0.005) of A0H1 in the Fleckvieh population. Consequently, throughout the rest of this paper, the less frequent putative QTL allele embedded in this less frequent haplotype is referred to as the minor QTL allele.", "Thirty-seven microsatellite markers (three markers BM6026, BMS610 and MNB71 showed extensive recombinations and were excluded) and the complex pedigree MSPED2089 were used for initial LDL mapping analyses. As shown in Figure 2, we observed a highly significant QTL effect (LRT = 20 to 22, i.e. P = 0.0000077 to 0.0000027), but were unable to improve the mapping accuracy because of the presence of two or three peaks.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 2089 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 37 microsatellites, a complex pedigree of 2 089 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nAccording to previous results [24], and to the results obtained in the first part of this study, we have assumed that haplotype A0H1 has only introduced one QTL into the mapping population. Therefore, we performed a second LDL analysis using the 21 most distal markers, and limited to GD and MGS families descending from A0 and known to carry A0H1, i.e. pedigree MSPED1038 (Figure 3). Unlike the analysis of pedigree MSPED2089, Figure 3 illustrates a single rather broad peak between positions 119.005 cM and 120.166 cM. However, this highly significant QTL (P = 0.000062 to 0.000021) is still mapped with a low accuracy, i.e. 1-LOD drop-off support intervals are 4.7 cM for FY1, 10.4 cM for PY1 and 11.5 cM for MY1.\nLDL analysis by variance component approach using microsatellites in a complex pedigree of 1 038 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 21 microsatellites covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 1 038 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in centiMorgan (cM) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. Solid triangles on the X-axis represent positions of markers included in the analysis.\nSince the confidence interval achieved by LDL analyses using pedigree MSPED1038 was still too large for a positional candidate gene approach, we analysed pedigree SNPPED723 using the LDL approach. The results were similar to those obtained with microsatellite markers and pedigree MSPED2089, namely, multiple peaks suggesting multiple QTL or no QTL (Figure 4).\nLDL analysis by variance component approach using SNP in a complex pedigree of 723 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 723 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nTo resolve this dilemma, we divided pedigree SNPPED723 into pedigree SNPPED421 consisting of all progeny-tested animals descending from ancestor A0, and pedigree SNPPED308 consisting of the remaining progeny-tested animals. The LDL analyses of SNPPED308 pedigree showed a moderately flat, non-significant test statistic along the investigated chromosomal segment (Figure 5). Only LRT values for FY1 reached an indicative level of 3.99 (P = 0.046). Conversely, it was possible to map a QTL with pedigree SNPPED421 whose minor allele is most probably originating from ancestor A0 (Figure 6). There were two distinct peaks; one with LRT values over 17 (P < 0.000037) for both MY1 and PY1 in a region of 0.5 Mb (from 118.107 to 118.606 Mb), and one with a very high LRT value for only PY1 (LRT = 20.72, P = 0.0000053) at position 122.115 Mb. Considering 1-LOD drop-off support intervals, the 97% confidence intervals were located between 117.962 Mb and 119.018 Mb (i.e. 1.056 Mb) for the QTL affecting MY1 and PY1, and between 121.800 Mb and 122.200 Mb (i.e. 0.400 Mb) for the QTL affecting only PY1. There were two additional peaks with LRT values over 15 in regions around the positions 115.650 and 116.300 Mb, but they were not included in the 97% confidence interval for PY1 and were not supported by the highly correlated MY1 trait.\nLDL analysis by variance component approach using SNP in a complex pedigree of 308 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covered the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 308 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis.\nLDL analysis by variance component approach using SNP in a complex pedigree of 421 animals. Joint linkage disequilibrium and linkage (LDL) analysis for three milk yield traits; Milk Yield (MY1), Milk Protein Yield (PY1) and Milk Fat Yield (FY1) during first lactation using 240 SNPs covering the most likely region containing the QTL in the distal part of bovine chromosome 5 (BTA5), a complex pedigree of 421 animals, EBV as phenotype and AIREML as implemented in LDLRAMS and LDL program. Chromosome length in Megabase (Mb) on the X-axis, log-likelihood ratio test (LRT) values on the Y-axis. The long homozygous region (~5 Mb) in ancestor A0 was shown (A0 Homo).\nThe two identified peaks (located between 118.107 Mb and 118.606 Mb and at 122.115 Mb, respectively) may be due to either the presence of more than one QTL, or the presence of one QTL with carryover effects to another region. Thus, a multiple-QTL analysis was performed. Two-QTL analyses using pedigree SNPPED421 for MY1 and PY1 fitting a QTL at position 118.202 Mb revealed a single QTL affecting only MY1 at this location and an additional QTL affecting PY1 at position 122.115 Mb (P = 0.019). However, two-QTL analyses accounting for the QTL at position 122.115 Mb did not rule out a possible second QTL affecting PY1 at position 118.202 Mb (P = 0.019).", "The aim of this study was to refine the position of a previously mapped QTL by increasing the marker density in the region, target sampling of additional families and adapting fine mapping methods. According to our previous results [24] and to results from the initial part of this study, we hypothesized the presence of a minor QTL allele with a strong effect, but at a very low frequency, in the Fleckvieh dual-purpose cattle breed. In such a situation, random sampling of additional families for confirmation and fine-mapping purposes can result in an increased frequency of the common QTL allele in the mapping design Thus, the capacity to differentiate between genetic background noise and the initially targeted QTL will be decreased. The reduced accuracy of QTL position estimates when using all genotyped animals (pedigrees MSPED2089 or SNPPED723) compared to a subset of animals (pedigrees MSPED1038 or SNPPED421) is counterintuitive to the general notion that the use of more information should result in better estimates. To further explore this unexpected result, we have investigated several possible explanations, including the effects of the haplotype distribution and the possibility of additional QTL. To study the haplotype distribution in the Fleckvieh population, 485 animals were genotyped with the Illumina 50 K SNP chip. Of these, a subset of 144 animals were not progeny-tested and not relevant for QTL mapping, but were very informative for the study of haplotype distribution. In particular, considering the putative QTL affecting MY1 and PY1 located within the 97% CI (between 117.962 Mb and 119.018 Mb), a haplotype of 25 markers (A0H1) covering this region was detected in 89 of 485 animals. This haplotype A0H1, most probably carrying the minor QTL allele, could be traced back to the ancestor A0 in all 89 cases (Figure 1). The alternative haplotype A0H2, most probably carrying the common QTL allele, was found in 13 cases but was traced back to the ancestor A0 only in three. A perfect LD between the minor QTL allele and A0H1 (and only A0H1) would result in a relatively low allele frequency (0.137) of the minor QTL allele in phenotyped animals of pedigree SNPPED723, and in a frequency about double (0.254) in pedigree SNPPED421. The mapping results did reflect this difference too.\nIn contrast, consider the six markers located within the 97% CI (between 121.800 Mb and 122.200 Mb) of the putative QTL region affecting only PY1. Ancestor A0 is homozygous for a very long segment of this region i.e. from positions 118.266 Mb to 123.347 Mb (three SNP telomeric to the main peak of QTL affecting MY1 and PY1). This segment of 5.080 Mb includes 109 informative markers in the Fleckvieh population. Comparison of mapping results from pedigrees SNPPED723\n(Figures 4), SNPPED421 (Figure 6),and SNPPED308 (Figure 5) revealed a highly significant QTL allele affecting PY1 only when the pedigree included families segregating for haplotype A0H1 (see comparison between Figures 4 and 6). Excluding these families yielded LRT values below 3.99 (P > 0.045) for all three milk yield traits and for the complete investigated region (Figure 5, between 113.500 Mb and 123.700 Mb). We therefore mainly used the linkage information in the SNPPED421 pedigree (A0H1 always traceable to A0), to map a QTL affecting both MY1 and PY1 in a 97% CI of 1 Mb.\nHaplotype and LDL analyses by microsatellite markers (Figures 2 and 3) and SNP (Figures 4 and 6) clearly suggest that the minor QTL allele associated with the putative QTL around the physical position 118 Mb (97% CI between 117.962 Mb to 119.018 Mb) has been introduced by ancestor A0 into the mapping population. The explanation of the second possible QTL that maps to the physical position 122.115 Mb and affects only PY1 is different. First, this QTL should also be associated with ancestor A0 haplotypes, i.e. absence of effect in the smaller SNPPED308 pedigree (Figure 5). Second, both ancestor haplotypes at the physical position 122.115 Mb are most probably identical by descent (i.e. homozygous for a 5.080 Mb segment with 109 informative SNP). Therefore, ancestor A0 is most probably homozygous for the putative QTL at this position too. Third, this part of the haplotype is not unique to A0, but also segregates in other families, i.e. there is LD information for mapping, too. The relatively sharp LRT peak at position 122.115 Mb and homozygosity of A0 suggest an essential contribution of LD to this mapping result. Fourth, analyses with the two-QTL model did not rule out the possibility of a second QTL affecting PY1 within the candidate region on BTA5. And finally, despite the overall presence of haplotypes with a high IBD to ancestor haplotypes around position 122.115 Mb, the complete absence of this peak in SNPPED308 pedigree can be explained by either a novel mutation in ancestor A0 or by the incapacity of the method and design used here to map it in a relatively small pedigree like SNPPED308. More reasonable explanations may be the lower statistical power of the pedigree SNPPED308, possible local inconsistencies in the map order (which was based on map release Btau_4.0), the presence of a strong QTL at position 118.000 Mb with carryover effects to other regions, or a combination of all these explanations.\nThe LDL analysis using SNPs and pedigree SNPPED723 indicate several peaks affecting MY1 and PY1 in the region investigated here. In principle, these results (Figure 4) are comparable to the fine-mapping results reported on BTA3 by Druet et al. [38]. In this study, the authors have also first carried out mapping by linkage analysis and finally ended up with LDL analyses and multiple LRT peaks. We used larger overlapping marker windows (80 SNP) than Druet et al. [38]. By dividing the data set according to the results of linkage and haplotype analyses, most of the multiple peaks were explained as genetic background noise in a larger family set. The multiple peak profile could be explained by the heterogeneous LD structure within the QTL region or by the use of LD in the model when there is no LD information at all [38]. This might be increased by possible local inconsistencies in the map order, which was based on the draft assembly, or on comparative map information. Moreover, the method and the data structure may not make it possible to discard some regions even though they do not harbour the QTL [38].\nTo check for possible effects of the data structure on the reported mapping results, we tested regression of EBV on genetic distance from ancestor A0 for all carriers of haplotype 1 (A0H1). The apparent lack of this regression suggests that we are looking at a real QTL effect, and not an artifact of pedigree-tracking.\nSearching the region between 117.900 and 119.100 Mb for candidate genes revealed 27 genes, 13 of which had no known function. Based on current biological information, the genes with partly known function could only be indirectly related to milk yield traits.", "In the present study, we have performed a haplotype-assisted extension of the mapping design and thus increased the allele frequency of the minor QTL allele in mapping families. Alternative analyses with family subsets resulted in a substantial reduction of the genetic background noise and an increased frequency of the minor QTL allele. Using these subsets, we succeeded in refining the map position of the previously detected QTL for milk production traits on BTA5 to a 1 Mb interval. In spite of implementing a two-QTL analysis, the possibility of a second QTL affecting only PY1 could not be ruled out. All in all, the results of both this study and the previous study by Awad et al. [24] support the presence of a QTL affecting both MY1 and PY1 that is close to the centromeric part of the long homozygous region (~5 Mb) in ancestor A0. Therefore, positional cloning and high-throughput sequencing of the candidate region located between 117.900 Mb and 119.100 Mb should now be considered, but should also not neglect the second possible QTL around position 122.115 Mb.", "The authors declare that they have no competing interests.", "AA carried out DNA extraction, microsatellite genotyping; AA and IM performed all data analysis and wrote the paper; IM and MF designed the study; IR performed SNP genotyping and partly performed sampling. All authors read and approved the final manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Animals", "Arachidonic Acid", "Cattle", "Cell Survival", "Cells, Cultured", "Corpus Luteum", "Cytokines", "Endothelial Cells", "Endothelin-1", "Epoxide Hydrolases", "Female", "Glutathione Transferase", "Hydroxyprostaglandin Dehydrogenases", "Interferon-gamma", "Intramolecular Oxidoreductases", "Luteal Cells", "Prostaglandin-E Synthases", "RNA, Messenger", "Tumor Necrosis Factor-alpha" ]

[ "Background", "Collection of CL for in vitro experiments", "Luteal endothelial cells isolation", "Experimental procedures", "Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization", "Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1)", "Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1)", "Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1)", "Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells", "Total RNA isolation", "Conventional PCR", "Immunofluorescence staining", "Real-time PCR quantification", "EIA determination", "Western blot analysis", "Statistical analysis", "Results", "Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization", "Experiment 2. Effect of cytokines on production and content of Arachidonic Acid metabolites in immortalized bovine luteal endothelial cells (EnCL-1)", "Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1)", "Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1)", "Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells", "Discussion", "Competing interests", "Authors' contributions" ]

[ "Although corpus luteum (CL) is a transient gland, it is one of the most vascularized tissues in the body [1], with endothelial cells representing greater than fifty percent of the total cells [2,3]. Angiogenesis is critical to CL development, whereas endothelial cells decline occurs during luteolysis [4]. On the other side, endothelial cells play a crucial role in a complex processes of tumor neovascularization [5], including CL cancers [6]. Because of these important and multiplex functions of endothelial cells in CL vascularity, the establishment of an experimental model of immortalized endothelial cells from bovine CL is a prerequisite for the study of cellular and molecular mechanism in this tissue. So far, the majority of studies have been conducted on fresh isolated or refrozen aliquots of bovine primary luteal endothelial cells [4,7-9] or cell line received not directly from CL [10]. Immortalized endothelial cells have been characterized in several kinds of bovine tissues, such as the pulmonary and coronary arteries, nevertheless no bovine luteal endothelial cell line is available [10]. There is a possibility that surface antigens and/or genetic programming differs for endothelial cells derived from various tissues, beside each kind of cell is strictly species dependent [11]. Thus, the stable bovine luteal endothelial cell line with determined fenotype and genotype would be the convenient and useful model for the future study.\nAmong mediators of interactions between different types of CL cells, including endothelial cells, the universal factors are immune cells and their secreted products, cytokines [12-14]. Endothelial cells are capable of tumor necrosis factor α (TNFα) synthesis and secretion [15]. Depending on the immediate microenvironment, TNFα may stimulate cell proliferation or induce apoptosis of luteal endothelial cells [4]. TNFα action in the bovine CL is a dose dependent: a low concentration of TNFα stimulates in vivo luteolytic factors, as well as induces apoptosis; whereas the high concentration of TNFα stimulates a survival pathway [16-19]. Moreover, TNFα induced apoptosis in cultured bovine luteal endothelial cells [20]. TNFα effect in the ovary was found to be more effective when TNFα acted synergistically with interferon γ (IFNγ; [13,19,21,22]). Reasonable is the generation of stable in vitro luteal endothelial cell culture for investigating the complex signaling pathway and transcriptional mechanisms regulated by cytokines in physiological and pathophysiological conditions in cattle.\nThe proper vascularization and endothelial cell activity per se are essential for normal CL function [23]. The effect of prostaglandins (PGs) on the vascularity of bovine CL is well known [9,21,23-26]. The ovarian blood flow has been shown to increase after PGE2 administration and decrease during spontaneous and PGF2α induced luteolysis in cows [27]. An acute increase in the luteal blood flow occurs as the first step of luteolysis in response to PGF2α [3]. Both the density and the number of blood vessels were higher in CLs obtained after PGF2α administration than in those without PGF2α treatment [21], which indicate that the number of blood vessels with smooth muscle in the regressing CL increased as a result of loosing steroidogenic cells and capillaries. A mitogenic effect and increased proliferation were observed after PGF2α treatment in bovine dispersed luteal endothelial cells [23]. Moreover, PG receptors, as well as leukotriene (LTs - other metabolites of arachidonic acid - aa) receptors are present on endothelial cells [23,26], which indicate that the endothelial cells of bovine CL are target for PGs and LTs. Leukotrienes are commonly known as the potential inflammatory factors that cause edema in respiratory tract diseases, but they also play the important role in reproduction and may enhance the action of PGs [28]. We hypothesize that cytokines are involved in aa metabolites action in bovine luteal endothelial cells.\nThe main goal of this study was to determine the effect of cytokines on aa metabolites action in bovine immortalized luteal endothelial cells. We examined: (i) the viability of EnCL-1 cells, (ii) mRNA expression for LTA4 hydrolase (LTA4H), LTC4 synthase (LTC4S), PGE2 and PGF2α synthases (PGES and PGFS) and endothelin-1 (EDN-1), (iii) protein expression for LTA4H, LTC4S, PGES, PGFS and EDN-1/2/3 and (iv) accumulation of LTB4 and C4, PGE2 and F2α and EDN-1 in the culture medium after TNFα/IFNγ stimulation.", "Healthy, normally cycling Holstein/Polish Black and White (75/25%; respectively) cows (3-4 lactation) were used for the collection of the ovaries with CL (in total 4 animals). The animals were eliminated by the owners (Experimental Animal Farm of Polish Academy of Sciences in Baranowo; Poland) from the dairy herds because of their lower milk production. The estrus of the cows was synchronized twice using an analogue of PGF2α (dinoprost, Dinolytic; Upjohn - Pharmacia N.V.S.A., Belgium) injections with an 11-day interval as recommended by a vendor. The onset of the estrus was determined by a veterinarian via per rectum USG examination using a DRAMINSKI ANIMALprofi Scanner [29] and confirmed by observing the signs of estrus (i.e. vaginal mucus, standing behavior). The onset of estrus was determined as Day 0. Only cows with signs of estrus were chosen for the study. Animals were slaughtered on Day 8-12 of the estrous cycle and the ovaries were obtained within 20 min of the exsanguinations and transported on ice to the laboratory.", "Endothelial cells isolation was proceeded according to the method described previously in details [8,26] using a Dynabeads kit (140.03 Dynabeads® M-450; Invitrogen, Oregon, USA). Briefly, the Dynabeads were coated with the specific antigen-lectin (BS-1; L 2380 Sigma; Chemical Co., St Louis, MO, USA). The beads coated with endothelial cells were attracted by a magnet to the well of the tube and the supernatant was removed. After washing with PBS, 1 ml of 0.1 M fucose (T 2252 Sigma) solution was added to break the connection of endothelial cells with beads. The free beads were then attracted by a magnet to the well of the tube and the supernatant with endothelial cells was collected. The obtained cell suspension contained more than 85% of luteal endothelial cells and only a few steroidogenic CL cells. The cells were suspended in Dulbecco's Modified Eagle's medium (DMEM/Ham's F-12; 1: 1(v/v); D 2906 Sigma) in a 3 ml culture flask in a humidified incubator at 37.5°C in 5% CO2/95% air atmosphere. After third passage the cells were trypsinized and placed at the concentration of 2 × 105 cells/ml into a 24-well culture plate. After 24-48 h of culture cells reached confluence and were proceeded the procedure of immortalization.\n[SUBTITLE] Experimental procedures [SUBSECTION] [SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\n[SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.", "[SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.", "The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).", "The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.", "The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].", "The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR", "The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.", "Total RNA was extracted from EnCL-1 cells using TRIZOL Reagent according to the manufacturer's instructions. One microgram of each sample of total RNA was reverse transcribed using the SuperScript First-Strand Synthesis System for RT-PCR (11904-018 Invitrogen), as described in the supplier's protocol.", "mRNA expression of the SV40 T-ag was confirmed by conventional PCR using primers for SV40 T-ag detailed in Table 1. The EnCL-1 cells cDNA was amplified with JumpStar™ REDTaq™ ReadyMix PCR Reaction Mix (P 0982; Sigma). The PCR conditions were as follows: 3 min, 95 °C and 30 sec, 58°C, and 30 sec 72°C for 30 cycles. The samples were electrophoresed on 1,5% agarose gel containing ethidium bromide.", "EnCL-1 were plated in 2- and 4- well chamber slides at a concentration 1 × 105 cells/ml and after 24 h the cells were fixed with 4% paraformaldehyde, washed 3x with PBS and blocked with PAV/10% NDS (normal donkey serum) 1 h. Then, the cells were incubated overnight with primary antibodies specific to von Willenbrand factor and VE-cadherin. Next, the cells were washed 3x with PBS and incubated 1 h room temp. with secondary antibodies conjugated with cyanine 3 (CY3; Jackson ImmunoResearch, West Grove, PA). Furthermore, the cells were counterstained with DAPI UltraCruz™ Mounting Medium (sc-24941; Santa Cruz Biotechnology). EnCL-1 cells were visualized with confocal imaging using a Nicon C1 confocal microscope.", "Quantitative fluorescence real-time PCR was performed using the Applied Biosystems 7300 System (850 Lincoln Centre Drive Foster City, CA 94404 USA) with a SYBR Green PCR master mix (4367659; Power SYBR® Green PCR Master Mix; Applied Biosystems) following the manufacturer's instructions. Real-time PCR (25 μl) included 12.5 μl SYBR Green PCR Master Mix, 0.5 μM sense and antisense primers each, and reverse-transcribed cDNA (1 μl of cDNA). Primer sequences are detailed in Table 1. For quantification, standard curves consisting of serial dilutions of the appropriate purified cDNA were plotted. Amplification was proceeded by an initial denaturation step (15 min at 95°C). The PCR programs for each gene were performed as follows: 40 cycles of denaturation (15 sec at 95°C), annealing (30 sec at 56°C), and elongation (60 sec at 72°C). After PCR, melting curves were acquired by stepwise increases at a temperature of 50-95°C to ensure that a single product was amplified in the reaction. The data obtained from the real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 were normalized against GAPDH. Control reactions in the absence of reverse transcriptase were performed to test for genomic DNA contamination. The specify of the PCR products for examined genes was confirmed by gel electrophoresis and sequencing. The sequence homology obtained in the experiment was 99%. Dissociation curve analysis was performed after each real-time experiment to confirm the presence of only one product and the absence of the formation of primer dimmers. The threshold cycle number (Ct) for each tested gene X was used to quantify the relative abundance of the gene; arbitrary units were calculated as: 2-{∆}Ct = 2- (Ct target gene-Ct housekeeping gene).", "Concentration of PGE2 in culture media was determined by EIA as previously described [31]. The PGE2 standard curve ranged from 0.07 to 20 ng/ml and the ID50 of the assay was 1.25 ng/ml. The intra- and inter-assay CV averaged 6.9% and 9.7% respectively. Concentration of PGF2α in culture media was determined by EIA as described previously [31].\nThe PGF2α standard curve ranged from 0.07 to 20 ng/ml, and the ID50 of the assay was 1.82 ng/ml. The intra- and inter-assay CV were 7.4% and 11.6% respectively.\nConcentration of EDN-1 in culture media was determined by EIA using commercially kit (385131, Endothelin-1 EIA Kit, Cayman, Cayman Chemical, Ann Arbor, MI, USA). The EDN-1 standard curve ranged from 0.39 to 250 pg/ml and the ID50 of the assay was 7.8 pg/ml. The intra- and inter-assay CV was 10.0%.\nThe concentrations of LTB4 and C4 in culture media were determined using commercially available EIA kits (520211, Leukotriene C4 EIA Kit, 520111, Leukotriene B4 EIA Kit; Cayman) according to Korzekwa et al. [19]. The LTB4 standard curve ranged from 1.96 to 1000 pg/ml, and the effective dose for 50% inhibition (ID50) of the assay was 2.5 pg/ml. The intra- and inter-assay coefficients of variation were on average 4.1% and 6.2%, respectively. The LTC4 standard curve ranged from 0.98 to 500 pg/ml and the effective dose for 50% inhibition (ID50) of the assay was 1.85 pg/ml. The intra- and inter-assay CV were on average 4.9% and 7.4%, respectively.", "The equal amounts (30 μg) of protein were dissolved in SDS gel-loading buffer (50 mM Tris-HCl, pH 6.8; 4% SDS, 20% glycerol and 2% β-mercaptoethanol), heated to 95°C for 4 min and separated on 10% (for GAPDH, PGFS, LTA4H, EDN-1/2/3) and 15% (for LTC4S and mPGES-1) SDS-PAGE. Separated proteins were electroblotted onto 0.2 μm nitrocellulose membrane in transfer buffer (20 mM Tris-HCl buffer, pH 8.2; 150 mM glycine, 20% methanol). After blocking in 5% non-fat dry milk in TBS-T buffer (Tris-buffered saline, containing 0.1% Tween-20) for 1.5 h at 25.6°C, the membranes were incubated overnight with 1:2000 anti-lung-type PGFS antiserum antiserum [32], or 1:1000 polyclonal anti-mPGES-1 antibodies (160140; Cayman) or 1:1000 polyclonal LTC4 synthase antibody (sc-20108; Santa Cruz Biotechnology, Inc, Heidelberg, Germany) or 1:1000 LTA4 hydrolase antibody (ab23677; Abcam, Cambridge, UK) or 1:200 EDN 1/2/3 (sc-98727; Santa Cruz Biotechnology, ET-1/2/3 = EDN-1, EDN-2, EDN-3) at 4°C. Protein expression of GAPDH (G 8795; Sigma, glyceraldehyde-3-phosphate dehydrogenase) as a reference was used. Subsequently, proteins were detected by incubating the membrane with 1:20000 dilution of secondary polyclonal anti-rabbit or anti-goat alkaline phosphatase-conjugated antibodies (A 3687; A 3562 Sigma) for 1.5 hour at 25.6°C. Immune complexes were visualized using standard alkaline phosphatase visualization procedure. Western blots were quantitated using Kodak 1D program (Eastman Kodak, Rochester, NY, USA).", "The statistical significance of differences between the control and treated groups were analyzed by one-way ANOVA followed by t-student's post-hoc test (ANOVA; GraphPAD PRISM Version 5.00, San Diego, CA; USA), if the initial ANOVA was significant (P < 0.05).", "[SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] Selected immortalized endothelial cell line (EnCL-1) was cultured till 50 passages without any sign of senescence, which allowed to receive clear immortalized line of cells with homogenous morphology and genotype, although from 10 passage the line of cells has no any percentage of primary luteal endothelial cells. Phase contrast microscopy revealed that immortalized EnCL-1 cells grew as confluent monolayers with typical cobblestone morphology of primary endothelial cells. These cells (13th passage) were homogenous, polygonal and had characteristic ovoid nuclei (Figure 1a,b). Moreover, immunofluorescence staining revealed the presence of endothelial cell markers: von Willebrand factor (Figure 1C,E,F) and VE-cadherin (Figure 1H,I) in EnCL-1 cells. All isolated colonies expressed transfected vector. Expression of SV40 T-ag gene in the cells was confirmed by RT-PCR (Figure 2).\nCharacterization of morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in EnCL-1 cells. Cells reveal cobblestone morphology and tendency to form monolayer on Petri dish (A-B, mag. x10, x40). Localization of cytoplasmatic protein - vWF (panel E, mag. x10, x40) and expression of VE-cadherin (panel H, mag. x10, x40). Cells (13th passage) were stained with specific antibodies and detected by secondary antibodies labeled with fluorescent dyes, and analyzed by confocal microscopy. The DNA was counterstained with DAPI (blue, panels D, G). Panels C, F and I merged pictures of two fluorescent dyes. Panel C mag. x100.\nExpression of SV40 T-ag mRNA in EnCL-1 cells by RT-PCR. Left - SV40 Tag positive control, Right-EnCL-1 cells.\nSelected immortalized endothelial cell line (EnCL-1) was cultured till 50 passages without any sign of senescence, which allowed to receive clear immortalized line of cells with homogenous morphology and genotype, although from 10 passage the line of cells has no any percentage of primary luteal endothelial cells. Phase contrast microscopy revealed that immortalized EnCL-1 cells grew as confluent monolayers with typical cobblestone morphology of primary endothelial cells. These cells (13th passage) were homogenous, polygonal and had characteristic ovoid nuclei (Figure 1a,b). Moreover, immunofluorescence staining revealed the presence of endothelial cell markers: von Willebrand factor (Figure 1C,E,F) and VE-cadherin (Figure 1H,I) in EnCL-1 cells. All isolated colonies expressed transfected vector. Expression of SV40 T-ag gene in the cells was confirmed by RT-PCR (Figure 2).\nCharacterization of morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in EnCL-1 cells. Cells reveal cobblestone morphology and tendency to form monolayer on Petri dish (A-B, mag. x10, x40). Localization of cytoplasmatic protein - vWF (panel E, mag. x10, x40) and expression of VE-cadherin (panel H, mag. x10, x40). Cells (13th passage) were stained with specific antibodies and detected by secondary antibodies labeled with fluorescent dyes, and analyzed by confocal microscopy. The DNA was counterstained with DAPI (blue, panels D, G). Panels C, F and I merged pictures of two fluorescent dyes. Panel C mag. x100.\nExpression of SV40 T-ag mRNA in EnCL-1 cells by RT-PCR. Left - SV40 Tag positive control, Right-EnCL-1 cells.\n[SUBTITLE] Experiment 2. Effect of cytokines on production and content of Arachidonic Acid metabolites in immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] [SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] TNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nTNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] TNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\nTNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] Cytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).\nCytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] TNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nTNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] TNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\nTNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] Cytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).\nCytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).", "Selected immortalized endothelial cell line (EnCL-1) was cultured till 50 passages without any sign of senescence, which allowed to receive clear immortalized line of cells with homogenous morphology and genotype, although from 10 passage the line of cells has no any percentage of primary luteal endothelial cells. Phase contrast microscopy revealed that immortalized EnCL-1 cells grew as confluent monolayers with typical cobblestone morphology of primary endothelial cells. These cells (13th passage) were homogenous, polygonal and had characteristic ovoid nuclei (Figure 1a,b). Moreover, immunofluorescence staining revealed the presence of endothelial cell markers: von Willebrand factor (Figure 1C,E,F) and VE-cadherin (Figure 1H,I) in EnCL-1 cells. All isolated colonies expressed transfected vector. Expression of SV40 T-ag gene in the cells was confirmed by RT-PCR (Figure 2).\nCharacterization of morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in EnCL-1 cells. Cells reveal cobblestone morphology and tendency to form monolayer on Petri dish (A-B, mag. x10, x40). Localization of cytoplasmatic protein - vWF (panel E, mag. x10, x40) and expression of VE-cadherin (panel H, mag. x10, x40). Cells (13th passage) were stained with specific antibodies and detected by secondary antibodies labeled with fluorescent dyes, and analyzed by confocal microscopy. The DNA was counterstained with DAPI (blue, panels D, G). Panels C, F and I merged pictures of two fluorescent dyes. Panel C mag. x100.\nExpression of SV40 T-ag mRNA in EnCL-1 cells by RT-PCR. Left - SV40 Tag positive control, Right-EnCL-1 cells.", "[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] TNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nTNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] TNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\nTNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] Cytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).\nCytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).", "TNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.", "TNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.", "Cytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).", "The presence of SV40 T-ag in EnCL-1 cells and repeated passage without the apparent senescence confirmed the permanent status of the selected cell line. VE-cadherin and von Willebrand factor - positive features support the endothelial phenotype of the cell line. T antigen present in SV40 virus blocks cell death by apoptosis and it also interacts with a cytoplasmic protein that contains BH3 domain [33]. The viral genes achieve immortalization by inactivating the tumor suppressor genes (p53, Rb) that induce a replicative senescent state in cells. SV40 T antigen also induces telomerase activity in the infected cells [33]. So far, each study was conducted on the primary luteal endothelial cells or line of endothelial cells but not received from the bovine CL, which could possess different surface antigens and at least physiology [10]. We established the line of endothelial cells from bovine CL. To our knowledge, this is the first report showing the morphological and physiological properties of immortalized endothelial cells collected from bovine CL. According to Davis et al. [4], there are five types of bovine luteal endothelial cells differ the presence of cytokeratin, expression of surface antigens and neuronal cell adhesion molecule, capable for the contact between steroidogenic and endothelial cells within CL. We did not determine the morphological type and surface antigens of received line of endothelial cells than further study are necessary.\nEndothelial cells posses receptors for TNFα [20,34,35] and IFNγ [36]. Several papers confirmed synergistic, antiproliferative and proapoptotic action of TNFα and IFNγ in the CL [4,10,19,21,22]. In this study, TNFα and IFNγ treatment of cells increased each of studied mRNA expressions. In addition, PGF2α secretion and its synthase protein expression were stimulated. Similar effect received Acosta et al. [8], TNFα elevated PGF2α content in fresh and unfrozen cells till 10th passage in primary luteal endothelial cells. Whereas, in the study of Cavicchio et al. [10], cytokines stimulation was unresponsive to PGF2α secretion in the luteal endothelial cells (passage 2). The role of cytokines in regression of CL and cytokine effect on the main luteolytic factor-PGF2α was considered, as enhancing PGF2α action in the functional and structural luteolysis [17,19,20,34]. We also received the stimulation of mRNA expression for PGES without the effect on its protein expression and the level of PGE2 after cytokines treatment. Such an effect may be the consequence of changes in intracellular regulation of EnCL-1 cells, especially in mitochondrial activity. PGE2 enhances cellular proliferation, promotes angiogenesis, inhibits apoptosis and suppresses immune responses in cancerogenesis [35]. EnCL-1 cells potentially are programmed genetically for proliferation, thus cytokines, typically causing apoptosis, may not cause such an effect in our study and simultaneously stimulate PGE2 mRNA expression as kind of the preparation for proliferative functions.\nFurthermore, we considered the role of cytokines in connection with another aa metabolites-LTs in EnCL-1 cells. mRNA expression, as well as protein expression for LTCS and LTC4 release were stimulated by cytokines in EnCL-1 cells. Cytokines also stimulated LTA4H mRNA expression but did not change LTA4H protein expression and LTB4 secretion in EnCL-1 cells. Recently we showed that primary bovine luteal endothelial cells show mRNA expression for LT receptors (for LTB4 and C4) [26]. So far LTs role in CL function focus on steroidogenic cells in vitro [30] or concerned with processes in bovine reproductive tract in vivo [37,38]. LTB4 plays luteotropic role in bovine CL, stimulating PGE2 secretion, whereas LTC4 stimulate PGF2α and thus acts as luteolytic factor in vivo [38]. There is lack of knowledge about LTs role in CL vascular processes like angiogenesis and angioregression. It is possible that cytokine effect in the ovary is modulated by LTs. The immune cells (macrophages, monocytes and leukocytes) infiltrate the ovary and secrete cytokines in process of ovulation [39]. Cytokines affect non-steroidogenic ovarian cells, causing the release of ovulation mediators, such as metabolites of aa: PGs and LTs [39]. Thus, cytokines are involved in ovarian processes during the estrous cycle such as differentiation of CL, ovulation, luteolysis and cooperate with PGs and LTs [26,30,37-39]. Beside, both PGs and LTs appear to act in parallel in the regulation of cell proliferation and neoangiogenesis [40].\nWe selected EDN-1 as one of the main factors produced in endothelial cells and checked the effect of cytokines action on edn-1 mRNA expression and EDN-1 release in EnCL-1 cells. Our results confirmed the earlier studies [8,20,41] because the cytokines stimulated both mRNA and its production in EnCL-1 cells. Protein expression of EDN-1/2/3 was also elevated in our study as summary of the expression for EDN-1, EDN-2 and EDN-3. EDN-1 mRNA and protein is expressed in luteal endothelial cells during all the estrous cycle and EDN-1 inhibits P4 production in late luteal phase [42,43], EDN-2, in the early CL [44], whereas EDN-3, on the contrary to mentioned EDN-1 and EDN-2, does not affect luteal steroidogenesis [45]. Our results indicate that cytokines enhance EDN-1/2/3 action and indicate on multiple characteristic functions of EnCL-1 cells.\nConcluding, cytokines modulate EnCL-1 cells function by up-regulation of PGES, PGFS, LTA4H, LTC4S and mRNA expression. Protein expression was elevated by cytokines for PGFS and LTCS and simultaneously the level of appropriate active metabolites of these synthases products (PGF2α and LTC4) were higher after cytokine stimulation. Protein expression for PGES and LTA4H was not changed and release of products of these syntases-PGE2 and LTB4 was also stable. Beside, mRNA expression, level of EDN-1 and protein expression for EDN1/2/3 were upregulated by cytokines, which suggest that EnCL-1 cells show multiple potency, both prolifereative and proapoptotic. In this respect, EnCL-1 cells may serve as an appropriate model for investigating the paradigm of counteraction of the luteolytic and luteotropic properties of bovine CL. The received line of immortalized EnCL-1 cells possess stable genotype and phenotype. However, an intricate molecular structure of the cells with multiple intervenient factors/hormones and actions deserve to be defined.", "The authors declare that they have no competing interests.", "AJK collected the experimental material, isolated primary endothelial CL cells, carried out the immunoassays, mRNA expression and MTT methods and performed the statistical analysis, designed, drafted and coordinated the study. GB carried out the immortalization of the primary endothelial cells, participated in the design of the study and draft the manuscript. JB assisted and helped technically in all steps of experimental procedures. AB and DJS helped to design and draft the manuscript. All authors read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Collection of CL for in vitro experiments", "Luteal endothelial cells isolation", "Experimental procedures", "Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization", "Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1)", "Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1)", "Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1)", "Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells", "Total RNA isolation", "Conventional PCR", "Immunofluorescence staining", "Real-time PCR quantification", "EIA determination", "Western blot analysis", "Statistical analysis", "Results", "Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization", "Experiment 2. Effect of cytokines on production and content of Arachidonic Acid metabolites in immortalized bovine luteal endothelial cells (EnCL-1)", "Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1)", "Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1)", "Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells", "Discussion", "Competing interests", "Authors' contributions" ]

[ "Although corpus luteum (CL) is a transient gland, it is one of the most vascularized tissues in the body [1], with endothelial cells representing greater than fifty percent of the total cells [2,3]. Angiogenesis is critical to CL development, whereas endothelial cells decline occurs during luteolysis [4]. On the other side, endothelial cells play a crucial role in a complex processes of tumor neovascularization [5], including CL cancers [6]. Because of these important and multiplex functions of endothelial cells in CL vascularity, the establishment of an experimental model of immortalized endothelial cells from bovine CL is a prerequisite for the study of cellular and molecular mechanism in this tissue. So far, the majority of studies have been conducted on fresh isolated or refrozen aliquots of bovine primary luteal endothelial cells [4,7-9] or cell line received not directly from CL [10]. Immortalized endothelial cells have been characterized in several kinds of bovine tissues, such as the pulmonary and coronary arteries, nevertheless no bovine luteal endothelial cell line is available [10]. There is a possibility that surface antigens and/or genetic programming differs for endothelial cells derived from various tissues, beside each kind of cell is strictly species dependent [11]. Thus, the stable bovine luteal endothelial cell line with determined fenotype and genotype would be the convenient and useful model for the future study.\nAmong mediators of interactions between different types of CL cells, including endothelial cells, the universal factors are immune cells and their secreted products, cytokines [12-14]. Endothelial cells are capable of tumor necrosis factor α (TNFα) synthesis and secretion [15]. Depending on the immediate microenvironment, TNFα may stimulate cell proliferation or induce apoptosis of luteal endothelial cells [4]. TNFα action in the bovine CL is a dose dependent: a low concentration of TNFα stimulates in vivo luteolytic factors, as well as induces apoptosis; whereas the high concentration of TNFα stimulates a survival pathway [16-19]. Moreover, TNFα induced apoptosis in cultured bovine luteal endothelial cells [20]. TNFα effect in the ovary was found to be more effective when TNFα acted synergistically with interferon γ (IFNγ; [13,19,21,22]). Reasonable is the generation of stable in vitro luteal endothelial cell culture for investigating the complex signaling pathway and transcriptional mechanisms regulated by cytokines in physiological and pathophysiological conditions in cattle.\nThe proper vascularization and endothelial cell activity per se are essential for normal CL function [23]. The effect of prostaglandins (PGs) on the vascularity of bovine CL is well known [9,21,23-26]. The ovarian blood flow has been shown to increase after PGE2 administration and decrease during spontaneous and PGF2α induced luteolysis in cows [27]. An acute increase in the luteal blood flow occurs as the first step of luteolysis in response to PGF2α [3]. Both the density and the number of blood vessels were higher in CLs obtained after PGF2α administration than in those without PGF2α treatment [21], which indicate that the number of blood vessels with smooth muscle in the regressing CL increased as a result of loosing steroidogenic cells and capillaries. A mitogenic effect and increased proliferation were observed after PGF2α treatment in bovine dispersed luteal endothelial cells [23]. Moreover, PG receptors, as well as leukotriene (LTs - other metabolites of arachidonic acid - aa) receptors are present on endothelial cells [23,26], which indicate that the endothelial cells of bovine CL are target for PGs and LTs. Leukotrienes are commonly known as the potential inflammatory factors that cause edema in respiratory tract diseases, but they also play the important role in reproduction and may enhance the action of PGs [28]. We hypothesize that cytokines are involved in aa metabolites action in bovine luteal endothelial cells.\nThe main goal of this study was to determine the effect of cytokines on aa metabolites action in bovine immortalized luteal endothelial cells. We examined: (i) the viability of EnCL-1 cells, (ii) mRNA expression for LTA4 hydrolase (LTA4H), LTC4 synthase (LTC4S), PGE2 and PGF2α synthases (PGES and PGFS) and endothelin-1 (EDN-1), (iii) protein expression for LTA4H, LTC4S, PGES, PGFS and EDN-1/2/3 and (iv) accumulation of LTB4 and C4, PGE2 and F2α and EDN-1 in the culture medium after TNFα/IFNγ stimulation.", "[SUBTITLE] Collection of CL for in vitro experiments [SUBSECTION] Healthy, normally cycling Holstein/Polish Black and White (75/25%; respectively) cows (3-4 lactation) were used for the collection of the ovaries with CL (in total 4 animals). The animals were eliminated by the owners (Experimental Animal Farm of Polish Academy of Sciences in Baranowo; Poland) from the dairy herds because of their lower milk production. The estrus of the cows was synchronized twice using an analogue of PGF2α (dinoprost, Dinolytic; Upjohn - Pharmacia N.V.S.A., Belgium) injections with an 11-day interval as recommended by a vendor. The onset of the estrus was determined by a veterinarian via per rectum USG examination using a DRAMINSKI ANIMALprofi Scanner [29] and confirmed by observing the signs of estrus (i.e. vaginal mucus, standing behavior). The onset of estrus was determined as Day 0. Only cows with signs of estrus were chosen for the study. Animals were slaughtered on Day 8-12 of the estrous cycle and the ovaries were obtained within 20 min of the exsanguinations and transported on ice to the laboratory.\nHealthy, normally cycling Holstein/Polish Black and White (75/25%; respectively) cows (3-4 lactation) were used for the collection of the ovaries with CL (in total 4 animals). The animals were eliminated by the owners (Experimental Animal Farm of Polish Academy of Sciences in Baranowo; Poland) from the dairy herds because of their lower milk production. The estrus of the cows was synchronized twice using an analogue of PGF2α (dinoprost, Dinolytic; Upjohn - Pharmacia N.V.S.A., Belgium) injections with an 11-day interval as recommended by a vendor. The onset of the estrus was determined by a veterinarian via per rectum USG examination using a DRAMINSKI ANIMALprofi Scanner [29] and confirmed by observing the signs of estrus (i.e. vaginal mucus, standing behavior). The onset of estrus was determined as Day 0. Only cows with signs of estrus were chosen for the study. Animals were slaughtered on Day 8-12 of the estrous cycle and the ovaries were obtained within 20 min of the exsanguinations and transported on ice to the laboratory.\n[SUBTITLE] Luteal endothelial cells isolation [SUBSECTION] Endothelial cells isolation was proceeded according to the method described previously in details [8,26] using a Dynabeads kit (140.03 Dynabeads® M-450; Invitrogen, Oregon, USA). Briefly, the Dynabeads were coated with the specific antigen-lectin (BS-1; L 2380 Sigma; Chemical Co., St Louis, MO, USA). The beads coated with endothelial cells were attracted by a magnet to the well of the tube and the supernatant was removed. After washing with PBS, 1 ml of 0.1 M fucose (T 2252 Sigma) solution was added to break the connection of endothelial cells with beads. The free beads were then attracted by a magnet to the well of the tube and the supernatant with endothelial cells was collected. The obtained cell suspension contained more than 85% of luteal endothelial cells and only a few steroidogenic CL cells. The cells were suspended in Dulbecco's Modified Eagle's medium (DMEM/Ham's F-12; 1: 1(v/v); D 2906 Sigma) in a 3 ml culture flask in a humidified incubator at 37.5°C in 5% CO2/95% air atmosphere. After third passage the cells were trypsinized and placed at the concentration of 2 × 105 cells/ml into a 24-well culture plate. After 24-48 h of culture cells reached confluence and were proceeded the procedure of immortalization.\n[SUBTITLE] Experimental procedures [SUBSECTION] [SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\n[SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nEndothelial cells isolation was proceeded according to the method described previously in details [8,26] using a Dynabeads kit (140.03 Dynabeads® M-450; Invitrogen, Oregon, USA). Briefly, the Dynabeads were coated with the specific antigen-lectin (BS-1; L 2380 Sigma; Chemical Co., St Louis, MO, USA). The beads coated with endothelial cells were attracted by a magnet to the well of the tube and the supernatant was removed. After washing with PBS, 1 ml of 0.1 M fucose (T 2252 Sigma) solution was added to break the connection of endothelial cells with beads. The free beads were then attracted by a magnet to the well of the tube and the supernatant with endothelial cells was collected. The obtained cell suspension contained more than 85% of luteal endothelial cells and only a few steroidogenic CL cells. The cells were suspended in Dulbecco's Modified Eagle's medium (DMEM/Ham's F-12; 1: 1(v/v); D 2906 Sigma) in a 3 ml culture flask in a humidified incubator at 37.5°C in 5% CO2/95% air atmosphere. After third passage the cells were trypsinized and placed at the concentration of 2 × 105 cells/ml into a 24-well culture plate. After 24-48 h of culture cells reached confluence and were proceeded the procedure of immortalization.\n[SUBTITLE] Experimental procedures [SUBSECTION] [SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\n[SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\n[SUBTITLE] Total RNA isolation [SUBSECTION] Total RNA was extracted from EnCL-1 cells using TRIZOL Reagent according to the manufacturer's instructions. One microgram of each sample of total RNA was reverse transcribed using the SuperScript First-Strand Synthesis System for RT-PCR (11904-018 Invitrogen), as described in the supplier's protocol.\nTotal RNA was extracted from EnCL-1 cells using TRIZOL Reagent according to the manufacturer's instructions. One microgram of each sample of total RNA was reverse transcribed using the SuperScript First-Strand Synthesis System for RT-PCR (11904-018 Invitrogen), as described in the supplier's protocol.\n[SUBTITLE] Conventional PCR [SUBSECTION] mRNA expression of the SV40 T-ag was confirmed by conventional PCR using primers for SV40 T-ag detailed in Table 1. The EnCL-1 cells cDNA was amplified with JumpStar™ REDTaq™ ReadyMix PCR Reaction Mix (P 0982; Sigma). The PCR conditions were as follows: 3 min, 95 °C and 30 sec, 58°C, and 30 sec 72°C for 30 cycles. The samples were electrophoresed on 1,5% agarose gel containing ethidium bromide.\nmRNA expression of the SV40 T-ag was confirmed by conventional PCR using primers for SV40 T-ag detailed in Table 1. The EnCL-1 cells cDNA was amplified with JumpStar™ REDTaq™ ReadyMix PCR Reaction Mix (P 0982; Sigma). The PCR conditions were as follows: 3 min, 95 °C and 30 sec, 58°C, and 30 sec 72°C for 30 cycles. The samples were electrophoresed on 1,5% agarose gel containing ethidium bromide.\n[SUBTITLE] Immunofluorescence staining [SUBSECTION] EnCL-1 were plated in 2- and 4- well chamber slides at a concentration 1 × 105 cells/ml and after 24 h the cells were fixed with 4% paraformaldehyde, washed 3x with PBS and blocked with PAV/10% NDS (normal donkey serum) 1 h. Then, the cells were incubated overnight with primary antibodies specific to von Willenbrand factor and VE-cadherin. Next, the cells were washed 3x with PBS and incubated 1 h room temp. with secondary antibodies conjugated with cyanine 3 (CY3; Jackson ImmunoResearch, West Grove, PA). Furthermore, the cells were counterstained with DAPI UltraCruz™ Mounting Medium (sc-24941; Santa Cruz Biotechnology). EnCL-1 cells were visualized with confocal imaging using a Nicon C1 confocal microscope.\nEnCL-1 were plated in 2- and 4- well chamber slides at a concentration 1 × 105 cells/ml and after 24 h the cells were fixed with 4% paraformaldehyde, washed 3x with PBS and blocked with PAV/10% NDS (normal donkey serum) 1 h. Then, the cells were incubated overnight with primary antibodies specific to von Willenbrand factor and VE-cadherin. Next, the cells were washed 3x with PBS and incubated 1 h room temp. with secondary antibodies conjugated with cyanine 3 (CY3; Jackson ImmunoResearch, West Grove, PA). Furthermore, the cells were counterstained with DAPI UltraCruz™ Mounting Medium (sc-24941; Santa Cruz Biotechnology). EnCL-1 cells were visualized with confocal imaging using a Nicon C1 confocal microscope.\n[SUBTITLE] Real-time PCR quantification [SUBSECTION] Quantitative fluorescence real-time PCR was performed using the Applied Biosystems 7300 System (850 Lincoln Centre Drive Foster City, CA 94404 USA) with a SYBR Green PCR master mix (4367659; Power SYBR® Green PCR Master Mix; Applied Biosystems) following the manufacturer's instructions. Real-time PCR (25 μl) included 12.5 μl SYBR Green PCR Master Mix, 0.5 μM sense and antisense primers each, and reverse-transcribed cDNA (1 μl of cDNA). Primer sequences are detailed in Table 1. For quantification, standard curves consisting of serial dilutions of the appropriate purified cDNA were plotted. Amplification was proceeded by an initial denaturation step (15 min at 95°C). The PCR programs for each gene were performed as follows: 40 cycles of denaturation (15 sec at 95°C), annealing (30 sec at 56°C), and elongation (60 sec at 72°C). After PCR, melting curves were acquired by stepwise increases at a temperature of 50-95°C to ensure that a single product was amplified in the reaction. The data obtained from the real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 were normalized against GAPDH. Control reactions in the absence of reverse transcriptase were performed to test for genomic DNA contamination. The specify of the PCR products for examined genes was confirmed by gel electrophoresis and sequencing. The sequence homology obtained in the experiment was 99%. Dissociation curve analysis was performed after each real-time experiment to confirm the presence of only one product and the absence of the formation of primer dimmers. The threshold cycle number (Ct) for each tested gene X was used to quantify the relative abundance of the gene; arbitrary units were calculated as: 2-{∆}Ct = 2- (Ct target gene-Ct housekeeping gene).\nQuantitative fluorescence real-time PCR was performed using the Applied Biosystems 7300 System (850 Lincoln Centre Drive Foster City, CA 94404 USA) with a SYBR Green PCR master mix (4367659; Power SYBR® Green PCR Master Mix; Applied Biosystems) following the manufacturer's instructions. Real-time PCR (25 μl) included 12.5 μl SYBR Green PCR Master Mix, 0.5 μM sense and antisense primers each, and reverse-transcribed cDNA (1 μl of cDNA). Primer sequences are detailed in Table 1. For quantification, standard curves consisting of serial dilutions of the appropriate purified cDNA were plotted. Amplification was proceeded by an initial denaturation step (15 min at 95°C). The PCR programs for each gene were performed as follows: 40 cycles of denaturation (15 sec at 95°C), annealing (30 sec at 56°C), and elongation (60 sec at 72°C). After PCR, melting curves were acquired by stepwise increases at a temperature of 50-95°C to ensure that a single product was amplified in the reaction. The data obtained from the real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 were normalized against GAPDH. Control reactions in the absence of reverse transcriptase were performed to test for genomic DNA contamination. The specify of the PCR products for examined genes was confirmed by gel electrophoresis and sequencing. The sequence homology obtained in the experiment was 99%. Dissociation curve analysis was performed after each real-time experiment to confirm the presence of only one product and the absence of the formation of primer dimmers. The threshold cycle number (Ct) for each tested gene X was used to quantify the relative abundance of the gene; arbitrary units were calculated as: 2-{∆}Ct = 2- (Ct target gene-Ct housekeeping gene).\n[SUBTITLE] EIA determination [SUBSECTION] Concentration of PGE2 in culture media was determined by EIA as previously described [31]. The PGE2 standard curve ranged from 0.07 to 20 ng/ml and the ID50 of the assay was 1.25 ng/ml. The intra- and inter-assay CV averaged 6.9% and 9.7% respectively. Concentration of PGF2α in culture media was determined by EIA as described previously [31].\nThe PGF2α standard curve ranged from 0.07 to 20 ng/ml, and the ID50 of the assay was 1.82 ng/ml. The intra- and inter-assay CV were 7.4% and 11.6% respectively.\nConcentration of EDN-1 in culture media was determined by EIA using commercially kit (385131, Endothelin-1 EIA Kit, Cayman, Cayman Chemical, Ann Arbor, MI, USA). The EDN-1 standard curve ranged from 0.39 to 250 pg/ml and the ID50 of the assay was 7.8 pg/ml. The intra- and inter-assay CV was 10.0%.\nThe concentrations of LTB4 and C4 in culture media were determined using commercially available EIA kits (520211, Leukotriene C4 EIA Kit, 520111, Leukotriene B4 EIA Kit; Cayman) according to Korzekwa et al. [19]. The LTB4 standard curve ranged from 1.96 to 1000 pg/ml, and the effective dose for 50% inhibition (ID50) of the assay was 2.5 pg/ml. The intra- and inter-assay coefficients of variation were on average 4.1% and 6.2%, respectively. The LTC4 standard curve ranged from 0.98 to 500 pg/ml and the effective dose for 50% inhibition (ID50) of the assay was 1.85 pg/ml. The intra- and inter-assay CV were on average 4.9% and 7.4%, respectively.\nConcentration of PGE2 in culture media was determined by EIA as previously described [31]. The PGE2 standard curve ranged from 0.07 to 20 ng/ml and the ID50 of the assay was 1.25 ng/ml. The intra- and inter-assay CV averaged 6.9% and 9.7% respectively. Concentration of PGF2α in culture media was determined by EIA as described previously [31].\nThe PGF2α standard curve ranged from 0.07 to 20 ng/ml, and the ID50 of the assay was 1.82 ng/ml. The intra- and inter-assay CV were 7.4% and 11.6% respectively.\nConcentration of EDN-1 in culture media was determined by EIA using commercially kit (385131, Endothelin-1 EIA Kit, Cayman, Cayman Chemical, Ann Arbor, MI, USA). The EDN-1 standard curve ranged from 0.39 to 250 pg/ml and the ID50 of the assay was 7.8 pg/ml. The intra- and inter-assay CV was 10.0%.\nThe concentrations of LTB4 and C4 in culture media were determined using commercially available EIA kits (520211, Leukotriene C4 EIA Kit, 520111, Leukotriene B4 EIA Kit; Cayman) according to Korzekwa et al. [19]. The LTB4 standard curve ranged from 1.96 to 1000 pg/ml, and the effective dose for 50% inhibition (ID50) of the assay was 2.5 pg/ml. The intra- and inter-assay coefficients of variation were on average 4.1% and 6.2%, respectively. The LTC4 standard curve ranged from 0.98 to 500 pg/ml and the effective dose for 50% inhibition (ID50) of the assay was 1.85 pg/ml. The intra- and inter-assay CV were on average 4.9% and 7.4%, respectively.\n[SUBTITLE] Western blot analysis [SUBSECTION] The equal amounts (30 μg) of protein were dissolved in SDS gel-loading buffer (50 mM Tris-HCl, pH 6.8; 4% SDS, 20% glycerol and 2% β-mercaptoethanol), heated to 95°C for 4 min and separated on 10% (for GAPDH, PGFS, LTA4H, EDN-1/2/3) and 15% (for LTC4S and mPGES-1) SDS-PAGE. Separated proteins were electroblotted onto 0.2 μm nitrocellulose membrane in transfer buffer (20 mM Tris-HCl buffer, pH 8.2; 150 mM glycine, 20% methanol). After blocking in 5% non-fat dry milk in TBS-T buffer (Tris-buffered saline, containing 0.1% Tween-20) for 1.5 h at 25.6°C, the membranes were incubated overnight with 1:2000 anti-lung-type PGFS antiserum antiserum [32], or 1:1000 polyclonal anti-mPGES-1 antibodies (160140; Cayman) or 1:1000 polyclonal LTC4 synthase antibody (sc-20108; Santa Cruz Biotechnology, Inc, Heidelberg, Germany) or 1:1000 LTA4 hydrolase antibody (ab23677; Abcam, Cambridge, UK) or 1:200 EDN 1/2/3 (sc-98727; Santa Cruz Biotechnology, ET-1/2/3 = EDN-1, EDN-2, EDN-3) at 4°C. Protein expression of GAPDH (G 8795; Sigma, glyceraldehyde-3-phosphate dehydrogenase) as a reference was used. Subsequently, proteins were detected by incubating the membrane with 1:20000 dilution of secondary polyclonal anti-rabbit or anti-goat alkaline phosphatase-conjugated antibodies (A 3687; A 3562 Sigma) for 1.5 hour at 25.6°C. Immune complexes were visualized using standard alkaline phosphatase visualization procedure. Western blots were quantitated using Kodak 1D program (Eastman Kodak, Rochester, NY, USA).\nThe equal amounts (30 μg) of protein were dissolved in SDS gel-loading buffer (50 mM Tris-HCl, pH 6.8; 4% SDS, 20% glycerol and 2% β-mercaptoethanol), heated to 95°C for 4 min and separated on 10% (for GAPDH, PGFS, LTA4H, EDN-1/2/3) and 15% (for LTC4S and mPGES-1) SDS-PAGE. Separated proteins were electroblotted onto 0.2 μm nitrocellulose membrane in transfer buffer (20 mM Tris-HCl buffer, pH 8.2; 150 mM glycine, 20% methanol). After blocking in 5% non-fat dry milk in TBS-T buffer (Tris-buffered saline, containing 0.1% Tween-20) for 1.5 h at 25.6°C, the membranes were incubated overnight with 1:2000 anti-lung-type PGFS antiserum antiserum [32], or 1:1000 polyclonal anti-mPGES-1 antibodies (160140; Cayman) or 1:1000 polyclonal LTC4 synthase antibody (sc-20108; Santa Cruz Biotechnology, Inc, Heidelberg, Germany) or 1:1000 LTA4 hydrolase antibody (ab23677; Abcam, Cambridge, UK) or 1:200 EDN 1/2/3 (sc-98727; Santa Cruz Biotechnology, ET-1/2/3 = EDN-1, EDN-2, EDN-3) at 4°C. Protein expression of GAPDH (G 8795; Sigma, glyceraldehyde-3-phosphate dehydrogenase) as a reference was used. Subsequently, proteins were detected by incubating the membrane with 1:20000 dilution of secondary polyclonal anti-rabbit or anti-goat alkaline phosphatase-conjugated antibodies (A 3687; A 3562 Sigma) for 1.5 hour at 25.6°C. Immune complexes were visualized using standard alkaline phosphatase visualization procedure. Western blots were quantitated using Kodak 1D program (Eastman Kodak, Rochester, NY, USA).\n[SUBTITLE] Statistical analysis [SUBSECTION] The statistical significance of differences between the control and treated groups were analyzed by one-way ANOVA followed by t-student's post-hoc test (ANOVA; GraphPAD PRISM Version 5.00, San Diego, CA; USA), if the initial ANOVA was significant (P < 0.05).\nThe statistical significance of differences between the control and treated groups were analyzed by one-way ANOVA followed by t-student's post-hoc test (ANOVA; GraphPAD PRISM Version 5.00, San Diego, CA; USA), if the initial ANOVA was significant (P < 0.05).", "Healthy, normally cycling Holstein/Polish Black and White (75/25%; respectively) cows (3-4 lactation) were used for the collection of the ovaries with CL (in total 4 animals). The animals were eliminated by the owners (Experimental Animal Farm of Polish Academy of Sciences in Baranowo; Poland) from the dairy herds because of their lower milk production. The estrus of the cows was synchronized twice using an analogue of PGF2α (dinoprost, Dinolytic; Upjohn - Pharmacia N.V.S.A., Belgium) injections with an 11-day interval as recommended by a vendor. The onset of the estrus was determined by a veterinarian via per rectum USG examination using a DRAMINSKI ANIMALprofi Scanner [29] and confirmed by observing the signs of estrus (i.e. vaginal mucus, standing behavior). The onset of estrus was determined as Day 0. Only cows with signs of estrus were chosen for the study. Animals were slaughtered on Day 8-12 of the estrous cycle and the ovaries were obtained within 20 min of the exsanguinations and transported on ice to the laboratory.", "Endothelial cells isolation was proceeded according to the method described previously in details [8,26] using a Dynabeads kit (140.03 Dynabeads® M-450; Invitrogen, Oregon, USA). Briefly, the Dynabeads were coated with the specific antigen-lectin (BS-1; L 2380 Sigma; Chemical Co., St Louis, MO, USA). The beads coated with endothelial cells were attracted by a magnet to the well of the tube and the supernatant was removed. After washing with PBS, 1 ml of 0.1 M fucose (T 2252 Sigma) solution was added to break the connection of endothelial cells with beads. The free beads were then attracted by a magnet to the well of the tube and the supernatant with endothelial cells was collected. The obtained cell suspension contained more than 85% of luteal endothelial cells and only a few steroidogenic CL cells. The cells were suspended in Dulbecco's Modified Eagle's medium (DMEM/Ham's F-12; 1: 1(v/v); D 2906 Sigma) in a 3 ml culture flask in a humidified incubator at 37.5°C in 5% CO2/95% air atmosphere. After third passage the cells were trypsinized and placed at the concentration of 2 × 105 cells/ml into a 24-well culture plate. After 24-48 h of culture cells reached confluence and were proceeded the procedure of immortalization.\n[SUBTITLE] Experimental procedures [SUBSECTION] [SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\n[SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.", "[SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\nThe primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).\n[SUBTITLE] Experiment 2. Effect of cytokines on production and secretion of Arachidonic Acid metabolites in immortalized bovine endothelial cells (EnCL-1) [SUBSECTION] The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\nThe concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\nThe aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\nThe aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.\nThe aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.", "The primary cultures of endothelial cells were immortalized by transfection with the vector carrying a Simian virus 40 T-antigen (SV40 T-ag) sequence. Lipofectamine LTX (11514-015, Invitrogen) was used as a transfection agent according to manufacturer's instruction. The fast proliferating colonies were selected and initially scanned for the presence of transfected vector - SV40 T-ag. Expression of SV40 T-ag gene was further confirmed by RT-PCR in selected passages of the cells (13, 18, 23, 31, 40, 50th). The selected immortalized cell line call EnCL-1 was cultured for next 50 passages without any sign of senescence. After 13th passage every 5th passage of EnCL-1 cells were pulled and frozen in -80°C in cryotube portions (1.5 ml at the concentration 2.0 × 105/ml). Viability of unfrozen cells was higher than 85% as assessed by trypan blue dye exclusion. Characterization of EnCL-1 cell line was provided by fluorescence morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in selected passages of EnCL-1 cells (13, 18, 23, 31, 40, 50th).", "The concentration of TNFα and IFNγ (each in the dose: 50 ng/ml), and exposure time (24 h) were determined on the basis of previous studies [18,19,22,26,30]. All treatments were conducted in triplicates, four experiments were performed.", "The aim of the experiment was to evaluate the percentage of live cells after stimulation with studied cytokines comparing with non-treated cells. The EnCL-1 cells (13th passage) were adjusted to 2.0 105/ml of medium: DMEM supplemented with 5% calf serum (F 7524 Sigma) and 20 μg/ml gentamycin. Cells were cultured in 96-culture plates in a humidified incubator at 37.5°C in 5% CO2 and 95% air atmosphere. After 24 h, the cells were washed with serum-free DMEM and the medium was replaced by fresh medium: DMEM/Ham's F-12 supplemented with 0.1% (w/v) BSA and containing 20 μg/ml gentamycin. Then, the cells were treated simultaneously with TNFα (recombinant human TNFα: HF-13; Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) and IFNγ (PBP007; AbD Serotec, Biokom, Warsaw, Poland) for 24 h. Cell viability (TOX-1; Sigma) was measured using commercially available colorimetric assay kits according to the manufacturer\"s instructions as previously described [18].", "The aim of the experiment was to examine whether TNFα and IFNγ affect mRNA and protein expression of selected factors. Dispersed, unfrozen EnCL-1 cells (13th passage) were seeded at 2.0 × 105 viable cells in 1 ml of cultured medium in 24-well culture plates (3524, Costar). After 18 h of culture in DMEM medium containing 5% CS, the medium was replaced with DMEM containing 0.1% BSA with or without TNFα/IFNγ.\nmRNA expression was quantitavely measured by real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 and protein expression was measured by western blotting for LTC4S, LTA4H, PGES, PGFS and EDN-1/2/3 as previously described [26,30]. The sequences of used primers are described in Table 1. The cells were disrupted with TRIZOL Reagent (15596, Invitrogen) and frozen at -80°C until processed for RNA isolation and Reverse Transcription-Polymerase Chain Reaction (RT-PCR). Separately, the cells were frozen at -80°C until processed for protein isolation.\nOligonucleotide sequences used for RT-PCR", "The aim of the experiment was to study the effect of TNFα and IFNγ on release of PGE2, PGF2α, LTB4, LTC4 and EDN-1 by EnCL-1 cells.\nAfter incubation, media from Experiment 2.2 were collected into tubes containing 10 μl of stabilizer [0.3 M EDTA and 1% aspirin, A 2093; Sigma] for each 500 μl of medium and stored at - 20°C until EIA determinations.", "Total RNA was extracted from EnCL-1 cells using TRIZOL Reagent according to the manufacturer's instructions. One microgram of each sample of total RNA was reverse transcribed using the SuperScript First-Strand Synthesis System for RT-PCR (11904-018 Invitrogen), as described in the supplier's protocol.", "mRNA expression of the SV40 T-ag was confirmed by conventional PCR using primers for SV40 T-ag detailed in Table 1. The EnCL-1 cells cDNA was amplified with JumpStar™ REDTaq™ ReadyMix PCR Reaction Mix (P 0982; Sigma). The PCR conditions were as follows: 3 min, 95 °C and 30 sec, 58°C, and 30 sec 72°C for 30 cycles. The samples were electrophoresed on 1,5% agarose gel containing ethidium bromide.", "EnCL-1 were plated in 2- and 4- well chamber slides at a concentration 1 × 105 cells/ml and after 24 h the cells were fixed with 4% paraformaldehyde, washed 3x with PBS and blocked with PAV/10% NDS (normal donkey serum) 1 h. Then, the cells were incubated overnight with primary antibodies specific to von Willenbrand factor and VE-cadherin. Next, the cells were washed 3x with PBS and incubated 1 h room temp. with secondary antibodies conjugated with cyanine 3 (CY3; Jackson ImmunoResearch, West Grove, PA). Furthermore, the cells were counterstained with DAPI UltraCruz™ Mounting Medium (sc-24941; Santa Cruz Biotechnology). EnCL-1 cells were visualized with confocal imaging using a Nicon C1 confocal microscope.", "Quantitative fluorescence real-time PCR was performed using the Applied Biosystems 7300 System (850 Lincoln Centre Drive Foster City, CA 94404 USA) with a SYBR Green PCR master mix (4367659; Power SYBR® Green PCR Master Mix; Applied Biosystems) following the manufacturer's instructions. Real-time PCR (25 μl) included 12.5 μl SYBR Green PCR Master Mix, 0.5 μM sense and antisense primers each, and reverse-transcribed cDNA (1 μl of cDNA). Primer sequences are detailed in Table 1. For quantification, standard curves consisting of serial dilutions of the appropriate purified cDNA were plotted. Amplification was proceeded by an initial denaturation step (15 min at 95°C). The PCR programs for each gene were performed as follows: 40 cycles of denaturation (15 sec at 95°C), annealing (30 sec at 56°C), and elongation (60 sec at 72°C). After PCR, melting curves were acquired by stepwise increases at a temperature of 50-95°C to ensure that a single product was amplified in the reaction. The data obtained from the real-time RT-PCR for LTC4S, LTA4H, PGES, PGFS and EDN-1 were normalized against GAPDH. Control reactions in the absence of reverse transcriptase were performed to test for genomic DNA contamination. The specify of the PCR products for examined genes was confirmed by gel electrophoresis and sequencing. The sequence homology obtained in the experiment was 99%. Dissociation curve analysis was performed after each real-time experiment to confirm the presence of only one product and the absence of the formation of primer dimmers. The threshold cycle number (Ct) for each tested gene X was used to quantify the relative abundance of the gene; arbitrary units were calculated as: 2-{∆}Ct = 2- (Ct target gene-Ct housekeeping gene).", "Concentration of PGE2 in culture media was determined by EIA as previously described [31]. The PGE2 standard curve ranged from 0.07 to 20 ng/ml and the ID50 of the assay was 1.25 ng/ml. The intra- and inter-assay CV averaged 6.9% and 9.7% respectively. Concentration of PGF2α in culture media was determined by EIA as described previously [31].\nThe PGF2α standard curve ranged from 0.07 to 20 ng/ml, and the ID50 of the assay was 1.82 ng/ml. The intra- and inter-assay CV were 7.4% and 11.6% respectively.\nConcentration of EDN-1 in culture media was determined by EIA using commercially kit (385131, Endothelin-1 EIA Kit, Cayman, Cayman Chemical, Ann Arbor, MI, USA). The EDN-1 standard curve ranged from 0.39 to 250 pg/ml and the ID50 of the assay was 7.8 pg/ml. The intra- and inter-assay CV was 10.0%.\nThe concentrations of LTB4 and C4 in culture media were determined using commercially available EIA kits (520211, Leukotriene C4 EIA Kit, 520111, Leukotriene B4 EIA Kit; Cayman) according to Korzekwa et al. [19]. The LTB4 standard curve ranged from 1.96 to 1000 pg/ml, and the effective dose for 50% inhibition (ID50) of the assay was 2.5 pg/ml. The intra- and inter-assay coefficients of variation were on average 4.1% and 6.2%, respectively. The LTC4 standard curve ranged from 0.98 to 500 pg/ml and the effective dose for 50% inhibition (ID50) of the assay was 1.85 pg/ml. The intra- and inter-assay CV were on average 4.9% and 7.4%, respectively.", "The equal amounts (30 μg) of protein were dissolved in SDS gel-loading buffer (50 mM Tris-HCl, pH 6.8; 4% SDS, 20% glycerol and 2% β-mercaptoethanol), heated to 95°C for 4 min and separated on 10% (for GAPDH, PGFS, LTA4H, EDN-1/2/3) and 15% (for LTC4S and mPGES-1) SDS-PAGE. Separated proteins were electroblotted onto 0.2 μm nitrocellulose membrane in transfer buffer (20 mM Tris-HCl buffer, pH 8.2; 150 mM glycine, 20% methanol). After blocking in 5% non-fat dry milk in TBS-T buffer (Tris-buffered saline, containing 0.1% Tween-20) for 1.5 h at 25.6°C, the membranes were incubated overnight with 1:2000 anti-lung-type PGFS antiserum antiserum [32], or 1:1000 polyclonal anti-mPGES-1 antibodies (160140; Cayman) or 1:1000 polyclonal LTC4 synthase antibody (sc-20108; Santa Cruz Biotechnology, Inc, Heidelberg, Germany) or 1:1000 LTA4 hydrolase antibody (ab23677; Abcam, Cambridge, UK) or 1:200 EDN 1/2/3 (sc-98727; Santa Cruz Biotechnology, ET-1/2/3 = EDN-1, EDN-2, EDN-3) at 4°C. Protein expression of GAPDH (G 8795; Sigma, glyceraldehyde-3-phosphate dehydrogenase) as a reference was used. Subsequently, proteins were detected by incubating the membrane with 1:20000 dilution of secondary polyclonal anti-rabbit or anti-goat alkaline phosphatase-conjugated antibodies (A 3687; A 3562 Sigma) for 1.5 hour at 25.6°C. Immune complexes were visualized using standard alkaline phosphatase visualization procedure. Western blots were quantitated using Kodak 1D program (Eastman Kodak, Rochester, NY, USA).", "The statistical significance of differences between the control and treated groups were analyzed by one-way ANOVA followed by t-student's post-hoc test (ANOVA; GraphPAD PRISM Version 5.00, San Diego, CA; USA), if the initial ANOVA was significant (P < 0.05).", "[SUBTITLE] Experiment 1. Establishment of immortalized bovine endothelial cell (EnCL-1) line and its phenotype characterization [SUBSECTION] Selected immortalized endothelial cell line (EnCL-1) was cultured till 50 passages without any sign of senescence, which allowed to receive clear immortalized line of cells with homogenous morphology and genotype, although from 10 passage the line of cells has no any percentage of primary luteal endothelial cells. Phase contrast microscopy revealed that immortalized EnCL-1 cells grew as confluent monolayers with typical cobblestone morphology of primary endothelial cells. These cells (13th passage) were homogenous, polygonal and had characteristic ovoid nuclei (Figure 1a,b). Moreover, immunofluorescence staining revealed the presence of endothelial cell markers: von Willebrand factor (Figure 1C,E,F) and VE-cadherin (Figure 1H,I) in EnCL-1 cells. All isolated colonies expressed transfected vector. Expression of SV40 T-ag gene in the cells was confirmed by RT-PCR (Figure 2).\nCharacterization of morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in EnCL-1 cells. Cells reveal cobblestone morphology and tendency to form monolayer on Petri dish (A-B, mag. x10, x40). Localization of cytoplasmatic protein - vWF (panel E, mag. x10, x40) and expression of VE-cadherin (panel H, mag. x10, x40). Cells (13th passage) were stained with specific antibodies and detected by secondary antibodies labeled with fluorescent dyes, and analyzed by confocal microscopy. The DNA was counterstained with DAPI (blue, panels D, G). Panels C, F and I merged pictures of two fluorescent dyes. Panel C mag. x100.\nExpression of SV40 T-ag mRNA in EnCL-1 cells by RT-PCR. Left - SV40 Tag positive control, Right-EnCL-1 cells.\nSelected immortalized endothelial cell line (EnCL-1) was cultured till 50 passages without any sign of senescence, which allowed to receive clear immortalized line of cells with homogenous morphology and genotype, although from 10 passage the line of cells has no any percentage of primary luteal endothelial cells. Phase contrast microscopy revealed that immortalized EnCL-1 cells grew as confluent monolayers with typical cobblestone morphology of primary endothelial cells. These cells (13th passage) were homogenous, polygonal and had characteristic ovoid nuclei (Figure 1a,b). Moreover, immunofluorescence staining revealed the presence of endothelial cell markers: von Willebrand factor (Figure 1C,E,F) and VE-cadherin (Figure 1H,I) in EnCL-1 cells. All isolated colonies expressed transfected vector. Expression of SV40 T-ag gene in the cells was confirmed by RT-PCR (Figure 2).\nCharacterization of morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in EnCL-1 cells. Cells reveal cobblestone morphology and tendency to form monolayer on Petri dish (A-B, mag. x10, x40). Localization of cytoplasmatic protein - vWF (panel E, mag. x10, x40) and expression of VE-cadherin (panel H, mag. x10, x40). Cells (13th passage) were stained with specific antibodies and detected by secondary antibodies labeled with fluorescent dyes, and analyzed by confocal microscopy. The DNA was counterstained with DAPI (blue, panels D, G). Panels C, F and I merged pictures of two fluorescent dyes. Panel C mag. x100.\nExpression of SV40 T-ag mRNA in EnCL-1 cells by RT-PCR. Left - SV40 Tag positive control, Right-EnCL-1 cells.\n[SUBTITLE] Experiment 2. Effect of cytokines on production and content of Arachidonic Acid metabolites in immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] [SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] TNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nTNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] TNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\nTNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] Cytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).\nCytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).\n[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] TNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nTNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] TNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\nTNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] Cytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).\nCytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).", "Selected immortalized endothelial cell line (EnCL-1) was cultured till 50 passages without any sign of senescence, which allowed to receive clear immortalized line of cells with homogenous morphology and genotype, although from 10 passage the line of cells has no any percentage of primary luteal endothelial cells. Phase contrast microscopy revealed that immortalized EnCL-1 cells grew as confluent monolayers with typical cobblestone morphology of primary endothelial cells. These cells (13th passage) were homogenous, polygonal and had characteristic ovoid nuclei (Figure 1a,b). Moreover, immunofluorescence staining revealed the presence of endothelial cell markers: von Willebrand factor (Figure 1C,E,F) and VE-cadherin (Figure 1H,I) in EnCL-1 cells. All isolated colonies expressed transfected vector. Expression of SV40 T-ag gene in the cells was confirmed by RT-PCR (Figure 2).\nCharacterization of morphology and cytoplasmatic protein - von Villenbrand factor and cell to cell adhesion - VE-cadherin in EnCL-1 cells. Cells reveal cobblestone morphology and tendency to form monolayer on Petri dish (A-B, mag. x10, x40). Localization of cytoplasmatic protein - vWF (panel E, mag. x10, x40) and expression of VE-cadherin (panel H, mag. x10, x40). Cells (13th passage) were stained with specific antibodies and detected by secondary antibodies labeled with fluorescent dyes, and analyzed by confocal microscopy. The DNA was counterstained with DAPI (blue, panels D, G). Panels C, F and I merged pictures of two fluorescent dyes. Panel C mag. x100.\nExpression of SV40 T-ag mRNA in EnCL-1 cells by RT-PCR. Left - SV40 Tag positive control, Right-EnCL-1 cells.", "[SUBTITLE] Experiment 2.1. Effect of TNFα and ifNγ on the viability of immortalized bovine luteal endothelial cells (EnCL-1) [SUBSECTION] TNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nTNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\n[SUBTITLE] Experiment 2.2. Effect of TNFα and ifNγ on mRNA and protein expression of LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and endothelin-1 in bovine endothelial immortalized cells (EnCL-1) [SUBSECTION] TNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\nTNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.\n[SUBTITLE] Experiment 2.3. Effect of TNFα and ifNγ on prostaglandins, leukotrienes and endothelin-1 release by EnCL-1 cells [SUBSECTION] Cytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).\nCytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).", "TNFα/IFNγ did not influence the viability of EnCL-1 cells after 24 h of incubation comparing to non-treated cells (P > 0.05; Figure 3).\nViability of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells. No statistical differences were observed between groups of non-treated and cytokine treated cells (P > 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.", "TNFα/IFNγ treatment of EnCL-1 cells resulted in increased mRNA expression of PGES, PGFS, LTA4H, LTC4S and EDN-1 in comparison to untreated cells (P < 0.05; Figure 4A).\nmRNA and protein expression for selected aa metabolites in non- and cytokine treated EnCL-1 cells. LTC4 synthase, LTA4 hydrolase, PGE2 synthase, PGF2α synthase and EDN-1 (A) mRNA, (B) protein expression of non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells and (C) representive immunoblots of studies factors: 1 lane: Positive Control-bovine lung tissue, 2 lane-Control, 3 lane-TNFα/IFNγ. Arrows indicate statistical differences in the respective mRNA quantitative expression between groups of non-treated and cytokine treated cells (P < 0.05), as determined by one-way ANOVA followed by t-student's multiple comparison test.\nPGES and LTA4H protein expression were not affected by cytokine treatment (P > 0.05), whereas PGFS, LTC4S and EDN-1/2/3 protein expression were stimulated by cytokines (P < 0.05; Figure 4B). Representive immunoblots of studies factors are presented in Figure 4C.", "Cytokine treatment did not change the levels of PGE2 and LTB4 in the medium (P > 0.05), whereas cytokines stimulated PGF2α, LTC4 and EDN-1 release by EnCL-1 cells (P < 0.05; Table 2).\nConcentration of PGE2, PGF2α;, LTB4, LTC4 and EDN-1 in non-treated and cytokine (TNFα/IFNγ) stimulated EnCL-1 cells (Mean ± SEM)\nSignificant differences (*** - P < 0.001), as determined by one-way ANOVA followed by t-student's multiple comparison test (GraphPad PRISM).", "The presence of SV40 T-ag in EnCL-1 cells and repeated passage without the apparent senescence confirmed the permanent status of the selected cell line. VE-cadherin and von Willebrand factor - positive features support the endothelial phenotype of the cell line. T antigen present in SV40 virus blocks cell death by apoptosis and it also interacts with a cytoplasmic protein that contains BH3 domain [33]. The viral genes achieve immortalization by inactivating the tumor suppressor genes (p53, Rb) that induce a replicative senescent state in cells. SV40 T antigen also induces telomerase activity in the infected cells [33]. So far, each study was conducted on the primary luteal endothelial cells or line of endothelial cells but not received from the bovine CL, which could possess different surface antigens and at least physiology [10]. We established the line of endothelial cells from bovine CL. To our knowledge, this is the first report showing the morphological and physiological properties of immortalized endothelial cells collected from bovine CL. According to Davis et al. [4], there are five types of bovine luteal endothelial cells differ the presence of cytokeratin, expression of surface antigens and neuronal cell adhesion molecule, capable for the contact between steroidogenic and endothelial cells within CL. We did not determine the morphological type and surface antigens of received line of endothelial cells than further study are necessary.\nEndothelial cells posses receptors for TNFα [20,34,35] and IFNγ [36]. Several papers confirmed synergistic, antiproliferative and proapoptotic action of TNFα and IFNγ in the CL [4,10,19,21,22]. In this study, TNFα and IFNγ treatment of cells increased each of studied mRNA expressions. In addition, PGF2α secretion and its synthase protein expression were stimulated. Similar effect received Acosta et al. [8], TNFα elevated PGF2α content in fresh and unfrozen cells till 10th passage in primary luteal endothelial cells. Whereas, in the study of Cavicchio et al. [10], cytokines stimulation was unresponsive to PGF2α secretion in the luteal endothelial cells (passage 2). The role of cytokines in regression of CL and cytokine effect on the main luteolytic factor-PGF2α was considered, as enhancing PGF2α action in the functional and structural luteolysis [17,19,20,34]. We also received the stimulation of mRNA expression for PGES without the effect on its protein expression and the level of PGE2 after cytokines treatment. Such an effect may be the consequence of changes in intracellular regulation of EnCL-1 cells, especially in mitochondrial activity. PGE2 enhances cellular proliferation, promotes angiogenesis, inhibits apoptosis and suppresses immune responses in cancerogenesis [35]. EnCL-1 cells potentially are programmed genetically for proliferation, thus cytokines, typically causing apoptosis, may not cause such an effect in our study and simultaneously stimulate PGE2 mRNA expression as kind of the preparation for proliferative functions.\nFurthermore, we considered the role of cytokines in connection with another aa metabolites-LTs in EnCL-1 cells. mRNA expression, as well as protein expression for LTCS and LTC4 release were stimulated by cytokines in EnCL-1 cells. Cytokines also stimulated LTA4H mRNA expression but did not change LTA4H protein expression and LTB4 secretion in EnCL-1 cells. Recently we showed that primary bovine luteal endothelial cells show mRNA expression for LT receptors (for LTB4 and C4) [26]. So far LTs role in CL function focus on steroidogenic cells in vitro [30] or concerned with processes in bovine reproductive tract in vivo [37,38]. LTB4 plays luteotropic role in bovine CL, stimulating PGE2 secretion, whereas LTC4 stimulate PGF2α and thus acts as luteolytic factor in vivo [38]. There is lack of knowledge about LTs role in CL vascular processes like angiogenesis and angioregression. It is possible that cytokine effect in the ovary is modulated by LTs. The immune cells (macrophages, monocytes and leukocytes) infiltrate the ovary and secrete cytokines in process of ovulation [39]. Cytokines affect non-steroidogenic ovarian cells, causing the release of ovulation mediators, such as metabolites of aa: PGs and LTs [39]. Thus, cytokines are involved in ovarian processes during the estrous cycle such as differentiation of CL, ovulation, luteolysis and cooperate with PGs and LTs [26,30,37-39]. Beside, both PGs and LTs appear to act in parallel in the regulation of cell proliferation and neoangiogenesis [40].\nWe selected EDN-1 as one of the main factors produced in endothelial cells and checked the effect of cytokines action on edn-1 mRNA expression and EDN-1 release in EnCL-1 cells. Our results confirmed the earlier studies [8,20,41] because the cytokines stimulated both mRNA and its production in EnCL-1 cells. Protein expression of EDN-1/2/3 was also elevated in our study as summary of the expression for EDN-1, EDN-2 and EDN-3. EDN-1 mRNA and protein is expressed in luteal endothelial cells during all the estrous cycle and EDN-1 inhibits P4 production in late luteal phase [42,43], EDN-2, in the early CL [44], whereas EDN-3, on the contrary to mentioned EDN-1 and EDN-2, does not affect luteal steroidogenesis [45]. Our results indicate that cytokines enhance EDN-1/2/3 action and indicate on multiple characteristic functions of EnCL-1 cells.\nConcluding, cytokines modulate EnCL-1 cells function by up-regulation of PGES, PGFS, LTA4H, LTC4S and mRNA expression. Protein expression was elevated by cytokines for PGFS and LTCS and simultaneously the level of appropriate active metabolites of these synthases products (PGF2α and LTC4) were higher after cytokine stimulation. Protein expression for PGES and LTA4H was not changed and release of products of these syntases-PGE2 and LTB4 was also stable. Beside, mRNA expression, level of EDN-1 and protein expression for EDN1/2/3 were upregulated by cytokines, which suggest that EnCL-1 cells show multiple potency, both prolifereative and proapoptotic. In this respect, EnCL-1 cells may serve as an appropriate model for investigating the paradigm of counteraction of the luteolytic and luteotropic properties of bovine CL. The received line of immortalized EnCL-1 cells possess stable genotype and phenotype. However, an intricate molecular structure of the cells with multiple intervenient factors/hormones and actions deserve to be defined.", "The authors declare that they have no competing interests.", "AJK collected the experimental material, isolated primary endothelial CL cells, carried out the immunoassays, mRNA expression and MTT methods and performed the statistical analysis, designed, drafted and coordinated the study. GB carried out the immortalization of the primary endothelial cells, participated in the design of the study and draft the manuscript. JB assisted and helped technically in all steps of experimental procedures. AB and DJS helped to design and draft the manuscript. All authors read and approved the final manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Case-Control Studies", "Endometriosis", "Female", "Gene Frequency", "Humans", "Infertility", "Mucin-4", "Polymerase Chain Reaction", "Polymorphism, Single Nucleotide", "Taiwan" ]

[ "Background", "Study population", "Clinical stages and association study", "Genomic DNA extraction and genotyping of SNPs in MUC4", "Statistical analysis", "Functional analyses and secondary structure predictions of MUC4 protein", "Results", "MUC4 gene polymorphisms and endometriosis", "Association of MUC4 gene polymorphisms and stages", "MUC4 gene polymorphisms and infertility", "MUC4 gene polymorphisms and amino acid substitutions", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Endometriosis is a common chronic gynecological disease characterized by the presence of endometrial gland and stroma outside the uterine cavity, affecting approximately 10% of reproductive age women [1,2]. The common clinical symptoms include pelvic pain, heavy menstrual bleeding, pelvic adhesion, bloating and fatigue. Notably, the prevalence of endometriosis is 0.5% to 5% in fertile women and 25% to 40% in infertile women [3], suggesting infertility as one possible consequence of endometriosis. To date, the implantation theory is widely accepted, stating that endometrial tissues pass through the fallopian tube, then attach and grow on pelvic tissue. However, this hypothesis cannot explain the existence of endometriosis outside the pelvis or how endometriosis progresses and invades other tissues. Additional factors such as genetic or immune differences have been suggested as possible contributors to trigger the formation of endometriosis [4-6]. Family history and genome-wide linkage studies also support genetic predisposition during the development of endometriosis [7-10]. These studies provide molecular evidence demonstrating endometriosis as a genetic disease, and it is desirable to explore more genetic variations associated with endometriosis.\nSimilarly to malignant diseases, extensive growth of endometrial cells on the peritoneal surface and invasion of the pelvic organ are very common during the development of endometriosis. This process is frequently associated with several mechanisms involved in angiogenesis and cellular adhesion. In fact, women who have endometriosis appear to be more at risk of developing several different kinds of ovarian cancers [8,11-13]. An epidemiological study showed that the prevalence rates of endometriosis in patients with endometrioid and clear cell ovarian carcinoma are 19 and 35.9, respectively [14]. These findings suggest that endometriosis and certain types of ovarian cancer may share several common genetic alterations during pathogenesis. Genes that regulate cell mobility and invasion in ovarian cancers are therefore possible candidates for playing roles in endometriosis.\nMucins are high-molecular-weight glycoproteins with the function of protecting and lubricating the epithelial surface of respiratory, gastrointestinal and reproductive tracts [15]. Among the mucin proteins, mucin 4 (MUC4) and mucin 1 (MUC1) are the major ones expressed in the endometrial epithelium [16,17]. In cancer studies, these two mucins have been shown to be aberrantly expressed in various malignancies and have been validated as novel targets for cancer diagnosis and therapy [18-20]. Distinct from MUC1, the extracellular domain of MUC4 can interact with human epidermal growth factor receptor 2 (HER2) on the cell surface and modulate downstream cell growth signaling by stabilizing and/or enhancing the activity of cell growth receptor complexes [18,21,22]. Consequently, changes of cytoarchitectures and cellular signaling may lead to the increase of cell mobility and tumor cell invasion.\nThe above findings provide us with clues to hypothesize that genetic variations in the extracellular domain of MUC4, especially those resulting in amino acid substitutions, may play roles involved in the development of endometriosis. With endometriosis as a possible cause of infertility in women, we also would like to study the association of MUC4 single-nucleotide polymorphisms (SNPs) with susceptibility to endometriosis-related infertility.", "A total of 140 individuals who underwent surgery for benign diseases and pathology-proven endometriosis were identified at China Medical University Hospital from 1998 to 2008 and were enrolled in this study. In general, these patients were diagnosed with ovarian cysts on the basis of sonography and had several clinical symptoms related to endometriosis, including dysmenorrhea, lower abdominal pain, infertility or abnormal menstruation. Study patients who failed to have pathology-proven endometriosis were excluded from this study. For the control group, blood samples of 150 healthy women were selected from a pool of individuals who received regular heath checkups at the same hospital and were identified as normal on the basis of the examinations conducted. A total of 142 controls were frequency-matched on the basis of age profile with the study patients (Additional file 1, Table S1). Controls who showed one of the endometriosis-associated symptoms, even though the results of their health checkups were normal, were excluded from this study. This study was approved by the institutional review board at China Medical University, with informed consent obtained from each patient.", "Clinical information on patients was collected from clinical notes, including clinical stage, lesion size, location, drug treatment and fertility (Additional file 1, Table S1). The definition of endometriosis staging was based on the classification of the American Society for Reproductive Medicine: Stage 1, minimal; Stage 2, mild; Stage 3, moderate; and Stage 4, severe [23].", "Genomic DNA was extracted from peripheral blood leukocytes according to standard protocols (Genomic DNA kit; Qiagen, Valencia, CA, USA). DNA fragments containing the target SNP sites were amplified by polymerase chain reaction (PCR) assay using the TaqMan SNP Genotyping Assay System from Applied Biosystems, Inc. (Carlsbad, CA, USA). The probe search and design are available on the Applied Biosystems, Inc. website [24].\nAdditional file 1, Table S2, lists probe identifications for the six SNPs tested. PCR amplification conditions consisted of initial denaturation at 95°C for 5 minutes followed by 40 cycles at 95°C for 10 seconds, 56°C for 10 seconds and 72°C for 20 seconds, with one additional cycle at 72°C for 5 minutes. Genetic variations were detected by reading the fluorescence signals of PCR products. A positive signal indicates a perfect match between the probe and the tested DNA, thus identifying the allele types. Ten percent of study participants were randomly chosen and genotyped in duplicate to confirm the concordance of the genotyping results. In our study, the call rates for these SNP probes were above 94% (Additional file 1, Table S3).", "The allelic frequency and genotype frequency distributions for the six polymorphisms of patients with endometriosis and the controls were performed by χ2 analysis using SPSS software (version 10.0; SPSS Inc., Chicago, IL, USA). An unordered, 2 df two-sided test was used for the statistical analyses of our genotyping results. P < 0.05 was considered statistically significant. Allelic and genotypic frequencies are expressed as percentages of total alleles and genotypes. Odds ratios (ORs) were calculated from allelic and genotype frequencies with 95% confident intervals (95% CIs). The major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. Adherence to the Hardy-Weinberg equilibrium (HWE) constant was tested using a χ2 test with 1 df. To study the association of the six SNPs with clinical stages and reproductive ability, Fisher's exact tests instead of χ2 tests were used because of the small number of participants tested.\nThe haplotypes of each individual were determined using the Bayesian statistical method available in the free download software program PHASE 2.1 [25]. This approach incorporates a priori expectations of haplotypic structure based on population genetics and coalescence theory. Lewontin's D' (|D'|) and the linkage disequilibrium (LD) coefficient r2 were determined between selected pairs of biallelic loci [26]. Haploview version 3.2 software (Whitehead Institute for Biomedical Research, Cambridge, MA, USA) was used to examine the structure of the LD block [27]. This program uses two-marker expectation maximization to estimate the maximum likelihood values of the four gamete frequencies from which the D' and log of odds (LOD) values are derived. The genetic effects of the inferred haplotypes were analyzed in the same way as the analysis of polymorphisms. The reported haplotype percentages are estimated on the basis of allele frequencies and LD. The P values are based on a comparison of a given haplotype with all other haplotypes combined.", "Functional characterization and annotation of MUC4 were performed by aligning the sequence with functional motifs and/or signatures in the PROSITE protein domain database [28]. To predict the secondary structure of the MUC4 sequence, the Chou and Fasman method was used [29]. An improved method was applied to increase the accuracy of the predictions by locating nucleation regions with refined wavelet transformation technology and by calculating propensity factors with larger data sets [30]. The program gives the propensity of each residue to be a part of an α-helix, a β-strand or a loop. We considered propensities Pα > 1.03 as significant for helices and propensities Pβ > 1.05 as significant for strands. Predicted regions with fewer than four contiguous residues were not considered secondary structure units. For a region with both helix and strand tendencies, the secondary structure conformed with higher propensity: Pα >Pβ or Pβ >Pα is predicted. To plot hydrophobicity and surface probability, the Kyte and Doolittle method [31] and Emini et al. surface accessibility prediction (SAP) [32] were used, respectively. We slid a window along the MUC4 sequence to assign a \"hydrophobicity\" or \"surface probability\" value to each amino acid. The values were summed in the window, and the results were plotted.", "[SUBTITLE] MUC4 gene polymorphisms and endometriosis [SUBSECTION] Six SNPs in the extracellular domain of the MUC4 gene with a frequency greater than 20% in Chinese Han Beijing were selected from International HapMap Project databank [33] (Additional file 1, Table S2). Genotypic analyses (Figure 1) indicated rs882605 as a unique SNP with a higher frequency of the TG genotype in patients than in controls (P = 0.04, OR = 1.97, 95% CI, 1.17 to 3.32) (Table 1), while allele type analyses of these SNPs showed no statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. To confirm the genetic impact of SNPs on endometriosis, the top two high-risk alleles at rs882605 and rs1104760 were selected for haplotype analyses. Significantly, the frequency of haplotype T-T was found to be higher in patients than in controls (P = 0.0353) (Table 2) (Additional file 2, Figure S1), suggesting the association of MUC4 SNPs with endometriosis development. The genotyping results were confirmed in duplicate, and the concordance of duplicates was 97.6%.\nAllelic discrimination plots of the six tested single-nucleotide polymorphisms (SNPs) in the mucin 4 (MUC4) gene. The DNA samples from patients and controls were genotyped by using the TaqMan SNP Genotyping Assay System. The major (also the wild-type) alleles were detected by 6-carboxyfluorescein (FAM)-labeled probes (blue), and the minor alleles were detected by 2'-chloro-7'-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC)-labeled probes. The genotyping results of the six SNPs in the MUC4 gene are presented as allelic discrimination plots. Of note, the intensity of FAM signals tended to be similar among samples in our assays, thus the dots for a wild-type genotype overlapped each other. \"X\" indicates the participant who failed to be genotyped.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in Taiwanese endometriosis patients and controlsa\naMUC4, mucin 4 gene; SNP, single-nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval; HWE, P values of deviation from the Hardy-Weinberg equilibrium constant. Allelic frequencies were determined by χ2 test using 2 × 2 contingency tables. Genotype frequencies were determined by χ2 test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients and controlsa\naMUC4, mucin 4 gene. The reported haplotype percents are estimated percents based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.\nSix SNPs in the extracellular domain of the MUC4 gene with a frequency greater than 20% in Chinese Han Beijing were selected from International HapMap Project databank [33] (Additional file 1, Table S2). Genotypic analyses (Figure 1) indicated rs882605 as a unique SNP with a higher frequency of the TG genotype in patients than in controls (P = 0.04, OR = 1.97, 95% CI, 1.17 to 3.32) (Table 1), while allele type analyses of these SNPs showed no statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. To confirm the genetic impact of SNPs on endometriosis, the top two high-risk alleles at rs882605 and rs1104760 were selected for haplotype analyses. Significantly, the frequency of haplotype T-T was found to be higher in patients than in controls (P = 0.0353) (Table 2) (Additional file 2, Figure S1), suggesting the association of MUC4 SNPs with endometriosis development. The genotyping results were confirmed in duplicate, and the concordance of duplicates was 97.6%.\nAllelic discrimination plots of the six tested single-nucleotide polymorphisms (SNPs) in the mucin 4 (MUC4) gene. The DNA samples from patients and controls were genotyped by using the TaqMan SNP Genotyping Assay System. The major (also the wild-type) alleles were detected by 6-carboxyfluorescein (FAM)-labeled probes (blue), and the minor alleles were detected by 2'-chloro-7'-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC)-labeled probes. The genotyping results of the six SNPs in the MUC4 gene are presented as allelic discrimination plots. Of note, the intensity of FAM signals tended to be similar among samples in our assays, thus the dots for a wild-type genotype overlapped each other. \"X\" indicates the participant who failed to be genotyped.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in Taiwanese endometriosis patients and controlsa\naMUC4, mucin 4 gene; SNP, single-nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval; HWE, P values of deviation from the Hardy-Weinberg equilibrium constant. Allelic frequencies were determined by χ2 test using 2 × 2 contingency tables. Genotype frequencies were determined by χ2 test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients and controlsa\naMUC4, mucin 4 gene. The reported haplotype percents are estimated percents based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.\n[SUBTITLE] Association of MUC4 gene polymorphisms and stages [SUBSECTION] We next asked whether MUC4 genetic variations could possibly associate with clinical stages. Patients were divided into two groups: the mild stage group with patients at stages 1 or 2 and the advanced group with patients at stages 3 or 4. Strikingly, genotype analyses revealed strong associations of CC type at rs2688513 (P = 0.04) and GG type at rs2246901 (P = 0.03), with more advanced endometriosis at stage 3 or 4 (Table 3). Dominant effects were found for other genetic variations at rs1104760 (CC + CT versus TT) and rs2258447 (AA + AG versus GG) during endometriosis progression. Allele type analyses suggested C allele at rs1104760, C allele at rs2688513, G allele at rs2246901 and A allele at rs2258447 as risk factors that correlated with more severe endometriosis.\nGenotype and allele distributions of SNPs in the MUC4 gene in endometriosis patients at different clinical stagesa\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bMild stage, patients at clinical stage1 or stage 2; cSevere stage, patients at clinical stage 3 or stage 4; dP < 0.05 was considered statistically significant.\nWe next asked whether MUC4 genetic variations could possibly associate with clinical stages. Patients were divided into two groups: the mild stage group with patients at stages 1 or 2 and the advanced group with patients at stages 3 or 4. Strikingly, genotype analyses revealed strong associations of CC type at rs2688513 (P = 0.04) and GG type at rs2246901 (P = 0.03), with more advanced endometriosis at stage 3 or 4 (Table 3). Dominant effects were found for other genetic variations at rs1104760 (CC + CT versus TT) and rs2258447 (AA + AG versus GG) during endometriosis progression. Allele type analyses suggested C allele at rs1104760, C allele at rs2688513, G allele at rs2246901 and A allele at rs2258447 as risk factors that correlated with more severe endometriosis.\nGenotype and allele distributions of SNPs in the MUC4 gene in endometriosis patients at different clinical stagesa\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bMild stage, patients at clinical stage1 or stage 2; cSevere stage, patients at clinical stage 3 or stage 4; dP < 0.05 was considered statistically significant.\n[SUBTITLE] MUC4 gene polymorphisms and infertility [SUBSECTION] Since endometriosis has been suspected as one potent factor leading to infertility in women [3], we also studied the possible linkage between MUC4 SNPs and infertility. Though no significant difference was found in our genotype association study, our data indicated T allele at rs882605 as a protective factor that associated with reduced frequency of infertility in patients with endometriosis (Table 4). Two other alleles, C at rs2688513 and G at rs2246901, showed similar protective effects, but the data did not reach statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. In our haplotype analyses, we thus sought to ascertain the impact of the genetic combination of these top three protective alleles. Table 5 indicates that patients with haplotype T-C-G did show a lower frequency of infertility, although the results did not show a statistically significant difference (P = 0.099). By contrast, haplotype G-T-T showed a strong association with infertility in patients (P = 0.012) (Table 5) (Additional file 2, Figure S2) and could be used as a risk indicator for patients at higher risk of developing severe complications such as infertility.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in endometriosis patients with different reproductive abilitya\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene; OR, odds ratio; 95% CI, 95% confidence interval. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients with infertilitya\naMUC4, mucin 4 gene. The reported haplotype percentages are estimated percentages based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.\nSince endometriosis has been suspected as one potent factor leading to infertility in women [3], we also studied the possible linkage between MUC4 SNPs and infertility. Though no significant difference was found in our genotype association study, our data indicated T allele at rs882605 as a protective factor that associated with reduced frequency of infertility in patients with endometriosis (Table 4). Two other alleles, C at rs2688513 and G at rs2246901, showed similar protective effects, but the data did not reach statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. In our haplotype analyses, we thus sought to ascertain the impact of the genetic combination of these top three protective alleles. Table 5 indicates that patients with haplotype T-C-G did show a lower frequency of infertility, although the results did not show a statistically significant difference (P = 0.099). By contrast, haplotype G-T-T showed a strong association with infertility in patients (P = 0.012) (Table 5) (Additional file 2, Figure S2) and could be used as a risk indicator for patients at higher risk of developing severe complications such as infertility.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in endometriosis patients with different reproductive abilitya\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene; OR, odds ratio; 95% CI, 95% confidence interval. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients with infertilitya\naMUC4, mucin 4 gene. The reported haplotype percentages are estimated percentages based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.\n[SUBTITLE] MUC4 gene polymorphisms and amino acid substitutions [SUBSECTION] Because these endometriosis-associated SNPs can cause amino acid substitutions (Additional file 1, Table S2), the biofunctions of MUC4 might be altered by changes in hydrophilicity and protein folding. Figure 2 illustrates the functional domains in MUC4 protein sequence and secondary structures that contain these SNPs. Our data show that genetic variations of rs882605 and rs2688513 cause amino acid substitutions in long-loop regions (> 10 residues) between secondary structure units (α-helices or β-strands) with high hydrophilicity and moderate surface probability (Figure 2). Amino acid composition analyses also revealed that these two loops (300-319 and 4,134-4,158) contain several negatively charged aspartate (D) residues and reverse turn elements, glycine (G) and proline (P), suggesting the importance of these loops in protein folding and functional regulation [34,35]. In addition, rs2246901 locates in a type D von Willebrand factor (VWFD) domain responsible for protein-protein interaction and cell adhesion and/or migration [36,37]. Our findings support the functional roles of MUC4 SNPs in regulating cellular mobility and invasion of endometrial cells during endometriosis development and progression.\nFunctional domains in the MUC4 protein sequence and the predicted secondary structures. Six functional domains and/or signatures (boxes) were annotated by aligning MUC4 protein sequence in PROSITE protein domain database [28]. The boundaries of each signature are listed. Among six SNPs tested in this study (stars), rs882605 and rs2688513 (black) were found in two different long-loop regions, 300-319 and 4,134-4,158, respectively (bold letters refer to amino acid substitution sites). The SNP rs2246901 (gray) was found in a type D von Willebrand factor (VWFD) domain. The MUC4 reference sequence can be located at National Center for Biotechnology Information databank NP_060876.4.\nBecause these endometriosis-associated SNPs can cause amino acid substitutions (Additional file 1, Table S2), the biofunctions of MUC4 might be altered by changes in hydrophilicity and protein folding. Figure 2 illustrates the functional domains in MUC4 protein sequence and secondary structures that contain these SNPs. Our data show that genetic variations of rs882605 and rs2688513 cause amino acid substitutions in long-loop regions (> 10 residues) between secondary structure units (α-helices or β-strands) with high hydrophilicity and moderate surface probability (Figure 2). Amino acid composition analyses also revealed that these two loops (300-319 and 4,134-4,158) contain several negatively charged aspartate (D) residues and reverse turn elements, glycine (G) and proline (P), suggesting the importance of these loops in protein folding and functional regulation [34,35]. In addition, rs2246901 locates in a type D von Willebrand factor (VWFD) domain responsible for protein-protein interaction and cell adhesion and/or migration [36,37]. Our findings support the functional roles of MUC4 SNPs in regulating cellular mobility and invasion of endometrial cells during endometriosis development and progression.\nFunctional domains in the MUC4 protein sequence and the predicted secondary structures. Six functional domains and/or signatures (boxes) were annotated by aligning MUC4 protein sequence in PROSITE protein domain database [28]. The boundaries of each signature are listed. Among six SNPs tested in this study (stars), rs882605 and rs2688513 (black) were found in two different long-loop regions, 300-319 and 4,134-4,158, respectively (bold letters refer to amino acid substitution sites). The SNP rs2246901 (gray) was found in a type D von Willebrand factor (VWFD) domain. The MUC4 reference sequence can be located at National Center for Biotechnology Information databank NP_060876.4.", "Six SNPs in the extracellular domain of the MUC4 gene with a frequency greater than 20% in Chinese Han Beijing were selected from International HapMap Project databank [33] (Additional file 1, Table S2). Genotypic analyses (Figure 1) indicated rs882605 as a unique SNP with a higher frequency of the TG genotype in patients than in controls (P = 0.04, OR = 1.97, 95% CI, 1.17 to 3.32) (Table 1), while allele type analyses of these SNPs showed no statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. To confirm the genetic impact of SNPs on endometriosis, the top two high-risk alleles at rs882605 and rs1104760 were selected for haplotype analyses. Significantly, the frequency of haplotype T-T was found to be higher in patients than in controls (P = 0.0353) (Table 2) (Additional file 2, Figure S1), suggesting the association of MUC4 SNPs with endometriosis development. The genotyping results were confirmed in duplicate, and the concordance of duplicates was 97.6%.\nAllelic discrimination plots of the six tested single-nucleotide polymorphisms (SNPs) in the mucin 4 (MUC4) gene. The DNA samples from patients and controls were genotyped by using the TaqMan SNP Genotyping Assay System. The major (also the wild-type) alleles were detected by 6-carboxyfluorescein (FAM)-labeled probes (blue), and the minor alleles were detected by 2'-chloro-7'-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC)-labeled probes. The genotyping results of the six SNPs in the MUC4 gene are presented as allelic discrimination plots. Of note, the intensity of FAM signals tended to be similar among samples in our assays, thus the dots for a wild-type genotype overlapped each other. \"X\" indicates the participant who failed to be genotyped.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in Taiwanese endometriosis patients and controlsa\naMUC4, mucin 4 gene; SNP, single-nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval; HWE, P values of deviation from the Hardy-Weinberg equilibrium constant. Allelic frequencies were determined by χ2 test using 2 × 2 contingency tables. Genotype frequencies were determined by χ2 test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients and controlsa\naMUC4, mucin 4 gene. The reported haplotype percents are estimated percents based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.", "We next asked whether MUC4 genetic variations could possibly associate with clinical stages. Patients were divided into two groups: the mild stage group with patients at stages 1 or 2 and the advanced group with patients at stages 3 or 4. Strikingly, genotype analyses revealed strong associations of CC type at rs2688513 (P = 0.04) and GG type at rs2246901 (P = 0.03), with more advanced endometriosis at stage 3 or 4 (Table 3). Dominant effects were found for other genetic variations at rs1104760 (CC + CT versus TT) and rs2258447 (AA + AG versus GG) during endometriosis progression. Allele type analyses suggested C allele at rs1104760, C allele at rs2688513, G allele at rs2246901 and A allele at rs2258447 as risk factors that correlated with more severe endometriosis.\nGenotype and allele distributions of SNPs in the MUC4 gene in endometriosis patients at different clinical stagesa\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bMild stage, patients at clinical stage1 or stage 2; cSevere stage, patients at clinical stage 3 or stage 4; dP < 0.05 was considered statistically significant.", "Since endometriosis has been suspected as one potent factor leading to infertility in women [3], we also studied the possible linkage between MUC4 SNPs and infertility. Though no significant difference was found in our genotype association study, our data indicated T allele at rs882605 as a protective factor that associated with reduced frequency of infertility in patients with endometriosis (Table 4). Two other alleles, C at rs2688513 and G at rs2246901, showed similar protective effects, but the data did not reach statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. In our haplotype analyses, we thus sought to ascertain the impact of the genetic combination of these top three protective alleles. Table 5 indicates that patients with haplotype T-C-G did show a lower frequency of infertility, although the results did not show a statistically significant difference (P = 0.099). By contrast, haplotype G-T-T showed a strong association with infertility in patients (P = 0.012) (Table 5) (Additional file 2, Figure S2) and could be used as a risk indicator for patients at higher risk of developing severe complications such as infertility.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in endometriosis patients with different reproductive abilitya\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene; OR, odds ratio; 95% CI, 95% confidence interval. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients with infertilitya\naMUC4, mucin 4 gene. The reported haplotype percentages are estimated percentages based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.", "Because these endometriosis-associated SNPs can cause amino acid substitutions (Additional file 1, Table S2), the biofunctions of MUC4 might be altered by changes in hydrophilicity and protein folding. Figure 2 illustrates the functional domains in MUC4 protein sequence and secondary structures that contain these SNPs. Our data show that genetic variations of rs882605 and rs2688513 cause amino acid substitutions in long-loop regions (> 10 residues) between secondary structure units (α-helices or β-strands) with high hydrophilicity and moderate surface probability (Figure 2). Amino acid composition analyses also revealed that these two loops (300-319 and 4,134-4,158) contain several negatively charged aspartate (D) residues and reverse turn elements, glycine (G) and proline (P), suggesting the importance of these loops in protein folding and functional regulation [34,35]. In addition, rs2246901 locates in a type D von Willebrand factor (VWFD) domain responsible for protein-protein interaction and cell adhesion and/or migration [36,37]. Our findings support the functional roles of MUC4 SNPs in regulating cellular mobility and invasion of endometrial cells during endometriosis development and progression.\nFunctional domains in the MUC4 protein sequence and the predicted secondary structures. Six functional domains and/or signatures (boxes) were annotated by aligning MUC4 protein sequence in PROSITE protein domain database [28]. The boundaries of each signature are listed. Among six SNPs tested in this study (stars), rs882605 and rs2688513 (black) were found in two different long-loop regions, 300-319 and 4,134-4,158, respectively (bold letters refer to amino acid substitution sites). The SNP rs2246901 (gray) was found in a type D von Willebrand factor (VWFD) domain. The MUC4 reference sequence can be located at National Center for Biotechnology Information databank NP_060876.4.", "Previous studies have shown polymorphism of cytokines and adhesion molecules which were associated with the pathogenesis of endometriosis [38,39]. To the best of our knowledge, however, no other study to date has investigated the possible association of MUC4 and endometriosis. The purpose of this study was to evaluate whether genetic variations in MUC4 associate with endometriosis in the Taiwanese population. Our data prove the association of MUC4 polymorphisms with advanced stages of endometriosis and the related infertility. Since the extracellular domain of MUC4 is critical for HER2 interaction and cell invasiveness, these defined SNPs located in putative functional domains of MUC4 may play important roles during endometriosis development and progression.\nThe development of endometriosis and certain types of ovarian cancer share several similar clinical features. For example, endometriosis could progressively invade pelvic viscera, resulting in adhesion, and could recur after medical treatment or surgery. Because of the functions involved in the acquisition of adhesion ligands or receptors and the loss of antiadhesion, proteins such as MUC4 and MUC1, the two major mucins present in endometrial epithelium, thus become suspect in endometriosis development [16,17]. In addition to gene overexpression, genetic variations in MUC1 have also been reported as risk factors contributing to cell mobility and the severity of cancer [40-43]. However, the influence of MUC4 genetic variations on cell behavior remains unclear. In this study, Pro4135Ser (rs2688513) and Ala4693Ser (2246901) substitutions in the putative functional domains of MUC4 were found to be associated with advanced stages of endometriosis. Since MUC4 is an emerging target for ovarian cancer [18,19,44,45], our study provides a new direction from which to address the roles of MUC4 in the development of gynecological disorders.\nTo study the genetic effects of mucin proteins by SNPs, the major interest focuses on the variable number of tandem repeat (VNTR) polymorphisms, which result in different-sized gene transcripts. For examples, MUC1 variations in the VNTR domain have been found to play roles in regulating Helicobacter pylori binding to gastric cells [46]. Other studies have also concluded that VNTR polymorphisms can influence T-antigen presentation and the local immune responses, which consequently have potential effects on gastric cancer development [47,48]. With regard to MUC4, a high degree of polymorphism in the VNTR domain was observed in human tissues, including the endometrial epithelium [49,50]. However, the different-sized MUC4 transcripts did not show association with embryo implantation or cancer development. By contrast, MUC4 can promote cell proliferation and antiapoptotic effects in cancer cells by interacting with HER2 on the cell surface [18,19,22], suggesting the potency of functional domains in the extracellular domain of MUC4. In this study, two SNPs (rs2688513 and rs2246901) that locate in a putative functional loop and the VWFD protein binding domain, respectively, were found to be associated with advanced stages of endometriosis. Further study may clarify whether these amino acid substitutions could change the interaction with HER2 and/or play crucial roles in regulating the cellular activity of the spread of endometriosis.\nEndometriosis could cause pelvic adhesion and tubal occlusion that may lead to infertility. However, among patients with endometriosis-related infertility, 50% to 60% of them were diagnosed with minimal or mild endometriosis [3]. Impaired folliculogenesis, bad oocyte quality and impaired implantation of embryo are therefore considered to be the possible mechanisms for endometriosis-related infertility. Changes in cytokines and growth factors in endometrium, follicular fluid and peritoneal fluid have been suggested as the key players for inducing the above-mentioned phenomena [5]. Recently, several studies have shown that MUC4 could promote cell migration, change the endometrial environment and create weak points in the epithelium, thus facilitating the failure of embryo implantation [50,51]. The Carraway et al. [52] study also showed that embryo implantation was associated with downregulation of MUC4 expression in an animal model. In this study, women with a T allele in rs882605 had a lower risk of endometriosis-related infertility, whereas rs2688513 and rs2246901 SNPs did not show any association with the reproductive ability of patients. The rs882605 SNP locates in a putative functional loop within the VNTR domain of MUC4 that may control T-cell antigen presentation and the local immune responses. Our findings may support the view that the regulation of local immunity, rather than uncontrolled cell proliferation, in the endometrium may play a more important role in the development of endometriosis-related infertility.\nOur study shows that the T/G genotype at rs882605, as compared with the T/T or G/G genotypes, is unique in patients with endometriosis. So far, our data cannot provide sufficient information to explain why the T/T genotype does not show higher risk of endometriosis than T/G. One reason could be the relatively small study group, while other possibilities could exist. For example, the SNPs analyzed are in tight LD with other unknown allele variants which have an opposite effect. In this case, only individuals with the T/G heterozygous genotype could be observed. Because this is a hospital-based study with a modest sample size, enrollment of a larger cohort based on a population approach could help to elucidate the functional role of MUC4 in endometriosis and the related infertility.", "In this study, we observed an association of MUC4 polymorphism with endometriosis development and endometriosis-related infertility in a Taiwanese population. However, the true mechanism of how MUC4 modulates the pathogenesis of endometriosis and infertility was not clearly understood. In addition, the risk SNPs pose for endometriosis stages and infertility differ, suggesting dissimilar molecular mechanisms for these clinical features. More detailed studies are needed to investigate the biochemical pathways regulated by MUC4 during the development of endometriosis.", "The authors declare that they have no competing interests.", "CYYC collected samples, carried out the clinical association study and drafted the manuscript. HWC participated in sample collection and the clinical association study. CMC carried out sample preparation and SNP analyses. CYL participated in SNP analyses. CPC participated in the clinical association study. CHL carried out statistical analyses. WYL carried out sample pretreatment and RNA extraction. JJCS carried out the experimental design and drafted the manuscript. FJT carried out the experimental design and revised the manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1741-7015/9/19/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study population", "Clinical stages and association study", "Genomic DNA extraction and genotyping of SNPs in MUC4", "Statistical analysis", "Functional analyses and secondary structure predictions of MUC4 protein", "Results", "MUC4 gene polymorphisms and endometriosis", "Association of MUC4 gene polymorphisms and stages", "MUC4 gene polymorphisms and infertility", "MUC4 gene polymorphisms and amino acid substitutions", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history", "Supplementary Material" ]

[ "Endometriosis is a common chronic gynecological disease characterized by the presence of endometrial gland and stroma outside the uterine cavity, affecting approximately 10% of reproductive age women [1,2]. The common clinical symptoms include pelvic pain, heavy menstrual bleeding, pelvic adhesion, bloating and fatigue. Notably, the prevalence of endometriosis is 0.5% to 5% in fertile women and 25% to 40% in infertile women [3], suggesting infertility as one possible consequence of endometriosis. To date, the implantation theory is widely accepted, stating that endometrial tissues pass through the fallopian tube, then attach and grow on pelvic tissue. However, this hypothesis cannot explain the existence of endometriosis outside the pelvis or how endometriosis progresses and invades other tissues. Additional factors such as genetic or immune differences have been suggested as possible contributors to trigger the formation of endometriosis [4-6]. Family history and genome-wide linkage studies also support genetic predisposition during the development of endometriosis [7-10]. These studies provide molecular evidence demonstrating endometriosis as a genetic disease, and it is desirable to explore more genetic variations associated with endometriosis.\nSimilarly to malignant diseases, extensive growth of endometrial cells on the peritoneal surface and invasion of the pelvic organ are very common during the development of endometriosis. This process is frequently associated with several mechanisms involved in angiogenesis and cellular adhesion. In fact, women who have endometriosis appear to be more at risk of developing several different kinds of ovarian cancers [8,11-13]. An epidemiological study showed that the prevalence rates of endometriosis in patients with endometrioid and clear cell ovarian carcinoma are 19 and 35.9, respectively [14]. These findings suggest that endometriosis and certain types of ovarian cancer may share several common genetic alterations during pathogenesis. Genes that regulate cell mobility and invasion in ovarian cancers are therefore possible candidates for playing roles in endometriosis.\nMucins are high-molecular-weight glycoproteins with the function of protecting and lubricating the epithelial surface of respiratory, gastrointestinal and reproductive tracts [15]. Among the mucin proteins, mucin 4 (MUC4) and mucin 1 (MUC1) are the major ones expressed in the endometrial epithelium [16,17]. In cancer studies, these two mucins have been shown to be aberrantly expressed in various malignancies and have been validated as novel targets for cancer diagnosis and therapy [18-20]. Distinct from MUC1, the extracellular domain of MUC4 can interact with human epidermal growth factor receptor 2 (HER2) on the cell surface and modulate downstream cell growth signaling by stabilizing and/or enhancing the activity of cell growth receptor complexes [18,21,22]. Consequently, changes of cytoarchitectures and cellular signaling may lead to the increase of cell mobility and tumor cell invasion.\nThe above findings provide us with clues to hypothesize that genetic variations in the extracellular domain of MUC4, especially those resulting in amino acid substitutions, may play roles involved in the development of endometriosis. With endometriosis as a possible cause of infertility in women, we also would like to study the association of MUC4 single-nucleotide polymorphisms (SNPs) with susceptibility to endometriosis-related infertility.", "[SUBTITLE] Study population [SUBSECTION] A total of 140 individuals who underwent surgery for benign diseases and pathology-proven endometriosis were identified at China Medical University Hospital from 1998 to 2008 and were enrolled in this study. In general, these patients were diagnosed with ovarian cysts on the basis of sonography and had several clinical symptoms related to endometriosis, including dysmenorrhea, lower abdominal pain, infertility or abnormal menstruation. Study patients who failed to have pathology-proven endometriosis were excluded from this study. For the control group, blood samples of 150 healthy women were selected from a pool of individuals who received regular heath checkups at the same hospital and were identified as normal on the basis of the examinations conducted. A total of 142 controls were frequency-matched on the basis of age profile with the study patients (Additional file 1, Table S1). Controls who showed one of the endometriosis-associated symptoms, even though the results of their health checkups were normal, were excluded from this study. This study was approved by the institutional review board at China Medical University, with informed consent obtained from each patient.\nA total of 140 individuals who underwent surgery for benign diseases and pathology-proven endometriosis were identified at China Medical University Hospital from 1998 to 2008 and were enrolled in this study. In general, these patients were diagnosed with ovarian cysts on the basis of sonography and had several clinical symptoms related to endometriosis, including dysmenorrhea, lower abdominal pain, infertility or abnormal menstruation. Study patients who failed to have pathology-proven endometriosis were excluded from this study. For the control group, blood samples of 150 healthy women were selected from a pool of individuals who received regular heath checkups at the same hospital and were identified as normal on the basis of the examinations conducted. A total of 142 controls were frequency-matched on the basis of age profile with the study patients (Additional file 1, Table S1). Controls who showed one of the endometriosis-associated symptoms, even though the results of their health checkups were normal, were excluded from this study. This study was approved by the institutional review board at China Medical University, with informed consent obtained from each patient.\n[SUBTITLE] Clinical stages and association study [SUBSECTION] Clinical information on patients was collected from clinical notes, including clinical stage, lesion size, location, drug treatment and fertility (Additional file 1, Table S1). The definition of endometriosis staging was based on the classification of the American Society for Reproductive Medicine: Stage 1, minimal; Stage 2, mild; Stage 3, moderate; and Stage 4, severe [23].\nClinical information on patients was collected from clinical notes, including clinical stage, lesion size, location, drug treatment and fertility (Additional file 1, Table S1). The definition of endometriosis staging was based on the classification of the American Society for Reproductive Medicine: Stage 1, minimal; Stage 2, mild; Stage 3, moderate; and Stage 4, severe [23].\n[SUBTITLE] Genomic DNA extraction and genotyping of SNPs in MUC4 [SUBSECTION] Genomic DNA was extracted from peripheral blood leukocytes according to standard protocols (Genomic DNA kit; Qiagen, Valencia, CA, USA). DNA fragments containing the target SNP sites were amplified by polymerase chain reaction (PCR) assay using the TaqMan SNP Genotyping Assay System from Applied Biosystems, Inc. (Carlsbad, CA, USA). The probe search and design are available on the Applied Biosystems, Inc. website [24].\nAdditional file 1, Table S2, lists probe identifications for the six SNPs tested. PCR amplification conditions consisted of initial denaturation at 95°C for 5 minutes followed by 40 cycles at 95°C for 10 seconds, 56°C for 10 seconds and 72°C for 20 seconds, with one additional cycle at 72°C for 5 minutes. Genetic variations were detected by reading the fluorescence signals of PCR products. A positive signal indicates a perfect match between the probe and the tested DNA, thus identifying the allele types. Ten percent of study participants were randomly chosen and genotyped in duplicate to confirm the concordance of the genotyping results. In our study, the call rates for these SNP probes were above 94% (Additional file 1, Table S3).\nGenomic DNA was extracted from peripheral blood leukocytes according to standard protocols (Genomic DNA kit; Qiagen, Valencia, CA, USA). DNA fragments containing the target SNP sites were amplified by polymerase chain reaction (PCR) assay using the TaqMan SNP Genotyping Assay System from Applied Biosystems, Inc. (Carlsbad, CA, USA). The probe search and design are available on the Applied Biosystems, Inc. website [24].\nAdditional file 1, Table S2, lists probe identifications for the six SNPs tested. PCR amplification conditions consisted of initial denaturation at 95°C for 5 minutes followed by 40 cycles at 95°C for 10 seconds, 56°C for 10 seconds and 72°C for 20 seconds, with one additional cycle at 72°C for 5 minutes. Genetic variations were detected by reading the fluorescence signals of PCR products. A positive signal indicates a perfect match between the probe and the tested DNA, thus identifying the allele types. Ten percent of study participants were randomly chosen and genotyped in duplicate to confirm the concordance of the genotyping results. In our study, the call rates for these SNP probes were above 94% (Additional file 1, Table S3).\n[SUBTITLE] Statistical analysis [SUBSECTION] The allelic frequency and genotype frequency distributions for the six polymorphisms of patients with endometriosis and the controls were performed by χ2 analysis using SPSS software (version 10.0; SPSS Inc., Chicago, IL, USA). An unordered, 2 df two-sided test was used for the statistical analyses of our genotyping results. P < 0.05 was considered statistically significant. Allelic and genotypic frequencies are expressed as percentages of total alleles and genotypes. Odds ratios (ORs) were calculated from allelic and genotype frequencies with 95% confident intervals (95% CIs). The major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. Adherence to the Hardy-Weinberg equilibrium (HWE) constant was tested using a χ2 test with 1 df. To study the association of the six SNPs with clinical stages and reproductive ability, Fisher's exact tests instead of χ2 tests were used because of the small number of participants tested.\nThe haplotypes of each individual were determined using the Bayesian statistical method available in the free download software program PHASE 2.1 [25]. This approach incorporates a priori expectations of haplotypic structure based on population genetics and coalescence theory. Lewontin's D' (|D'|) and the linkage disequilibrium (LD) coefficient r2 were determined between selected pairs of biallelic loci [26]. Haploview version 3.2 software (Whitehead Institute for Biomedical Research, Cambridge, MA, USA) was used to examine the structure of the LD block [27]. This program uses two-marker expectation maximization to estimate the maximum likelihood values of the four gamete frequencies from which the D' and log of odds (LOD) values are derived. The genetic effects of the inferred haplotypes were analyzed in the same way as the analysis of polymorphisms. The reported haplotype percentages are estimated on the basis of allele frequencies and LD. The P values are based on a comparison of a given haplotype with all other haplotypes combined.\nThe allelic frequency and genotype frequency distributions for the six polymorphisms of patients with endometriosis and the controls were performed by χ2 analysis using SPSS software (version 10.0; SPSS Inc., Chicago, IL, USA). An unordered, 2 df two-sided test was used for the statistical analyses of our genotyping results. P < 0.05 was considered statistically significant. Allelic and genotypic frequencies are expressed as percentages of total alleles and genotypes. Odds ratios (ORs) were calculated from allelic and genotype frequencies with 95% confident intervals (95% CIs). The major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. Adherence to the Hardy-Weinberg equilibrium (HWE) constant was tested using a χ2 test with 1 df. To study the association of the six SNPs with clinical stages and reproductive ability, Fisher's exact tests instead of χ2 tests were used because of the small number of participants tested.\nThe haplotypes of each individual were determined using the Bayesian statistical method available in the free download software program PHASE 2.1 [25]. This approach incorporates a priori expectations of haplotypic structure based on population genetics and coalescence theory. Lewontin's D' (|D'|) and the linkage disequilibrium (LD) coefficient r2 were determined between selected pairs of biallelic loci [26]. Haploview version 3.2 software (Whitehead Institute for Biomedical Research, Cambridge, MA, USA) was used to examine the structure of the LD block [27]. This program uses two-marker expectation maximization to estimate the maximum likelihood values of the four gamete frequencies from which the D' and log of odds (LOD) values are derived. The genetic effects of the inferred haplotypes were analyzed in the same way as the analysis of polymorphisms. The reported haplotype percentages are estimated on the basis of allele frequencies and LD. The P values are based on a comparison of a given haplotype with all other haplotypes combined.\n[SUBTITLE] Functional analyses and secondary structure predictions of MUC4 protein [SUBSECTION] Functional characterization and annotation of MUC4 were performed by aligning the sequence with functional motifs and/or signatures in the PROSITE protein domain database [28]. To predict the secondary structure of the MUC4 sequence, the Chou and Fasman method was used [29]. An improved method was applied to increase the accuracy of the predictions by locating nucleation regions with refined wavelet transformation technology and by calculating propensity factors with larger data sets [30]. The program gives the propensity of each residue to be a part of an α-helix, a β-strand or a loop. We considered propensities Pα > 1.03 as significant for helices and propensities Pβ > 1.05 as significant for strands. Predicted regions with fewer than four contiguous residues were not considered secondary structure units. For a region with both helix and strand tendencies, the secondary structure conformed with higher propensity: Pα >Pβ or Pβ >Pα is predicted. To plot hydrophobicity and surface probability, the Kyte and Doolittle method [31] and Emini et al. surface accessibility prediction (SAP) [32] were used, respectively. We slid a window along the MUC4 sequence to assign a \"hydrophobicity\" or \"surface probability\" value to each amino acid. The values were summed in the window, and the results were plotted.\nFunctional characterization and annotation of MUC4 were performed by aligning the sequence with functional motifs and/or signatures in the PROSITE protein domain database [28]. To predict the secondary structure of the MUC4 sequence, the Chou and Fasman method was used [29]. An improved method was applied to increase the accuracy of the predictions by locating nucleation regions with refined wavelet transformation technology and by calculating propensity factors with larger data sets [30]. The program gives the propensity of each residue to be a part of an α-helix, a β-strand or a loop. We considered propensities Pα > 1.03 as significant for helices and propensities Pβ > 1.05 as significant for strands. Predicted regions with fewer than four contiguous residues were not considered secondary structure units. For a region with both helix and strand tendencies, the secondary structure conformed with higher propensity: Pα >Pβ or Pβ >Pα is predicted. To plot hydrophobicity and surface probability, the Kyte and Doolittle method [31] and Emini et al. surface accessibility prediction (SAP) [32] were used, respectively. We slid a window along the MUC4 sequence to assign a \"hydrophobicity\" or \"surface probability\" value to each amino acid. The values were summed in the window, and the results were plotted.", "A total of 140 individuals who underwent surgery for benign diseases and pathology-proven endometriosis were identified at China Medical University Hospital from 1998 to 2008 and were enrolled in this study. In general, these patients were diagnosed with ovarian cysts on the basis of sonography and had several clinical symptoms related to endometriosis, including dysmenorrhea, lower abdominal pain, infertility or abnormal menstruation. Study patients who failed to have pathology-proven endometriosis were excluded from this study. For the control group, blood samples of 150 healthy women were selected from a pool of individuals who received regular heath checkups at the same hospital and were identified as normal on the basis of the examinations conducted. A total of 142 controls were frequency-matched on the basis of age profile with the study patients (Additional file 1, Table S1). Controls who showed one of the endometriosis-associated symptoms, even though the results of their health checkups were normal, were excluded from this study. This study was approved by the institutional review board at China Medical University, with informed consent obtained from each patient.", "Clinical information on patients was collected from clinical notes, including clinical stage, lesion size, location, drug treatment and fertility (Additional file 1, Table S1). The definition of endometriosis staging was based on the classification of the American Society for Reproductive Medicine: Stage 1, minimal; Stage 2, mild; Stage 3, moderate; and Stage 4, severe [23].", "Genomic DNA was extracted from peripheral blood leukocytes according to standard protocols (Genomic DNA kit; Qiagen, Valencia, CA, USA). DNA fragments containing the target SNP sites were amplified by polymerase chain reaction (PCR) assay using the TaqMan SNP Genotyping Assay System from Applied Biosystems, Inc. (Carlsbad, CA, USA). The probe search and design are available on the Applied Biosystems, Inc. website [24].\nAdditional file 1, Table S2, lists probe identifications for the six SNPs tested. PCR amplification conditions consisted of initial denaturation at 95°C for 5 minutes followed by 40 cycles at 95°C for 10 seconds, 56°C for 10 seconds and 72°C for 20 seconds, with one additional cycle at 72°C for 5 minutes. Genetic variations were detected by reading the fluorescence signals of PCR products. A positive signal indicates a perfect match between the probe and the tested DNA, thus identifying the allele types. Ten percent of study participants were randomly chosen and genotyped in duplicate to confirm the concordance of the genotyping results. In our study, the call rates for these SNP probes were above 94% (Additional file 1, Table S3).", "The allelic frequency and genotype frequency distributions for the six polymorphisms of patients with endometriosis and the controls were performed by χ2 analysis using SPSS software (version 10.0; SPSS Inc., Chicago, IL, USA). An unordered, 2 df two-sided test was used for the statistical analyses of our genotyping results. P < 0.05 was considered statistically significant. Allelic and genotypic frequencies are expressed as percentages of total alleles and genotypes. Odds ratios (ORs) were calculated from allelic and genotype frequencies with 95% confident intervals (95% CIs). The major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. Adherence to the Hardy-Weinberg equilibrium (HWE) constant was tested using a χ2 test with 1 df. To study the association of the six SNPs with clinical stages and reproductive ability, Fisher's exact tests instead of χ2 tests were used because of the small number of participants tested.\nThe haplotypes of each individual were determined using the Bayesian statistical method available in the free download software program PHASE 2.1 [25]. This approach incorporates a priori expectations of haplotypic structure based on population genetics and coalescence theory. Lewontin's D' (|D'|) and the linkage disequilibrium (LD) coefficient r2 were determined between selected pairs of biallelic loci [26]. Haploview version 3.2 software (Whitehead Institute for Biomedical Research, Cambridge, MA, USA) was used to examine the structure of the LD block [27]. This program uses two-marker expectation maximization to estimate the maximum likelihood values of the four gamete frequencies from which the D' and log of odds (LOD) values are derived. The genetic effects of the inferred haplotypes were analyzed in the same way as the analysis of polymorphisms. The reported haplotype percentages are estimated on the basis of allele frequencies and LD. The P values are based on a comparison of a given haplotype with all other haplotypes combined.", "Functional characterization and annotation of MUC4 were performed by aligning the sequence with functional motifs and/or signatures in the PROSITE protein domain database [28]. To predict the secondary structure of the MUC4 sequence, the Chou and Fasman method was used [29]. An improved method was applied to increase the accuracy of the predictions by locating nucleation regions with refined wavelet transformation technology and by calculating propensity factors with larger data sets [30]. The program gives the propensity of each residue to be a part of an α-helix, a β-strand or a loop. We considered propensities Pα > 1.03 as significant for helices and propensities Pβ > 1.05 as significant for strands. Predicted regions with fewer than four contiguous residues were not considered secondary structure units. For a region with both helix and strand tendencies, the secondary structure conformed with higher propensity: Pα >Pβ or Pβ >Pα is predicted. To plot hydrophobicity and surface probability, the Kyte and Doolittle method [31] and Emini et al. surface accessibility prediction (SAP) [32] were used, respectively. We slid a window along the MUC4 sequence to assign a \"hydrophobicity\" or \"surface probability\" value to each amino acid. The values were summed in the window, and the results were plotted.", "[SUBTITLE] MUC4 gene polymorphisms and endometriosis [SUBSECTION] Six SNPs in the extracellular domain of the MUC4 gene with a frequency greater than 20% in Chinese Han Beijing were selected from International HapMap Project databank [33] (Additional file 1, Table S2). Genotypic analyses (Figure 1) indicated rs882605 as a unique SNP with a higher frequency of the TG genotype in patients than in controls (P = 0.04, OR = 1.97, 95% CI, 1.17 to 3.32) (Table 1), while allele type analyses of these SNPs showed no statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. To confirm the genetic impact of SNPs on endometriosis, the top two high-risk alleles at rs882605 and rs1104760 were selected for haplotype analyses. Significantly, the frequency of haplotype T-T was found to be higher in patients than in controls (P = 0.0353) (Table 2) (Additional file 2, Figure S1), suggesting the association of MUC4 SNPs with endometriosis development. The genotyping results were confirmed in duplicate, and the concordance of duplicates was 97.6%.\nAllelic discrimination plots of the six tested single-nucleotide polymorphisms (SNPs) in the mucin 4 (MUC4) gene. The DNA samples from patients and controls were genotyped by using the TaqMan SNP Genotyping Assay System. The major (also the wild-type) alleles were detected by 6-carboxyfluorescein (FAM)-labeled probes (blue), and the minor alleles were detected by 2'-chloro-7'-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC)-labeled probes. The genotyping results of the six SNPs in the MUC4 gene are presented as allelic discrimination plots. Of note, the intensity of FAM signals tended to be similar among samples in our assays, thus the dots for a wild-type genotype overlapped each other. \"X\" indicates the participant who failed to be genotyped.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in Taiwanese endometriosis patients and controlsa\naMUC4, mucin 4 gene; SNP, single-nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval; HWE, P values of deviation from the Hardy-Weinberg equilibrium constant. Allelic frequencies were determined by χ2 test using 2 × 2 contingency tables. Genotype frequencies were determined by χ2 test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients and controlsa\naMUC4, mucin 4 gene. The reported haplotype percents are estimated percents based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.\nSix SNPs in the extracellular domain of the MUC4 gene with a frequency greater than 20% in Chinese Han Beijing were selected from International HapMap Project databank [33] (Additional file 1, Table S2). Genotypic analyses (Figure 1) indicated rs882605 as a unique SNP with a higher frequency of the TG genotype in patients than in controls (P = 0.04, OR = 1.97, 95% CI, 1.17 to 3.32) (Table 1), while allele type analyses of these SNPs showed no statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. To confirm the genetic impact of SNPs on endometriosis, the top two high-risk alleles at rs882605 and rs1104760 were selected for haplotype analyses. Significantly, the frequency of haplotype T-T was found to be higher in patients than in controls (P = 0.0353) (Table 2) (Additional file 2, Figure S1), suggesting the association of MUC4 SNPs with endometriosis development. The genotyping results were confirmed in duplicate, and the concordance of duplicates was 97.6%.\nAllelic discrimination plots of the six tested single-nucleotide polymorphisms (SNPs) in the mucin 4 (MUC4) gene. The DNA samples from patients and controls were genotyped by using the TaqMan SNP Genotyping Assay System. The major (also the wild-type) alleles were detected by 6-carboxyfluorescein (FAM)-labeled probes (blue), and the minor alleles were detected by 2'-chloro-7'-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC)-labeled probes. The genotyping results of the six SNPs in the MUC4 gene are presented as allelic discrimination plots. Of note, the intensity of FAM signals tended to be similar among samples in our assays, thus the dots for a wild-type genotype overlapped each other. \"X\" indicates the participant who failed to be genotyped.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in Taiwanese endometriosis patients and controlsa\naMUC4, mucin 4 gene; SNP, single-nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval; HWE, P values of deviation from the Hardy-Weinberg equilibrium constant. Allelic frequencies were determined by χ2 test using 2 × 2 contingency tables. Genotype frequencies were determined by χ2 test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients and controlsa\naMUC4, mucin 4 gene. The reported haplotype percents are estimated percents based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.\n[SUBTITLE] Association of MUC4 gene polymorphisms and stages [SUBSECTION] We next asked whether MUC4 genetic variations could possibly associate with clinical stages. Patients were divided into two groups: the mild stage group with patients at stages 1 or 2 and the advanced group with patients at stages 3 or 4. Strikingly, genotype analyses revealed strong associations of CC type at rs2688513 (P = 0.04) and GG type at rs2246901 (P = 0.03), with more advanced endometriosis at stage 3 or 4 (Table 3). Dominant effects were found for other genetic variations at rs1104760 (CC + CT versus TT) and rs2258447 (AA + AG versus GG) during endometriosis progression. Allele type analyses suggested C allele at rs1104760, C allele at rs2688513, G allele at rs2246901 and A allele at rs2258447 as risk factors that correlated with more severe endometriosis.\nGenotype and allele distributions of SNPs in the MUC4 gene in endometriosis patients at different clinical stagesa\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bMild stage, patients at clinical stage1 or stage 2; cSevere stage, patients at clinical stage 3 or stage 4; dP < 0.05 was considered statistically significant.\nWe next asked whether MUC4 genetic variations could possibly associate with clinical stages. Patients were divided into two groups: the mild stage group with patients at stages 1 or 2 and the advanced group with patients at stages 3 or 4. Strikingly, genotype analyses revealed strong associations of CC type at rs2688513 (P = 0.04) and GG type at rs2246901 (P = 0.03), with more advanced endometriosis at stage 3 or 4 (Table 3). Dominant effects were found for other genetic variations at rs1104760 (CC + CT versus TT) and rs2258447 (AA + AG versus GG) during endometriosis progression. Allele type analyses suggested C allele at rs1104760, C allele at rs2688513, G allele at rs2246901 and A allele at rs2258447 as risk factors that correlated with more severe endometriosis.\nGenotype and allele distributions of SNPs in the MUC4 gene in endometriosis patients at different clinical stagesa\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bMild stage, patients at clinical stage1 or stage 2; cSevere stage, patients at clinical stage 3 or stage 4; dP < 0.05 was considered statistically significant.\n[SUBTITLE] MUC4 gene polymorphisms and infertility [SUBSECTION] Since endometriosis has been suspected as one potent factor leading to infertility in women [3], we also studied the possible linkage between MUC4 SNPs and infertility. Though no significant difference was found in our genotype association study, our data indicated T allele at rs882605 as a protective factor that associated with reduced frequency of infertility in patients with endometriosis (Table 4). Two other alleles, C at rs2688513 and G at rs2246901, showed similar protective effects, but the data did not reach statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. In our haplotype analyses, we thus sought to ascertain the impact of the genetic combination of these top three protective alleles. Table 5 indicates that patients with haplotype T-C-G did show a lower frequency of infertility, although the results did not show a statistically significant difference (P = 0.099). By contrast, haplotype G-T-T showed a strong association with infertility in patients (P = 0.012) (Table 5) (Additional file 2, Figure S2) and could be used as a risk indicator for patients at higher risk of developing severe complications such as infertility.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in endometriosis patients with different reproductive abilitya\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene; OR, odds ratio; 95% CI, 95% confidence interval. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients with infertilitya\naMUC4, mucin 4 gene. The reported haplotype percentages are estimated percentages based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.\nSince endometriosis has been suspected as one potent factor leading to infertility in women [3], we also studied the possible linkage between MUC4 SNPs and infertility. Though no significant difference was found in our genotype association study, our data indicated T allele at rs882605 as a protective factor that associated with reduced frequency of infertility in patients with endometriosis (Table 4). Two other alleles, C at rs2688513 and G at rs2246901, showed similar protective effects, but the data did not reach statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. In our haplotype analyses, we thus sought to ascertain the impact of the genetic combination of these top three protective alleles. Table 5 indicates that patients with haplotype T-C-G did show a lower frequency of infertility, although the results did not show a statistically significant difference (P = 0.099). By contrast, haplotype G-T-T showed a strong association with infertility in patients (P = 0.012) (Table 5) (Additional file 2, Figure S2) and could be used as a risk indicator for patients at higher risk of developing severe complications such as infertility.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in endometriosis patients with different reproductive abilitya\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene; OR, odds ratio; 95% CI, 95% confidence interval. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients with infertilitya\naMUC4, mucin 4 gene. The reported haplotype percentages are estimated percentages based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.\n[SUBTITLE] MUC4 gene polymorphisms and amino acid substitutions [SUBSECTION] Because these endometriosis-associated SNPs can cause amino acid substitutions (Additional file 1, Table S2), the biofunctions of MUC4 might be altered by changes in hydrophilicity and protein folding. Figure 2 illustrates the functional domains in MUC4 protein sequence and secondary structures that contain these SNPs. Our data show that genetic variations of rs882605 and rs2688513 cause amino acid substitutions in long-loop regions (> 10 residues) between secondary structure units (α-helices or β-strands) with high hydrophilicity and moderate surface probability (Figure 2). Amino acid composition analyses also revealed that these two loops (300-319 and 4,134-4,158) contain several negatively charged aspartate (D) residues and reverse turn elements, glycine (G) and proline (P), suggesting the importance of these loops in protein folding and functional regulation [34,35]. In addition, rs2246901 locates in a type D von Willebrand factor (VWFD) domain responsible for protein-protein interaction and cell adhesion and/or migration [36,37]. Our findings support the functional roles of MUC4 SNPs in regulating cellular mobility and invasion of endometrial cells during endometriosis development and progression.\nFunctional domains in the MUC4 protein sequence and the predicted secondary structures. Six functional domains and/or signatures (boxes) were annotated by aligning MUC4 protein sequence in PROSITE protein domain database [28]. The boundaries of each signature are listed. Among six SNPs tested in this study (stars), rs882605 and rs2688513 (black) were found in two different long-loop regions, 300-319 and 4,134-4,158, respectively (bold letters refer to amino acid substitution sites). The SNP rs2246901 (gray) was found in a type D von Willebrand factor (VWFD) domain. The MUC4 reference sequence can be located at National Center for Biotechnology Information databank NP_060876.4.\nBecause these endometriosis-associated SNPs can cause amino acid substitutions (Additional file 1, Table S2), the biofunctions of MUC4 might be altered by changes in hydrophilicity and protein folding. Figure 2 illustrates the functional domains in MUC4 protein sequence and secondary structures that contain these SNPs. Our data show that genetic variations of rs882605 and rs2688513 cause amino acid substitutions in long-loop regions (> 10 residues) between secondary structure units (α-helices or β-strands) with high hydrophilicity and moderate surface probability (Figure 2). Amino acid composition analyses also revealed that these two loops (300-319 and 4,134-4,158) contain several negatively charged aspartate (D) residues and reverse turn elements, glycine (G) and proline (P), suggesting the importance of these loops in protein folding and functional regulation [34,35]. In addition, rs2246901 locates in a type D von Willebrand factor (VWFD) domain responsible for protein-protein interaction and cell adhesion and/or migration [36,37]. Our findings support the functional roles of MUC4 SNPs in regulating cellular mobility and invasion of endometrial cells during endometriosis development and progression.\nFunctional domains in the MUC4 protein sequence and the predicted secondary structures. Six functional domains and/or signatures (boxes) were annotated by aligning MUC4 protein sequence in PROSITE protein domain database [28]. The boundaries of each signature are listed. Among six SNPs tested in this study (stars), rs882605 and rs2688513 (black) were found in two different long-loop regions, 300-319 and 4,134-4,158, respectively (bold letters refer to amino acid substitution sites). The SNP rs2246901 (gray) was found in a type D von Willebrand factor (VWFD) domain. The MUC4 reference sequence can be located at National Center for Biotechnology Information databank NP_060876.4.", "Six SNPs in the extracellular domain of the MUC4 gene with a frequency greater than 20% in Chinese Han Beijing were selected from International HapMap Project databank [33] (Additional file 1, Table S2). Genotypic analyses (Figure 1) indicated rs882605 as a unique SNP with a higher frequency of the TG genotype in patients than in controls (P = 0.04, OR = 1.97, 95% CI, 1.17 to 3.32) (Table 1), while allele type analyses of these SNPs showed no statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. To confirm the genetic impact of SNPs on endometriosis, the top two high-risk alleles at rs882605 and rs1104760 were selected for haplotype analyses. Significantly, the frequency of haplotype T-T was found to be higher in patients than in controls (P = 0.0353) (Table 2) (Additional file 2, Figure S1), suggesting the association of MUC4 SNPs with endometriosis development. The genotyping results were confirmed in duplicate, and the concordance of duplicates was 97.6%.\nAllelic discrimination plots of the six tested single-nucleotide polymorphisms (SNPs) in the mucin 4 (MUC4) gene. The DNA samples from patients and controls were genotyped by using the TaqMan SNP Genotyping Assay System. The major (also the wild-type) alleles were detected by 6-carboxyfluorescein (FAM)-labeled probes (blue), and the minor alleles were detected by 2'-chloro-7'-phenyl-1,4-dichloro-6-carboxyfluorescein (VIC)-labeled probes. The genotyping results of the six SNPs in the MUC4 gene are presented as allelic discrimination plots. Of note, the intensity of FAM signals tended to be similar among samples in our assays, thus the dots for a wild-type genotype overlapped each other. \"X\" indicates the participant who failed to be genotyped.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in Taiwanese endometriosis patients and controlsa\naMUC4, mucin 4 gene; SNP, single-nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval; HWE, P values of deviation from the Hardy-Weinberg equilibrium constant. Allelic frequencies were determined by χ2 test using 2 × 2 contingency tables. Genotype frequencies were determined by χ2 test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients and controlsa\naMUC4, mucin 4 gene. The reported haplotype percents are estimated percents based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.", "We next asked whether MUC4 genetic variations could possibly associate with clinical stages. Patients were divided into two groups: the mild stage group with patients at stages 1 or 2 and the advanced group with patients at stages 3 or 4. Strikingly, genotype analyses revealed strong associations of CC type at rs2688513 (P = 0.04) and GG type at rs2246901 (P = 0.03), with more advanced endometriosis at stage 3 or 4 (Table 3). Dominant effects were found for other genetic variations at rs1104760 (CC + CT versus TT) and rs2258447 (AA + AG versus GG) during endometriosis progression. Allele type analyses suggested C allele at rs1104760, C allele at rs2688513, G allele at rs2246901 and A allele at rs2258447 as risk factors that correlated with more severe endometriosis.\nGenotype and allele distributions of SNPs in the MUC4 gene in endometriosis patients at different clinical stagesa\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bMild stage, patients at clinical stage1 or stage 2; cSevere stage, patients at clinical stage 3 or stage 4; dP < 0.05 was considered statistically significant.", "Since endometriosis has been suspected as one potent factor leading to infertility in women [3], we also studied the possible linkage between MUC4 SNPs and infertility. Though no significant difference was found in our genotype association study, our data indicated T allele at rs882605 as a protective factor that associated with reduced frequency of infertility in patients with endometriosis (Table 4). Two other alleles, C at rs2688513 and G at rs2246901, showed similar protective effects, but the data did not reach statistical significance. Of note, the major (also the wild-type) allele was used as the reference for the allelic analyses. For the genotypic analyses, the homozygous major allele genotype was used as the referent group. In our haplotype analyses, we thus sought to ascertain the impact of the genetic combination of these top three protective alleles. Table 5 indicates that patients with haplotype T-C-G did show a lower frequency of infertility, although the results did not show a statistically significant difference (P = 0.099). By contrast, haplotype G-T-T showed a strong association with infertility in patients (P = 0.012) (Table 5) (Additional file 2, Figure S2) and could be used as a risk indicator for patients at higher risk of developing severe complications such as infertility.\nGenotype and allele distributions of the six SNPs in the MUC4 gene in endometriosis patients with different reproductive abilitya\naSNP, single-nucleotide polymorphism; MUC4, mucin 4 gene; OR, odds ratio; 95% CI, 95% confidence interval. Allelic frequencies were determined by Fisher's exact test using 2 × 2 contingency tables. Genotype frequencies were determined by Fisher's exact test using 2 × 3 contingency tables. bP < 0.05 was considered statistically significant.\nHaplotype frequencies of MUC4 polymorphisms in endometriosis patients with infertilitya\naMUC4, mucin 4 gene. The reported haplotype percentages are estimated percentages based on allele frequencies and the linkage disequilibrium. The P values are based on a comparison of a given haplotype with all other haplotypes combined. bP < 0.05 was considered statistically significant.", "Because these endometriosis-associated SNPs can cause amino acid substitutions (Additional file 1, Table S2), the biofunctions of MUC4 might be altered by changes in hydrophilicity and protein folding. Figure 2 illustrates the functional domains in MUC4 protein sequence and secondary structures that contain these SNPs. Our data show that genetic variations of rs882605 and rs2688513 cause amino acid substitutions in long-loop regions (> 10 residues) between secondary structure units (α-helices or β-strands) with high hydrophilicity and moderate surface probability (Figure 2). Amino acid composition analyses also revealed that these two loops (300-319 and 4,134-4,158) contain several negatively charged aspartate (D) residues and reverse turn elements, glycine (G) and proline (P), suggesting the importance of these loops in protein folding and functional regulation [34,35]. In addition, rs2246901 locates in a type D von Willebrand factor (VWFD) domain responsible for protein-protein interaction and cell adhesion and/or migration [36,37]. Our findings support the functional roles of MUC4 SNPs in regulating cellular mobility and invasion of endometrial cells during endometriosis development and progression.\nFunctional domains in the MUC4 protein sequence and the predicted secondary structures. Six functional domains and/or signatures (boxes) were annotated by aligning MUC4 protein sequence in PROSITE protein domain database [28]. The boundaries of each signature are listed. Among six SNPs tested in this study (stars), rs882605 and rs2688513 (black) were found in two different long-loop regions, 300-319 and 4,134-4,158, respectively (bold letters refer to amino acid substitution sites). The SNP rs2246901 (gray) was found in a type D von Willebrand factor (VWFD) domain. The MUC4 reference sequence can be located at National Center for Biotechnology Information databank NP_060876.4.", "Previous studies have shown polymorphism of cytokines and adhesion molecules which were associated with the pathogenesis of endometriosis [38,39]. To the best of our knowledge, however, no other study to date has investigated the possible association of MUC4 and endometriosis. The purpose of this study was to evaluate whether genetic variations in MUC4 associate with endometriosis in the Taiwanese population. Our data prove the association of MUC4 polymorphisms with advanced stages of endometriosis and the related infertility. Since the extracellular domain of MUC4 is critical for HER2 interaction and cell invasiveness, these defined SNPs located in putative functional domains of MUC4 may play important roles during endometriosis development and progression.\nThe development of endometriosis and certain types of ovarian cancer share several similar clinical features. For example, endometriosis could progressively invade pelvic viscera, resulting in adhesion, and could recur after medical treatment or surgery. Because of the functions involved in the acquisition of adhesion ligands or receptors and the loss of antiadhesion, proteins such as MUC4 and MUC1, the two major mucins present in endometrial epithelium, thus become suspect in endometriosis development [16,17]. In addition to gene overexpression, genetic variations in MUC1 have also been reported as risk factors contributing to cell mobility and the severity of cancer [40-43]. However, the influence of MUC4 genetic variations on cell behavior remains unclear. In this study, Pro4135Ser (rs2688513) and Ala4693Ser (2246901) substitutions in the putative functional domains of MUC4 were found to be associated with advanced stages of endometriosis. Since MUC4 is an emerging target for ovarian cancer [18,19,44,45], our study provides a new direction from which to address the roles of MUC4 in the development of gynecological disorders.\nTo study the genetic effects of mucin proteins by SNPs, the major interest focuses on the variable number of tandem repeat (VNTR) polymorphisms, which result in different-sized gene transcripts. For examples, MUC1 variations in the VNTR domain have been found to play roles in regulating Helicobacter pylori binding to gastric cells [46]. Other studies have also concluded that VNTR polymorphisms can influence T-antigen presentation and the local immune responses, which consequently have potential effects on gastric cancer development [47,48]. With regard to MUC4, a high degree of polymorphism in the VNTR domain was observed in human tissues, including the endometrial epithelium [49,50]. However, the different-sized MUC4 transcripts did not show association with embryo implantation or cancer development. By contrast, MUC4 can promote cell proliferation and antiapoptotic effects in cancer cells by interacting with HER2 on the cell surface [18,19,22], suggesting the potency of functional domains in the extracellular domain of MUC4. In this study, two SNPs (rs2688513 and rs2246901) that locate in a putative functional loop and the VWFD protein binding domain, respectively, were found to be associated with advanced stages of endometriosis. Further study may clarify whether these amino acid substitutions could change the interaction with HER2 and/or play crucial roles in regulating the cellular activity of the spread of endometriosis.\nEndometriosis could cause pelvic adhesion and tubal occlusion that may lead to infertility. However, among patients with endometriosis-related infertility, 50% to 60% of them were diagnosed with minimal or mild endometriosis [3]. Impaired folliculogenesis, bad oocyte quality and impaired implantation of embryo are therefore considered to be the possible mechanisms for endometriosis-related infertility. Changes in cytokines and growth factors in endometrium, follicular fluid and peritoneal fluid have been suggested as the key players for inducing the above-mentioned phenomena [5]. Recently, several studies have shown that MUC4 could promote cell migration, change the endometrial environment and create weak points in the epithelium, thus facilitating the failure of embryo implantation [50,51]. The Carraway et al. [52] study also showed that embryo implantation was associated with downregulation of MUC4 expression in an animal model. In this study, women with a T allele in rs882605 had a lower risk of endometriosis-related infertility, whereas rs2688513 and rs2246901 SNPs did not show any association with the reproductive ability of patients. The rs882605 SNP locates in a putative functional loop within the VNTR domain of MUC4 that may control T-cell antigen presentation and the local immune responses. Our findings may support the view that the regulation of local immunity, rather than uncontrolled cell proliferation, in the endometrium may play a more important role in the development of endometriosis-related infertility.\nOur study shows that the T/G genotype at rs882605, as compared with the T/T or G/G genotypes, is unique in patients with endometriosis. So far, our data cannot provide sufficient information to explain why the T/T genotype does not show higher risk of endometriosis than T/G. One reason could be the relatively small study group, while other possibilities could exist. For example, the SNPs analyzed are in tight LD with other unknown allele variants which have an opposite effect. In this case, only individuals with the T/G heterozygous genotype could be observed. Because this is a hospital-based study with a modest sample size, enrollment of a larger cohort based on a population approach could help to elucidate the functional role of MUC4 in endometriosis and the related infertility.", "In this study, we observed an association of MUC4 polymorphism with endometriosis development and endometriosis-related infertility in a Taiwanese population. However, the true mechanism of how MUC4 modulates the pathogenesis of endometriosis and infertility was not clearly understood. In addition, the risk SNPs pose for endometriosis stages and infertility differ, suggesting dissimilar molecular mechanisms for these clinical features. More detailed studies are needed to investigate the biochemical pathways regulated by MUC4 during the development of endometriosis.", "The authors declare that they have no competing interests.", "CYYC collected samples, carried out the clinical association study and drafted the manuscript. HWC participated in sample collection and the clinical association study. CMC carried out sample preparation and SNP analyses. CYL participated in SNP analyses. CPC participated in the clinical association study. CHL carried out statistical analyses. WYL carried out sample pretreatment and RNA extraction. JJCS carried out the experimental design and drafted the manuscript. FJT carried out the experimental design and revised the manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1741-7015/9/19/prepub\n", "Supplementary Tables S1 to S3.\nClick here for file\nSupplementary Figures S1 and S2.\nClick here for file" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Amino Acid Substitution", "Arginine", "Biomarkers, Tumor", "Carcinoma", "Case-Control Studies", "Cell Transformation, Neoplastic", "Disease Progression", "Glycine", "Humans", "Male", "Polymorphism, Single Nucleotide", "Prostatic Neoplasms", "Receptor, Fibroblast Growth Factor, Type 4" ]

[ "Background", "Publication search", "Inclusion, exclusion criteria and data abstraction", "Statistical analysis", "Results", "Study characteristics", "Quantitative synthesis", "Sensitivity analysis", "Publication bias", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Prostate cancer is the most frequently diagnosed solid tumor and the second leading cause of cancer-related death among American men, with an estimated 192,280 new cases and 27,360 deaths in the United States in 2009[1]. The etiology of human prostate cancer is complex and largely remains unknown.\nFibroblast growth factor receptor 4 (FGFR4) belongs to the family of fibroblast growth factor receptors (FGFR1-4), which display multiple biological activities, including mitogenic and angiogenic activity, with a consequent crucial role in cell differentiation, development, hormonal and proliferative signaling in response to more than 20 known ligands[2,3]. In light of its involvement in the regulation of essential biologic mechanisms, FGF signaling is also likely to play a role in tumor growth and progression; indeed, dysreglation of this pathway has been demonstrated in several tumor types[3]. Recently, FGFR4 was found to be more abundantly expressed in malignant than benign prostate cells and in vitro suppression of FGFR4 expression effectively blocked prostate cancer proliferation and invasion[4]. Moreover, strong expression of FGFR4 in prostate cancer cells, as assessed by immunohistochemistry, is significantly associated with increased clinical stage and tumor grade and decreased patient survival[5].\nA germ line polymorphism in FGFR4 gene, resulting in different expression of FGFR4 containing either glycine (Gly388) or arginine (Arg388) at codon 388 in the transmembrane domain of the receptor was identified several years ago. In addition, the FGFR4 Arg388 allele may predispose cancer patients to disease progression, based on the reported significant association between FGFR4 genotype and tumor aggressiveness or patients' survival in several cancers[6,7].\nTo date, several studies had been reported to focus on the association between this polymorphism and incidence and aggressiveness of prostate cancer[4,8-12]. However, a single study may be too underpowered to detect a possible small effect of the polymorphism on prostate cancer, especially when the sample size is relatively small. Hence, we carried out a meta-analysis of all eligible case-control studies to derive a more precise estimation of the association of FGFR4 Gly388Arg polymorphism with prostate cancer.", "PubMed and EMBASE were searched (the last search update on the 1st Nov. 2010) using the search terms: 'FGFR4 or fibroblast growth factor receptor 4', 'polymorphism', 'Gly388Arg or rs351855' and 'prostate cancer or prostate neoplasm'. All published English language papers with available full text matching the eligible criteria were retrieved. In addition, we checked all the references of relevant reviews and eligible articles that our search retrieved. Two investigators (BX and SQC) searched the literature and extracted data independently.", "For inclusion in the meta-analysis, the identified articles had to provide information on: (1) evaluation of FGFR4 Gly388Arg polymorphism and prostate cancer risk, (2) using a case-control design and (3) containing information about available genotype frequency that can help infer the results in the papers. Major reasons for the exclusion of studies were: (1) no control population; (2) no usable data reported; (3) duplicates. For each of the eligible case-control studies, the following data were collected: the first author's last name, year of publication, country of origin, ethnicity, numbers of genotyped cases and controls, genotyping methods.", "The strength of the association between the FGFR4 Gly388Arg polymorphism and prostate cancer risk was measured by ORs with 95% confidence intervals (CIs). We explored the association between allele Arg388 and prostate cancer development and progression, as well as homozygote comparison (Arg/Arg vs. Gly/Gly), dominant genetic model [(Gly/Arg+Arg/Arg) vs. Gly/Gly] and recessive model [Arg/Arg vs. (Gly/Gly+ Gly/Arg)]. Heterogeneity assumption was checked by a chi-square-based Q-test[13]. A P-value of more than 0.05 for the Q-test indicated a lack of heterogeneity among the studies, so the summary OR estimate of each study was calculated by the fixed-effects model (the Mantel-Haenszel method). Otherwise, the random effects model (DerSimonian and Laird method) was used[14,15]. The significance of the pooled OR was determined by the Z-test, and P < 0.05 was considered as statistically significant. To evaluate the ethnic-specific effect, subgroup analysis was conducted on the basis of different ethnicities.\nEvidence of publication bias was determined using Begg's[16] and Egger's[17] formal statistical test and by visual inspection of the funnel plot. All statistical analyses were performed with Stata software (version 10.0; StataCorp LP, College Station, TX), using two-sided P values.", "[SUBTITLE] Study characteristics [SUBSECTION] Using the searching terms, seven papers were reviewed in the two online databases. The study of Spinola et al.[18] was focused on the association between FGFR4 Gly388Arg polymorphism and lung cancer risk, and the studies of Wang et al.[12] and Sahadevan et al.[4] were not epidemiological association studies, so they were all excluded in present study. In the four papers left, FitzGerald et al.[8] and Wang et al.[11] provided data on both African-American and Caucasian. Overall, four articles (six studies) with 2618 prostate cancer cases and 2305 controls were retrieved based on the search criteria for prostate cancer susceptibility related to the FGFR4 Gly388Arg polymorphism. Study characteristics are summarized in Table 1. All the studies used frequency-matched controls to the cases by the age, sex or ethnicity, and the distribution of genotypes in the controls was consistent with Hardy-Weinberg equilibrium in all studies. Moreover, among the four articles, three[8,10,11] mentioned the association between FGFR4 Gly388Arg polymorphism and progression of prostate cancer. The stratifications for pathological parameters of the cases in the three articles were also shown in Table 1. The cases of the three articles were all stratified by Gleason score and tumor stage. However, the classification standard of Gleason score was not uniform; thus, we only focused on the association between FGFR4 Gly388Arg polymorphism and tumor stage (advanced vs. localized). Advanced stage corresponded to T3 stage in the study of Wang et al.[11], regional/distant stage in the study of FitzGerald et al.[8], and stage C+D in the study of Ma et al.[10], respectively. And localized stage meant T2 stage in the study of Wang et al., local stage in the study of FitzGerald et al., and stage A+B in the study of Ma et al., respectively.\nMain characteristics of selected studies\na Stage A (T1a-bN0M0), StageB (T1c-2N0M0), Stage C (T3-4N0M0) and Stage D (T1-4N1M0-1 orT1-4N0-1M1) by the modified Whitmore-Jewett system.\nUsing the searching terms, seven papers were reviewed in the two online databases. The study of Spinola et al.[18] was focused on the association between FGFR4 Gly388Arg polymorphism and lung cancer risk, and the studies of Wang et al.[12] and Sahadevan et al.[4] were not epidemiological association studies, so they were all excluded in present study. In the four papers left, FitzGerald et al.[8] and Wang et al.[11] provided data on both African-American and Caucasian. Overall, four articles (six studies) with 2618 prostate cancer cases and 2305 controls were retrieved based on the search criteria for prostate cancer susceptibility related to the FGFR4 Gly388Arg polymorphism. Study characteristics are summarized in Table 1. All the studies used frequency-matched controls to the cases by the age, sex or ethnicity, and the distribution of genotypes in the controls was consistent with Hardy-Weinberg equilibrium in all studies. Moreover, among the four articles, three[8,10,11] mentioned the association between FGFR4 Gly388Arg polymorphism and progression of prostate cancer. The stratifications for pathological parameters of the cases in the three articles were also shown in Table 1. The cases of the three articles were all stratified by Gleason score and tumor stage. However, the classification standard of Gleason score was not uniform; thus, we only focused on the association between FGFR4 Gly388Arg polymorphism and tumor stage (advanced vs. localized). Advanced stage corresponded to T3 stage in the study of Wang et al.[11], regional/distant stage in the study of FitzGerald et al.[8], and stage C+D in the study of Ma et al.[10], respectively. And localized stage meant T2 stage in the study of Wang et al., local stage in the study of FitzGerald et al., and stage A+B in the study of Ma et al., respectively.\nMain characteristics of selected studies\na Stage A (T1a-bN0M0), StageB (T1c-2N0M0), Stage C (T3-4N0M0) and Stage D (T1-4N1M0-1 orT1-4N0-1M1) by the modified Whitmore-Jewett system.\n[SUBTITLE] Quantitative synthesis [SUBSECTION] We observed a wide variation of Arg388 allele frequencies across different ethnicities. The frequency of Arg388 allele was 28.90% among Caucasian controls and 41.57% among Asian controls, which were significantly higher than that in African-American controls (11.49%, P < 0.01).\nOverall, the combined result based on all studies showed the evidence of an association between the increased risk of prostate cancer and the variant genotypes in different genetic models. As shown in Table 2 and Figure 1, the Arg388 allele increased overall prostate cancer risk compared with Gly388 allele (OR = 1.17, 95% CI = 1.07-1.29). Significant main effects were also observed in dominate genetic model (OR = 1.21, 95% CI = 1.08-1.36).\nStratified analyses of the FGFR4 Gly388Arg polymorphism on cancer risk\naNumber of comparisons; bP value of Q-test for heterogeneity test; cRandom-effects model was used when P value for heterogeneity test < 0.05; otherwise, fixed-effects model was used;\nForest plot of prostate cancer risk associated with FGFR4 Gly388Arg polymorphism (Arg allele vs. Gly allele). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\nWhen stratifying for race, results were similar. Specially, significantly increased risk was found among Caucasian populations (Arg388 and Gly388 comparison: OR = 1.21, 95% CI: 1.00-1.47; dominant genetic model: OR = 1.23, 95% CI: 1.08-1.40) and Asian population (Arg388 and Gly388 comparison: OR = 1.24, 95% CI: 1.02-1.51; homozygote comparison: OR = 1.52, 95% CI: 1.05-2.22; recessive genetic model: OR = 1.53, 95% CI: 1.10-2.14). Although the effect in African-Americans was in the same direction as for other groups, the difference was not statistically significant (Arg388 and Gly388 comparison: OR = 1.15, 95% CI: 0.73-1.82; homozygote comparison: OR = 2.17, 95% CI: 0.20-23.14; dominant genetic model: OR = 1.11, 95% CI: 0.66-1.86 and recessive genetic model: OR = 2.21, 95% CI: 0.18-26.83).\nIn addition, when concerning tumor stage and FGFR4 Gly388Arg polymorphism, patients with prostate cancer with Arg/Arg genotype had a 1.34-fold increased risk of advanced or metastatic prostate cancer (95% CI: 1.03-1.74) compared with the Gly/Gly+Gly/Arg genotype (seen Figure 2).\nForest plot of prostate cancer progression associated with FGFR4 Gly388Arg polymorphism (Arg/Arg vs. Gly/Gly+Gly/Arg). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\nWe observed a wide variation of Arg388 allele frequencies across different ethnicities. The frequency of Arg388 allele was 28.90% among Caucasian controls and 41.57% among Asian controls, which were significantly higher than that in African-American controls (11.49%, P < 0.01).\nOverall, the combined result based on all studies showed the evidence of an association between the increased risk of prostate cancer and the variant genotypes in different genetic models. As shown in Table 2 and Figure 1, the Arg388 allele increased overall prostate cancer risk compared with Gly388 allele (OR = 1.17, 95% CI = 1.07-1.29). Significant main effects were also observed in dominate genetic model (OR = 1.21, 95% CI = 1.08-1.36).\nStratified analyses of the FGFR4 Gly388Arg polymorphism on cancer risk\naNumber of comparisons; bP value of Q-test for heterogeneity test; cRandom-effects model was used when P value for heterogeneity test < 0.05; otherwise, fixed-effects model was used;\nForest plot of prostate cancer risk associated with FGFR4 Gly388Arg polymorphism (Arg allele vs. Gly allele). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\nWhen stratifying for race, results were similar. Specially, significantly increased risk was found among Caucasian populations (Arg388 and Gly388 comparison: OR = 1.21, 95% CI: 1.00-1.47; dominant genetic model: OR = 1.23, 95% CI: 1.08-1.40) and Asian population (Arg388 and Gly388 comparison: OR = 1.24, 95% CI: 1.02-1.51; homozygote comparison: OR = 1.52, 95% CI: 1.05-2.22; recessive genetic model: OR = 1.53, 95% CI: 1.10-2.14). Although the effect in African-Americans was in the same direction as for other groups, the difference was not statistically significant (Arg388 and Gly388 comparison: OR = 1.15, 95% CI: 0.73-1.82; homozygote comparison: OR = 2.17, 95% CI: 0.20-23.14; dominant genetic model: OR = 1.11, 95% CI: 0.66-1.86 and recessive genetic model: OR = 2.21, 95% CI: 0.18-26.83).\nIn addition, when concerning tumor stage and FGFR4 Gly388Arg polymorphism, patients with prostate cancer with Arg/Arg genotype had a 1.34-fold increased risk of advanced or metastatic prostate cancer (95% CI: 1.03-1.74) compared with the Gly/Gly+Gly/Arg genotype (seen Figure 2).\nForest plot of prostate cancer progression associated with FGFR4 Gly388Arg polymorphism (Arg/Arg vs. Gly/Gly+Gly/Arg). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\n[SUBTITLE] Sensitivity analysis [SUBSECTION] Sensitivity analysis was performed by sequential omission of individual studies. The pooled 95% CI for Arg388 vs. Gly388 was consistently over 1.0, indicating that the results of this meta-analysis are stable.\nSensitivity analysis was performed by sequential omission of individual studies. The pooled 95% CI for Arg388 vs. Gly388 was consistently over 1.0, indicating that the results of this meta-analysis are stable.\n[SUBTITLE] Publication bias [SUBSECTION] Begg's funnel plot and Egger's test were performed to assess the publication bias. The shape of the funnel plots seemed symmetrical in the comparison of the Arg388 vs. Gly388 (Figure 3). Furthermore, Egger's test was used to provide statistical evidence for funnel plot symmetry (t = 1.30, P = 0.26), suggesting that no publication bias was exist.\nBegg's funnel plot for publication bias test. Each point represents a separate study for the indicated association. Log[OR], natural logarithm of OR. Horizontal line, mean effect size.\nBegg's funnel plot and Egger's test were performed to assess the publication bias. The shape of the funnel plots seemed symmetrical in the comparison of the Arg388 vs. Gly388 (Figure 3). Furthermore, Egger's test was used to provide statistical evidence for funnel plot symmetry (t = 1.30, P = 0.26), suggesting that no publication bias was exist.\nBegg's funnel plot for publication bias test. Each point represents a separate study for the indicated association. Log[OR], natural logarithm of OR. Horizontal line, mean effect size.", "Using the searching terms, seven papers were reviewed in the two online databases. The study of Spinola et al.[18] was focused on the association between FGFR4 Gly388Arg polymorphism and lung cancer risk, and the studies of Wang et al.[12] and Sahadevan et al.[4] were not epidemiological association studies, so they were all excluded in present study. In the four papers left, FitzGerald et al.[8] and Wang et al.[11] provided data on both African-American and Caucasian. Overall, four articles (six studies) with 2618 prostate cancer cases and 2305 controls were retrieved based on the search criteria for prostate cancer susceptibility related to the FGFR4 Gly388Arg polymorphism. Study characteristics are summarized in Table 1. All the studies used frequency-matched controls to the cases by the age, sex or ethnicity, and the distribution of genotypes in the controls was consistent with Hardy-Weinberg equilibrium in all studies. Moreover, among the four articles, three[8,10,11] mentioned the association between FGFR4 Gly388Arg polymorphism and progression of prostate cancer. The stratifications for pathological parameters of the cases in the three articles were also shown in Table 1. The cases of the three articles were all stratified by Gleason score and tumor stage. However, the classification standard of Gleason score was not uniform; thus, we only focused on the association between FGFR4 Gly388Arg polymorphism and tumor stage (advanced vs. localized). Advanced stage corresponded to T3 stage in the study of Wang et al.[11], regional/distant stage in the study of FitzGerald et al.[8], and stage C+D in the study of Ma et al.[10], respectively. And localized stage meant T2 stage in the study of Wang et al., local stage in the study of FitzGerald et al., and stage A+B in the study of Ma et al., respectively.\nMain characteristics of selected studies\na Stage A (T1a-bN0M0), StageB (T1c-2N0M0), Stage C (T3-4N0M0) and Stage D (T1-4N1M0-1 orT1-4N0-1M1) by the modified Whitmore-Jewett system.", "We observed a wide variation of Arg388 allele frequencies across different ethnicities. The frequency of Arg388 allele was 28.90% among Caucasian controls and 41.57% among Asian controls, which were significantly higher than that in African-American controls (11.49%, P < 0.01).\nOverall, the combined result based on all studies showed the evidence of an association between the increased risk of prostate cancer and the variant genotypes in different genetic models. As shown in Table 2 and Figure 1, the Arg388 allele increased overall prostate cancer risk compared with Gly388 allele (OR = 1.17, 95% CI = 1.07-1.29). Significant main effects were also observed in dominate genetic model (OR = 1.21, 95% CI = 1.08-1.36).\nStratified analyses of the FGFR4 Gly388Arg polymorphism on cancer risk\naNumber of comparisons; bP value of Q-test for heterogeneity test; cRandom-effects model was used when P value for heterogeneity test < 0.05; otherwise, fixed-effects model was used;\nForest plot of prostate cancer risk associated with FGFR4 Gly388Arg polymorphism (Arg allele vs. Gly allele). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\nWhen stratifying for race, results were similar. Specially, significantly increased risk was found among Caucasian populations (Arg388 and Gly388 comparison: OR = 1.21, 95% CI: 1.00-1.47; dominant genetic model: OR = 1.23, 95% CI: 1.08-1.40) and Asian population (Arg388 and Gly388 comparison: OR = 1.24, 95% CI: 1.02-1.51; homozygote comparison: OR = 1.52, 95% CI: 1.05-2.22; recessive genetic model: OR = 1.53, 95% CI: 1.10-2.14). Although the effect in African-Americans was in the same direction as for other groups, the difference was not statistically significant (Arg388 and Gly388 comparison: OR = 1.15, 95% CI: 0.73-1.82; homozygote comparison: OR = 2.17, 95% CI: 0.20-23.14; dominant genetic model: OR = 1.11, 95% CI: 0.66-1.86 and recessive genetic model: OR = 2.21, 95% CI: 0.18-26.83).\nIn addition, when concerning tumor stage and FGFR4 Gly388Arg polymorphism, patients with prostate cancer with Arg/Arg genotype had a 1.34-fold increased risk of advanced or metastatic prostate cancer (95% CI: 1.03-1.74) compared with the Gly/Gly+Gly/Arg genotype (seen Figure 2).\nForest plot of prostate cancer progression associated with FGFR4 Gly388Arg polymorphism (Arg/Arg vs. Gly/Gly+Gly/Arg). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.", "Sensitivity analysis was performed by sequential omission of individual studies. The pooled 95% CI for Arg388 vs. Gly388 was consistently over 1.0, indicating that the results of this meta-analysis are stable.", "Begg's funnel plot and Egger's test were performed to assess the publication bias. The shape of the funnel plots seemed symmetrical in the comparison of the Arg388 vs. Gly388 (Figure 3). Furthermore, Egger's test was used to provide statistical evidence for funnel plot symmetry (t = 1.30, P = 0.26), suggesting that no publication bias was exist.\nBegg's funnel plot for publication bias test. Each point represents a separate study for the indicated association. Log[OR], natural logarithm of OR. Horizontal line, mean effect size.", "The present meta-analysis, including 2,618 cases and 2,305 controls from six published studies, explored the association between FGFR4 Gly388Arg polymorphism and development and progression of prostate cancer. To the best of our knowledge, this is the first meta-analysis to explore FGFR4 Gly388Arg polymorphism in development and progression of prostate cancer. The results indicated that FGFR4 Arg388 allele is a potential risk factor for developing and progressing prostate cancer. These findings may be biologically plausible. The FGFR4 Gly388Arg polymorphism results in an amino acid change in the transmembrane domain of the receptor, which may alter the activity of the receptor. FGFR4 is the activator of the MAPK signaling cascade, yet it is a principal receptor for key mitogenic FGFs in prostate cancer cells[19-21]. There was evidence that FGFR4 contributed to progression in liver, lung, colon tumors[22] and prostate cancer[12]. The effects of FGFR4 Arg388 allele may also predispose cancer patients to disease progression, based on the reported significant association between FGFR4 genotype and tumor aggressiveness (lymph node involvement, advanced stage) or patients' survival[6,7], and the results about its biological role on cancer cell motility and invasiveness[11]. In our meta-analysis, we found that subjects carrying Arg388 were associated with higher risk of developing and progressing prostate cancer than those with the wild-type allele, which confirmed the hypothesis described above.\nSome limitations of this meta-analysis should be acknowledged. First of all, the control populations were not uniform. Healthy populations as well as non-cancer patients like BPH patients were included. Some individuals in the control group are likely to develop cancer in subsequent years though they had no clinical symptoms at the time of investigation. Misclassification bias results in deviation of genotype distribution in the controls. Second, prostate cancer, as a complex disease, was considered as the result of combined effects of multi-factors, including inherited and environmental factors[23], however, no such data was observed in previous studies. Thus, our result was only based on unadjusted estimates. Lacking of the information for the data analysis may cause serious confounding bias. Third, the effect of the polymorphism was relatively trivial with small ORs. We need further studies with larger number participants to confirm the effect.\nOur meta-analysis also had some advantages. First, disease progression status as tumor stage was taken into account in present study. Second, data in present study were pooled from different studies, which significantly increased statistical power of the analysis. Third, the quality of studies included in our meta-analysis was satisfactory and cruelly met our inclusion criterion. Fourth, the distribution of genotypes in the controls was consistent with Hardy-Weinberg equilibrium in all studies. We further performed sensitivity analysis to detect the stability of the meta-analysis, and the results did not alter the pattern of association and revealed that the risk effect of Arg388 was stable. In addition, publication bias was not detected in present study, indicating that our findings seemed not to be due to biased publications.", "Our meta-analysis showed the evidence that FGFR4 Arg388 allele was associated with an increased risk of prostate cancer development and progression, suggesting that FGFR4 Gly388Arg polymorphism could be a marker for prostate cancer development and progression. Based on the limitations of present study list above, further prospective researches using standardized unbiased methods, and larger numbers of worldwide participants are expected to examine the association to confirm our results, and other possible confounding risk factors like age, life style, and familial history should also be controlled when it was assessed. Moreover, gene-gene and gene-environment interactions should also be considered.", "FGFR4: Fibroblast growth factor receptor 4; Gly: glycine; Arg: arginine; OR: odds ratio; CI: confidence interval.", "The authors declare that they have no competing interests.", "BX participated in study design and drafted the manuscript.\nSQC participated in collection of data and manuscript preparation.\nNT and ZDZ performed the statistical analysis and participated in the critical revision of the manuscript.\nZJW critically revised the manuscript.\nMC and LXH participated in its design.\nAll authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2407/11/84/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Publication search", "Inclusion, exclusion criteria and data abstraction", "Statistical analysis", "Results", "Study characteristics", "Quantitative synthesis", "Sensitivity analysis", "Publication bias", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Prostate cancer is the most frequently diagnosed solid tumor and the second leading cause of cancer-related death among American men, with an estimated 192,280 new cases and 27,360 deaths in the United States in 2009[1]. The etiology of human prostate cancer is complex and largely remains unknown.\nFibroblast growth factor receptor 4 (FGFR4) belongs to the family of fibroblast growth factor receptors (FGFR1-4), which display multiple biological activities, including mitogenic and angiogenic activity, with a consequent crucial role in cell differentiation, development, hormonal and proliferative signaling in response to more than 20 known ligands[2,3]. In light of its involvement in the regulation of essential biologic mechanisms, FGF signaling is also likely to play a role in tumor growth and progression; indeed, dysreglation of this pathway has been demonstrated in several tumor types[3]. Recently, FGFR4 was found to be more abundantly expressed in malignant than benign prostate cells and in vitro suppression of FGFR4 expression effectively blocked prostate cancer proliferation and invasion[4]. Moreover, strong expression of FGFR4 in prostate cancer cells, as assessed by immunohistochemistry, is significantly associated with increased clinical stage and tumor grade and decreased patient survival[5].\nA germ line polymorphism in FGFR4 gene, resulting in different expression of FGFR4 containing either glycine (Gly388) or arginine (Arg388) at codon 388 in the transmembrane domain of the receptor was identified several years ago. In addition, the FGFR4 Arg388 allele may predispose cancer patients to disease progression, based on the reported significant association between FGFR4 genotype and tumor aggressiveness or patients' survival in several cancers[6,7].\nTo date, several studies had been reported to focus on the association between this polymorphism and incidence and aggressiveness of prostate cancer[4,8-12]. However, a single study may be too underpowered to detect a possible small effect of the polymorphism on prostate cancer, especially when the sample size is relatively small. Hence, we carried out a meta-analysis of all eligible case-control studies to derive a more precise estimation of the association of FGFR4 Gly388Arg polymorphism with prostate cancer.", "[SUBTITLE] Publication search [SUBSECTION] PubMed and EMBASE were searched (the last search update on the 1st Nov. 2010) using the search terms: 'FGFR4 or fibroblast growth factor receptor 4', 'polymorphism', 'Gly388Arg or rs351855' and 'prostate cancer or prostate neoplasm'. All published English language papers with available full text matching the eligible criteria were retrieved. In addition, we checked all the references of relevant reviews and eligible articles that our search retrieved. Two investigators (BX and SQC) searched the literature and extracted data independently.\nPubMed and EMBASE were searched (the last search update on the 1st Nov. 2010) using the search terms: 'FGFR4 or fibroblast growth factor receptor 4', 'polymorphism', 'Gly388Arg or rs351855' and 'prostate cancer or prostate neoplasm'. All published English language papers with available full text matching the eligible criteria were retrieved. In addition, we checked all the references of relevant reviews and eligible articles that our search retrieved. Two investigators (BX and SQC) searched the literature and extracted data independently.\n[SUBTITLE] Inclusion, exclusion criteria and data abstraction [SUBSECTION] For inclusion in the meta-analysis, the identified articles had to provide information on: (1) evaluation of FGFR4 Gly388Arg polymorphism and prostate cancer risk, (2) using a case-control design and (3) containing information about available genotype frequency that can help infer the results in the papers. Major reasons for the exclusion of studies were: (1) no control population; (2) no usable data reported; (3) duplicates. For each of the eligible case-control studies, the following data were collected: the first author's last name, year of publication, country of origin, ethnicity, numbers of genotyped cases and controls, genotyping methods.\nFor inclusion in the meta-analysis, the identified articles had to provide information on: (1) evaluation of FGFR4 Gly388Arg polymorphism and prostate cancer risk, (2) using a case-control design and (3) containing information about available genotype frequency that can help infer the results in the papers. Major reasons for the exclusion of studies were: (1) no control population; (2) no usable data reported; (3) duplicates. For each of the eligible case-control studies, the following data were collected: the first author's last name, year of publication, country of origin, ethnicity, numbers of genotyped cases and controls, genotyping methods.\n[SUBTITLE] Statistical analysis [SUBSECTION] The strength of the association between the FGFR4 Gly388Arg polymorphism and prostate cancer risk was measured by ORs with 95% confidence intervals (CIs). We explored the association between allele Arg388 and prostate cancer development and progression, as well as homozygote comparison (Arg/Arg vs. Gly/Gly), dominant genetic model [(Gly/Arg+Arg/Arg) vs. Gly/Gly] and recessive model [Arg/Arg vs. (Gly/Gly+ Gly/Arg)]. Heterogeneity assumption was checked by a chi-square-based Q-test[13]. A P-value of more than 0.05 for the Q-test indicated a lack of heterogeneity among the studies, so the summary OR estimate of each study was calculated by the fixed-effects model (the Mantel-Haenszel method). Otherwise, the random effects model (DerSimonian and Laird method) was used[14,15]. The significance of the pooled OR was determined by the Z-test, and P < 0.05 was considered as statistically significant. To evaluate the ethnic-specific effect, subgroup analysis was conducted on the basis of different ethnicities.\nEvidence of publication bias was determined using Begg's[16] and Egger's[17] formal statistical test and by visual inspection of the funnel plot. All statistical analyses were performed with Stata software (version 10.0; StataCorp LP, College Station, TX), using two-sided P values.\nThe strength of the association between the FGFR4 Gly388Arg polymorphism and prostate cancer risk was measured by ORs with 95% confidence intervals (CIs). We explored the association between allele Arg388 and prostate cancer development and progression, as well as homozygote comparison (Arg/Arg vs. Gly/Gly), dominant genetic model [(Gly/Arg+Arg/Arg) vs. Gly/Gly] and recessive model [Arg/Arg vs. (Gly/Gly+ Gly/Arg)]. Heterogeneity assumption was checked by a chi-square-based Q-test[13]. A P-value of more than 0.05 for the Q-test indicated a lack of heterogeneity among the studies, so the summary OR estimate of each study was calculated by the fixed-effects model (the Mantel-Haenszel method). Otherwise, the random effects model (DerSimonian and Laird method) was used[14,15]. The significance of the pooled OR was determined by the Z-test, and P < 0.05 was considered as statistically significant. To evaluate the ethnic-specific effect, subgroup analysis was conducted on the basis of different ethnicities.\nEvidence of publication bias was determined using Begg's[16] and Egger's[17] formal statistical test and by visual inspection of the funnel plot. All statistical analyses were performed with Stata software (version 10.0; StataCorp LP, College Station, TX), using two-sided P values.", "PubMed and EMBASE were searched (the last search update on the 1st Nov. 2010) using the search terms: 'FGFR4 or fibroblast growth factor receptor 4', 'polymorphism', 'Gly388Arg or rs351855' and 'prostate cancer or prostate neoplasm'. All published English language papers with available full text matching the eligible criteria were retrieved. In addition, we checked all the references of relevant reviews and eligible articles that our search retrieved. Two investigators (BX and SQC) searched the literature and extracted data independently.", "For inclusion in the meta-analysis, the identified articles had to provide information on: (1) evaluation of FGFR4 Gly388Arg polymorphism and prostate cancer risk, (2) using a case-control design and (3) containing information about available genotype frequency that can help infer the results in the papers. Major reasons for the exclusion of studies were: (1) no control population; (2) no usable data reported; (3) duplicates. For each of the eligible case-control studies, the following data were collected: the first author's last name, year of publication, country of origin, ethnicity, numbers of genotyped cases and controls, genotyping methods.", "The strength of the association between the FGFR4 Gly388Arg polymorphism and prostate cancer risk was measured by ORs with 95% confidence intervals (CIs). We explored the association between allele Arg388 and prostate cancer development and progression, as well as homozygote comparison (Arg/Arg vs. Gly/Gly), dominant genetic model [(Gly/Arg+Arg/Arg) vs. Gly/Gly] and recessive model [Arg/Arg vs. (Gly/Gly+ Gly/Arg)]. Heterogeneity assumption was checked by a chi-square-based Q-test[13]. A P-value of more than 0.05 for the Q-test indicated a lack of heterogeneity among the studies, so the summary OR estimate of each study was calculated by the fixed-effects model (the Mantel-Haenszel method). Otherwise, the random effects model (DerSimonian and Laird method) was used[14,15]. The significance of the pooled OR was determined by the Z-test, and P < 0.05 was considered as statistically significant. To evaluate the ethnic-specific effect, subgroup analysis was conducted on the basis of different ethnicities.\nEvidence of publication bias was determined using Begg's[16] and Egger's[17] formal statistical test and by visual inspection of the funnel plot. All statistical analyses were performed with Stata software (version 10.0; StataCorp LP, College Station, TX), using two-sided P values.", "[SUBTITLE] Study characteristics [SUBSECTION] Using the searching terms, seven papers were reviewed in the two online databases. The study of Spinola et al.[18] was focused on the association between FGFR4 Gly388Arg polymorphism and lung cancer risk, and the studies of Wang et al.[12] and Sahadevan et al.[4] were not epidemiological association studies, so they were all excluded in present study. In the four papers left, FitzGerald et al.[8] and Wang et al.[11] provided data on both African-American and Caucasian. Overall, four articles (six studies) with 2618 prostate cancer cases and 2305 controls were retrieved based on the search criteria for prostate cancer susceptibility related to the FGFR4 Gly388Arg polymorphism. Study characteristics are summarized in Table 1. All the studies used frequency-matched controls to the cases by the age, sex or ethnicity, and the distribution of genotypes in the controls was consistent with Hardy-Weinberg equilibrium in all studies. Moreover, among the four articles, three[8,10,11] mentioned the association between FGFR4 Gly388Arg polymorphism and progression of prostate cancer. The stratifications for pathological parameters of the cases in the three articles were also shown in Table 1. The cases of the three articles were all stratified by Gleason score and tumor stage. However, the classification standard of Gleason score was not uniform; thus, we only focused on the association between FGFR4 Gly388Arg polymorphism and tumor stage (advanced vs. localized). Advanced stage corresponded to T3 stage in the study of Wang et al.[11], regional/distant stage in the study of FitzGerald et al.[8], and stage C+D in the study of Ma et al.[10], respectively. And localized stage meant T2 stage in the study of Wang et al., local stage in the study of FitzGerald et al., and stage A+B in the study of Ma et al., respectively.\nMain characteristics of selected studies\na Stage A (T1a-bN0M0), StageB (T1c-2N0M0), Stage C (T3-4N0M0) and Stage D (T1-4N1M0-1 orT1-4N0-1M1) by the modified Whitmore-Jewett system.\nUsing the searching terms, seven papers were reviewed in the two online databases. The study of Spinola et al.[18] was focused on the association between FGFR4 Gly388Arg polymorphism and lung cancer risk, and the studies of Wang et al.[12] and Sahadevan et al.[4] were not epidemiological association studies, so they were all excluded in present study. In the four papers left, FitzGerald et al.[8] and Wang et al.[11] provided data on both African-American and Caucasian. Overall, four articles (six studies) with 2618 prostate cancer cases and 2305 controls were retrieved based on the search criteria for prostate cancer susceptibility related to the FGFR4 Gly388Arg polymorphism. Study characteristics are summarized in Table 1. All the studies used frequency-matched controls to the cases by the age, sex or ethnicity, and the distribution of genotypes in the controls was consistent with Hardy-Weinberg equilibrium in all studies. Moreover, among the four articles, three[8,10,11] mentioned the association between FGFR4 Gly388Arg polymorphism and progression of prostate cancer. The stratifications for pathological parameters of the cases in the three articles were also shown in Table 1. The cases of the three articles were all stratified by Gleason score and tumor stage. However, the classification standard of Gleason score was not uniform; thus, we only focused on the association between FGFR4 Gly388Arg polymorphism and tumor stage (advanced vs. localized). Advanced stage corresponded to T3 stage in the study of Wang et al.[11], regional/distant stage in the study of FitzGerald et al.[8], and stage C+D in the study of Ma et al.[10], respectively. And localized stage meant T2 stage in the study of Wang et al., local stage in the study of FitzGerald et al., and stage A+B in the study of Ma et al., respectively.\nMain characteristics of selected studies\na Stage A (T1a-bN0M0), StageB (T1c-2N0M0), Stage C (T3-4N0M0) and Stage D (T1-4N1M0-1 orT1-4N0-1M1) by the modified Whitmore-Jewett system.\n[SUBTITLE] Quantitative synthesis [SUBSECTION] We observed a wide variation of Arg388 allele frequencies across different ethnicities. The frequency of Arg388 allele was 28.90% among Caucasian controls and 41.57% among Asian controls, which were significantly higher than that in African-American controls (11.49%, P < 0.01).\nOverall, the combined result based on all studies showed the evidence of an association between the increased risk of prostate cancer and the variant genotypes in different genetic models. As shown in Table 2 and Figure 1, the Arg388 allele increased overall prostate cancer risk compared with Gly388 allele (OR = 1.17, 95% CI = 1.07-1.29). Significant main effects were also observed in dominate genetic model (OR = 1.21, 95% CI = 1.08-1.36).\nStratified analyses of the FGFR4 Gly388Arg polymorphism on cancer risk\naNumber of comparisons; bP value of Q-test for heterogeneity test; cRandom-effects model was used when P value for heterogeneity test < 0.05; otherwise, fixed-effects model was used;\nForest plot of prostate cancer risk associated with FGFR4 Gly388Arg polymorphism (Arg allele vs. Gly allele). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\nWhen stratifying for race, results were similar. Specially, significantly increased risk was found among Caucasian populations (Arg388 and Gly388 comparison: OR = 1.21, 95% CI: 1.00-1.47; dominant genetic model: OR = 1.23, 95% CI: 1.08-1.40) and Asian population (Arg388 and Gly388 comparison: OR = 1.24, 95% CI: 1.02-1.51; homozygote comparison: OR = 1.52, 95% CI: 1.05-2.22; recessive genetic model: OR = 1.53, 95% CI: 1.10-2.14). Although the effect in African-Americans was in the same direction as for other groups, the difference was not statistically significant (Arg388 and Gly388 comparison: OR = 1.15, 95% CI: 0.73-1.82; homozygote comparison: OR = 2.17, 95% CI: 0.20-23.14; dominant genetic model: OR = 1.11, 95% CI: 0.66-1.86 and recessive genetic model: OR = 2.21, 95% CI: 0.18-26.83).\nIn addition, when concerning tumor stage and FGFR4 Gly388Arg polymorphism, patients with prostate cancer with Arg/Arg genotype had a 1.34-fold increased risk of advanced or metastatic prostate cancer (95% CI: 1.03-1.74) compared with the Gly/Gly+Gly/Arg genotype (seen Figure 2).\nForest plot of prostate cancer progression associated with FGFR4 Gly388Arg polymorphism (Arg/Arg vs. Gly/Gly+Gly/Arg). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\nWe observed a wide variation of Arg388 allele frequencies across different ethnicities. The frequency of Arg388 allele was 28.90% among Caucasian controls and 41.57% among Asian controls, which were significantly higher than that in African-American controls (11.49%, P < 0.01).\nOverall, the combined result based on all studies showed the evidence of an association between the increased risk of prostate cancer and the variant genotypes in different genetic models. As shown in Table 2 and Figure 1, the Arg388 allele increased overall prostate cancer risk compared with Gly388 allele (OR = 1.17, 95% CI = 1.07-1.29). Significant main effects were also observed in dominate genetic model (OR = 1.21, 95% CI = 1.08-1.36).\nStratified analyses of the FGFR4 Gly388Arg polymorphism on cancer risk\naNumber of comparisons; bP value of Q-test for heterogeneity test; cRandom-effects model was used when P value for heterogeneity test < 0.05; otherwise, fixed-effects model was used;\nForest plot of prostate cancer risk associated with FGFR4 Gly388Arg polymorphism (Arg allele vs. Gly allele). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\nWhen stratifying for race, results were similar. Specially, significantly increased risk was found among Caucasian populations (Arg388 and Gly388 comparison: OR = 1.21, 95% CI: 1.00-1.47; dominant genetic model: OR = 1.23, 95% CI: 1.08-1.40) and Asian population (Arg388 and Gly388 comparison: OR = 1.24, 95% CI: 1.02-1.51; homozygote comparison: OR = 1.52, 95% CI: 1.05-2.22; recessive genetic model: OR = 1.53, 95% CI: 1.10-2.14). Although the effect in African-Americans was in the same direction as for other groups, the difference was not statistically significant (Arg388 and Gly388 comparison: OR = 1.15, 95% CI: 0.73-1.82; homozygote comparison: OR = 2.17, 95% CI: 0.20-23.14; dominant genetic model: OR = 1.11, 95% CI: 0.66-1.86 and recessive genetic model: OR = 2.21, 95% CI: 0.18-26.83).\nIn addition, when concerning tumor stage and FGFR4 Gly388Arg polymorphism, patients with prostate cancer with Arg/Arg genotype had a 1.34-fold increased risk of advanced or metastatic prostate cancer (95% CI: 1.03-1.74) compared with the Gly/Gly+Gly/Arg genotype (seen Figure 2).\nForest plot of prostate cancer progression associated with FGFR4 Gly388Arg polymorphism (Arg/Arg vs. Gly/Gly+Gly/Arg). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\n[SUBTITLE] Sensitivity analysis [SUBSECTION] Sensitivity analysis was performed by sequential omission of individual studies. The pooled 95% CI for Arg388 vs. Gly388 was consistently over 1.0, indicating that the results of this meta-analysis are stable.\nSensitivity analysis was performed by sequential omission of individual studies. The pooled 95% CI for Arg388 vs. Gly388 was consistently over 1.0, indicating that the results of this meta-analysis are stable.\n[SUBTITLE] Publication bias [SUBSECTION] Begg's funnel plot and Egger's test were performed to assess the publication bias. The shape of the funnel plots seemed symmetrical in the comparison of the Arg388 vs. Gly388 (Figure 3). Furthermore, Egger's test was used to provide statistical evidence for funnel plot symmetry (t = 1.30, P = 0.26), suggesting that no publication bias was exist.\nBegg's funnel plot for publication bias test. Each point represents a separate study for the indicated association. Log[OR], natural logarithm of OR. Horizontal line, mean effect size.\nBegg's funnel plot and Egger's test were performed to assess the publication bias. The shape of the funnel plots seemed symmetrical in the comparison of the Arg388 vs. Gly388 (Figure 3). Furthermore, Egger's test was used to provide statistical evidence for funnel plot symmetry (t = 1.30, P = 0.26), suggesting that no publication bias was exist.\nBegg's funnel plot for publication bias test. Each point represents a separate study for the indicated association. Log[OR], natural logarithm of OR. Horizontal line, mean effect size.", "Using the searching terms, seven papers were reviewed in the two online databases. The study of Spinola et al.[18] was focused on the association between FGFR4 Gly388Arg polymorphism and lung cancer risk, and the studies of Wang et al.[12] and Sahadevan et al.[4] were not epidemiological association studies, so they were all excluded in present study. In the four papers left, FitzGerald et al.[8] and Wang et al.[11] provided data on both African-American and Caucasian. Overall, four articles (six studies) with 2618 prostate cancer cases and 2305 controls were retrieved based on the search criteria for prostate cancer susceptibility related to the FGFR4 Gly388Arg polymorphism. Study characteristics are summarized in Table 1. All the studies used frequency-matched controls to the cases by the age, sex or ethnicity, and the distribution of genotypes in the controls was consistent with Hardy-Weinberg equilibrium in all studies. Moreover, among the four articles, three[8,10,11] mentioned the association between FGFR4 Gly388Arg polymorphism and progression of prostate cancer. The stratifications for pathological parameters of the cases in the three articles were also shown in Table 1. The cases of the three articles were all stratified by Gleason score and tumor stage. However, the classification standard of Gleason score was not uniform; thus, we only focused on the association between FGFR4 Gly388Arg polymorphism and tumor stage (advanced vs. localized). Advanced stage corresponded to T3 stage in the study of Wang et al.[11], regional/distant stage in the study of FitzGerald et al.[8], and stage C+D in the study of Ma et al.[10], respectively. And localized stage meant T2 stage in the study of Wang et al., local stage in the study of FitzGerald et al., and stage A+B in the study of Ma et al., respectively.\nMain characteristics of selected studies\na Stage A (T1a-bN0M0), StageB (T1c-2N0M0), Stage C (T3-4N0M0) and Stage D (T1-4N1M0-1 orT1-4N0-1M1) by the modified Whitmore-Jewett system.", "We observed a wide variation of Arg388 allele frequencies across different ethnicities. The frequency of Arg388 allele was 28.90% among Caucasian controls and 41.57% among Asian controls, which were significantly higher than that in African-American controls (11.49%, P < 0.01).\nOverall, the combined result based on all studies showed the evidence of an association between the increased risk of prostate cancer and the variant genotypes in different genetic models. As shown in Table 2 and Figure 1, the Arg388 allele increased overall prostate cancer risk compared with Gly388 allele (OR = 1.17, 95% CI = 1.07-1.29). Significant main effects were also observed in dominate genetic model (OR = 1.21, 95% CI = 1.08-1.36).\nStratified analyses of the FGFR4 Gly388Arg polymorphism on cancer risk\naNumber of comparisons; bP value of Q-test for heterogeneity test; cRandom-effects model was used when P value for heterogeneity test < 0.05; otherwise, fixed-effects model was used;\nForest plot of prostate cancer risk associated with FGFR4 Gly388Arg polymorphism (Arg allele vs. Gly allele). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.\nWhen stratifying for race, results were similar. Specially, significantly increased risk was found among Caucasian populations (Arg388 and Gly388 comparison: OR = 1.21, 95% CI: 1.00-1.47; dominant genetic model: OR = 1.23, 95% CI: 1.08-1.40) and Asian population (Arg388 and Gly388 comparison: OR = 1.24, 95% CI: 1.02-1.51; homozygote comparison: OR = 1.52, 95% CI: 1.05-2.22; recessive genetic model: OR = 1.53, 95% CI: 1.10-2.14). Although the effect in African-Americans was in the same direction as for other groups, the difference was not statistically significant (Arg388 and Gly388 comparison: OR = 1.15, 95% CI: 0.73-1.82; homozygote comparison: OR = 2.17, 95% CI: 0.20-23.14; dominant genetic model: OR = 1.11, 95% CI: 0.66-1.86 and recessive genetic model: OR = 2.21, 95% CI: 0.18-26.83).\nIn addition, when concerning tumor stage and FGFR4 Gly388Arg polymorphism, patients with prostate cancer with Arg/Arg genotype had a 1.34-fold increased risk of advanced or metastatic prostate cancer (95% CI: 1.03-1.74) compared with the Gly/Gly+Gly/Arg genotype (seen Figure 2).\nForest plot of prostate cancer progression associated with FGFR4 Gly388Arg polymorphism (Arg/Arg vs. Gly/Gly+Gly/Arg). The squares and horizontal lines correspond to the study-specific OR and 95% CI. The area of the squares reflects the weight (inverse of the variance). The diamond represents the summary OR and 95% CI.", "Sensitivity analysis was performed by sequential omission of individual studies. The pooled 95% CI for Arg388 vs. Gly388 was consistently over 1.0, indicating that the results of this meta-analysis are stable.", "Begg's funnel plot and Egger's test were performed to assess the publication bias. The shape of the funnel plots seemed symmetrical in the comparison of the Arg388 vs. Gly388 (Figure 3). Furthermore, Egger's test was used to provide statistical evidence for funnel plot symmetry (t = 1.30, P = 0.26), suggesting that no publication bias was exist.\nBegg's funnel plot for publication bias test. Each point represents a separate study for the indicated association. Log[OR], natural logarithm of OR. Horizontal line, mean effect size.", "The present meta-analysis, including 2,618 cases and 2,305 controls from six published studies, explored the association between FGFR4 Gly388Arg polymorphism and development and progression of prostate cancer. To the best of our knowledge, this is the first meta-analysis to explore FGFR4 Gly388Arg polymorphism in development and progression of prostate cancer. The results indicated that FGFR4 Arg388 allele is a potential risk factor for developing and progressing prostate cancer. These findings may be biologically plausible. The FGFR4 Gly388Arg polymorphism results in an amino acid change in the transmembrane domain of the receptor, which may alter the activity of the receptor. FGFR4 is the activator of the MAPK signaling cascade, yet it is a principal receptor for key mitogenic FGFs in prostate cancer cells[19-21]. There was evidence that FGFR4 contributed to progression in liver, lung, colon tumors[22] and prostate cancer[12]. The effects of FGFR4 Arg388 allele may also predispose cancer patients to disease progression, based on the reported significant association between FGFR4 genotype and tumor aggressiveness (lymph node involvement, advanced stage) or patients' survival[6,7], and the results about its biological role on cancer cell motility and invasiveness[11]. In our meta-analysis, we found that subjects carrying Arg388 were associated with higher risk of developing and progressing prostate cancer than those with the wild-type allele, which confirmed the hypothesis described above.\nSome limitations of this meta-analysis should be acknowledged. First of all, the control populations were not uniform. Healthy populations as well as non-cancer patients like BPH patients were included. Some individuals in the control group are likely to develop cancer in subsequent years though they had no clinical symptoms at the time of investigation. Misclassification bias results in deviation of genotype distribution in the controls. Second, prostate cancer, as a complex disease, was considered as the result of combined effects of multi-factors, including inherited and environmental factors[23], however, no such data was observed in previous studies. Thus, our result was only based on unadjusted estimates. Lacking of the information for the data analysis may cause serious confounding bias. Third, the effect of the polymorphism was relatively trivial with small ORs. We need further studies with larger number participants to confirm the effect.\nOur meta-analysis also had some advantages. First, disease progression status as tumor stage was taken into account in present study. Second, data in present study were pooled from different studies, which significantly increased statistical power of the analysis. Third, the quality of studies included in our meta-analysis was satisfactory and cruelly met our inclusion criterion. Fourth, the distribution of genotypes in the controls was consistent with Hardy-Weinberg equilibrium in all studies. We further performed sensitivity analysis to detect the stability of the meta-analysis, and the results did not alter the pattern of association and revealed that the risk effect of Arg388 was stable. In addition, publication bias was not detected in present study, indicating that our findings seemed not to be due to biased publications.", "Our meta-analysis showed the evidence that FGFR4 Arg388 allele was associated with an increased risk of prostate cancer development and progression, suggesting that FGFR4 Gly388Arg polymorphism could be a marker for prostate cancer development and progression. Based on the limitations of present study list above, further prospective researches using standardized unbiased methods, and larger numbers of worldwide participants are expected to examine the association to confirm our results, and other possible confounding risk factors like age, life style, and familial history should also be controlled when it was assessed. Moreover, gene-gene and gene-environment interactions should also be considered.", "FGFR4: Fibroblast growth factor receptor 4; Gly: glycine; Arg: arginine; OR: odds ratio; CI: confidence interval.", "The authors declare that they have no competing interests.", "BX participated in study design and drafted the manuscript.\nSQC participated in collection of data and manuscript preparation.\nNT and ZDZ performed the statistical analysis and participated in the critical revision of the manuscript.\nZJW critically revised the manuscript.\nMC and LXH participated in its design.\nAll authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2407/11/84/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Aged", "Aged, 80 and over", "Case-Control Studies", "Cohort Studies", "Diabetes Mellitus, Type 2", "Female", "Genetic Variation", "Humans", "Hypoglycemic Agents", "Insulin", "Insulin Secretion", "Male", "Sulfonylurea Compounds", "Transcription Factor 7-Like 2 Protein" ]

[ "Background", "Subjects", "Genotyping of rs7903146", "Statistics", "Results", "Secondary confirmatory analyses", "Discussion", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "The TCF7L2-gene (TCF7L2; Transcription factor 7-like 2) encodes a transcription factor (Tcf-4) that is involved in the regulation of cellular proliferation and differentiation [1]. Variants in the TCF7L2 have initially been shown to be associated with an increased risk for type 2 diabetes (T2D) in a genome-wide analysis of the isolate population of Iceland [2]. The strongest associations with T2D with a clear gene dose effect were reported for the rs7903146 variant [3]. The initial findings have been replicated in independent studies in multiple ethnic populations and were summarized in a large global meta-analysis [4]. The risk alleles actually predicted the progression from impaired glucose tolerance to diabetes prospectively [5] and an increased severity of the disease [6] in adults. Also, TCF7L2 variants conferred a higher risk for early impairment of glucose metabolism emerging as soon as in childhood and adolescence [7]. Some clinical data suggested that the polymorphisms affected the capacity of pancreatic β-cells to secrete insulin rather than aggravating insulin resistance [5,8-13], possibly by impaired β-cell proinsulin-processing [14]. This was further supported by expression data suggesting a putative role of TCF7L2 in β-cell differentiation [12]. Considering the role of TCF7L2 risk variants in insulin secretion, Pearson et al. [15] hypothesized that patients with diabetes risk alleles at rs12255372 and rs7903146 have an altered hypoglycaemic response to sulfonylureas (SUs) due to decreased β-cell function. They could indeed show that carriers of the diabetes risk alleles from the GoDART (Genetics of Diabetes Audit and Research Tayside) study were less likely to respond to SUs [15]. This study suggested that genetic variation in TCF7L2 can alter response to therapy in T2D. Since a causal phenotype-genotype relationship can not be established with one initial report, replication studies are the backbone to the genetic epidemiology of complex diseases [16], and play a crucial role in pharmacogenomics as well.\nTherefore, we sought to evaluate the association between genetic variants in TCF7L2 with SU treatment failure in an independent cohort from Germany. We recruited 189 patients with T2D being treated with SU agents and determined the TCF7L2 rs7903146 diabetes risk genotype, which has been reported as having the strongest association with T2D [3]. We used a logistic regression with secondary SU failure defined as an A1C ≥7.0% after 6 months of SU treatment.", "One hundred and eighty-nine patients with T2D, all of them being treated with SU agents, were recruited at the medical department of the Klinikum Lippe-Detmold, a large tertiary care hospital in East Westphalia, Germany, between 1 January 2000 and 30.11.2009. As the only hospital in the area, the one at Lippe-Detmold is responsible for the inpatient and outpatient management of all emergencies in the region. All patients had been treated with the SU drugs glimepiride (n = 147), glibenclamide (n = 39) and gliquidon (n = 3). Ninety-seven patients failed to respond to SU treatment according to our definition of A1C ≥7.0% after 6 months of treatment (76 patients treated with glimepiride, 19 with glibenclamide and 2 with gliquidon). Forty six patients were additionally treated with insulin. The mean (± SD) daily dose of SU agents was comparable between subjects who failed to respond to SUs and the controls (5.0 ± 3.7 mg vs. 6.8 ± 3.7 mg, P = 0.13 for glibenclamide; 2.5 ± 1.6 mg vs. 2.5 ± 1.4 mg for glimepiride, P = 0.99). As our patients were recruited within the framework of a study originally investigating the risk of hypoglycaemia [17], eighty nine patients had experienced a severe hypoglycaemia, which was defined as a symptomatic event requiring treatment with intravenous glucose and was confirmed by a blood glucose measurement of <50 mg/dl (<2.8 mmol/l). Seventy-two subjects were additionally treated with insulin sensitizing drug metformin (32 patients in the control group and 40 patients in the group of patients with SU treatment failure; P = 0.36, Table 1). The protocol was approved by the Ethics Committee of the University of Münster School of Medicine and by the Ethics Committee of the University of Leipzig, School of Medicine. All patients gave written informed consent to participate in the study.\nClinical characteristics of all participants.\nData are mean ± SD; P- values for comparisons between genotypic groups by ANOVA statistics;\n*- χ2 test.", "Genotyping of rs7903146 in all study subjects was done using the TaqMan allelic discrimination assay (Assays-on-Demand (TM), SNP Genotyping Products; Applied Biosystems, Inc.) on an ABI PRISM 7500 sequence detector (Applied Biosystems Inc.) according to the manufacturer's protocol. The genotype distribution was consistent with Hardy-Weinberg equilibrium (P > 0.05). Genotyping success rate was >99%, and duplicate genotyping concordance was 100%.", "Standard descriptive and comparative statistics (χ2 test, t-test) were used to characterize and compare clinical parameters in different groups (controls, cases). Logistic regression analyses were used to calculate the effects of investigated factors on SU treatment failure, which were reported as odds ratio with 95% CI (confidence intervals). In the additive model, homozygotes for the major allele, heterozygotes and homozygotes for the minor allele were coded to a continuous numeric variable for genotype (as 0, 1, 2). Data were analyzed using the SPSS software package (version 15.0; SPSS, Inc., Chicago, IL).", "Clinical characteristics of all study participants are given in Table 1. As expected, the subjects with failure of SU treatment had a higher A1C than the controls (Table 1). However, both groups were comparable in regard to the age, gender, age at onset of diabetes, duration of diabetes and creatinine clearance and co-medication presented as number of drugs taken by the patient (Table 1). Also the number of patients additionally treated with metformin was similar between the groups (Table 1).\nIn the univariate logistic regression analyses, the TCF7L2 genotype was the only predictor of SU treatment failure (Table 2). The rs7903146 T allele was significantly more frequent in the group of patients who failed to respond to SU (36%) than in the control group (26%) [P = 0.046; odds ratio (OR): 1.57 (1.01-2.45) in an additive mode of inheritance] (Table 3). In the control group, 56.0% of subjects had the CC, 36.3% had the CT and 7.7% had the TT genotype. Among patients who failed to respond to SUs, 41.2% were CC homozygous, 46.4% were CT heterozygous and 12.4% were TT homozygous (Table 3).\nUnivariate regression analyses of predictors on failure of sulfonylurea treatment in patients with type 2 diabetes.\nEffect of rs7903146 genotype on sulfonylurea treatment failure under logistic regression analysis.\nTo investigate whether the rs7903146 effect is specific to the mechanism of action of SUs, we evaluated the genotype effects on response to a non-insulin secretagogue metformin. By analysing 72 metformin-treated individuals only, no effect of the genotype on SU treatment failure was found [P = 0.98; OR 1.01 (0.50-2.03)].\n[SUBTITLE] Secondary confirmatory analyses [SUBSECTION] In secondary analyses we used a logistic regression with secondary SU failure defined as the addition of insulin after at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0%. Based on these criteria 46 patients from our cohort failed to respond to SU treatment and were additionally treated with insulin.\nIn the univariate logistic regression analyses including 46 patients who failed to respond to SU treatment and 143 control subjects, diabetes duration (<5 yrs vs. >5 yrs) appeared to be the strongest predictor of SU treatment failure [OR: 4.06 (1.50-11.01), P = 0.006]. We also assessed the effect of rs7903146 on SU treatment failure. The rs7903146 T allele was significantly more frequent in the group of patients additionally treated with insulin (40%) than in the control group treated only with SUs (28%) [P = 0.03; odds ratio (OR): 1.73 (1.06-2.84) in an additive mode of inheritance]. In the control group, 53% of subjects had the CC genotype, 39% had CT and 8% had TT. Among patients treated with insulin, 35% were CC homozygous, 50% were CT heterozygous and 15% had the TT genotype. The results remained materially unchanged even after including diabetes duration as a strong predictor of SU treatment failure in these analyses, thus indicating an independent effect of the genotype [OR: 1.66 (0.99-2.79), P = 0.06 in additive model].\nIn secondary analyses we used a logistic regression with secondary SU failure defined as the addition of insulin after at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0%. Based on these criteria 46 patients from our cohort failed to respond to SU treatment and were additionally treated with insulin.\nIn the univariate logistic regression analyses including 46 patients who failed to respond to SU treatment and 143 control subjects, diabetes duration (<5 yrs vs. >5 yrs) appeared to be the strongest predictor of SU treatment failure [OR: 4.06 (1.50-11.01), P = 0.006]. We also assessed the effect of rs7903146 on SU treatment failure. The rs7903146 T allele was significantly more frequent in the group of patients additionally treated with insulin (40%) than in the control group treated only with SUs (28%) [P = 0.03; odds ratio (OR): 1.73 (1.06-2.84) in an additive mode of inheritance]. In the control group, 53% of subjects had the CC genotype, 39% had CT and 8% had TT. Among patients treated with insulin, 35% were CC homozygous, 50% were CT heterozygous and 15% had the TT genotype. The results remained materially unchanged even after including diabetes duration as a strong predictor of SU treatment failure in these analyses, thus indicating an independent effect of the genotype [OR: 1.66 (0.99-2.79), P = 0.06 in additive model].", "In secondary analyses we used a logistic regression with secondary SU failure defined as the addition of insulin after at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0%. Based on these criteria 46 patients from our cohort failed to respond to SU treatment and were additionally treated with insulin.\nIn the univariate logistic regression analyses including 46 patients who failed to respond to SU treatment and 143 control subjects, diabetes duration (<5 yrs vs. >5 yrs) appeared to be the strongest predictor of SU treatment failure [OR: 4.06 (1.50-11.01), P = 0.006]. We also assessed the effect of rs7903146 on SU treatment failure. The rs7903146 T allele was significantly more frequent in the group of patients additionally treated with insulin (40%) than in the control group treated only with SUs (28%) [P = 0.03; odds ratio (OR): 1.73 (1.06-2.84) in an additive mode of inheritance]. In the control group, 53% of subjects had the CC genotype, 39% had CT and 8% had TT. Among patients treated with insulin, 35% were CC homozygous, 50% were CT heterozygous and 15% had the TT genotype. The results remained materially unchanged even after including diabetes duration as a strong predictor of SU treatment failure in these analyses, thus indicating an independent effect of the genotype [OR: 1.66 (0.99-2.79), P = 0.06 in additive model].", "In the present study, we investigated the effect of TCF7L2 diabetes risk T-allele at rs7903146 on therapeutic response to SUs. In univariate regression analyses, TCF7L2 genotype was the only predictor of SU treatment failure. The rs7903146 T-allele conferred a higher risk for sulfonylurea treatment failure as it was significantly more frequent in the group of patients who failed to respond to SUs (36%) than in the control group (26%). After adjusting for diabetes duration the odds ratio did not change (OR = 1.57) and the P-value reduced just minimally (from P = 0.046 to P = 0.057), thus indicating independent effect of the TCF7L2 genotype. Despite the smaller sample size and so, limited statistical power, our data are in line with findings reported by Pearson et al. [15], suggesting that variation in TCF7L2 influences therapeutic response to SUs. Pearson et al. observed that homozygous carriers of the TCF7L2 risk alleles (rs1225372 and rs7903146) were twice as likely not to respond to SUs as patients homozygous for the non-risk alleles [15]. Considering pretreatment A1C levels as covariate in logistic regression analyses even strengthened the association between sulfonylurea response and genotype at rs7903146 [15]. Even though the findings are consistent between the present study and that reported by Pearson et al., several differences should be noted. First, Pearson et al. investigated a total of 911 SU users of 4,469 patients with T2D from the DARTS/MEMO (Diabetes Audit and Research Tayside/Medicines Monitoring Unit) collaboration database, who were recruited to GoDARTS between 1997 and 2006 [15]. Second, SU failure was defined very restrictively as an A1C >7% within 3-12 months after treatment initiation. According to these criteria, 42% of SU users failed to respond to the therapy [15]. In our study we chose a comparable, overlapping definition of secondary SU failure with at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0%. According to this definition, 51% of our patients failed to respond to SU therapy. We are aware that the difference in SU treatment failure frequency may be due to various study designs but it is noteworthy that to date, there is no widely accepted definition of secondary SU failure. Due to the lack of uniform definition, the frequency of SU failure varies considerably between 22% and 50% after 12 and 36 months of treatment, respectively [18,19]. The decreasing effectiveness of SUs results from progressive loss of β-cell function but also from patient related factors (dietary incompliance, weight gain, lack of exercise). Despite the above mentioned differences between the studies, frequencies of TT homozygous subjects in the groups of patients who failed to respond to the therapy (independent of definition) were significantly higher than in the control groups and were comparable between both studies (12% vs. 8% in the German patients and 16% vs. 8% in the GoDARTS study). Also, genotype distribution of the rs7903146 was similar, with 10% and 11% of diabetic population with 2 copies of the T-allele in our study and the GoDARTS study, respectively. Finally, similarly to the GoDARTS study, our data suggest that carriers of the T allele were 57% more likely to fail SU treatment; TT homozygotes were twice as likely as CC homozygotes.\nIt is noteworthy that an alternate definition of SU treatment failure in our cohort based on addition of insulin after at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0% yielded similar results. Even though not independent from the previous analyses, these findings provide further support for the role of TCF7L2 genotypes in altered hypoglycaemic response to SUs. Interestingly, when using this definition of SU treatment failure, diabetes duration appeared to be a predictor of treatment failure along with the TCF7L2 genotype. Nevertheless, the genotype effect was independent as even after adjustment for diabetes duration, the results remained materially unchanged. Although the P-value went from 0.04 to 0.06 the odds ratio reduced only minimally (from 1.73 to 1.66).\nIndirectly, our findings also support studies that favour the role of TCF7L2 in the regulation of insulin secretion. However, we are aware that since the TCF7L2 variants increase progression from IGT to diabetes [5], additional models considering diabetes therapy, particularly including a control group having been treated with a different antidiabetic agent - e.g., metformin, would be desirable to clarify whether the observed data reflect pharmacogenetic effects specific to SUs or rather a disease-genetic process. Indeed, such a control group treated with metformin was included in the GoDARTS study [15]. The study suggested pharmacogenetic effects of TCF7L2 SNPs influencing therapeutic response to sulfonylureas but not metformin, since no association was seen between metformin response and TCF7L2 variants [15]. Even though limited by the small sample size (N = 72), we also failed to observe any influence of TCF7L2 genotypes on the response to metformin, as a non-insulin secretagogue, thus further supporting the notion that TCF7L2 effect is specific to the mechanism of action of SUs.\nOne of the major limitations of our study is the low sample size and so, limited statistical power. Taking into account genotype frequencies and the sample size in our study we had a statistical power of 80% (at α = 0.05) to detect genetic risk (odds ratio) of 1.8 for treatment failure in additive mode of inheritance (software Quanto version 1.2.2) [20]. In contrast, the GoDARTS study by Pearson et al. [15] had 80% power to detect risk (OR) as low as 1.3. Also noteworthy, given the TCF7L2 diabetes risk genotypes make patients more resistant to the action of sulfonylureas, one would expect that carriers of the risk genotype should be less likely to become hypoglycaemic. However, in our study we could not observe any differences in the frequency of the diabetes risk allele between patients with and without hypoglycaemia (P = 0.30).", "In conclusion, present data strengthen previously reported findings suggesting altered therapeutic response to SUs in patients with T2D carrying the diabetes risk alleles at TCF7L2 variants.", "The authors declare that they have no competing interests.", "MH was responsible for clinical characteristics of study subjects. AK and MS edited the manuscript and contributed to the discussion. AH and PK conceived and designed the study and wrote the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2350/12/30/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Subjects", "Genotyping of rs7903146", "Statistics", "Results", "Secondary confirmatory analyses", "Discussion", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "The TCF7L2-gene (TCF7L2; Transcription factor 7-like 2) encodes a transcription factor (Tcf-4) that is involved in the regulation of cellular proliferation and differentiation [1]. Variants in the TCF7L2 have initially been shown to be associated with an increased risk for type 2 diabetes (T2D) in a genome-wide analysis of the isolate population of Iceland [2]. The strongest associations with T2D with a clear gene dose effect were reported for the rs7903146 variant [3]. The initial findings have been replicated in independent studies in multiple ethnic populations and were summarized in a large global meta-analysis [4]. The risk alleles actually predicted the progression from impaired glucose tolerance to diabetes prospectively [5] and an increased severity of the disease [6] in adults. Also, TCF7L2 variants conferred a higher risk for early impairment of glucose metabolism emerging as soon as in childhood and adolescence [7]. Some clinical data suggested that the polymorphisms affected the capacity of pancreatic β-cells to secrete insulin rather than aggravating insulin resistance [5,8-13], possibly by impaired β-cell proinsulin-processing [14]. This was further supported by expression data suggesting a putative role of TCF7L2 in β-cell differentiation [12]. Considering the role of TCF7L2 risk variants in insulin secretion, Pearson et al. [15] hypothesized that patients with diabetes risk alleles at rs12255372 and rs7903146 have an altered hypoglycaemic response to sulfonylureas (SUs) due to decreased β-cell function. They could indeed show that carriers of the diabetes risk alleles from the GoDART (Genetics of Diabetes Audit and Research Tayside) study were less likely to respond to SUs [15]. This study suggested that genetic variation in TCF7L2 can alter response to therapy in T2D. Since a causal phenotype-genotype relationship can not be established with one initial report, replication studies are the backbone to the genetic epidemiology of complex diseases [16], and play a crucial role in pharmacogenomics as well.\nTherefore, we sought to evaluate the association between genetic variants in TCF7L2 with SU treatment failure in an independent cohort from Germany. We recruited 189 patients with T2D being treated with SU agents and determined the TCF7L2 rs7903146 diabetes risk genotype, which has been reported as having the strongest association with T2D [3]. We used a logistic regression with secondary SU failure defined as an A1C ≥7.0% after 6 months of SU treatment.", "[SUBTITLE] Subjects [SUBSECTION] One hundred and eighty-nine patients with T2D, all of them being treated with SU agents, were recruited at the medical department of the Klinikum Lippe-Detmold, a large tertiary care hospital in East Westphalia, Germany, between 1 January 2000 and 30.11.2009. As the only hospital in the area, the one at Lippe-Detmold is responsible for the inpatient and outpatient management of all emergencies in the region. All patients had been treated with the SU drugs glimepiride (n = 147), glibenclamide (n = 39) and gliquidon (n = 3). Ninety-seven patients failed to respond to SU treatment according to our definition of A1C ≥7.0% after 6 months of treatment (76 patients treated with glimepiride, 19 with glibenclamide and 2 with gliquidon). Forty six patients were additionally treated with insulin. The mean (± SD) daily dose of SU agents was comparable between subjects who failed to respond to SUs and the controls (5.0 ± 3.7 mg vs. 6.8 ± 3.7 mg, P = 0.13 for glibenclamide; 2.5 ± 1.6 mg vs. 2.5 ± 1.4 mg for glimepiride, P = 0.99). As our patients were recruited within the framework of a study originally investigating the risk of hypoglycaemia [17], eighty nine patients had experienced a severe hypoglycaemia, which was defined as a symptomatic event requiring treatment with intravenous glucose and was confirmed by a blood glucose measurement of <50 mg/dl (<2.8 mmol/l). Seventy-two subjects were additionally treated with insulin sensitizing drug metformin (32 patients in the control group and 40 patients in the group of patients with SU treatment failure; P = 0.36, Table 1). The protocol was approved by the Ethics Committee of the University of Münster School of Medicine and by the Ethics Committee of the University of Leipzig, School of Medicine. All patients gave written informed consent to participate in the study.\nClinical characteristics of all participants.\nData are mean ± SD; P- values for comparisons between genotypic groups by ANOVA statistics;\n*- χ2 test.\nOne hundred and eighty-nine patients with T2D, all of them being treated with SU agents, were recruited at the medical department of the Klinikum Lippe-Detmold, a large tertiary care hospital in East Westphalia, Germany, between 1 January 2000 and 30.11.2009. As the only hospital in the area, the one at Lippe-Detmold is responsible for the inpatient and outpatient management of all emergencies in the region. All patients had been treated with the SU drugs glimepiride (n = 147), glibenclamide (n = 39) and gliquidon (n = 3). Ninety-seven patients failed to respond to SU treatment according to our definition of A1C ≥7.0% after 6 months of treatment (76 patients treated with glimepiride, 19 with glibenclamide and 2 with gliquidon). Forty six patients were additionally treated with insulin. The mean (± SD) daily dose of SU agents was comparable between subjects who failed to respond to SUs and the controls (5.0 ± 3.7 mg vs. 6.8 ± 3.7 mg, P = 0.13 for glibenclamide; 2.5 ± 1.6 mg vs. 2.5 ± 1.4 mg for glimepiride, P = 0.99). As our patients were recruited within the framework of a study originally investigating the risk of hypoglycaemia [17], eighty nine patients had experienced a severe hypoglycaemia, which was defined as a symptomatic event requiring treatment with intravenous glucose and was confirmed by a blood glucose measurement of <50 mg/dl (<2.8 mmol/l). Seventy-two subjects were additionally treated with insulin sensitizing drug metformin (32 patients in the control group and 40 patients in the group of patients with SU treatment failure; P = 0.36, Table 1). The protocol was approved by the Ethics Committee of the University of Münster School of Medicine and by the Ethics Committee of the University of Leipzig, School of Medicine. All patients gave written informed consent to participate in the study.\nClinical characteristics of all participants.\nData are mean ± SD; P- values for comparisons between genotypic groups by ANOVA statistics;\n*- χ2 test.\n[SUBTITLE] Genotyping of rs7903146 [SUBSECTION] Genotyping of rs7903146 in all study subjects was done using the TaqMan allelic discrimination assay (Assays-on-Demand (TM), SNP Genotyping Products; Applied Biosystems, Inc.) on an ABI PRISM 7500 sequence detector (Applied Biosystems Inc.) according to the manufacturer's protocol. The genotype distribution was consistent with Hardy-Weinberg equilibrium (P > 0.05). Genotyping success rate was >99%, and duplicate genotyping concordance was 100%.\nGenotyping of rs7903146 in all study subjects was done using the TaqMan allelic discrimination assay (Assays-on-Demand (TM), SNP Genotyping Products; Applied Biosystems, Inc.) on an ABI PRISM 7500 sequence detector (Applied Biosystems Inc.) according to the manufacturer's protocol. The genotype distribution was consistent with Hardy-Weinberg equilibrium (P > 0.05). Genotyping success rate was >99%, and duplicate genotyping concordance was 100%.\n[SUBTITLE] Statistics [SUBSECTION] Standard descriptive and comparative statistics (χ2 test, t-test) were used to characterize and compare clinical parameters in different groups (controls, cases). Logistic regression analyses were used to calculate the effects of investigated factors on SU treatment failure, which were reported as odds ratio with 95% CI (confidence intervals). In the additive model, homozygotes for the major allele, heterozygotes and homozygotes for the minor allele were coded to a continuous numeric variable for genotype (as 0, 1, 2). Data were analyzed using the SPSS software package (version 15.0; SPSS, Inc., Chicago, IL).\nStandard descriptive and comparative statistics (χ2 test, t-test) were used to characterize and compare clinical parameters in different groups (controls, cases). Logistic regression analyses were used to calculate the effects of investigated factors on SU treatment failure, which were reported as odds ratio with 95% CI (confidence intervals). In the additive model, homozygotes for the major allele, heterozygotes and homozygotes for the minor allele were coded to a continuous numeric variable for genotype (as 0, 1, 2). Data were analyzed using the SPSS software package (version 15.0; SPSS, Inc., Chicago, IL).", "One hundred and eighty-nine patients with T2D, all of them being treated with SU agents, were recruited at the medical department of the Klinikum Lippe-Detmold, a large tertiary care hospital in East Westphalia, Germany, between 1 January 2000 and 30.11.2009. As the only hospital in the area, the one at Lippe-Detmold is responsible for the inpatient and outpatient management of all emergencies in the region. All patients had been treated with the SU drugs glimepiride (n = 147), glibenclamide (n = 39) and gliquidon (n = 3). Ninety-seven patients failed to respond to SU treatment according to our definition of A1C ≥7.0% after 6 months of treatment (76 patients treated with glimepiride, 19 with glibenclamide and 2 with gliquidon). Forty six patients were additionally treated with insulin. The mean (± SD) daily dose of SU agents was comparable between subjects who failed to respond to SUs and the controls (5.0 ± 3.7 mg vs. 6.8 ± 3.7 mg, P = 0.13 for glibenclamide; 2.5 ± 1.6 mg vs. 2.5 ± 1.4 mg for glimepiride, P = 0.99). As our patients were recruited within the framework of a study originally investigating the risk of hypoglycaemia [17], eighty nine patients had experienced a severe hypoglycaemia, which was defined as a symptomatic event requiring treatment with intravenous glucose and was confirmed by a blood glucose measurement of <50 mg/dl (<2.8 mmol/l). Seventy-two subjects were additionally treated with insulin sensitizing drug metformin (32 patients in the control group and 40 patients in the group of patients with SU treatment failure; P = 0.36, Table 1). The protocol was approved by the Ethics Committee of the University of Münster School of Medicine and by the Ethics Committee of the University of Leipzig, School of Medicine. All patients gave written informed consent to participate in the study.\nClinical characteristics of all participants.\nData are mean ± SD; P- values for comparisons between genotypic groups by ANOVA statistics;\n*- χ2 test.", "Genotyping of rs7903146 in all study subjects was done using the TaqMan allelic discrimination assay (Assays-on-Demand (TM), SNP Genotyping Products; Applied Biosystems, Inc.) on an ABI PRISM 7500 sequence detector (Applied Biosystems Inc.) according to the manufacturer's protocol. The genotype distribution was consistent with Hardy-Weinberg equilibrium (P > 0.05). Genotyping success rate was >99%, and duplicate genotyping concordance was 100%.", "Standard descriptive and comparative statistics (χ2 test, t-test) were used to characterize and compare clinical parameters in different groups (controls, cases). Logistic regression analyses were used to calculate the effects of investigated factors on SU treatment failure, which were reported as odds ratio with 95% CI (confidence intervals). In the additive model, homozygotes for the major allele, heterozygotes and homozygotes for the minor allele were coded to a continuous numeric variable for genotype (as 0, 1, 2). Data were analyzed using the SPSS software package (version 15.0; SPSS, Inc., Chicago, IL).", "Clinical characteristics of all study participants are given in Table 1. As expected, the subjects with failure of SU treatment had a higher A1C than the controls (Table 1). However, both groups were comparable in regard to the age, gender, age at onset of diabetes, duration of diabetes and creatinine clearance and co-medication presented as number of drugs taken by the patient (Table 1). Also the number of patients additionally treated with metformin was similar between the groups (Table 1).\nIn the univariate logistic regression analyses, the TCF7L2 genotype was the only predictor of SU treatment failure (Table 2). The rs7903146 T allele was significantly more frequent in the group of patients who failed to respond to SU (36%) than in the control group (26%) [P = 0.046; odds ratio (OR): 1.57 (1.01-2.45) in an additive mode of inheritance] (Table 3). In the control group, 56.0% of subjects had the CC, 36.3% had the CT and 7.7% had the TT genotype. Among patients who failed to respond to SUs, 41.2% were CC homozygous, 46.4% were CT heterozygous and 12.4% were TT homozygous (Table 3).\nUnivariate regression analyses of predictors on failure of sulfonylurea treatment in patients with type 2 diabetes.\nEffect of rs7903146 genotype on sulfonylurea treatment failure under logistic regression analysis.\nTo investigate whether the rs7903146 effect is specific to the mechanism of action of SUs, we evaluated the genotype effects on response to a non-insulin secretagogue metformin. By analysing 72 metformin-treated individuals only, no effect of the genotype on SU treatment failure was found [P = 0.98; OR 1.01 (0.50-2.03)].\n[SUBTITLE] Secondary confirmatory analyses [SUBSECTION] In secondary analyses we used a logistic regression with secondary SU failure defined as the addition of insulin after at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0%. Based on these criteria 46 patients from our cohort failed to respond to SU treatment and were additionally treated with insulin.\nIn the univariate logistic regression analyses including 46 patients who failed to respond to SU treatment and 143 control subjects, diabetes duration (<5 yrs vs. >5 yrs) appeared to be the strongest predictor of SU treatment failure [OR: 4.06 (1.50-11.01), P = 0.006]. We also assessed the effect of rs7903146 on SU treatment failure. The rs7903146 T allele was significantly more frequent in the group of patients additionally treated with insulin (40%) than in the control group treated only with SUs (28%) [P = 0.03; odds ratio (OR): 1.73 (1.06-2.84) in an additive mode of inheritance]. In the control group, 53% of subjects had the CC genotype, 39% had CT and 8% had TT. Among patients treated with insulin, 35% were CC homozygous, 50% were CT heterozygous and 15% had the TT genotype. The results remained materially unchanged even after including diabetes duration as a strong predictor of SU treatment failure in these analyses, thus indicating an independent effect of the genotype [OR: 1.66 (0.99-2.79), P = 0.06 in additive model].\nIn secondary analyses we used a logistic regression with secondary SU failure defined as the addition of insulin after at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0%. Based on these criteria 46 patients from our cohort failed to respond to SU treatment and were additionally treated with insulin.\nIn the univariate logistic regression analyses including 46 patients who failed to respond to SU treatment and 143 control subjects, diabetes duration (<5 yrs vs. >5 yrs) appeared to be the strongest predictor of SU treatment failure [OR: 4.06 (1.50-11.01), P = 0.006]. We also assessed the effect of rs7903146 on SU treatment failure. The rs7903146 T allele was significantly more frequent in the group of patients additionally treated with insulin (40%) than in the control group treated only with SUs (28%) [P = 0.03; odds ratio (OR): 1.73 (1.06-2.84) in an additive mode of inheritance]. In the control group, 53% of subjects had the CC genotype, 39% had CT and 8% had TT. Among patients treated with insulin, 35% were CC homozygous, 50% were CT heterozygous and 15% had the TT genotype. The results remained materially unchanged even after including diabetes duration as a strong predictor of SU treatment failure in these analyses, thus indicating an independent effect of the genotype [OR: 1.66 (0.99-2.79), P = 0.06 in additive model].", "In secondary analyses we used a logistic regression with secondary SU failure defined as the addition of insulin after at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0%. Based on these criteria 46 patients from our cohort failed to respond to SU treatment and were additionally treated with insulin.\nIn the univariate logistic regression analyses including 46 patients who failed to respond to SU treatment and 143 control subjects, diabetes duration (<5 yrs vs. >5 yrs) appeared to be the strongest predictor of SU treatment failure [OR: 4.06 (1.50-11.01), P = 0.006]. We also assessed the effect of rs7903146 on SU treatment failure. The rs7903146 T allele was significantly more frequent in the group of patients additionally treated with insulin (40%) than in the control group treated only with SUs (28%) [P = 0.03; odds ratio (OR): 1.73 (1.06-2.84) in an additive mode of inheritance]. In the control group, 53% of subjects had the CC genotype, 39% had CT and 8% had TT. Among patients treated with insulin, 35% were CC homozygous, 50% were CT heterozygous and 15% had the TT genotype. The results remained materially unchanged even after including diabetes duration as a strong predictor of SU treatment failure in these analyses, thus indicating an independent effect of the genotype [OR: 1.66 (0.99-2.79), P = 0.06 in additive model].", "In the present study, we investigated the effect of TCF7L2 diabetes risk T-allele at rs7903146 on therapeutic response to SUs. In univariate regression analyses, TCF7L2 genotype was the only predictor of SU treatment failure. The rs7903146 T-allele conferred a higher risk for sulfonylurea treatment failure as it was significantly more frequent in the group of patients who failed to respond to SUs (36%) than in the control group (26%). After adjusting for diabetes duration the odds ratio did not change (OR = 1.57) and the P-value reduced just minimally (from P = 0.046 to P = 0.057), thus indicating independent effect of the TCF7L2 genotype. Despite the smaller sample size and so, limited statistical power, our data are in line with findings reported by Pearson et al. [15], suggesting that variation in TCF7L2 influences therapeutic response to SUs. Pearson et al. observed that homozygous carriers of the TCF7L2 risk alleles (rs1225372 and rs7903146) were twice as likely not to respond to SUs as patients homozygous for the non-risk alleles [15]. Considering pretreatment A1C levels as covariate in logistic regression analyses even strengthened the association between sulfonylurea response and genotype at rs7903146 [15]. Even though the findings are consistent between the present study and that reported by Pearson et al., several differences should be noted. First, Pearson et al. investigated a total of 911 SU users of 4,469 patients with T2D from the DARTS/MEMO (Diabetes Audit and Research Tayside/Medicines Monitoring Unit) collaboration database, who were recruited to GoDARTS between 1997 and 2006 [15]. Second, SU failure was defined very restrictively as an A1C >7% within 3-12 months after treatment initiation. According to these criteria, 42% of SU users failed to respond to the therapy [15]. In our study we chose a comparable, overlapping definition of secondary SU failure with at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0%. According to this definition, 51% of our patients failed to respond to SU therapy. We are aware that the difference in SU treatment failure frequency may be due to various study designs but it is noteworthy that to date, there is no widely accepted definition of secondary SU failure. Due to the lack of uniform definition, the frequency of SU failure varies considerably between 22% and 50% after 12 and 36 months of treatment, respectively [18,19]. The decreasing effectiveness of SUs results from progressive loss of β-cell function but also from patient related factors (dietary incompliance, weight gain, lack of exercise). Despite the above mentioned differences between the studies, frequencies of TT homozygous subjects in the groups of patients who failed to respond to the therapy (independent of definition) were significantly higher than in the control groups and were comparable between both studies (12% vs. 8% in the German patients and 16% vs. 8% in the GoDARTS study). Also, genotype distribution of the rs7903146 was similar, with 10% and 11% of diabetic population with 2 copies of the T-allele in our study and the GoDARTS study, respectively. Finally, similarly to the GoDARTS study, our data suggest that carriers of the T allele were 57% more likely to fail SU treatment; TT homozygotes were twice as likely as CC homozygotes.\nIt is noteworthy that an alternate definition of SU treatment failure in our cohort based on addition of insulin after at least 6 months of SU therapy and corresponding A1C measurement of ≥7.0% yielded similar results. Even though not independent from the previous analyses, these findings provide further support for the role of TCF7L2 genotypes in altered hypoglycaemic response to SUs. Interestingly, when using this definition of SU treatment failure, diabetes duration appeared to be a predictor of treatment failure along with the TCF7L2 genotype. Nevertheless, the genotype effect was independent as even after adjustment for diabetes duration, the results remained materially unchanged. Although the P-value went from 0.04 to 0.06 the odds ratio reduced only minimally (from 1.73 to 1.66).\nIndirectly, our findings also support studies that favour the role of TCF7L2 in the regulation of insulin secretion. However, we are aware that since the TCF7L2 variants increase progression from IGT to diabetes [5], additional models considering diabetes therapy, particularly including a control group having been treated with a different antidiabetic agent - e.g., metformin, would be desirable to clarify whether the observed data reflect pharmacogenetic effects specific to SUs or rather a disease-genetic process. Indeed, such a control group treated with metformin was included in the GoDARTS study [15]. The study suggested pharmacogenetic effects of TCF7L2 SNPs influencing therapeutic response to sulfonylureas but not metformin, since no association was seen between metformin response and TCF7L2 variants [15]. Even though limited by the small sample size (N = 72), we also failed to observe any influence of TCF7L2 genotypes on the response to metformin, as a non-insulin secretagogue, thus further supporting the notion that TCF7L2 effect is specific to the mechanism of action of SUs.\nOne of the major limitations of our study is the low sample size and so, limited statistical power. Taking into account genotype frequencies and the sample size in our study we had a statistical power of 80% (at α = 0.05) to detect genetic risk (odds ratio) of 1.8 for treatment failure in additive mode of inheritance (software Quanto version 1.2.2) [20]. In contrast, the GoDARTS study by Pearson et al. [15] had 80% power to detect risk (OR) as low as 1.3. Also noteworthy, given the TCF7L2 diabetes risk genotypes make patients more resistant to the action of sulfonylureas, one would expect that carriers of the risk genotype should be less likely to become hypoglycaemic. However, in our study we could not observe any differences in the frequency of the diabetes risk allele between patients with and without hypoglycaemia (P = 0.30).", "In conclusion, present data strengthen previously reported findings suggesting altered therapeutic response to SUs in patients with T2D carrying the diabetes risk alleles at TCF7L2 variants.", "The authors declare that they have no competing interests.", "MH was responsible for clinical characteristics of study subjects. AK and MS edited the manuscript and contributed to the discussion. AH and PK conceived and designed the study and wrote the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2350/12/30/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null ]

[]

[ "Aged", "Antibodies, Blocking", "Cell Line, Tumor", "Cell Movement", "Cell Proliferation", "Chemokine CXCL12", "Colorectal Neoplasms", "Enzyme-Linked Immunosorbent Assay", "Female", "Gene Expression Regulation, Neoplastic", "Gene Knockdown Techniques", "Gene Silencing", "Humans", "Male", "Middle Aged", "Organ Specificity", "RNA, Messenger", "Receptors, CXCR4" ]

[ "Background", "Materials and patients", "Tissue preparation", "Single-strand cDNA synthesis", "Real-time PCR", "Isolation of total protein", "Sandwich-Type Enzyme-Linked Immunosorbent Assay", "Immunohistochemistry/Immunocytochemistry", "Cell migration assays", "Cell proliferation assays", "Cell inhibition assays", "siRNA assays", "Western Blot Analysis", "Calculations and Statistical Methods", "Results", "CXCL12/CXCR4 expression in CRC tissues", "Organ-specific expression of CXCL12", "CXCL12/CXCR4 expression in CRC cell lines", "CXCR4 protein inhibition abrogates migration of colorectal cancer cells", "CXC receptor-4 gene silencing abrogates migration of colorectal cancer cells", "Proliferative rate of colorectal cancer cells after CXCR4 blockage by mRNA silencing and inhibition antibodies", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Colorectal cancer (CRC) represents one of the most frequent malignancies worldwide with distant recurrence primarily affecting the liver as the predominant cause of CRC related mortality. The 5-year survival rate of 90% in patients with tumor restricted to the colon decreases to 10% in the presence of distant metastasis [1].\nRecently, various cancer-related studies demonstrated that specific chemokines and their receptors may be involved in the molecular mechanisms that control metastasis in the early stages of cancer development [2]. In this respect, the homeostatic chemokine CXCL12, a non-ELR+ CXC chemokine, has been implicated in promoting angiogenesis and metastasis [3,4]. CXCL12, also known as stromal derived factor 1 (SDF-1), is the only chemokine that is essential for survival [5] and a highly efficacious chemoattractant for T cells and thymocytes [6,7]. It is expressed by stromal cells such as fibroblasts and endothelial cells and signals exclusively via its G-protein-linked transmembrane receptor CXCR4 [8]. Expression of functional CXCR4 has been reported in various types of cancer cells [9-12], but also in immune cells such as peripheral blood lymphocytes, unprimed T cells, dendritic cells and lymphocytic leukemia B cells, where CXCR4 mediates spontaneous migration beneath bone marrow and stromal cells [13]. In infectious disease CXCR4 is employed by the human immunodefiency virus (HIV) to gain entry to cells [14] and in stem cell localization CXCR4 plays an important role for B-cell lymphopoiesis and bone marrow myelopoiesis [5]. In breast cancer, CXCR4 signaling was shown to play a crucial role in distant recurrence by mediating actin polymerisation and pseudopodia formation, thus, leading to chemotactic and invasive responses [3].\nRecently, CXCR4 expression was associated with advanced tumor stage and the development and recurrence of lymphatic or distant colorectal liver metastases (CRLM) [15-17]. Thus, it was reported that CXCR4 is differentially expressed in CRC and significantly correlates with survival, recurrence and liver metastasis [15,18]. Moreover, it was shown that concomitant and high expression of CXCR4 and VEGF is a strong and independent predictor of early distant relapse in CRC [19] and recent evidence indicates that CXCR4 may also play a role in tumor angiogenesis of CRC [20].\nDespite increasing knowledge about the expression of CXCL12/CXCR4 in CRC and its involvement of CXCR4 in the invasion and dissemination of CRC the precise mechanisms of CXCL12/CXCR4 interactions and subsequent metastasis are not entirely known. We therefore performed a comparative CXCL12/CXCR4 expression analysis and investigated how external addition of CXCL12 would promote CXCR4-mediated migration of CRC cell lines with different metastatic capabilities and how inhibition of CXCR4 by either CXCR4 siRNAs or neutralizing anti-CXCR4 antibodies would influence their migration potential.", "Surgical specimens and corresponding normal tissue from the same samples were collected from patients who underwent surgical resection at our department between 2003 and 2006. No patient underwent any specific cancer therapy prior to the resection. Our patient collectives comprised patients with CRC of different cancer stages (n = 50). In cases of organ confined CRC patients underwent resection of the primary tumor. Adjacent, disease free tissue samples of the colon and rectum served as control groups, respectively. Further, 10 specimens from patients with liver ruptures, focal nodular hyperplasia and haemangiomas as well as 10 unaffected tumor neighbor tissues from primary esophageal, pancreatic, gastric and colorectal carcinoma, respectively, were assessed. Informed consent for tissue procurement was obtained from all patients. The study was approved by the ethics commission of the Ärztekammer of the Saarland. The clinical variables presented in Table 1 were obtained from the clinical and pathological records and are in accordance with the UICC/TNM classification [21]. Cell culture assays were performed on non-metastatic cell line Caco-2 and metastatic cell lines SW480 and HT-29. Their metastatic potential has been investigated in murine liver metastasis models [22-24].\nClinical characteristics of patients with colorectal carcinomas\n1Tumor-node-metastasis; 2Colorectal carcinoma;\n3Median with range in parentheses.", "Tissue samples were collected immediately after resection, snap frozen in liquid nitrogen and then stored at -80°C until they were processed under nucleic acid sterile conditions for RNA and protein extraction. Tumor samples were taken from vital areas of histopathologically confirmed adenocarcinomas. As corresponding normal tissue we used adjacent unaffected mucosa, 2-3 cm distal to the resection margin from the same resected adenocarcinoma. All tissues obtained were reviewed by an experienced pathologist (M.W.) and examined for the presence of tumor cells. As minimum criteria for usefulness for our studies we only chose tumor tissues in which tumor cells occupied a major component (>75%) of the tumor sample.", "Total RNA was isolated using RNeasy columns from Qiagen (Hilden, Germany) according to the manufacturer's instructions. RNA integrity was confirmed spectrophotometrically and by electrophoresis on 1% agarose gels. For cDNA synthesis 5 μg of each patient total RNA sample were reverse-transcribed in a final reaction volume of 50 μL containing 1× TaqMan RT buffer, 2.5 μM/L random hexamers, 500 μM/L each dNTP, 5.5 mM/L MgCl2, 0.4 U/μl RNase inhibitor, and 1.25 U/μL Multiscribe RT. All RT-PCR reagents were purchased from Applied Biosystems (Foster City, CA). The reaction conditions were 10 min at 25°C, 30 min at 48°C, and 5 min at 95°C.", "All Q-RT PCR assays containing the primer and probe mix were purchased from Applied Biosystems, (Applied Biosystems, Foster City, CA) and utilized according to the manufacturer's instructions. PCR reactions were carried out using 10 μL 2× Taqman PCR Universal Master Mix No AmpErase® UNG and 1 μL gene assay (Applied Biosystems, Foster City, CA), 8 μL Rnase-free water and 1 μL cDNA template (50 mg/L). The theoretical basis of the qRT assays is described in detail elsewhere [25]. All reactions were run in triplicate along with no template controls and an additional reaction in which reverse transcriptase was omitted to assure absence of genomic DNA contamination in each RNA sample. For the signal detection, ABI Prism 7900 sequence detector was programmed to an initial step of 10 min at 95°C, followed by 40 thermal cycles of 15 s at 95°C and 10 min at 60°C and the log-linear phase of amplification was monitored to obtain CT values for each RNA sample.\nGene expression of all target genes was analyzed in relation to the levels of the slope matched housekeeping genes phosphomannomutase (PMM1) and β2-Microglobulin (β2 M) [26]. Data analysis was performed according to the relative standard curve method. Data are presented in relation to the respective housekeeping genes.", "Protein lysates from frozen tissues were extracted with the radioimmunoprecipitation (RIPA) cell lysis and extraction buffer from Pierce (Pierce, Rockford, USA). Total protein content was assessed by using the Pierce BCA protein assay reagent kit (Pierce, Rockford, USA).", "The chemokine protein levels in the different tissue lysates were determined by sandwich-type enzyme-linked immunosorbent assays (ELISA) according to the manufacturer's instructions. Samples were assayed in duplicate with all values calculated as the mean of the two measurements. CXCL12 levels were assayed using a validated commercial ELISA (Duo Set R&D Systems, DY350, Minneapolis, Minn., USA). The absorbance was read at 450 nm in a 96-well microtiter plate reader. The chemokine concentration from each tissue lysate was normalized to the total protein content of each sample.", "Operative specimens were routinely fixed in formalin and subsequently embedded in paraffin. Before staining, 4-μm thick paraffin-embedded tissue sections were mounted on Superfrost Plus slides, deparaffinized and rehydrated in graded ethanol to deionized water. The sections were microwaved with an antigen retrieval solution (Target Retrieval, Dakocytomation, Carpinteria, CA, USA) and after blocking of endogenous peroxidase activity with 3% hydrogen peroxide, the sections were sequential treated with avidin and biotin (Avidin/Biotin blocking kit, Vector Laboratories Inc., Burlingame, CA, USA). In addition the specimens were further blocked for 30 min at room temperature with normal rabbit serum. Overnight incubation at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK) was followed by incubation of secondary biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA). After colour reaction with aminoethylcarbazol solution (Merck, Darmstadt, Germany), tissues were counterstained with haematoxylin. Immunocytochemical staining of CRC cells was performed using culture slides from Becton Dickinson (Becton Dickinson, Heidelberg, Germany). After blocking of endogenous peroxidase activity and blocking for 30 min at room temperature with normal rabbit serum, cells were overnight incubated at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK). Following incubation of secondary biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA) and all further steps were carried out as described above for the immohistochemical staining procedure.\nNegative controls were conducted in all cases omitting primary antibody. For evaluation of immunocytochemical staining the total number of cells per 5 high-power fields (using ×40 -HPF objective magnification) was determined. Cells were considered positive, when they demonstrated strong and exclusive labelling.", "Cell migration assays were performed in triplicate using BD Falcon cell culture inserts containing polyethylene terephthalate membranes (8 μm pore size) from BD Biosciences (Bedford, MA, USA). Caco-2, HT-29 and SW480 cells at a concentration of 50000 cells per ml in 500 μl medium with antibiotics but without FCS were placed in the top of a two-chamber assay system and incubated for 48 hours with and without CXCL12 (350-NS-10, R&D Systems, Wiesbaden, Germany) in 800 μl medium without FCS placed in the lower chamber. CXCL12 concentrations in the receiver wells were 10 ng, 50 ng and 100 ng per ml, respectively. After the incubation period, the non-migrated cells on the upper surface of the filters were removed using a cotton swab. Migrated cells, which are adherent to the lower surface of the filters, were stained with a 0.1% crystal violet solution (Merck, Darmstadt, Germany). The number of these migrated cells was quantified microscopically by counting them in 10 high-power microscopic fields. Medium with 20% FCS placed in the lower chamber of the two-chamber assay system served as a positive control and reference in the migration assays.", "Cell proliferation assays (Roche Diagnostics GmbH, Mannheim, Germany) were performed in triplicate using flat-bottomed 96-well microtiter plates for culturing Caco-2, HT-29 and SW480 cells in the presence of CXCL12 at 37°C for 48 hours. Subsequently, bromodeoxyuridine (BrdU) was added to the cells and the cells were reincubated for 24 hours. During this labeling period the pyrimidine analogue BrdU is incorporated in place of thymidine into the DNA of proliferating cells. After removing the culture media, cells were fixed and anti-BrdU-POD was added to bind to the BrdU incorporated in newly synthesized cellular DNA. Immune complexes were detected by the subsequent substrate reaction and the reaction product was quantified by photometrically measuring the absorbance of the developed color.", "Inhibition assays were performed in analogy to the cell migration assays using BD Falcon cell culture inserts from BD Biosciences. Similar to the migration assays Caco-2, HT-29 and SW480 cells in 500 μl medium without FCS were mixed with anti-CXCR4 antibodies (MAB171, R&D systems) applied in a final concentration of 6 μg per ml and placed in the upper chamber of a two-chamber assay system. After incubation for 48 hours with and without CXCL12 in 800 μl medium without FCS placed in the lower chamber steps were continued as mentioned above for the migration assays and migrated CRC cells were quantified microscopically by counting the cells that had migrated into the filters.", "CXCR4 siRNA transfection of Caco-2, HT-29 and SW480 cells was performed according to the HiPerFect Transfection Reagent Handbook from Qiagen (Qiagen, Hilden, Germany) and the Dharmacon DharmaFECT siRNA transfection protocol (Thermo Fisher Scientific, Waltham, USA). Briefly, cells were trypsinized, counted and then diluted in antibiotic-free complete medium containing serum to achieve the appropriate plating density in 100 μl of solution. On average, 6 × 104 cells per well of a 24-well plate or 5 × 105 cells per well of a 6-well plate, respectively, were seeded and overnight incubated under their normal growth conditions. On the day of transfection, four different CXCR4 siRNAs (Qiagen, SI00052220 (1), SI00052227 (2), SI02664235 (7), SI02664242 (8)) were diluted to a final concentration of 10 nM in 100 μl culture medium without serum, respectively. 3 μl of HiPerFect (Qiagen) for the transfection of SW480 cells and 5 μl of Dharmacon DharmaFECT siRNA transfection reagent (Thermo Fisher Scientific) for the transfection of Caco-2 and HT-29 cells were added to the diluted siRNA reactions, respectively. Samples were incubated for 5-10 minutes at room temperature to allow formation of transfection complexes. Hence, the complexes were added drop-wise to the cells and incubated for three hours under their normal growth conditions. Gene silencing was monitored after 48 and 72 hours after transfection at the mRNA and at the protein level using Realtime PCR and western blotting, respectively. Further, gene silencing was monitored visually by observing the effect on a cell death control. Further, control samples with untransfected cells, a mock-transfection with only transfection reagent, a negative control siRNA and control samples with MAPK1 siRNA were used. A detailed overview of siRNA control experiments is presented in Table 2. Subsequent migration assays were performed as described above.\nApplied RNAi control experiments - overview", "Gene silencing was monitored 72 hours after transfection at the protein level using western blotting. Total protein (25 μg/lane) was separated by SDS-PAGE using a 10% gel and blotted onto nitrocellulose membranes (Hybond ECL, Amersham Biosciences, Piscataway, NJ, USA). Membranes were blocked by incubation in Tris-buffered saline (TBS) containing 5% non fat dry milk and 0.1% Tween 20 for 2 h at room temperature and then incubated overnight at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK). Blots were then washed and incubated at room temperature for 1 h with donkey anti-goat HRP antibody (diluted 1:5000, sc-2056, Santa Cruz Biotechnology, Santa Cruz, CA USA). Bands were visualized by ECL Western blotting analysis systems (Amersham Biosciences, Piscataway, NJ, USA). The human cell lysate HL-60 (sc-2209, Santa Cruz Biotechnology, Santa Cruz, CA, USA) served as positive control. Quantification of band intensities has been performed on three independent samples using image J software.", "Chemokine receptor expression profiles in the different groups are shown as mean and standard error of the mean (SEM). Statistical calculations were done with the MedCalc software package. Where appropriate, either the Student's t-test or the Wilcoxon's rank sum test was applied to test for group differences of continuous variables. A P value of 0.05 or less was considered significant.", "[SUBTITLE] CXCL12/CXCR4 expression in CRC tissues [SUBSECTION] Q-RT-PCR analysis of CRC tissue specimens revealed significant up-regulation of CXCR4 and significant down-regulation of CXCL12 compared to corresponding normal tissues (Figure 1A). These results were verified for CXCL12 on the protein level as shown by ELISA assays which demonstrated significant CXCL12 down-regulation in CRC tissues compared to the corresponding normal tissues (Figure 1B). Detection of CXCR4 expression was assessed by immunohistochemical staining. Immunostaining of the CRC tissues revealed positive cytoplasmic staining in 28 out of 50 CRC specimens (56%) as shown for a representative example in Figure 1C. However, no significant difference in CXCR4 expression was detected according to the metastatic behaviour. Patients with synchronous or metachronous colorectal liver metastases expressed CXCR4 in their primary tumor with no significant difference with respect to CRC patients who did not develop metastases.\nCXC12/CXCR4 expression in colorectal cancer (CRC) tissue specimens as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunohistochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05, n = 50. Fold increase above 1 indicates gene overexpression in affected tissues related to unaffected neighbor tissues. (B) Detection of CXCL12 protein concentrations (pg/ml pro mg total protein) in total cell lysates of CRC and adjacent normal tissues from CRC patients (n = 50). Protein data are expressed as mean +/- SEM, *P < 0.05. (C) Detection of CXCR4 protein expression in representative CRC specimens as assessed by immunohistochemical staining with CXCR4-specific antibodies showing positive cytoplasmic staining in CRC and in unaffected corresponding tissues (original magnification × 200 and × 400).\nQ-RT-PCR analysis of CRC tissue specimens revealed significant up-regulation of CXCR4 and significant down-regulation of CXCL12 compared to corresponding normal tissues (Figure 1A). These results were verified for CXCL12 on the protein level as shown by ELISA assays which demonstrated significant CXCL12 down-regulation in CRC tissues compared to the corresponding normal tissues (Figure 1B). Detection of CXCR4 expression was assessed by immunohistochemical staining. Immunostaining of the CRC tissues revealed positive cytoplasmic staining in 28 out of 50 CRC specimens (56%) as shown for a representative example in Figure 1C. However, no significant difference in CXCR4 expression was detected according to the metastatic behaviour. Patients with synchronous or metachronous colorectal liver metastases expressed CXCR4 in their primary tumor with no significant difference with respect to CRC patients who did not develop metastases.\nCXC12/CXCR4 expression in colorectal cancer (CRC) tissue specimens as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunohistochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05, n = 50. Fold increase above 1 indicates gene overexpression in affected tissues related to unaffected neighbor tissues. (B) Detection of CXCL12 protein concentrations (pg/ml pro mg total protein) in total cell lysates of CRC and adjacent normal tissues from CRC patients (n = 50). Protein data are expressed as mean +/- SEM, *P < 0.05. (C) Detection of CXCR4 protein expression in representative CRC specimens as assessed by immunohistochemical staining with CXCR4-specific antibodies showing positive cytoplasmic staining in CRC and in unaffected corresponding tissues (original magnification × 200 and × 400).\n[SUBTITLE] Organ-specific expression of CXCL12 [SUBSECTION] CXCL12 mRNA and protein expression as determined by RT-PCR and ELISA was investigated in different human organs such as pancreas, stomach, colon/rectum and esophagus related to the liver. On the protein level CXCL12 was found to exhibit peak levels of expression in the liver compared to stomach, esophagus, pancreas, colon and rectum (Figure 2). However, on the mRNA level CXCL12 did not show a markedly higher expression in the liver in comparison to other gastrointestinal organs or glands corresponding to previous results [27].\nCXCL12 protein expression in different human organs related to the liver as determined by ELISA. CXCL12 protein concentrations (pg/ml per mg total protein) were measured in total cell lysates of normal liver, pancreas, stomach, colon/rectum and esophagus tissues of 10 patients, respectively (n = 10). Protein data are expressed as mean +/- SEM, *P < 0.05.\nCXCL12 mRNA and protein expression as determined by RT-PCR and ELISA was investigated in different human organs such as pancreas, stomach, colon/rectum and esophagus related to the liver. On the protein level CXCL12 was found to exhibit peak levels of expression in the liver compared to stomach, esophagus, pancreas, colon and rectum (Figure 2). However, on the mRNA level CXCL12 did not show a markedly higher expression in the liver in comparison to other gastrointestinal organs or glands corresponding to previous results [27].\nCXCL12 protein expression in different human organs related to the liver as determined by ELISA. CXCL12 protein concentrations (pg/ml per mg total protein) were measured in total cell lysates of normal liver, pancreas, stomach, colon/rectum and esophagus tissues of 10 patients, respectively (n = 10). Protein data are expressed as mean +/- SEM, *P < 0.05.\n[SUBTITLE] CXCL12/CXCR4 expression in CRC cell lines [SUBSECTION] In non-metastatic cell line Caco-2 low CXCL12 expression was detected on the mRNA level by PCR and on the protein level in the cell culture supernatant by ELISA (Figure 3A and 3B, respectively). In metastatic cell lines SW480 and HT-29 CXCL12 mRNA and protein expression was below detection limit (Figure 3A and 3B, respectively). In contrast, CXCR4 mRNA expression was detected in HT-29 and SW480 cells with the latter displaying significant overexpression relative to B2M (Figure 3A). When CXCR4 protein expression was assessed by immunocytochemical staining, CXCR4 positive staining was observed in 22% of Caco-2 cells, 74% of HT-29 cells and 80% of SW480 cells as quantified microscopically and shown for a representative example in Figure 3C. Likewise, CXCR4 protein expression as determined by western blot analysis revealed similar expression rates in HT-29 and SW480 cells. Variation in CXCR4 mRNA and protein data may be due to posttranscriptional and posttranslational modifications as well as to the relative amplification modus of the qRT-PCR results with respect to housekeeping gene B2M.\nCXC12/CXCR4 expression in Caco-2, HT-29 and SW480 cells as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunocytochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05. Fold increase above 1 indicates gene overexpression related to housekeeping gene B2M. (B) Detection of CXCL12 protein concentrations in pg/ml in cell culture supernatant of Caco-2, HT-29 and SW480 cells. Protein data are expressed as mean +/- SEM. (C) Detection of CXCR4 protein expression in representative cell culture slides of Caco-2, HT-29 and SW480 cells as assessed by immunocytochemical staining with CXCR4-specific antibodies.\nIn non-metastatic cell line Caco-2 low CXCL12 expression was detected on the mRNA level by PCR and on the protein level in the cell culture supernatant by ELISA (Figure 3A and 3B, respectively). In metastatic cell lines SW480 and HT-29 CXCL12 mRNA and protein expression was below detection limit (Figure 3A and 3B, respectively). In contrast, CXCR4 mRNA expression was detected in HT-29 and SW480 cells with the latter displaying significant overexpression relative to B2M (Figure 3A). When CXCR4 protein expression was assessed by immunocytochemical staining, CXCR4 positive staining was observed in 22% of Caco-2 cells, 74% of HT-29 cells and 80% of SW480 cells as quantified microscopically and shown for a representative example in Figure 3C. Likewise, CXCR4 protein expression as determined by western blot analysis revealed similar expression rates in HT-29 and SW480 cells. Variation in CXCR4 mRNA and protein data may be due to posttranscriptional and posttranslational modifications as well as to the relative amplification modus of the qRT-PCR results with respect to housekeeping gene B2M.\nCXC12/CXCR4 expression in Caco-2, HT-29 and SW480 cells as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunocytochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05. Fold increase above 1 indicates gene overexpression related to housekeeping gene B2M. (B) Detection of CXCL12 protein concentrations in pg/ml in cell culture supernatant of Caco-2, HT-29 and SW480 cells. Protein data are expressed as mean +/- SEM. (C) Detection of CXCR4 protein expression in representative cell culture slides of Caco-2, HT-29 and SW480 cells as assessed by immunocytochemical staining with CXCR4-specific antibodies.\n[SUBTITLE] CXCR4 protein inhibition abrogates migration of colorectal cancer cells [SUBSECTION] All three cell lines were stimulated with 10, 50 and 100 ng/ml CXCL12, respectively, and incubated for 48 hours. While Caco-2 cells were not stimulated by any concentration of CXCL12 (Figure 4A), we observed a significant CXCL12 dose-independent stimulation of migration for HT-29 and SW480 cells (Figure 4B and 4C, respectively) (P < 0.05). Thus, CXCL12 was shown to be chemotactic for HT-29 and SW480 cells.\nCXCR4 protein inhibition abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to a positive control containing 20% FCS (PC) without inhibition and after inhibition with anti-CXCR4 antibody. (B) Percent HT-29 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody. (C) Percent SW480 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody.\nWhen we added neutralizing anti-CXCR4 antibody prior to performance of cell migration assays, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated at all CXCL12 concentrations under investigation (Figure 4B and 4C, respectively) (P < 0.05). Application of anti-CXCR4 antibodies had no impact on the migration potential of Caco-2 cells (Figure 4A).\nAll three cell lines were stimulated with 10, 50 and 100 ng/ml CXCL12, respectively, and incubated for 48 hours. While Caco-2 cells were not stimulated by any concentration of CXCL12 (Figure 4A), we observed a significant CXCL12 dose-independent stimulation of migration for HT-29 and SW480 cells (Figure 4B and 4C, respectively) (P < 0.05). Thus, CXCL12 was shown to be chemotactic for HT-29 and SW480 cells.\nCXCR4 protein inhibition abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to a positive control containing 20% FCS (PC) without inhibition and after inhibition with anti-CXCR4 antibody. (B) Percent HT-29 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody. (C) Percent SW480 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody.\nWhen we added neutralizing anti-CXCR4 antibody prior to performance of cell migration assays, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated at all CXCL12 concentrations under investigation (Figure 4B and 4C, respectively) (P < 0.05). Application of anti-CXCR4 antibodies had no impact on the migration potential of Caco-2 cells (Figure 4A).\n[SUBTITLE] CXC receptor-4 gene silencing abrogates migration of colorectal cancer cells [SUBSECTION] To analyze if the chemotactic effect of CXCL12 on the migration of HT-29 and SW480 cells could be counteracted by the down-regulation of the mRNA of corresponding receptor CXCR4, we applied four different CXCR4 siRNAs. The mean percent CXCR4 knockdown related to the siRNA negative control was 78% for Caco-2 cells, 80% HT-29 cells and 84% for SW480 cells as determined by RT-PCR 48 hours after transfection on the mRNA level (Figure 5A). On the protein level gene silencing was monitored 72 hours after transfection in western blot analyses (Figure 5B). Application of CXCR4 siRNAs had no impact on the migration potential of Caco-2 cells (Figure 6A). In contrast, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated by all four different CXCR4 siRNAs at all CXCL12 concentrations under investigation (Figure 6B and 6C, respectively) (P < 0.05).\nDemonstration of CXCR4 knockdown as determined by mRNA and protein analysis. (A) Mean percent CXCR4 knockdown related to siRNA negative control (NC) as determined by mRNA analysis in Caco-2, HT-29 and SW480 cells. (B) CXCR4 protein expression after CXCR4 siRNA silencing as determined by Western Blot analysis in Caco-2, HT-29 and SW480 cells related to untransfected total cell lysates. Total cell lysates of untransfected cells and one representative example of CXCR4 siRNA transfected cells of each cell line were immunoblotted with antibodies specifically recognizing chemokine receptor CXCR4. Acute leucemia cell line HL60 served as positive control for the detection of CXCR4.\nCXCR4 mRNA silencing abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to positive control containing 20% FCS (PC) without transfection and after transfection with 4 different CXCR4 siRNAs. (B) Percent HT-29 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs. (C) Percent SW480 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs.\nTo analyze if the chemotactic effect of CXCL12 on the migration of HT-29 and SW480 cells could be counteracted by the down-regulation of the mRNA of corresponding receptor CXCR4, we applied four different CXCR4 siRNAs. The mean percent CXCR4 knockdown related to the siRNA negative control was 78% for Caco-2 cells, 80% HT-29 cells and 84% for SW480 cells as determined by RT-PCR 48 hours after transfection on the mRNA level (Figure 5A). On the protein level gene silencing was monitored 72 hours after transfection in western blot analyses (Figure 5B). Application of CXCR4 siRNAs had no impact on the migration potential of Caco-2 cells (Figure 6A). In contrast, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated by all four different CXCR4 siRNAs at all CXCL12 concentrations under investigation (Figure 6B and 6C, respectively) (P < 0.05).\nDemonstration of CXCR4 knockdown as determined by mRNA and protein analysis. (A) Mean percent CXCR4 knockdown related to siRNA negative control (NC) as determined by mRNA analysis in Caco-2, HT-29 and SW480 cells. (B) CXCR4 protein expression after CXCR4 siRNA silencing as determined by Western Blot analysis in Caco-2, HT-29 and SW480 cells related to untransfected total cell lysates. Total cell lysates of untransfected cells and one representative example of CXCR4 siRNA transfected cells of each cell line were immunoblotted with antibodies specifically recognizing chemokine receptor CXCR4. Acute leucemia cell line HL60 served as positive control for the detection of CXCR4.\nCXCR4 mRNA silencing abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to positive control containing 20% FCS (PC) without transfection and after transfection with 4 different CXCR4 siRNAs. (B) Percent HT-29 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs. (C) Percent SW480 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs.\n[SUBTITLE] Proliferative rate of colorectal cancer cells after CXCR4 blockage by mRNA silencing and inhibition antibodies [SUBSECTION] CXCR4 blockage by mRNA silencing or anti-CXCR4 antibodies might restrain the proliferation of CRC cell lines, so that the decreased migration capacity of cancer cells might result from a lower proliferative rate. To ensure that the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing by all four different CXCR4 siRNAs is not resulting from a lower proliferative rate with respect to the untransfected cells, we investigated their proliferative capacity before and after siRNA transfection. In addition, we also included cell line Caco-2 in this survey. Likewise, we investigated the proliferative capacity of Caco-2, HT-29 and SW480 cells before and after CXCR4 blockage by neutralizing anti-CXCR4 antibody. Here, cells without anti-CXCR4 antibody served as reference. As presented in figure 7 there is no marked difference between the proliferation rates of Caco-2 (Figure 7A), HT-29 (Figure 7B) and SW480 (Figure 7C) cells before or after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody. Thus, the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody is not resulting from a reduced proliferation rate.\nProliferative rate of colorectal cancer cells after CXCR4 mRNA silencing or CXCR4 blockage by inhibition antibodies. Knockdown of endogeneous CXCR4 expression by four different CXCR4 siRNAs or CXCR4 blockage by ant-CXCR4 inhibition antibodies does not decelerate the proliferation rate of (A) Caco-2 cells, (B) HT-29 cells and (C) SW480 cells in relation to untransfected cells without inhibition antibody (control).\nCXCR4 blockage by mRNA silencing or anti-CXCR4 antibodies might restrain the proliferation of CRC cell lines, so that the decreased migration capacity of cancer cells might result from a lower proliferative rate. To ensure that the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing by all four different CXCR4 siRNAs is not resulting from a lower proliferative rate with respect to the untransfected cells, we investigated their proliferative capacity before and after siRNA transfection. In addition, we also included cell line Caco-2 in this survey. Likewise, we investigated the proliferative capacity of Caco-2, HT-29 and SW480 cells before and after CXCR4 blockage by neutralizing anti-CXCR4 antibody. Here, cells without anti-CXCR4 antibody served as reference. As presented in figure 7 there is no marked difference between the proliferation rates of Caco-2 (Figure 7A), HT-29 (Figure 7B) and SW480 (Figure 7C) cells before or after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody. Thus, the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody is not resulting from a reduced proliferation rate.\nProliferative rate of colorectal cancer cells after CXCR4 mRNA silencing or CXCR4 blockage by inhibition antibodies. Knockdown of endogeneous CXCR4 expression by four different CXCR4 siRNAs or CXCR4 blockage by ant-CXCR4 inhibition antibodies does not decelerate the proliferation rate of (A) Caco-2 cells, (B) HT-29 cells and (C) SW480 cells in relation to untransfected cells without inhibition antibody (control).", "Q-RT-PCR analysis of CRC tissue specimens revealed significant up-regulation of CXCR4 and significant down-regulation of CXCL12 compared to corresponding normal tissues (Figure 1A). These results were verified for CXCL12 on the protein level as shown by ELISA assays which demonstrated significant CXCL12 down-regulation in CRC tissues compared to the corresponding normal tissues (Figure 1B). Detection of CXCR4 expression was assessed by immunohistochemical staining. Immunostaining of the CRC tissues revealed positive cytoplasmic staining in 28 out of 50 CRC specimens (56%) as shown for a representative example in Figure 1C. However, no significant difference in CXCR4 expression was detected according to the metastatic behaviour. Patients with synchronous or metachronous colorectal liver metastases expressed CXCR4 in their primary tumor with no significant difference with respect to CRC patients who did not develop metastases.\nCXC12/CXCR4 expression in colorectal cancer (CRC) tissue specimens as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunohistochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05, n = 50. Fold increase above 1 indicates gene overexpression in affected tissues related to unaffected neighbor tissues. (B) Detection of CXCL12 protein concentrations (pg/ml pro mg total protein) in total cell lysates of CRC and adjacent normal tissues from CRC patients (n = 50). Protein data are expressed as mean +/- SEM, *P < 0.05. (C) Detection of CXCR4 protein expression in representative CRC specimens as assessed by immunohistochemical staining with CXCR4-specific antibodies showing positive cytoplasmic staining in CRC and in unaffected corresponding tissues (original magnification × 200 and × 400).", "CXCL12 mRNA and protein expression as determined by RT-PCR and ELISA was investigated in different human organs such as pancreas, stomach, colon/rectum and esophagus related to the liver. On the protein level CXCL12 was found to exhibit peak levels of expression in the liver compared to stomach, esophagus, pancreas, colon and rectum (Figure 2). However, on the mRNA level CXCL12 did not show a markedly higher expression in the liver in comparison to other gastrointestinal organs or glands corresponding to previous results [27].\nCXCL12 protein expression in different human organs related to the liver as determined by ELISA. CXCL12 protein concentrations (pg/ml per mg total protein) were measured in total cell lysates of normal liver, pancreas, stomach, colon/rectum and esophagus tissues of 10 patients, respectively (n = 10). Protein data are expressed as mean +/- SEM, *P < 0.05.", "In non-metastatic cell line Caco-2 low CXCL12 expression was detected on the mRNA level by PCR and on the protein level in the cell culture supernatant by ELISA (Figure 3A and 3B, respectively). In metastatic cell lines SW480 and HT-29 CXCL12 mRNA and protein expression was below detection limit (Figure 3A and 3B, respectively). In contrast, CXCR4 mRNA expression was detected in HT-29 and SW480 cells with the latter displaying significant overexpression relative to B2M (Figure 3A). When CXCR4 protein expression was assessed by immunocytochemical staining, CXCR4 positive staining was observed in 22% of Caco-2 cells, 74% of HT-29 cells and 80% of SW480 cells as quantified microscopically and shown for a representative example in Figure 3C. Likewise, CXCR4 protein expression as determined by western blot analysis revealed similar expression rates in HT-29 and SW480 cells. Variation in CXCR4 mRNA and protein data may be due to posttranscriptional and posttranslational modifications as well as to the relative amplification modus of the qRT-PCR results with respect to housekeeping gene B2M.\nCXC12/CXCR4 expression in Caco-2, HT-29 and SW480 cells as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunocytochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05. Fold increase above 1 indicates gene overexpression related to housekeeping gene B2M. (B) Detection of CXCL12 protein concentrations in pg/ml in cell culture supernatant of Caco-2, HT-29 and SW480 cells. Protein data are expressed as mean +/- SEM. (C) Detection of CXCR4 protein expression in representative cell culture slides of Caco-2, HT-29 and SW480 cells as assessed by immunocytochemical staining with CXCR4-specific antibodies.", "All three cell lines were stimulated with 10, 50 and 100 ng/ml CXCL12, respectively, and incubated for 48 hours. While Caco-2 cells were not stimulated by any concentration of CXCL12 (Figure 4A), we observed a significant CXCL12 dose-independent stimulation of migration for HT-29 and SW480 cells (Figure 4B and 4C, respectively) (P < 0.05). Thus, CXCL12 was shown to be chemotactic for HT-29 and SW480 cells.\nCXCR4 protein inhibition abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to a positive control containing 20% FCS (PC) without inhibition and after inhibition with anti-CXCR4 antibody. (B) Percent HT-29 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody. (C) Percent SW480 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody.\nWhen we added neutralizing anti-CXCR4 antibody prior to performance of cell migration assays, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated at all CXCL12 concentrations under investigation (Figure 4B and 4C, respectively) (P < 0.05). Application of anti-CXCR4 antibodies had no impact on the migration potential of Caco-2 cells (Figure 4A).", "To analyze if the chemotactic effect of CXCL12 on the migration of HT-29 and SW480 cells could be counteracted by the down-regulation of the mRNA of corresponding receptor CXCR4, we applied four different CXCR4 siRNAs. The mean percent CXCR4 knockdown related to the siRNA negative control was 78% for Caco-2 cells, 80% HT-29 cells and 84% for SW480 cells as determined by RT-PCR 48 hours after transfection on the mRNA level (Figure 5A). On the protein level gene silencing was monitored 72 hours after transfection in western blot analyses (Figure 5B). Application of CXCR4 siRNAs had no impact on the migration potential of Caco-2 cells (Figure 6A). In contrast, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated by all four different CXCR4 siRNAs at all CXCL12 concentrations under investigation (Figure 6B and 6C, respectively) (P < 0.05).\nDemonstration of CXCR4 knockdown as determined by mRNA and protein analysis. (A) Mean percent CXCR4 knockdown related to siRNA negative control (NC) as determined by mRNA analysis in Caco-2, HT-29 and SW480 cells. (B) CXCR4 protein expression after CXCR4 siRNA silencing as determined by Western Blot analysis in Caco-2, HT-29 and SW480 cells related to untransfected total cell lysates. Total cell lysates of untransfected cells and one representative example of CXCR4 siRNA transfected cells of each cell line were immunoblotted with antibodies specifically recognizing chemokine receptor CXCR4. Acute leucemia cell line HL60 served as positive control for the detection of CXCR4.\nCXCR4 mRNA silencing abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to positive control containing 20% FCS (PC) without transfection and after transfection with 4 different CXCR4 siRNAs. (B) Percent HT-29 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs. (C) Percent SW480 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs.", "CXCR4 blockage by mRNA silencing or anti-CXCR4 antibodies might restrain the proliferation of CRC cell lines, so that the decreased migration capacity of cancer cells might result from a lower proliferative rate. To ensure that the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing by all four different CXCR4 siRNAs is not resulting from a lower proliferative rate with respect to the untransfected cells, we investigated their proliferative capacity before and after siRNA transfection. In addition, we also included cell line Caco-2 in this survey. Likewise, we investigated the proliferative capacity of Caco-2, HT-29 and SW480 cells before and after CXCR4 blockage by neutralizing anti-CXCR4 antibody. Here, cells without anti-CXCR4 antibody served as reference. As presented in figure 7 there is no marked difference between the proliferation rates of Caco-2 (Figure 7A), HT-29 (Figure 7B) and SW480 (Figure 7C) cells before or after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody. Thus, the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody is not resulting from a reduced proliferation rate.\nProliferative rate of colorectal cancer cells after CXCR4 mRNA silencing or CXCR4 blockage by inhibition antibodies. Knockdown of endogeneous CXCR4 expression by four different CXCR4 siRNAs or CXCR4 blockage by ant-CXCR4 inhibition antibodies does not decelerate the proliferation rate of (A) Caco-2 cells, (B) HT-29 cells and (C) SW480 cells in relation to untransfected cells without inhibition antibody (control).", "The importance of the CXCL12/CXCR4 axis in tumor progression and metastasis has been emphasized. Recent evidence also suggests that the CXCL12/CXCR4 system is involved in the progression and metastasis of CRC [15,20,28].\nThus, the aim of this study was to conduct a comparative CXCL12/CXCR4 expression analysis and to investigate the functional role of CXCR4 in metastatic and non-metastatic CRC derived cell lines. For this purpose the migration potential of these cell lines was tested under conditions where CXCR4 was inhibited on the mRNA as well as on the protein level. On the mRNA level, CXCR4 silencing was achieved by CXCR4 RNA interference while on the protein level inhibition of CXCR4 protein activity was achieved using an anti-CXCR4 antibody.\nAs shown initially, CXCL12 protein is significantly higher expressed in the liver in relation to other intestinal organs or glands, thus supporting potential CXCL12 involvement in the homing of CRC cells to the liver. Subsequently, we observed significant up-regulation of CXCR4 and significant down-regulation of CXCL12 in CRC tissues. This inverse expression pattern is supported by previous studies, where CXCR4 was shown to be highly up-regulated in all T-stages of CRC tissues [16] and CXCL12 expression levels were shown to be low at the base of the crypts and increased in the more differentiated apical intestinal epithelial cells [29,30]. In compliance with the low CXCL12 expression status in tissues we observed low or no CXCL12 expression, respectively, in three human CRC derived cell lines, non-metastatic cell line Caco-2 and metastatic cell lines SW480 and HT-29. While low CXCL12 expression was detected in Caco-2 cells on the mRNA and on the protein level, CXCL12 was below the detection limit in HT-29 and SW480. It was hypothesized that changes in epithelial CXCL12 expression might contribute to CRC disease progression by allowing CRC cells to more readily sense CXCL12 from exogeneous sources hereby promoting metastasis [31]. Wendt et al. have shown that the reduced CXCL12 expression pattern in CRC tissues and cells is due to DNA hypermethylation in primary CRC and carcinoma-derived cell lines. Thus, it was demonstrated that inhibition or ablation of DNA methyltransferases prevent promoter methylation and restore CXCL12 expression. In addition, re-expression of functional, endogeneous CXCL12 in CRC cells was shown to reduce metastatic tumor formation significantly while silencing CXCL12 greatly enhanced the metastatic potential of CRC cells [31]. Moreover, endogeneous CXCL12 was shown to provide a barrier to metastasis by increasing anoikis via activation of a Bim-mediated intrinsic apoptotic pathway [32,33]. These results may constitute a plausible explanation for our observation of significantly reduced CXCL12 expression rates in CRC tissues and carcinoma-derived cell lines. Concordantly, an elevated migratory signaling response to ectopic CXCL12 was also shown to contribute to the metastatic potential of CXCR4-expressing mammary carcinoma cells, subsequent to epigenetic silencing of autocrine CXCL12 [34].\nFor CXCR4 we observed an inverse expression pattern in the three CRC-derived cell lines. Thus, expression in non-metastatic cell line Caco-2 was significantly lower compared to the metastatic cell lines under investigation. Also drug-resistant invasive HT-29 cells with a metastatic behaviour in immunodeficient mice were shown to exhibit high CXCR4 expression [35]. Responsible for promoting invasion in drug-resistant colon carcinoma cells is the autocrine CXCR4 ligand macrophage migration-inhibitory factor (MIF) as shown by Dessein et al. Impairing the MIF-CXCR4 signaling pathway, e.g. by silencing CXCR4, abolished this aggressive phenotype. To date, in various cancer entities aberrantly expressed chemokine receptors were demonstrated to contribute directly to the development of organ selective distant metastasis by interaction with their organ selectively expressed chemokine ligands. As a consequence of the low CXCR4 expression, Caco-2 cells showed no increase in migration in response to CXCL12 stimulation. In contrast, CXCL12 significantly increased cell migration in the CXCR4 expressing metastatic cell lines SW480 and HT-29. While a stimulative effect of CXCL12 on migration of CRC cell lines was described before [36], these effects were not significant and refer only to one CXCL12 concentration (100 ng/ml). Moreover, these tests were performed on CRC cell lines LS174T, SW620 and SW480 but not on cell lines HT-29 and Caco-2. To further investigate the functional role of CXCR4 we performed inhibition assays with anti-CXCR4 antibodies added previously to CXCL12 stimulation. The addition of the inhibition antibodies significantly blocked the CXCL12-dependent stimulation of HT-29 and SW480 cell migration but had no impact on Caco-2 migration. Our results correlate with inhibition studies performed on SW480 cells [37], where chemotaxis induced by CXCL12 was also shown to be blocked, although not as complete as we have shown for both SW480 and HT-29 cells. Likewise, Kim et al [38] reported on a partial inhibitory effect of anti-CXCR4 antibodies on CXCL12 stimulated melanoma and CRC cell migration (38%). However, to date no study analyzed the effect of CXCR4 gene silencing on CXCL12 mediated cell migration of CRC cells. Thus, we investigated if the chemotactic effect of CXCL12 on the migration of HT-29 and SW480 cells could be counteracted by the down-regulation of the CXCR4 mRNA. Applying four different CXCR4 siRNAs we achieved a mean percent CXCR4 knockdown of 78-84% and a significant abrogation of CXCL12 stimulated migration of HT-29 and SW480 for all four different CXCR4 siRNAs at all CXCL12 concentrations under investigation. On the other hand, application of CXCR4 siRNAs had no impact on the migration potential of Caco-2 cells. Thus, we have demonstrated the functional status of the CXCR4 receptor on CRC cell lines in response to CXCL12 on the mRNA level.\nIn CRC patients, no significant difference in CXCR4 expression was detected according to the metastatic behaviour. Based on our in vitro results CRC patients with metastasis may be expected to express more CXCR4 in their tumor cells compared to CRC patients without metastasis. However, many other important factors contribute to the metastatic properties of tumor tissues and influence their metastatic potential. When a CRC cell is turned metastatic, CXCR4 may be an important factor for the homing of such a tumor cell to its favourite organ destination, the liver.\nOur results are well in line with recent findings demonstrating a role of CXCL12 in promoting cell migration and tumor growth of CRC metastasis in vivo in a murine model [39]. However, while we observed no CXCR4-dependent proliferation of CRC cells, Shen et al. demonstrated CXCR4-induced proliferation for pancreatic cancer cells, where it was linked to AKT and ERK dependent pathways [40]. Moreover, CXCR4 knockdown by small interfering RNA was shown to inhibit cell proliferation and invasion of oral squamous cell carcinoma cells [41]. A recent study demonstrated that CXCL12/CXCR4 interactions may also promote early extravasation of liver metastatic epithelial tumor cells which determines a critical step in formation of organ-specific metastases [42]. Currently, various small-molecule chemokine receptor antagonist compounds are undergoing development in phase I to III studies in infectious and autoimmune diseases and a CCR5 inhibitor is already in the clinic for the treatment of HIV-infected patients.", "In conclusion, our results provide evidence that CXCR4 is up-regulated in CRC and stimulation of CXCR4 bearing cancer cells with CXCL12 led to increased migration, an effect which could be inhibited both by CXCR4 siRNA and neutralizing CXCR4 antibodies. Interestingly, CXCR4 was predominantly expressed in cell lines with metastatic potential and consequently, these cell lines showed increased migration after external CXCL12 stimulation. Our results suggest that the metastatic potential of CRC cells may be associated with the aberrant expression of CXCR4 and subsequently the ability of cells to interact with CXCL12 via autocrine and/or paracrine mechanisms.", "The authors declare that they have no competing interests.", "All authors read and approved the final manuscript. CR is responsible for the study concept and design and drafted the manuscript. VOF took part in the acquisition, analysis and interpretation of the data. PG is responsible for the critical assessment and revision of the manuscript. MW examined the tissue sections for the presence of tumor cells and histopathologically confirmed all tissues under investigation. CJ provided clinical information and SG participated in the statistical analysis. SKF performed the siRNA experiments and BV contributed to scientific discussion. OK provided clinical information. MKS is responsible for the provision of all the patient material and participated in the critical revision of the manuscript for important intellectual content." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Materials and patients", "Tissue preparation", "Single-strand cDNA synthesis", "Real-time PCR", "Isolation of total protein", "Sandwich-Type Enzyme-Linked Immunosorbent Assay", "Immunohistochemistry/Immunocytochemistry", "Cell migration assays", "Cell proliferation assays", "Cell inhibition assays", "siRNA assays", "Western Blot Analysis", "Calculations and Statistical Methods", "Results", "CXCL12/CXCR4 expression in CRC tissues", "Organ-specific expression of CXCL12", "CXCL12/CXCR4 expression in CRC cell lines", "CXCR4 protein inhibition abrogates migration of colorectal cancer cells", "CXC receptor-4 gene silencing abrogates migration of colorectal cancer cells", "Proliferative rate of colorectal cancer cells after CXCR4 blockage by mRNA silencing and inhibition antibodies", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Colorectal cancer (CRC) represents one of the most frequent malignancies worldwide with distant recurrence primarily affecting the liver as the predominant cause of CRC related mortality. The 5-year survival rate of 90% in patients with tumor restricted to the colon decreases to 10% in the presence of distant metastasis [1].\nRecently, various cancer-related studies demonstrated that specific chemokines and their receptors may be involved in the molecular mechanisms that control metastasis in the early stages of cancer development [2]. In this respect, the homeostatic chemokine CXCL12, a non-ELR+ CXC chemokine, has been implicated in promoting angiogenesis and metastasis [3,4]. CXCL12, also known as stromal derived factor 1 (SDF-1), is the only chemokine that is essential for survival [5] and a highly efficacious chemoattractant for T cells and thymocytes [6,7]. It is expressed by stromal cells such as fibroblasts and endothelial cells and signals exclusively via its G-protein-linked transmembrane receptor CXCR4 [8]. Expression of functional CXCR4 has been reported in various types of cancer cells [9-12], but also in immune cells such as peripheral blood lymphocytes, unprimed T cells, dendritic cells and lymphocytic leukemia B cells, where CXCR4 mediates spontaneous migration beneath bone marrow and stromal cells [13]. In infectious disease CXCR4 is employed by the human immunodefiency virus (HIV) to gain entry to cells [14] and in stem cell localization CXCR4 plays an important role for B-cell lymphopoiesis and bone marrow myelopoiesis [5]. In breast cancer, CXCR4 signaling was shown to play a crucial role in distant recurrence by mediating actin polymerisation and pseudopodia formation, thus, leading to chemotactic and invasive responses [3].\nRecently, CXCR4 expression was associated with advanced tumor stage and the development and recurrence of lymphatic or distant colorectal liver metastases (CRLM) [15-17]. Thus, it was reported that CXCR4 is differentially expressed in CRC and significantly correlates with survival, recurrence and liver metastasis [15,18]. Moreover, it was shown that concomitant and high expression of CXCR4 and VEGF is a strong and independent predictor of early distant relapse in CRC [19] and recent evidence indicates that CXCR4 may also play a role in tumor angiogenesis of CRC [20].\nDespite increasing knowledge about the expression of CXCL12/CXCR4 in CRC and its involvement of CXCR4 in the invasion and dissemination of CRC the precise mechanisms of CXCL12/CXCR4 interactions and subsequent metastasis are not entirely known. We therefore performed a comparative CXCL12/CXCR4 expression analysis and investigated how external addition of CXCL12 would promote CXCR4-mediated migration of CRC cell lines with different metastatic capabilities and how inhibition of CXCR4 by either CXCR4 siRNAs or neutralizing anti-CXCR4 antibodies would influence their migration potential.", "[SUBTITLE] Materials and patients [SUBSECTION] Surgical specimens and corresponding normal tissue from the same samples were collected from patients who underwent surgical resection at our department between 2003 and 2006. No patient underwent any specific cancer therapy prior to the resection. Our patient collectives comprised patients with CRC of different cancer stages (n = 50). In cases of organ confined CRC patients underwent resection of the primary tumor. Adjacent, disease free tissue samples of the colon and rectum served as control groups, respectively. Further, 10 specimens from patients with liver ruptures, focal nodular hyperplasia and haemangiomas as well as 10 unaffected tumor neighbor tissues from primary esophageal, pancreatic, gastric and colorectal carcinoma, respectively, were assessed. Informed consent for tissue procurement was obtained from all patients. The study was approved by the ethics commission of the Ärztekammer of the Saarland. The clinical variables presented in Table 1 were obtained from the clinical and pathological records and are in accordance with the UICC/TNM classification [21]. Cell culture assays were performed on non-metastatic cell line Caco-2 and metastatic cell lines SW480 and HT-29. Their metastatic potential has been investigated in murine liver metastasis models [22-24].\nClinical characteristics of patients with colorectal carcinomas\n1Tumor-node-metastasis; 2Colorectal carcinoma;\n3Median with range in parentheses.\nSurgical specimens and corresponding normal tissue from the same samples were collected from patients who underwent surgical resection at our department between 2003 and 2006. No patient underwent any specific cancer therapy prior to the resection. Our patient collectives comprised patients with CRC of different cancer stages (n = 50). In cases of organ confined CRC patients underwent resection of the primary tumor. Adjacent, disease free tissue samples of the colon and rectum served as control groups, respectively. Further, 10 specimens from patients with liver ruptures, focal nodular hyperplasia and haemangiomas as well as 10 unaffected tumor neighbor tissues from primary esophageal, pancreatic, gastric and colorectal carcinoma, respectively, were assessed. Informed consent for tissue procurement was obtained from all patients. The study was approved by the ethics commission of the Ärztekammer of the Saarland. The clinical variables presented in Table 1 were obtained from the clinical and pathological records and are in accordance with the UICC/TNM classification [21]. Cell culture assays were performed on non-metastatic cell line Caco-2 and metastatic cell lines SW480 and HT-29. Their metastatic potential has been investigated in murine liver metastasis models [22-24].\nClinical characteristics of patients with colorectal carcinomas\n1Tumor-node-metastasis; 2Colorectal carcinoma;\n3Median with range in parentheses.\n[SUBTITLE] Tissue preparation [SUBSECTION] Tissue samples were collected immediately after resection, snap frozen in liquid nitrogen and then stored at -80°C until they were processed under nucleic acid sterile conditions for RNA and protein extraction. Tumor samples were taken from vital areas of histopathologically confirmed adenocarcinomas. As corresponding normal tissue we used adjacent unaffected mucosa, 2-3 cm distal to the resection margin from the same resected adenocarcinoma. All tissues obtained were reviewed by an experienced pathologist (M.W.) and examined for the presence of tumor cells. As minimum criteria for usefulness for our studies we only chose tumor tissues in which tumor cells occupied a major component (>75%) of the tumor sample.\nTissue samples were collected immediately after resection, snap frozen in liquid nitrogen and then stored at -80°C until they were processed under nucleic acid sterile conditions for RNA and protein extraction. Tumor samples were taken from vital areas of histopathologically confirmed adenocarcinomas. As corresponding normal tissue we used adjacent unaffected mucosa, 2-3 cm distal to the resection margin from the same resected adenocarcinoma. All tissues obtained were reviewed by an experienced pathologist (M.W.) and examined for the presence of tumor cells. As minimum criteria for usefulness for our studies we only chose tumor tissues in which tumor cells occupied a major component (>75%) of the tumor sample.\n[SUBTITLE] Single-strand cDNA synthesis [SUBSECTION] Total RNA was isolated using RNeasy columns from Qiagen (Hilden, Germany) according to the manufacturer's instructions. RNA integrity was confirmed spectrophotometrically and by electrophoresis on 1% agarose gels. For cDNA synthesis 5 μg of each patient total RNA sample were reverse-transcribed in a final reaction volume of 50 μL containing 1× TaqMan RT buffer, 2.5 μM/L random hexamers, 500 μM/L each dNTP, 5.5 mM/L MgCl2, 0.4 U/μl RNase inhibitor, and 1.25 U/μL Multiscribe RT. All RT-PCR reagents were purchased from Applied Biosystems (Foster City, CA). The reaction conditions were 10 min at 25°C, 30 min at 48°C, and 5 min at 95°C.\nTotal RNA was isolated using RNeasy columns from Qiagen (Hilden, Germany) according to the manufacturer's instructions. RNA integrity was confirmed spectrophotometrically and by electrophoresis on 1% agarose gels. For cDNA synthesis 5 μg of each patient total RNA sample were reverse-transcribed in a final reaction volume of 50 μL containing 1× TaqMan RT buffer, 2.5 μM/L random hexamers, 500 μM/L each dNTP, 5.5 mM/L MgCl2, 0.4 U/μl RNase inhibitor, and 1.25 U/μL Multiscribe RT. All RT-PCR reagents were purchased from Applied Biosystems (Foster City, CA). The reaction conditions were 10 min at 25°C, 30 min at 48°C, and 5 min at 95°C.\n[SUBTITLE] Real-time PCR [SUBSECTION] All Q-RT PCR assays containing the primer and probe mix were purchased from Applied Biosystems, (Applied Biosystems, Foster City, CA) and utilized according to the manufacturer's instructions. PCR reactions were carried out using 10 μL 2× Taqman PCR Universal Master Mix No AmpErase® UNG and 1 μL gene assay (Applied Biosystems, Foster City, CA), 8 μL Rnase-free water and 1 μL cDNA template (50 mg/L). The theoretical basis of the qRT assays is described in detail elsewhere [25]. All reactions were run in triplicate along with no template controls and an additional reaction in which reverse transcriptase was omitted to assure absence of genomic DNA contamination in each RNA sample. For the signal detection, ABI Prism 7900 sequence detector was programmed to an initial step of 10 min at 95°C, followed by 40 thermal cycles of 15 s at 95°C and 10 min at 60°C and the log-linear phase of amplification was monitored to obtain CT values for each RNA sample.\nGene expression of all target genes was analyzed in relation to the levels of the slope matched housekeeping genes phosphomannomutase (PMM1) and β2-Microglobulin (β2 M) [26]. Data analysis was performed according to the relative standard curve method. Data are presented in relation to the respective housekeeping genes.\nAll Q-RT PCR assays containing the primer and probe mix were purchased from Applied Biosystems, (Applied Biosystems, Foster City, CA) and utilized according to the manufacturer's instructions. PCR reactions were carried out using 10 μL 2× Taqman PCR Universal Master Mix No AmpErase® UNG and 1 μL gene assay (Applied Biosystems, Foster City, CA), 8 μL Rnase-free water and 1 μL cDNA template (50 mg/L). The theoretical basis of the qRT assays is described in detail elsewhere [25]. All reactions were run in triplicate along with no template controls and an additional reaction in which reverse transcriptase was omitted to assure absence of genomic DNA contamination in each RNA sample. For the signal detection, ABI Prism 7900 sequence detector was programmed to an initial step of 10 min at 95°C, followed by 40 thermal cycles of 15 s at 95°C and 10 min at 60°C and the log-linear phase of amplification was monitored to obtain CT values for each RNA sample.\nGene expression of all target genes was analyzed in relation to the levels of the slope matched housekeeping genes phosphomannomutase (PMM1) and β2-Microglobulin (β2 M) [26]. Data analysis was performed according to the relative standard curve method. Data are presented in relation to the respective housekeeping genes.\n[SUBTITLE] Isolation of total protein [SUBSECTION] Protein lysates from frozen tissues were extracted with the radioimmunoprecipitation (RIPA) cell lysis and extraction buffer from Pierce (Pierce, Rockford, USA). Total protein content was assessed by using the Pierce BCA protein assay reagent kit (Pierce, Rockford, USA).\nProtein lysates from frozen tissues were extracted with the radioimmunoprecipitation (RIPA) cell lysis and extraction buffer from Pierce (Pierce, Rockford, USA). Total protein content was assessed by using the Pierce BCA protein assay reagent kit (Pierce, Rockford, USA).\n[SUBTITLE] Sandwich-Type Enzyme-Linked Immunosorbent Assay [SUBSECTION] The chemokine protein levels in the different tissue lysates were determined by sandwich-type enzyme-linked immunosorbent assays (ELISA) according to the manufacturer's instructions. Samples were assayed in duplicate with all values calculated as the mean of the two measurements. CXCL12 levels were assayed using a validated commercial ELISA (Duo Set R&D Systems, DY350, Minneapolis, Minn., USA). The absorbance was read at 450 nm in a 96-well microtiter plate reader. The chemokine concentration from each tissue lysate was normalized to the total protein content of each sample.\nThe chemokine protein levels in the different tissue lysates were determined by sandwich-type enzyme-linked immunosorbent assays (ELISA) according to the manufacturer's instructions. Samples were assayed in duplicate with all values calculated as the mean of the two measurements. CXCL12 levels were assayed using a validated commercial ELISA (Duo Set R&D Systems, DY350, Minneapolis, Minn., USA). The absorbance was read at 450 nm in a 96-well microtiter plate reader. The chemokine concentration from each tissue lysate was normalized to the total protein content of each sample.\n[SUBTITLE] Immunohistochemistry/Immunocytochemistry [SUBSECTION] Operative specimens were routinely fixed in formalin and subsequently embedded in paraffin. Before staining, 4-μm thick paraffin-embedded tissue sections were mounted on Superfrost Plus slides, deparaffinized and rehydrated in graded ethanol to deionized water. The sections were microwaved with an antigen retrieval solution (Target Retrieval, Dakocytomation, Carpinteria, CA, USA) and after blocking of endogenous peroxidase activity with 3% hydrogen peroxide, the sections were sequential treated with avidin and biotin (Avidin/Biotin blocking kit, Vector Laboratories Inc., Burlingame, CA, USA). In addition the specimens were further blocked for 30 min at room temperature with normal rabbit serum. Overnight incubation at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK) was followed by incubation of secondary biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA). After colour reaction with aminoethylcarbazol solution (Merck, Darmstadt, Germany), tissues were counterstained with haematoxylin. Immunocytochemical staining of CRC cells was performed using culture slides from Becton Dickinson (Becton Dickinson, Heidelberg, Germany). After blocking of endogenous peroxidase activity and blocking for 30 min at room temperature with normal rabbit serum, cells were overnight incubated at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK). Following incubation of secondary biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA) and all further steps were carried out as described above for the immohistochemical staining procedure.\nNegative controls were conducted in all cases omitting primary antibody. For evaluation of immunocytochemical staining the total number of cells per 5 high-power fields (using ×40 -HPF objective magnification) was determined. Cells were considered positive, when they demonstrated strong and exclusive labelling.\nOperative specimens were routinely fixed in formalin and subsequently embedded in paraffin. Before staining, 4-μm thick paraffin-embedded tissue sections were mounted on Superfrost Plus slides, deparaffinized and rehydrated in graded ethanol to deionized water. The sections were microwaved with an antigen retrieval solution (Target Retrieval, Dakocytomation, Carpinteria, CA, USA) and after blocking of endogenous peroxidase activity with 3% hydrogen peroxide, the sections were sequential treated with avidin and biotin (Avidin/Biotin blocking kit, Vector Laboratories Inc., Burlingame, CA, USA). In addition the specimens were further blocked for 30 min at room temperature with normal rabbit serum. Overnight incubation at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK) was followed by incubation of secondary biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA). After colour reaction with aminoethylcarbazol solution (Merck, Darmstadt, Germany), tissues were counterstained with haematoxylin. Immunocytochemical staining of CRC cells was performed using culture slides from Becton Dickinson (Becton Dickinson, Heidelberg, Germany). After blocking of endogenous peroxidase activity and blocking for 30 min at room temperature with normal rabbit serum, cells were overnight incubated at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK). Following incubation of secondary biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA) and all further steps were carried out as described above for the immohistochemical staining procedure.\nNegative controls were conducted in all cases omitting primary antibody. For evaluation of immunocytochemical staining the total number of cells per 5 high-power fields (using ×40 -HPF objective magnification) was determined. Cells were considered positive, when they demonstrated strong and exclusive labelling.\n[SUBTITLE] Cell migration assays [SUBSECTION] Cell migration assays were performed in triplicate using BD Falcon cell culture inserts containing polyethylene terephthalate membranes (8 μm pore size) from BD Biosciences (Bedford, MA, USA). Caco-2, HT-29 and SW480 cells at a concentration of 50000 cells per ml in 500 μl medium with antibiotics but without FCS were placed in the top of a two-chamber assay system and incubated for 48 hours with and without CXCL12 (350-NS-10, R&D Systems, Wiesbaden, Germany) in 800 μl medium without FCS placed in the lower chamber. CXCL12 concentrations in the receiver wells were 10 ng, 50 ng and 100 ng per ml, respectively. After the incubation period, the non-migrated cells on the upper surface of the filters were removed using a cotton swab. Migrated cells, which are adherent to the lower surface of the filters, were stained with a 0.1% crystal violet solution (Merck, Darmstadt, Germany). The number of these migrated cells was quantified microscopically by counting them in 10 high-power microscopic fields. Medium with 20% FCS placed in the lower chamber of the two-chamber assay system served as a positive control and reference in the migration assays.\nCell migration assays were performed in triplicate using BD Falcon cell culture inserts containing polyethylene terephthalate membranes (8 μm pore size) from BD Biosciences (Bedford, MA, USA). Caco-2, HT-29 and SW480 cells at a concentration of 50000 cells per ml in 500 μl medium with antibiotics but without FCS were placed in the top of a two-chamber assay system and incubated for 48 hours with and without CXCL12 (350-NS-10, R&D Systems, Wiesbaden, Germany) in 800 μl medium without FCS placed in the lower chamber. CXCL12 concentrations in the receiver wells were 10 ng, 50 ng and 100 ng per ml, respectively. After the incubation period, the non-migrated cells on the upper surface of the filters were removed using a cotton swab. Migrated cells, which are adherent to the lower surface of the filters, were stained with a 0.1% crystal violet solution (Merck, Darmstadt, Germany). The number of these migrated cells was quantified microscopically by counting them in 10 high-power microscopic fields. Medium with 20% FCS placed in the lower chamber of the two-chamber assay system served as a positive control and reference in the migration assays.\n[SUBTITLE] Cell proliferation assays [SUBSECTION] Cell proliferation assays (Roche Diagnostics GmbH, Mannheim, Germany) were performed in triplicate using flat-bottomed 96-well microtiter plates for culturing Caco-2, HT-29 and SW480 cells in the presence of CXCL12 at 37°C for 48 hours. Subsequently, bromodeoxyuridine (BrdU) was added to the cells and the cells were reincubated for 24 hours. During this labeling period the pyrimidine analogue BrdU is incorporated in place of thymidine into the DNA of proliferating cells. After removing the culture media, cells were fixed and anti-BrdU-POD was added to bind to the BrdU incorporated in newly synthesized cellular DNA. Immune complexes were detected by the subsequent substrate reaction and the reaction product was quantified by photometrically measuring the absorbance of the developed color.\nCell proliferation assays (Roche Diagnostics GmbH, Mannheim, Germany) were performed in triplicate using flat-bottomed 96-well microtiter plates for culturing Caco-2, HT-29 and SW480 cells in the presence of CXCL12 at 37°C for 48 hours. Subsequently, bromodeoxyuridine (BrdU) was added to the cells and the cells were reincubated for 24 hours. During this labeling period the pyrimidine analogue BrdU is incorporated in place of thymidine into the DNA of proliferating cells. After removing the culture media, cells were fixed and anti-BrdU-POD was added to bind to the BrdU incorporated in newly synthesized cellular DNA. Immune complexes were detected by the subsequent substrate reaction and the reaction product was quantified by photometrically measuring the absorbance of the developed color.\n[SUBTITLE] Cell inhibition assays [SUBSECTION] Inhibition assays were performed in analogy to the cell migration assays using BD Falcon cell culture inserts from BD Biosciences. Similar to the migration assays Caco-2, HT-29 and SW480 cells in 500 μl medium without FCS were mixed with anti-CXCR4 antibodies (MAB171, R&D systems) applied in a final concentration of 6 μg per ml and placed in the upper chamber of a two-chamber assay system. After incubation for 48 hours with and without CXCL12 in 800 μl medium without FCS placed in the lower chamber steps were continued as mentioned above for the migration assays and migrated CRC cells were quantified microscopically by counting the cells that had migrated into the filters.\nInhibition assays were performed in analogy to the cell migration assays using BD Falcon cell culture inserts from BD Biosciences. Similar to the migration assays Caco-2, HT-29 and SW480 cells in 500 μl medium without FCS were mixed with anti-CXCR4 antibodies (MAB171, R&D systems) applied in a final concentration of 6 μg per ml and placed in the upper chamber of a two-chamber assay system. After incubation for 48 hours with and without CXCL12 in 800 μl medium without FCS placed in the lower chamber steps were continued as mentioned above for the migration assays and migrated CRC cells were quantified microscopically by counting the cells that had migrated into the filters.\n[SUBTITLE] siRNA assays [SUBSECTION] CXCR4 siRNA transfection of Caco-2, HT-29 and SW480 cells was performed according to the HiPerFect Transfection Reagent Handbook from Qiagen (Qiagen, Hilden, Germany) and the Dharmacon DharmaFECT siRNA transfection protocol (Thermo Fisher Scientific, Waltham, USA). Briefly, cells were trypsinized, counted and then diluted in antibiotic-free complete medium containing serum to achieve the appropriate plating density in 100 μl of solution. On average, 6 × 104 cells per well of a 24-well plate or 5 × 105 cells per well of a 6-well plate, respectively, were seeded and overnight incubated under their normal growth conditions. On the day of transfection, four different CXCR4 siRNAs (Qiagen, SI00052220 (1), SI00052227 (2), SI02664235 (7), SI02664242 (8)) were diluted to a final concentration of 10 nM in 100 μl culture medium without serum, respectively. 3 μl of HiPerFect (Qiagen) for the transfection of SW480 cells and 5 μl of Dharmacon DharmaFECT siRNA transfection reagent (Thermo Fisher Scientific) for the transfection of Caco-2 and HT-29 cells were added to the diluted siRNA reactions, respectively. Samples were incubated for 5-10 minutes at room temperature to allow formation of transfection complexes. Hence, the complexes were added drop-wise to the cells and incubated for three hours under their normal growth conditions. Gene silencing was monitored after 48 and 72 hours after transfection at the mRNA and at the protein level using Realtime PCR and western blotting, respectively. Further, gene silencing was monitored visually by observing the effect on a cell death control. Further, control samples with untransfected cells, a mock-transfection with only transfection reagent, a negative control siRNA and control samples with MAPK1 siRNA were used. A detailed overview of siRNA control experiments is presented in Table 2. Subsequent migration assays were performed as described above.\nApplied RNAi control experiments - overview\nCXCR4 siRNA transfection of Caco-2, HT-29 and SW480 cells was performed according to the HiPerFect Transfection Reagent Handbook from Qiagen (Qiagen, Hilden, Germany) and the Dharmacon DharmaFECT siRNA transfection protocol (Thermo Fisher Scientific, Waltham, USA). Briefly, cells were trypsinized, counted and then diluted in antibiotic-free complete medium containing serum to achieve the appropriate plating density in 100 μl of solution. On average, 6 × 104 cells per well of a 24-well plate or 5 × 105 cells per well of a 6-well plate, respectively, were seeded and overnight incubated under their normal growth conditions. On the day of transfection, four different CXCR4 siRNAs (Qiagen, SI00052220 (1), SI00052227 (2), SI02664235 (7), SI02664242 (8)) were diluted to a final concentration of 10 nM in 100 μl culture medium without serum, respectively. 3 μl of HiPerFect (Qiagen) for the transfection of SW480 cells and 5 μl of Dharmacon DharmaFECT siRNA transfection reagent (Thermo Fisher Scientific) for the transfection of Caco-2 and HT-29 cells were added to the diluted siRNA reactions, respectively. Samples were incubated for 5-10 minutes at room temperature to allow formation of transfection complexes. Hence, the complexes were added drop-wise to the cells and incubated for three hours under their normal growth conditions. Gene silencing was monitored after 48 and 72 hours after transfection at the mRNA and at the protein level using Realtime PCR and western blotting, respectively. Further, gene silencing was monitored visually by observing the effect on a cell death control. Further, control samples with untransfected cells, a mock-transfection with only transfection reagent, a negative control siRNA and control samples with MAPK1 siRNA were used. A detailed overview of siRNA control experiments is presented in Table 2. Subsequent migration assays were performed as described above.\nApplied RNAi control experiments - overview\n[SUBTITLE] Western Blot Analysis [SUBSECTION] Gene silencing was monitored 72 hours after transfection at the protein level using western blotting. Total protein (25 μg/lane) was separated by SDS-PAGE using a 10% gel and blotted onto nitrocellulose membranes (Hybond ECL, Amersham Biosciences, Piscataway, NJ, USA). Membranes were blocked by incubation in Tris-buffered saline (TBS) containing 5% non fat dry milk and 0.1% Tween 20 for 2 h at room temperature and then incubated overnight at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK). Blots were then washed and incubated at room temperature for 1 h with donkey anti-goat HRP antibody (diluted 1:5000, sc-2056, Santa Cruz Biotechnology, Santa Cruz, CA USA). Bands were visualized by ECL Western blotting analysis systems (Amersham Biosciences, Piscataway, NJ, USA). The human cell lysate HL-60 (sc-2209, Santa Cruz Biotechnology, Santa Cruz, CA, USA) served as positive control. Quantification of band intensities has been performed on three independent samples using image J software.\nGene silencing was monitored 72 hours after transfection at the protein level using western blotting. Total protein (25 μg/lane) was separated by SDS-PAGE using a 10% gel and blotted onto nitrocellulose membranes (Hybond ECL, Amersham Biosciences, Piscataway, NJ, USA). Membranes were blocked by incubation in Tris-buffered saline (TBS) containing 5% non fat dry milk and 0.1% Tween 20 for 2 h at room temperature and then incubated overnight at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK). Blots were then washed and incubated at room temperature for 1 h with donkey anti-goat HRP antibody (diluted 1:5000, sc-2056, Santa Cruz Biotechnology, Santa Cruz, CA USA). Bands were visualized by ECL Western blotting analysis systems (Amersham Biosciences, Piscataway, NJ, USA). The human cell lysate HL-60 (sc-2209, Santa Cruz Biotechnology, Santa Cruz, CA, USA) served as positive control. Quantification of band intensities has been performed on three independent samples using image J software.\n[SUBTITLE] Calculations and Statistical Methods [SUBSECTION] Chemokine receptor expression profiles in the different groups are shown as mean and standard error of the mean (SEM). Statistical calculations were done with the MedCalc software package. Where appropriate, either the Student's t-test or the Wilcoxon's rank sum test was applied to test for group differences of continuous variables. A P value of 0.05 or less was considered significant.\nChemokine receptor expression profiles in the different groups are shown as mean and standard error of the mean (SEM). Statistical calculations were done with the MedCalc software package. Where appropriate, either the Student's t-test or the Wilcoxon's rank sum test was applied to test for group differences of continuous variables. A P value of 0.05 or less was considered significant.", "Surgical specimens and corresponding normal tissue from the same samples were collected from patients who underwent surgical resection at our department between 2003 and 2006. No patient underwent any specific cancer therapy prior to the resection. Our patient collectives comprised patients with CRC of different cancer stages (n = 50). In cases of organ confined CRC patients underwent resection of the primary tumor. Adjacent, disease free tissue samples of the colon and rectum served as control groups, respectively. Further, 10 specimens from patients with liver ruptures, focal nodular hyperplasia and haemangiomas as well as 10 unaffected tumor neighbor tissues from primary esophageal, pancreatic, gastric and colorectal carcinoma, respectively, were assessed. Informed consent for tissue procurement was obtained from all patients. The study was approved by the ethics commission of the Ärztekammer of the Saarland. The clinical variables presented in Table 1 were obtained from the clinical and pathological records and are in accordance with the UICC/TNM classification [21]. Cell culture assays were performed on non-metastatic cell line Caco-2 and metastatic cell lines SW480 and HT-29. Their metastatic potential has been investigated in murine liver metastasis models [22-24].\nClinical characteristics of patients with colorectal carcinomas\n1Tumor-node-metastasis; 2Colorectal carcinoma;\n3Median with range in parentheses.", "Tissue samples were collected immediately after resection, snap frozen in liquid nitrogen and then stored at -80°C until they were processed under nucleic acid sterile conditions for RNA and protein extraction. Tumor samples were taken from vital areas of histopathologically confirmed adenocarcinomas. As corresponding normal tissue we used adjacent unaffected mucosa, 2-3 cm distal to the resection margin from the same resected adenocarcinoma. All tissues obtained were reviewed by an experienced pathologist (M.W.) and examined for the presence of tumor cells. As minimum criteria for usefulness for our studies we only chose tumor tissues in which tumor cells occupied a major component (>75%) of the tumor sample.", "Total RNA was isolated using RNeasy columns from Qiagen (Hilden, Germany) according to the manufacturer's instructions. RNA integrity was confirmed spectrophotometrically and by electrophoresis on 1% agarose gels. For cDNA synthesis 5 μg of each patient total RNA sample were reverse-transcribed in a final reaction volume of 50 μL containing 1× TaqMan RT buffer, 2.5 μM/L random hexamers, 500 μM/L each dNTP, 5.5 mM/L MgCl2, 0.4 U/μl RNase inhibitor, and 1.25 U/μL Multiscribe RT. All RT-PCR reagents were purchased from Applied Biosystems (Foster City, CA). The reaction conditions were 10 min at 25°C, 30 min at 48°C, and 5 min at 95°C.", "All Q-RT PCR assays containing the primer and probe mix were purchased from Applied Biosystems, (Applied Biosystems, Foster City, CA) and utilized according to the manufacturer's instructions. PCR reactions were carried out using 10 μL 2× Taqman PCR Universal Master Mix No AmpErase® UNG and 1 μL gene assay (Applied Biosystems, Foster City, CA), 8 μL Rnase-free water and 1 μL cDNA template (50 mg/L). The theoretical basis of the qRT assays is described in detail elsewhere [25]. All reactions were run in triplicate along with no template controls and an additional reaction in which reverse transcriptase was omitted to assure absence of genomic DNA contamination in each RNA sample. For the signal detection, ABI Prism 7900 sequence detector was programmed to an initial step of 10 min at 95°C, followed by 40 thermal cycles of 15 s at 95°C and 10 min at 60°C and the log-linear phase of amplification was monitored to obtain CT values for each RNA sample.\nGene expression of all target genes was analyzed in relation to the levels of the slope matched housekeeping genes phosphomannomutase (PMM1) and β2-Microglobulin (β2 M) [26]. Data analysis was performed according to the relative standard curve method. Data are presented in relation to the respective housekeeping genes.", "Protein lysates from frozen tissues were extracted with the radioimmunoprecipitation (RIPA) cell lysis and extraction buffer from Pierce (Pierce, Rockford, USA). Total protein content was assessed by using the Pierce BCA protein assay reagent kit (Pierce, Rockford, USA).", "The chemokine protein levels in the different tissue lysates were determined by sandwich-type enzyme-linked immunosorbent assays (ELISA) according to the manufacturer's instructions. Samples were assayed in duplicate with all values calculated as the mean of the two measurements. CXCL12 levels were assayed using a validated commercial ELISA (Duo Set R&D Systems, DY350, Minneapolis, Minn., USA). The absorbance was read at 450 nm in a 96-well microtiter plate reader. The chemokine concentration from each tissue lysate was normalized to the total protein content of each sample.", "Operative specimens were routinely fixed in formalin and subsequently embedded in paraffin. Before staining, 4-μm thick paraffin-embedded tissue sections were mounted on Superfrost Plus slides, deparaffinized and rehydrated in graded ethanol to deionized water. The sections were microwaved with an antigen retrieval solution (Target Retrieval, Dakocytomation, Carpinteria, CA, USA) and after blocking of endogenous peroxidase activity with 3% hydrogen peroxide, the sections were sequential treated with avidin and biotin (Avidin/Biotin blocking kit, Vector Laboratories Inc., Burlingame, CA, USA). In addition the specimens were further blocked for 30 min at room temperature with normal rabbit serum. Overnight incubation at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK) was followed by incubation of secondary biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA). After colour reaction with aminoethylcarbazol solution (Merck, Darmstadt, Germany), tissues were counterstained with haematoxylin. Immunocytochemical staining of CRC cells was performed using culture slides from Becton Dickinson (Becton Dickinson, Heidelberg, Germany). After blocking of endogenous peroxidase activity and blocking for 30 min at room temperature with normal rabbit serum, cells were overnight incubated at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK). Following incubation of secondary biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA) and all further steps were carried out as described above for the immohistochemical staining procedure.\nNegative controls were conducted in all cases omitting primary antibody. For evaluation of immunocytochemical staining the total number of cells per 5 high-power fields (using ×40 -HPF objective magnification) was determined. Cells were considered positive, when they demonstrated strong and exclusive labelling.", "Cell migration assays were performed in triplicate using BD Falcon cell culture inserts containing polyethylene terephthalate membranes (8 μm pore size) from BD Biosciences (Bedford, MA, USA). Caco-2, HT-29 and SW480 cells at a concentration of 50000 cells per ml in 500 μl medium with antibiotics but without FCS were placed in the top of a two-chamber assay system and incubated for 48 hours with and without CXCL12 (350-NS-10, R&D Systems, Wiesbaden, Germany) in 800 μl medium without FCS placed in the lower chamber. CXCL12 concentrations in the receiver wells were 10 ng, 50 ng and 100 ng per ml, respectively. After the incubation period, the non-migrated cells on the upper surface of the filters were removed using a cotton swab. Migrated cells, which are adherent to the lower surface of the filters, were stained with a 0.1% crystal violet solution (Merck, Darmstadt, Germany). The number of these migrated cells was quantified microscopically by counting them in 10 high-power microscopic fields. Medium with 20% FCS placed in the lower chamber of the two-chamber assay system served as a positive control and reference in the migration assays.", "Cell proliferation assays (Roche Diagnostics GmbH, Mannheim, Germany) were performed in triplicate using flat-bottomed 96-well microtiter plates for culturing Caco-2, HT-29 and SW480 cells in the presence of CXCL12 at 37°C for 48 hours. Subsequently, bromodeoxyuridine (BrdU) was added to the cells and the cells were reincubated for 24 hours. During this labeling period the pyrimidine analogue BrdU is incorporated in place of thymidine into the DNA of proliferating cells. After removing the culture media, cells were fixed and anti-BrdU-POD was added to bind to the BrdU incorporated in newly synthesized cellular DNA. Immune complexes were detected by the subsequent substrate reaction and the reaction product was quantified by photometrically measuring the absorbance of the developed color.", "Inhibition assays were performed in analogy to the cell migration assays using BD Falcon cell culture inserts from BD Biosciences. Similar to the migration assays Caco-2, HT-29 and SW480 cells in 500 μl medium without FCS were mixed with anti-CXCR4 antibodies (MAB171, R&D systems) applied in a final concentration of 6 μg per ml and placed in the upper chamber of a two-chamber assay system. After incubation for 48 hours with and without CXCL12 in 800 μl medium without FCS placed in the lower chamber steps were continued as mentioned above for the migration assays and migrated CRC cells were quantified microscopically by counting the cells that had migrated into the filters.", "CXCR4 siRNA transfection of Caco-2, HT-29 and SW480 cells was performed according to the HiPerFect Transfection Reagent Handbook from Qiagen (Qiagen, Hilden, Germany) and the Dharmacon DharmaFECT siRNA transfection protocol (Thermo Fisher Scientific, Waltham, USA). Briefly, cells were trypsinized, counted and then diluted in antibiotic-free complete medium containing serum to achieve the appropriate plating density in 100 μl of solution. On average, 6 × 104 cells per well of a 24-well plate or 5 × 105 cells per well of a 6-well plate, respectively, were seeded and overnight incubated under their normal growth conditions. On the day of transfection, four different CXCR4 siRNAs (Qiagen, SI00052220 (1), SI00052227 (2), SI02664235 (7), SI02664242 (8)) were diluted to a final concentration of 10 nM in 100 μl culture medium without serum, respectively. 3 μl of HiPerFect (Qiagen) for the transfection of SW480 cells and 5 μl of Dharmacon DharmaFECT siRNA transfection reagent (Thermo Fisher Scientific) for the transfection of Caco-2 and HT-29 cells were added to the diluted siRNA reactions, respectively. Samples were incubated for 5-10 minutes at room temperature to allow formation of transfection complexes. Hence, the complexes were added drop-wise to the cells and incubated for three hours under their normal growth conditions. Gene silencing was monitored after 48 and 72 hours after transfection at the mRNA and at the protein level using Realtime PCR and western blotting, respectively. Further, gene silencing was monitored visually by observing the effect on a cell death control. Further, control samples with untransfected cells, a mock-transfection with only transfection reagent, a negative control siRNA and control samples with MAPK1 siRNA were used. A detailed overview of siRNA control experiments is presented in Table 2. Subsequent migration assays were performed as described above.\nApplied RNAi control experiments - overview", "Gene silencing was monitored 72 hours after transfection at the protein level using western blotting. Total protein (25 μg/lane) was separated by SDS-PAGE using a 10% gel and blotted onto nitrocellulose membranes (Hybond ECL, Amersham Biosciences, Piscataway, NJ, USA). Membranes were blocked by incubation in Tris-buffered saline (TBS) containing 5% non fat dry milk and 0.1% Tween 20 for 2 h at room temperature and then incubated overnight at 4°C with primary goat polyclonal anti-human CXCR4 antibody (1:300, AB1671, Abcam, Cambridge, UK). Blots were then washed and incubated at room temperature for 1 h with donkey anti-goat HRP antibody (diluted 1:5000, sc-2056, Santa Cruz Biotechnology, Santa Cruz, CA USA). Bands were visualized by ECL Western blotting analysis systems (Amersham Biosciences, Piscataway, NJ, USA). The human cell lysate HL-60 (sc-2209, Santa Cruz Biotechnology, Santa Cruz, CA, USA) served as positive control. Quantification of band intensities has been performed on three independent samples using image J software.", "Chemokine receptor expression profiles in the different groups are shown as mean and standard error of the mean (SEM). Statistical calculations were done with the MedCalc software package. Where appropriate, either the Student's t-test or the Wilcoxon's rank sum test was applied to test for group differences of continuous variables. A P value of 0.05 or less was considered significant.", "[SUBTITLE] CXCL12/CXCR4 expression in CRC tissues [SUBSECTION] Q-RT-PCR analysis of CRC tissue specimens revealed significant up-regulation of CXCR4 and significant down-regulation of CXCL12 compared to corresponding normal tissues (Figure 1A). These results were verified for CXCL12 on the protein level as shown by ELISA assays which demonstrated significant CXCL12 down-regulation in CRC tissues compared to the corresponding normal tissues (Figure 1B). Detection of CXCR4 expression was assessed by immunohistochemical staining. Immunostaining of the CRC tissues revealed positive cytoplasmic staining in 28 out of 50 CRC specimens (56%) as shown for a representative example in Figure 1C. However, no significant difference in CXCR4 expression was detected according to the metastatic behaviour. Patients with synchronous or metachronous colorectal liver metastases expressed CXCR4 in their primary tumor with no significant difference with respect to CRC patients who did not develop metastases.\nCXC12/CXCR4 expression in colorectal cancer (CRC) tissue specimens as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunohistochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05, n = 50. Fold increase above 1 indicates gene overexpression in affected tissues related to unaffected neighbor tissues. (B) Detection of CXCL12 protein concentrations (pg/ml pro mg total protein) in total cell lysates of CRC and adjacent normal tissues from CRC patients (n = 50). Protein data are expressed as mean +/- SEM, *P < 0.05. (C) Detection of CXCR4 protein expression in representative CRC specimens as assessed by immunohistochemical staining with CXCR4-specific antibodies showing positive cytoplasmic staining in CRC and in unaffected corresponding tissues (original magnification × 200 and × 400).\nQ-RT-PCR analysis of CRC tissue specimens revealed significant up-regulation of CXCR4 and significant down-regulation of CXCL12 compared to corresponding normal tissues (Figure 1A). These results were verified for CXCL12 on the protein level as shown by ELISA assays which demonstrated significant CXCL12 down-regulation in CRC tissues compared to the corresponding normal tissues (Figure 1B). Detection of CXCR4 expression was assessed by immunohistochemical staining. Immunostaining of the CRC tissues revealed positive cytoplasmic staining in 28 out of 50 CRC specimens (56%) as shown for a representative example in Figure 1C. However, no significant difference in CXCR4 expression was detected according to the metastatic behaviour. Patients with synchronous or metachronous colorectal liver metastases expressed CXCR4 in their primary tumor with no significant difference with respect to CRC patients who did not develop metastases.\nCXC12/CXCR4 expression in colorectal cancer (CRC) tissue specimens as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunohistochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05, n = 50. Fold increase above 1 indicates gene overexpression in affected tissues related to unaffected neighbor tissues. (B) Detection of CXCL12 protein concentrations (pg/ml pro mg total protein) in total cell lysates of CRC and adjacent normal tissues from CRC patients (n = 50). Protein data are expressed as mean +/- SEM, *P < 0.05. (C) Detection of CXCR4 protein expression in representative CRC specimens as assessed by immunohistochemical staining with CXCR4-specific antibodies showing positive cytoplasmic staining in CRC and in unaffected corresponding tissues (original magnification × 200 and × 400).\n[SUBTITLE] Organ-specific expression of CXCL12 [SUBSECTION] CXCL12 mRNA and protein expression as determined by RT-PCR and ELISA was investigated in different human organs such as pancreas, stomach, colon/rectum and esophagus related to the liver. On the protein level CXCL12 was found to exhibit peak levels of expression in the liver compared to stomach, esophagus, pancreas, colon and rectum (Figure 2). However, on the mRNA level CXCL12 did not show a markedly higher expression in the liver in comparison to other gastrointestinal organs or glands corresponding to previous results [27].\nCXCL12 protein expression in different human organs related to the liver as determined by ELISA. CXCL12 protein concentrations (pg/ml per mg total protein) were measured in total cell lysates of normal liver, pancreas, stomach, colon/rectum and esophagus tissues of 10 patients, respectively (n = 10). Protein data are expressed as mean +/- SEM, *P < 0.05.\nCXCL12 mRNA and protein expression as determined by RT-PCR and ELISA was investigated in different human organs such as pancreas, stomach, colon/rectum and esophagus related to the liver. On the protein level CXCL12 was found to exhibit peak levels of expression in the liver compared to stomach, esophagus, pancreas, colon and rectum (Figure 2). However, on the mRNA level CXCL12 did not show a markedly higher expression in the liver in comparison to other gastrointestinal organs or glands corresponding to previous results [27].\nCXCL12 protein expression in different human organs related to the liver as determined by ELISA. CXCL12 protein concentrations (pg/ml per mg total protein) were measured in total cell lysates of normal liver, pancreas, stomach, colon/rectum and esophagus tissues of 10 patients, respectively (n = 10). Protein data are expressed as mean +/- SEM, *P < 0.05.\n[SUBTITLE] CXCL12/CXCR4 expression in CRC cell lines [SUBSECTION] In non-metastatic cell line Caco-2 low CXCL12 expression was detected on the mRNA level by PCR and on the protein level in the cell culture supernatant by ELISA (Figure 3A and 3B, respectively). In metastatic cell lines SW480 and HT-29 CXCL12 mRNA and protein expression was below detection limit (Figure 3A and 3B, respectively). In contrast, CXCR4 mRNA expression was detected in HT-29 and SW480 cells with the latter displaying significant overexpression relative to B2M (Figure 3A). When CXCR4 protein expression was assessed by immunocytochemical staining, CXCR4 positive staining was observed in 22% of Caco-2 cells, 74% of HT-29 cells and 80% of SW480 cells as quantified microscopically and shown for a representative example in Figure 3C. Likewise, CXCR4 protein expression as determined by western blot analysis revealed similar expression rates in HT-29 and SW480 cells. Variation in CXCR4 mRNA and protein data may be due to posttranscriptional and posttranslational modifications as well as to the relative amplification modus of the qRT-PCR results with respect to housekeeping gene B2M.\nCXC12/CXCR4 expression in Caco-2, HT-29 and SW480 cells as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunocytochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05. Fold increase above 1 indicates gene overexpression related to housekeeping gene B2M. (B) Detection of CXCL12 protein concentrations in pg/ml in cell culture supernatant of Caco-2, HT-29 and SW480 cells. Protein data are expressed as mean +/- SEM. (C) Detection of CXCR4 protein expression in representative cell culture slides of Caco-2, HT-29 and SW480 cells as assessed by immunocytochemical staining with CXCR4-specific antibodies.\nIn non-metastatic cell line Caco-2 low CXCL12 expression was detected on the mRNA level by PCR and on the protein level in the cell culture supernatant by ELISA (Figure 3A and 3B, respectively). In metastatic cell lines SW480 and HT-29 CXCL12 mRNA and protein expression was below detection limit (Figure 3A and 3B, respectively). In contrast, CXCR4 mRNA expression was detected in HT-29 and SW480 cells with the latter displaying significant overexpression relative to B2M (Figure 3A). When CXCR4 protein expression was assessed by immunocytochemical staining, CXCR4 positive staining was observed in 22% of Caco-2 cells, 74% of HT-29 cells and 80% of SW480 cells as quantified microscopically and shown for a representative example in Figure 3C. Likewise, CXCR4 protein expression as determined by western blot analysis revealed similar expression rates in HT-29 and SW480 cells. Variation in CXCR4 mRNA and protein data may be due to posttranscriptional and posttranslational modifications as well as to the relative amplification modus of the qRT-PCR results with respect to housekeeping gene B2M.\nCXC12/CXCR4 expression in Caco-2, HT-29 and SW480 cells as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunocytochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05. Fold increase above 1 indicates gene overexpression related to housekeeping gene B2M. (B) Detection of CXCL12 protein concentrations in pg/ml in cell culture supernatant of Caco-2, HT-29 and SW480 cells. Protein data are expressed as mean +/- SEM. (C) Detection of CXCR4 protein expression in representative cell culture slides of Caco-2, HT-29 and SW480 cells as assessed by immunocytochemical staining with CXCR4-specific antibodies.\n[SUBTITLE] CXCR4 protein inhibition abrogates migration of colorectal cancer cells [SUBSECTION] All three cell lines were stimulated with 10, 50 and 100 ng/ml CXCL12, respectively, and incubated for 48 hours. While Caco-2 cells were not stimulated by any concentration of CXCL12 (Figure 4A), we observed a significant CXCL12 dose-independent stimulation of migration for HT-29 and SW480 cells (Figure 4B and 4C, respectively) (P < 0.05). Thus, CXCL12 was shown to be chemotactic for HT-29 and SW480 cells.\nCXCR4 protein inhibition abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to a positive control containing 20% FCS (PC) without inhibition and after inhibition with anti-CXCR4 antibody. (B) Percent HT-29 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody. (C) Percent SW480 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody.\nWhen we added neutralizing anti-CXCR4 antibody prior to performance of cell migration assays, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated at all CXCL12 concentrations under investigation (Figure 4B and 4C, respectively) (P < 0.05). Application of anti-CXCR4 antibodies had no impact on the migration potential of Caco-2 cells (Figure 4A).\nAll three cell lines were stimulated with 10, 50 and 100 ng/ml CXCL12, respectively, and incubated for 48 hours. While Caco-2 cells were not stimulated by any concentration of CXCL12 (Figure 4A), we observed a significant CXCL12 dose-independent stimulation of migration for HT-29 and SW480 cells (Figure 4B and 4C, respectively) (P < 0.05). Thus, CXCL12 was shown to be chemotactic for HT-29 and SW480 cells.\nCXCR4 protein inhibition abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to a positive control containing 20% FCS (PC) without inhibition and after inhibition with anti-CXCR4 antibody. (B) Percent HT-29 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody. (C) Percent SW480 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody.\nWhen we added neutralizing anti-CXCR4 antibody prior to performance of cell migration assays, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated at all CXCL12 concentrations under investigation (Figure 4B and 4C, respectively) (P < 0.05). Application of anti-CXCR4 antibodies had no impact on the migration potential of Caco-2 cells (Figure 4A).\n[SUBTITLE] CXC receptor-4 gene silencing abrogates migration of colorectal cancer cells [SUBSECTION] To analyze if the chemotactic effect of CXCL12 on the migration of HT-29 and SW480 cells could be counteracted by the down-regulation of the mRNA of corresponding receptor CXCR4, we applied four different CXCR4 siRNAs. The mean percent CXCR4 knockdown related to the siRNA negative control was 78% for Caco-2 cells, 80% HT-29 cells and 84% for SW480 cells as determined by RT-PCR 48 hours after transfection on the mRNA level (Figure 5A). On the protein level gene silencing was monitored 72 hours after transfection in western blot analyses (Figure 5B). Application of CXCR4 siRNAs had no impact on the migration potential of Caco-2 cells (Figure 6A). In contrast, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated by all four different CXCR4 siRNAs at all CXCL12 concentrations under investigation (Figure 6B and 6C, respectively) (P < 0.05).\nDemonstration of CXCR4 knockdown as determined by mRNA and protein analysis. (A) Mean percent CXCR4 knockdown related to siRNA negative control (NC) as determined by mRNA analysis in Caco-2, HT-29 and SW480 cells. (B) CXCR4 protein expression after CXCR4 siRNA silencing as determined by Western Blot analysis in Caco-2, HT-29 and SW480 cells related to untransfected total cell lysates. Total cell lysates of untransfected cells and one representative example of CXCR4 siRNA transfected cells of each cell line were immunoblotted with antibodies specifically recognizing chemokine receptor CXCR4. Acute leucemia cell line HL60 served as positive control for the detection of CXCR4.\nCXCR4 mRNA silencing abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to positive control containing 20% FCS (PC) without transfection and after transfection with 4 different CXCR4 siRNAs. (B) Percent HT-29 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs. (C) Percent SW480 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs.\nTo analyze if the chemotactic effect of CXCL12 on the migration of HT-29 and SW480 cells could be counteracted by the down-regulation of the mRNA of corresponding receptor CXCR4, we applied four different CXCR4 siRNAs. The mean percent CXCR4 knockdown related to the siRNA negative control was 78% for Caco-2 cells, 80% HT-29 cells and 84% for SW480 cells as determined by RT-PCR 48 hours after transfection on the mRNA level (Figure 5A). On the protein level gene silencing was monitored 72 hours after transfection in western blot analyses (Figure 5B). Application of CXCR4 siRNAs had no impact on the migration potential of Caco-2 cells (Figure 6A). In contrast, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated by all four different CXCR4 siRNAs at all CXCL12 concentrations under investigation (Figure 6B and 6C, respectively) (P < 0.05).\nDemonstration of CXCR4 knockdown as determined by mRNA and protein analysis. (A) Mean percent CXCR4 knockdown related to siRNA negative control (NC) as determined by mRNA analysis in Caco-2, HT-29 and SW480 cells. (B) CXCR4 protein expression after CXCR4 siRNA silencing as determined by Western Blot analysis in Caco-2, HT-29 and SW480 cells related to untransfected total cell lysates. Total cell lysates of untransfected cells and one representative example of CXCR4 siRNA transfected cells of each cell line were immunoblotted with antibodies specifically recognizing chemokine receptor CXCR4. Acute leucemia cell line HL60 served as positive control for the detection of CXCR4.\nCXCR4 mRNA silencing abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to positive control containing 20% FCS (PC) without transfection and after transfection with 4 different CXCR4 siRNAs. (B) Percent HT-29 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs. (C) Percent SW480 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs.\n[SUBTITLE] Proliferative rate of colorectal cancer cells after CXCR4 blockage by mRNA silencing and inhibition antibodies [SUBSECTION] CXCR4 blockage by mRNA silencing or anti-CXCR4 antibodies might restrain the proliferation of CRC cell lines, so that the decreased migration capacity of cancer cells might result from a lower proliferative rate. To ensure that the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing by all four different CXCR4 siRNAs is not resulting from a lower proliferative rate with respect to the untransfected cells, we investigated their proliferative capacity before and after siRNA transfection. In addition, we also included cell line Caco-2 in this survey. Likewise, we investigated the proliferative capacity of Caco-2, HT-29 and SW480 cells before and after CXCR4 blockage by neutralizing anti-CXCR4 antibody. Here, cells without anti-CXCR4 antibody served as reference. As presented in figure 7 there is no marked difference between the proliferation rates of Caco-2 (Figure 7A), HT-29 (Figure 7B) and SW480 (Figure 7C) cells before or after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody. Thus, the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody is not resulting from a reduced proliferation rate.\nProliferative rate of colorectal cancer cells after CXCR4 mRNA silencing or CXCR4 blockage by inhibition antibodies. Knockdown of endogeneous CXCR4 expression by four different CXCR4 siRNAs or CXCR4 blockage by ant-CXCR4 inhibition antibodies does not decelerate the proliferation rate of (A) Caco-2 cells, (B) HT-29 cells and (C) SW480 cells in relation to untransfected cells without inhibition antibody (control).\nCXCR4 blockage by mRNA silencing or anti-CXCR4 antibodies might restrain the proliferation of CRC cell lines, so that the decreased migration capacity of cancer cells might result from a lower proliferative rate. To ensure that the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing by all four different CXCR4 siRNAs is not resulting from a lower proliferative rate with respect to the untransfected cells, we investigated their proliferative capacity before and after siRNA transfection. In addition, we also included cell line Caco-2 in this survey. Likewise, we investigated the proliferative capacity of Caco-2, HT-29 and SW480 cells before and after CXCR4 blockage by neutralizing anti-CXCR4 antibody. Here, cells without anti-CXCR4 antibody served as reference. As presented in figure 7 there is no marked difference between the proliferation rates of Caco-2 (Figure 7A), HT-29 (Figure 7B) and SW480 (Figure 7C) cells before or after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody. Thus, the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody is not resulting from a reduced proliferation rate.\nProliferative rate of colorectal cancer cells after CXCR4 mRNA silencing or CXCR4 blockage by inhibition antibodies. Knockdown of endogeneous CXCR4 expression by four different CXCR4 siRNAs or CXCR4 blockage by ant-CXCR4 inhibition antibodies does not decelerate the proliferation rate of (A) Caco-2 cells, (B) HT-29 cells and (C) SW480 cells in relation to untransfected cells without inhibition antibody (control).", "Q-RT-PCR analysis of CRC tissue specimens revealed significant up-regulation of CXCR4 and significant down-regulation of CXCL12 compared to corresponding normal tissues (Figure 1A). These results were verified for CXCL12 on the protein level as shown by ELISA assays which demonstrated significant CXCL12 down-regulation in CRC tissues compared to the corresponding normal tissues (Figure 1B). Detection of CXCR4 expression was assessed by immunohistochemical staining. Immunostaining of the CRC tissues revealed positive cytoplasmic staining in 28 out of 50 CRC specimens (56%) as shown for a representative example in Figure 1C. However, no significant difference in CXCR4 expression was detected according to the metastatic behaviour. Patients with synchronous or metachronous colorectal liver metastases expressed CXCR4 in their primary tumor with no significant difference with respect to CRC patients who did not develop metastases.\nCXC12/CXCR4 expression in colorectal cancer (CRC) tissue specimens as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunohistochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05, n = 50. Fold increase above 1 indicates gene overexpression in affected tissues related to unaffected neighbor tissues. (B) Detection of CXCL12 protein concentrations (pg/ml pro mg total protein) in total cell lysates of CRC and adjacent normal tissues from CRC patients (n = 50). Protein data are expressed as mean +/- SEM, *P < 0.05. (C) Detection of CXCR4 protein expression in representative CRC specimens as assessed by immunohistochemical staining with CXCR4-specific antibodies showing positive cytoplasmic staining in CRC and in unaffected corresponding tissues (original magnification × 200 and × 400).", "CXCL12 mRNA and protein expression as determined by RT-PCR and ELISA was investigated in different human organs such as pancreas, stomach, colon/rectum and esophagus related to the liver. On the protein level CXCL12 was found to exhibit peak levels of expression in the liver compared to stomach, esophagus, pancreas, colon and rectum (Figure 2). However, on the mRNA level CXCL12 did not show a markedly higher expression in the liver in comparison to other gastrointestinal organs or glands corresponding to previous results [27].\nCXCL12 protein expression in different human organs related to the liver as determined by ELISA. CXCL12 protein concentrations (pg/ml per mg total protein) were measured in total cell lysates of normal liver, pancreas, stomach, colon/rectum and esophagus tissues of 10 patients, respectively (n = 10). Protein data are expressed as mean +/- SEM, *P < 0.05.", "In non-metastatic cell line Caco-2 low CXCL12 expression was detected on the mRNA level by PCR and on the protein level in the cell culture supernatant by ELISA (Figure 3A and 3B, respectively). In metastatic cell lines SW480 and HT-29 CXCL12 mRNA and protein expression was below detection limit (Figure 3A and 3B, respectively). In contrast, CXCR4 mRNA expression was detected in HT-29 and SW480 cells with the latter displaying significant overexpression relative to B2M (Figure 3A). When CXCR4 protein expression was assessed by immunocytochemical staining, CXCR4 positive staining was observed in 22% of Caco-2 cells, 74% of HT-29 cells and 80% of SW480 cells as quantified microscopically and shown for a representative example in Figure 3C. Likewise, CXCR4 protein expression as determined by western blot analysis revealed similar expression rates in HT-29 and SW480 cells. Variation in CXCR4 mRNA and protein data may be due to posttranscriptional and posttranslational modifications as well as to the relative amplification modus of the qRT-PCR results with respect to housekeeping gene B2M.\nCXC12/CXCR4 expression in Caco-2, HT-29 and SW480 cells as determined by (A) Q-RT-PCR for CXCL12 and CXCR4, (B) ELISA analysis for CXCL12 and (C) immunocytochemistry for CXCR4. (A) Q-RT-PCR data are expressed as mean +/- standard error of the mean (SEM), *P < 0.05. Fold increase above 1 indicates gene overexpression related to housekeeping gene B2M. (B) Detection of CXCL12 protein concentrations in pg/ml in cell culture supernatant of Caco-2, HT-29 and SW480 cells. Protein data are expressed as mean +/- SEM. (C) Detection of CXCR4 protein expression in representative cell culture slides of Caco-2, HT-29 and SW480 cells as assessed by immunocytochemical staining with CXCR4-specific antibodies.", "All three cell lines were stimulated with 10, 50 and 100 ng/ml CXCL12, respectively, and incubated for 48 hours. While Caco-2 cells were not stimulated by any concentration of CXCL12 (Figure 4A), we observed a significant CXCL12 dose-independent stimulation of migration for HT-29 and SW480 cells (Figure 4B and 4C, respectively) (P < 0.05). Thus, CXCL12 was shown to be chemotactic for HT-29 and SW480 cells.\nCXCR4 protein inhibition abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to a positive control containing 20% FCS (PC) without inhibition and after inhibition with anti-CXCR4 antibody. (B) Percent HT-29 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody. (C) Percent SW480 cell migration related to PC without inhibition and after inhibition with anti-CXCR4 antibody.\nWhen we added neutralizing anti-CXCR4 antibody prior to performance of cell migration assays, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated at all CXCL12 concentrations under investigation (Figure 4B and 4C, respectively) (P < 0.05). Application of anti-CXCR4 antibodies had no impact on the migration potential of Caco-2 cells (Figure 4A).", "To analyze if the chemotactic effect of CXCL12 on the migration of HT-29 and SW480 cells could be counteracted by the down-regulation of the mRNA of corresponding receptor CXCR4, we applied four different CXCR4 siRNAs. The mean percent CXCR4 knockdown related to the siRNA negative control was 78% for Caco-2 cells, 80% HT-29 cells and 84% for SW480 cells as determined by RT-PCR 48 hours after transfection on the mRNA level (Figure 5A). On the protein level gene silencing was monitored 72 hours after transfection in western blot analyses (Figure 5B). Application of CXCR4 siRNAs had no impact on the migration potential of Caco-2 cells (Figure 6A). In contrast, the CXCL12 stimulated migration of HT-29 and SW480 cells was significantly abrogated by all four different CXCR4 siRNAs at all CXCL12 concentrations under investigation (Figure 6B and 6C, respectively) (P < 0.05).\nDemonstration of CXCR4 knockdown as determined by mRNA and protein analysis. (A) Mean percent CXCR4 knockdown related to siRNA negative control (NC) as determined by mRNA analysis in Caco-2, HT-29 and SW480 cells. (B) CXCR4 protein expression after CXCR4 siRNA silencing as determined by Western Blot analysis in Caco-2, HT-29 and SW480 cells related to untransfected total cell lysates. Total cell lysates of untransfected cells and one representative example of CXCR4 siRNA transfected cells of each cell line were immunoblotted with antibodies specifically recognizing chemokine receptor CXCR4. Acute leucemia cell line HL60 served as positive control for the detection of CXCR4.\nCXCR4 mRNA silencing abrogates migration of colorectal cancer cells. (A) Percent Caco-2 cell migration related to positive control containing 20% FCS (PC) without transfection and after transfection with 4 different CXCR4 siRNAs. (B) Percent HT-29 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs. (C) Percent SW480 cell migration related to PC without transfection and after transfection with 4 different CXCR4 siRNAs.", "CXCR4 blockage by mRNA silencing or anti-CXCR4 antibodies might restrain the proliferation of CRC cell lines, so that the decreased migration capacity of cancer cells might result from a lower proliferative rate. To ensure that the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing by all four different CXCR4 siRNAs is not resulting from a lower proliferative rate with respect to the untransfected cells, we investigated their proliferative capacity before and after siRNA transfection. In addition, we also included cell line Caco-2 in this survey. Likewise, we investigated the proliferative capacity of Caco-2, HT-29 and SW480 cells before and after CXCR4 blockage by neutralizing anti-CXCR4 antibody. Here, cells without anti-CXCR4 antibody served as reference. As presented in figure 7 there is no marked difference between the proliferation rates of Caco-2 (Figure 7A), HT-29 (Figure 7B) and SW480 (Figure 7C) cells before or after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody. Thus, the significantly abrogated migration capacity of HT-29 and SW480 cells after CXCR4 mRNA silencing or CXCR4 blockage by anti-CXCR4 antibody is not resulting from a reduced proliferation rate.\nProliferative rate of colorectal cancer cells after CXCR4 mRNA silencing or CXCR4 blockage by inhibition antibodies. Knockdown of endogeneous CXCR4 expression by four different CXCR4 siRNAs or CXCR4 blockage by ant-CXCR4 inhibition antibodies does not decelerate the proliferation rate of (A) Caco-2 cells, (B) HT-29 cells and (C) SW480 cells in relation to untransfected cells without inhibition antibody (control).", "The importance of the CXCL12/CXCR4 axis in tumor progression and metastasis has been emphasized. Recent evidence also suggests that the CXCL12/CXCR4 system is involved in the progression and metastasis of CRC [15,20,28].\nThus, the aim of this study was to conduct a comparative CXCL12/CXCR4 expression analysis and to investigate the functional role of CXCR4 in metastatic and non-metastatic CRC derived cell lines. For this purpose the migration potential of these cell lines was tested under conditions where CXCR4 was inhibited on the mRNA as well as on the protein level. On the mRNA level, CXCR4 silencing was achieved by CXCR4 RNA interference while on the protein level inhibition of CXCR4 protein activity was achieved using an anti-CXCR4 antibody.\nAs shown initially, CXCL12 protein is significantly higher expressed in the liver in relation to other intestinal organs or glands, thus supporting potential CXCL12 involvement in the homing of CRC cells to the liver. Subsequently, we observed significant up-regulation of CXCR4 and significant down-regulation of CXCL12 in CRC tissues. This inverse expression pattern is supported by previous studies, where CXCR4 was shown to be highly up-regulated in all T-stages of CRC tissues [16] and CXCL12 expression levels were shown to be low at the base of the crypts and increased in the more differentiated apical intestinal epithelial cells [29,30]. In compliance with the low CXCL12 expression status in tissues we observed low or no CXCL12 expression, respectively, in three human CRC derived cell lines, non-metastatic cell line Caco-2 and metastatic cell lines SW480 and HT-29. While low CXCL12 expression was detected in Caco-2 cells on the mRNA and on the protein level, CXCL12 was below the detection limit in HT-29 and SW480. It was hypothesized that changes in epithelial CXCL12 expression might contribute to CRC disease progression by allowing CRC cells to more readily sense CXCL12 from exogeneous sources hereby promoting metastasis [31]. Wendt et al. have shown that the reduced CXCL12 expression pattern in CRC tissues and cells is due to DNA hypermethylation in primary CRC and carcinoma-derived cell lines. Thus, it was demonstrated that inhibition or ablation of DNA methyltransferases prevent promoter methylation and restore CXCL12 expression. In addition, re-expression of functional, endogeneous CXCL12 in CRC cells was shown to reduce metastatic tumor formation significantly while silencing CXCL12 greatly enhanced the metastatic potential of CRC cells [31]. Moreover, endogeneous CXCL12 was shown to provide a barrier to metastasis by increasing anoikis via activation of a Bim-mediated intrinsic apoptotic pathway [32,33]. These results may constitute a plausible explanation for our observation of significantly reduced CXCL12 expression rates in CRC tissues and carcinoma-derived cell lines. Concordantly, an elevated migratory signaling response to ectopic CXCL12 was also shown to contribute to the metastatic potential of CXCR4-expressing mammary carcinoma cells, subsequent to epigenetic silencing of autocrine CXCL12 [34].\nFor CXCR4 we observed an inverse expression pattern in the three CRC-derived cell lines. Thus, expression in non-metastatic cell line Caco-2 was significantly lower compared to the metastatic cell lines under investigation. Also drug-resistant invasive HT-29 cells with a metastatic behaviour in immunodeficient mice were shown to exhibit high CXCR4 expression [35]. Responsible for promoting invasion in drug-resistant colon carcinoma cells is the autocrine CXCR4 ligand macrophage migration-inhibitory factor (MIF) as shown by Dessein et al. Impairing the MIF-CXCR4 signaling pathway, e.g. by silencing CXCR4, abolished this aggressive phenotype. To date, in various cancer entities aberrantly expressed chemokine receptors were demonstrated to contribute directly to the development of organ selective distant metastasis by interaction with their organ selectively expressed chemokine ligands. As a consequence of the low CXCR4 expression, Caco-2 cells showed no increase in migration in response to CXCL12 stimulation. In contrast, CXCL12 significantly increased cell migration in the CXCR4 expressing metastatic cell lines SW480 and HT-29. While a stimulative effect of CXCL12 on migration of CRC cell lines was described before [36], these effects were not significant and refer only to one CXCL12 concentration (100 ng/ml). Moreover, these tests were performed on CRC cell lines LS174T, SW620 and SW480 but not on cell lines HT-29 and Caco-2. To further investigate the functional role of CXCR4 we performed inhibition assays with anti-CXCR4 antibodies added previously to CXCL12 stimulation. The addition of the inhibition antibodies significantly blocked the CXCL12-dependent stimulation of HT-29 and SW480 cell migration but had no impact on Caco-2 migration. Our results correlate with inhibition studies performed on SW480 cells [37], where chemotaxis induced by CXCL12 was also shown to be blocked, although not as complete as we have shown for both SW480 and HT-29 cells. Likewise, Kim et al [38] reported on a partial inhibitory effect of anti-CXCR4 antibodies on CXCL12 stimulated melanoma and CRC cell migration (38%). However, to date no study analyzed the effect of CXCR4 gene silencing on CXCL12 mediated cell migration of CRC cells. Thus, we investigated if the chemotactic effect of CXCL12 on the migration of HT-29 and SW480 cells could be counteracted by the down-regulation of the CXCR4 mRNA. Applying four different CXCR4 siRNAs we achieved a mean percent CXCR4 knockdown of 78-84% and a significant abrogation of CXCL12 stimulated migration of HT-29 and SW480 for all four different CXCR4 siRNAs at all CXCL12 concentrations under investigation. On the other hand, application of CXCR4 siRNAs had no impact on the migration potential of Caco-2 cells. Thus, we have demonstrated the functional status of the CXCR4 receptor on CRC cell lines in response to CXCL12 on the mRNA level.\nIn CRC patients, no significant difference in CXCR4 expression was detected according to the metastatic behaviour. Based on our in vitro results CRC patients with metastasis may be expected to express more CXCR4 in their tumor cells compared to CRC patients without metastasis. However, many other important factors contribute to the metastatic properties of tumor tissues and influence their metastatic potential. When a CRC cell is turned metastatic, CXCR4 may be an important factor for the homing of such a tumor cell to its favourite organ destination, the liver.\nOur results are well in line with recent findings demonstrating a role of CXCL12 in promoting cell migration and tumor growth of CRC metastasis in vivo in a murine model [39]. However, while we observed no CXCR4-dependent proliferation of CRC cells, Shen et al. demonstrated CXCR4-induced proliferation for pancreatic cancer cells, where it was linked to AKT and ERK dependent pathways [40]. Moreover, CXCR4 knockdown by small interfering RNA was shown to inhibit cell proliferation and invasion of oral squamous cell carcinoma cells [41]. A recent study demonstrated that CXCL12/CXCR4 interactions may also promote early extravasation of liver metastatic epithelial tumor cells which determines a critical step in formation of organ-specific metastases [42]. Currently, various small-molecule chemokine receptor antagonist compounds are undergoing development in phase I to III studies in infectious and autoimmune diseases and a CCR5 inhibitor is already in the clinic for the treatment of HIV-infected patients.", "In conclusion, our results provide evidence that CXCR4 is up-regulated in CRC and stimulation of CXCR4 bearing cancer cells with CXCL12 led to increased migration, an effect which could be inhibited both by CXCR4 siRNA and neutralizing CXCR4 antibodies. Interestingly, CXCR4 was predominantly expressed in cell lines with metastatic potential and consequently, these cell lines showed increased migration after external CXCL12 stimulation. Our results suggest that the metastatic potential of CRC cells may be associated with the aberrant expression of CXCR4 and subsequently the ability of cells to interact with CXCL12 via autocrine and/or paracrine mechanisms.", "The authors declare that they have no competing interests.", "All authors read and approved the final manuscript. CR is responsible for the study concept and design and drafted the manuscript. VOF took part in the acquisition, analysis and interpretation of the data. PG is responsible for the critical assessment and revision of the manuscript. MW examined the tissue sections for the presence of tumor cells and histopathologically confirmed all tissues under investigation. CJ provided clinical information and SG participated in the statistical analysis. SKF performed the siRNA experiments and BV contributed to scientific discussion. OK provided clinical information. MKS is responsible for the provision of all the patient material and participated in the critical revision of the manuscript for important intellectual content." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Cross-Sectional Studies", "Forecasting", "Hospitals", "Humans", "Lebanon", "Medical Errors", "Organizational Culture", "Outcome Assessment, Health Care", "Safety Management", "Surveys and Questionnaires" ]

[ "Background", "Lebanese Context", "Study Design, Setting, and Sample", "Survey Measures and Outcome Variables", "Bivariate Analysis", "Multivariate Regression Analysis", "Results", "Comparison of Means for the Frequency of Events Reported and the Overall Perception of Safety across Respondent and Hospital Characteristics", "Correlations between Patient Safety Culture Composites", "Comparison of Means between Outcome Variables and Patient Safety Composites", "Distribution of the Patient Safety Grade and the Number of Events Reported across Respondent and Hospital Characteristics", "Patient Safety Grade", "Number of Events Reported", "Generalized Estimating Equations for the Patient Safety Composite Scores and Respondent and Hospital Characteristics against the Patient Safety Grade and the Number of Events Reported (See Table 5)", "Patient Safety Grade", "Number of Events Reported", "Linear Mixed Regression for the Patient Safety Composite Scores and Respondent and Hospital Characteristics against the Frequency of Events Reported and the Overall Perception of Safety (See Table 6)", "Frequency of Events Reported", "Overall Perception of Safety", "Discussion", "Event Reporting", "Communication", "Patient safety leadership and management", "Staffing", "Hospital size and accreditation", "Limitations", "Conclusion", "List of abbreviations used", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Developing a patient safety culture was one of the recommendations made by the Institute of Medicine to assist hospitals in improving patient safety [1,2]. Assessing the organization's existing safety culture is the first stage of developing a safety culture [3]. Patient safety culture assessments, required by international accreditation organizations, allow healthcare organizations to obtain a clear view of the patient safety aspects requiring urgent attention, identify the strengths and weaknesses of their safety culture [4], help care giving units identify their existing patient safety problems [5], and benchmark their scores with other hospitals [6].\nAccording to literature, the major predictors of a positive patient safety culture in healthcare organizations specifically hospitals include communication founded on mutual trust, good information flow, shared perception of the importance of safety, organizational learning, commitment from management and leadership, and the presence of a non-punitive approach to incident and error reporting [7]. Patient safety culture outcomes include the staff members' perception of safety, the willingness of staff members to report events, the number of events reported, and an overall patient safety grade given by staff members to their units [8].\nA multitude of evidence has been published in the area of patient safety culture in recent years. Some of the available evidence tackles patient safety culture issues that require attention, factors affecting incident reporting by hospital staff, the role of workplace environment in shaping safety, and steps that can be followed to improve safety.\nDespite the wealth of evidence on patient safety culture, limited evidence still exists about the linkage between predictors and outcomes of patient safety culture especially in countries of the Eastern Mediterranean Region. One of the first efforts to assess the culture of safety in hospitals in the region was conducted in Lebanon by El-Jardali et al. [9].\n[SUBTITLE] Lebanese Context [SUBSECTION] The study by El-Jardali et al. (2010) entitled \"The Current state of Patient Safety Culture in Lebanese Hospitals: A study at Baseline\" [9] utilized an Arabic translated version of the Hospital Survey on Patient Safety Culture (HSOPSC) [8]. It aimed at identifying the most critical issues related to patient safety culture and potential strategies to implement the patient safety accreditation standards in the light of a newly added chapter to the Lebanese handbook of hospital accreditation [10].\nThe HSOPSC measures 12 patient safety culture composites representing several patient safety culture predictors (See Box 1). The HSOPSC also requires respondents to give their work area/unit a patient safety grade and to answer a question on the number of events reported in the past 12 months [8].\nCalculating the percentage of positive responses for each composite revealed that the composites with the highest positive ratings were teamwork within units, hospital management support for patient safety, and organizational learning and continuous improvement. However, composites with the lowest ratings were teamwork across hospital units, hospital handoffs and transitions, staffing, and non-punitive response to error [9]. Approximately 60% of respondents reported not completing any event reports in the past 12 months and over 70% gave their units an \"Excellent/Very Good\" patient safety grade. Bivariate and multivariate analyses revealed significant differences across hospitals of different size and accreditation status [9].\nStudy findings outlined above represent the first component of data analysis for that stage [9]. Findings provided evidence that communication across units, staffing, event reporting, and the culture of response to error were major patient safety culture issues [9]. This paper, though, will further explore the association between patient safety culture predictors and outcomes, taking into consideration respondent and hospital characteristics. In addition, it will examine the correlation between the patient safety culture composites. Thus, the objective of this paper is to address the afore-mentioned objectives using bivariate as well as multivariate analyses.\nThe study by El-Jardali et al. (2010) entitled \"The Current state of Patient Safety Culture in Lebanese Hospitals: A study at Baseline\" [9] utilized an Arabic translated version of the Hospital Survey on Patient Safety Culture (HSOPSC) [8]. It aimed at identifying the most critical issues related to patient safety culture and potential strategies to implement the patient safety accreditation standards in the light of a newly added chapter to the Lebanese handbook of hospital accreditation [10].\nThe HSOPSC measures 12 patient safety culture composites representing several patient safety culture predictors (See Box 1). The HSOPSC also requires respondents to give their work area/unit a patient safety grade and to answer a question on the number of events reported in the past 12 months [8].\nCalculating the percentage of positive responses for each composite revealed that the composites with the highest positive ratings were teamwork within units, hospital management support for patient safety, and organizational learning and continuous improvement. However, composites with the lowest ratings were teamwork across hospital units, hospital handoffs and transitions, staffing, and non-punitive response to error [9]. Approximately 60% of respondents reported not completing any event reports in the past 12 months and over 70% gave their units an \"Excellent/Very Good\" patient safety grade. Bivariate and multivariate analyses revealed significant differences across hospitals of different size and accreditation status [9].\nStudy findings outlined above represent the first component of data analysis for that stage [9]. Findings provided evidence that communication across units, staffing, event reporting, and the culture of response to error were major patient safety culture issues [9]. This paper, though, will further explore the association between patient safety culture predictors and outcomes, taking into consideration respondent and hospital characteristics. In addition, it will examine the correlation between the patient safety culture composites. Thus, the objective of this paper is to address the afore-mentioned objectives using bivariate as well as multivariate analyses.", "The study by El-Jardali et al. (2010) entitled \"The Current state of Patient Safety Culture in Lebanese Hospitals: A study at Baseline\" [9] utilized an Arabic translated version of the Hospital Survey on Patient Safety Culture (HSOPSC) [8]. It aimed at identifying the most critical issues related to patient safety culture and potential strategies to implement the patient safety accreditation standards in the light of a newly added chapter to the Lebanese handbook of hospital accreditation [10].\nThe HSOPSC measures 12 patient safety culture composites representing several patient safety culture predictors (See Box 1). The HSOPSC also requires respondents to give their work area/unit a patient safety grade and to answer a question on the number of events reported in the past 12 months [8].\nCalculating the percentage of positive responses for each composite revealed that the composites with the highest positive ratings were teamwork within units, hospital management support for patient safety, and organizational learning and continuous improvement. However, composites with the lowest ratings were teamwork across hospital units, hospital handoffs and transitions, staffing, and non-punitive response to error [9]. Approximately 60% of respondents reported not completing any event reports in the past 12 months and over 70% gave their units an \"Excellent/Very Good\" patient safety grade. Bivariate and multivariate analyses revealed significant differences across hospitals of different size and accreditation status [9].\nStudy findings outlined above represent the first component of data analysis for that stage [9]. Findings provided evidence that communication across units, staffing, event reporting, and the culture of response to error were major patient safety culture issues [9]. This paper, though, will further explore the association between patient safety culture predictors and outcomes, taking into consideration respondent and hospital characteristics. In addition, it will examine the correlation between the patient safety culture composites. Thus, the objective of this paper is to address the afore-mentioned objectives using bivariate as well as multivariate analyses.", "The study adopted a cross sectional research design utilizing a customized version of the HSOPSC. A pilot testing phase preceded data collection in order to ensure the validity and reliability of the Arabic translated version of the questionnaire. Of the 126 hospitals registered in the Lebanese Syndicate of Private Hospitals that were contacted and asked to participate, 68 consented [9]. The survey targeted hospital employees including physicians, nurses, clinical and non-clinical staff, pharmacy and laboratory staff, dietary and radiology staff, supervisors, and hospital managers. The questionnaire was distributed to 12,250 hospital employees and 6807 were returned yielding an overall response rate of 55.56%. Additional details on the study methodology can be found in El-Jardali et al. [9].", "The HSOPSC is composed of 42 items that measure 12 composites of patient safety culture (See Box1, Additional file1 ). Items were scored using a five-point scale reflecting the agreement rate on a five-point frequency scale. The percentage of positive responses for each item was calculated; negatively worded items were reversed when computing percent positive response. Composite level scores were computed by summation of the items within the composite scales and dividing by the number of items with non-missing values [9].\nTwo of the composites (frequency of events reported and overall perception of safety) are two of the four patient safety culture outcome variables [8]. The remaining two outcome variables are the patient safety grade and the number of events reported [8].", "Bivariate analyses were used to examine the associations between patient safety culture composites and differences across hospitals of different size and accreditation status.\nPearson correlations were used to examine the association between the patient safety culture composites. ANOVA f-test with multiple comparison corrected using the Bonferroni method was used to examine the association between the two outcome variables (patient safety grade and the number of events reported) with the remaining patient safety culture composites.\nStudent T-Test and ANOVA f-test with multiple comparison corrected using the Bonferroni method were then used to examine how trends in the outcome variables (frequency of events reported and overall perception of safety) differ across hospital and respondent characteristics. Finally, cross tables were constructed and chi-square tests were used to assess how trends in the outcome variables (patient safety grade and the number of events reported) differed across respondent and hospital characteristics.", "The four outcome variables were regressed against the 10 composite scores, respondent's position in the hospital, accreditation status, and hospital size. Since the data was clustered by hospital, we used appropriate statistical techniques to control for this effect. Four regression models were constructed, two adopted Generalized Estimating Equations (the two categorical outcome variables: number of events reported and patient safety grade) and the other two models followed a linear mixed regression model (the two composites for frequency of events reported and overall perception of safety).", "[SUBTITLE] Comparison of Means for the Frequency of Events Reported and the Overall Perception of Safety across Respondent and Hospital Characteristics [SUBSECTION] Significant differences were observed between units and positions when comparing results across respondent and hospital characteristics. For both the frequency of events reported and the overall perception of safety, significantly higher means were observed for diagnostics (mean = 3.97 ± 1.01; mean = 3.96 ± 0.68) as compared to surgical (mean = 3.78 ± 1.05; mean = 3.82 ± 0.67) and medical units (mean = 3.93 ± 0.99; mean = 3.83 ± 0.68) (See Table 1).\nComparison of means for two outcome composite scores across hospital and respondent characteristics (identical letters represent significance between indicated groups)\nSignificantly higher means were observed for nurses (mean = 3.89 ± 1.00; mean = 3.80 ± 0.66) than physicians (mean = 3.78 ± 0.92; mean = 3.69 ± 0.75) on the frequency of events reported and the overall perception of safety respectively (See Table 1).\nFor frequency of events reported, higher means (and thus more events) were observed for increasing years of experience whereas the opposite trend prevailed for overall perception of safety (See Table 1). Small hospitals (mean = 3.95 ± 1.00) had a significantly higher frequency of events reported in comparison to medium (mean = 3.87 ± 1.03) and large hospitals (mean = 3.81 ± 1.08). Small hospitals (mean = 3.84 ± 0.67) also had a better overall perception of safety than large hospitals (mean = 3.76 ± 0.71). Accredited hospitals (mean = 3.91 ± 1.03) had a significantly higher frequency of events reported than non-accredited hospitals (mean = 3.84 ± 1.03). Accredited hospitals (mean = 3.84 ± 0.69) also had a better overall perception of safety than non-accredited hospitals (mean = 3.71 ± 0.68) (See Table 1).\nSignificant differences were observed between units and positions when comparing results across respondent and hospital characteristics. For both the frequency of events reported and the overall perception of safety, significantly higher means were observed for diagnostics (mean = 3.97 ± 1.01; mean = 3.96 ± 0.68) as compared to surgical (mean = 3.78 ± 1.05; mean = 3.82 ± 0.67) and medical units (mean = 3.93 ± 0.99; mean = 3.83 ± 0.68) (See Table 1).\nComparison of means for two outcome composite scores across hospital and respondent characteristics (identical letters represent significance between indicated groups)\nSignificantly higher means were observed for nurses (mean = 3.89 ± 1.00; mean = 3.80 ± 0.66) than physicians (mean = 3.78 ± 0.92; mean = 3.69 ± 0.75) on the frequency of events reported and the overall perception of safety respectively (See Table 1).\nFor frequency of events reported, higher means (and thus more events) were observed for increasing years of experience whereas the opposite trend prevailed for overall perception of safety (See Table 1). Small hospitals (mean = 3.95 ± 1.00) had a significantly higher frequency of events reported in comparison to medium (mean = 3.87 ± 1.03) and large hospitals (mean = 3.81 ± 1.08). Small hospitals (mean = 3.84 ± 0.67) also had a better overall perception of safety than large hospitals (mean = 3.76 ± 0.71). Accredited hospitals (mean = 3.91 ± 1.03) had a significantly higher frequency of events reported than non-accredited hospitals (mean = 3.84 ± 1.03). Accredited hospitals (mean = 3.84 ± 0.69) also had a better overall perception of safety than non-accredited hospitals (mean = 3.71 ± 0.68) (See Table 1).\n[SUBTITLE] Correlations between Patient Safety Culture Composites [SUBSECTION] While significant correlations were observed among all patient safety culture composites, differences in the strength of the correlation were observed (See Table 2). The composite measuring staffing showed the weakest correlation with the outcome on frequency of event reporting (Pearson r = 0.107). The composite measuring feedback and communication about errors had the strongest correlation with this outcome variable (Pearson r = 0.378) (See Table 2).\nCorrelation between patient safety culture composites\n*Correlation is significant at the 0.01 level (2-tailed).\nThe composite with the weakest correlation with the outcome variable measuring perceptions of patient safety was non-punitive response to error (Pearson r = 0.179) while that with the strongest correlation was the composite measuring supervisor/manager expectations and actions promoting safety (Pearson r = 0.371) (See Table 2).\nWhile significant correlations were observed among all patient safety culture composites, differences in the strength of the correlation were observed (See Table 2). The composite measuring staffing showed the weakest correlation with the outcome on frequency of event reporting (Pearson r = 0.107). The composite measuring feedback and communication about errors had the strongest correlation with this outcome variable (Pearson r = 0.378) (See Table 2).\nCorrelation between patient safety culture composites\n*Correlation is significant at the 0.01 level (2-tailed).\nThe composite with the weakest correlation with the outcome variable measuring perceptions of patient safety was non-punitive response to error (Pearson r = 0.179) while that with the strongest correlation was the composite measuring supervisor/manager expectations and actions promoting safety (Pearson r = 0.371) (See Table 2).\n[SUBTITLE] Comparison of Means between Outcome Variables and Patient Safety Composites [SUBSECTION] Significantly different means were reported for patient safety grade with all patient safety composite scores. Respondents who gave \"Excellent/Very Good\" patient safety grades had the highest mean scores for patient safety composites (See Table 3). The number of events reported was significantly associated with the composites measuring communication openness, feedback and communication about errors, non-punitive response to error, hospital handoffs and transitions, and teamwork across hospital units. The highest means were observed when reporting more than 5 events for the composites on communication openness, feedback and communication about errors, and non-punitive response to error. The opposite was observed for the composites measuring hospital handoffs and transitions and teamwork across hospital units where the highest means were observed when no events were reported (see Table 3).\nComparison of means between patient safety grade and number of events reported with patient safety culture composite scores\nPatient Safety Grade\na. Significant difference between \"Poor or Failing\" and \"Acceptable\"\nb. Significant difference between \"Poor or Failing\" and \"Excellent/Very Good\"\nc. Significant difference between \"Acceptable\" and \"Excellent/Very Good\"\nNumber of Events Reported\na. Significant difference between \"No events reported\" and \"1 to 5 events reported\"\nb. Significant difference between \"No events reported\" and \"> 5 events reported\"\nc. Significant difference between \"1 to 5 events reported\" and \"> 5 events reported\"\nSignificantly different means were reported for patient safety grade with all patient safety composite scores. Respondents who gave \"Excellent/Very Good\" patient safety grades had the highest mean scores for patient safety composites (See Table 3). The number of events reported was significantly associated with the composites measuring communication openness, feedback and communication about errors, non-punitive response to error, hospital handoffs and transitions, and teamwork across hospital units. The highest means were observed when reporting more than 5 events for the composites on communication openness, feedback and communication about errors, and non-punitive response to error. The opposite was observed for the composites measuring hospital handoffs and transitions and teamwork across hospital units where the highest means were observed when no events were reported (see Table 3).\nComparison of means between patient safety grade and number of events reported with patient safety culture composite scores\nPatient Safety Grade\na. Significant difference between \"Poor or Failing\" and \"Acceptable\"\nb. Significant difference between \"Poor or Failing\" and \"Excellent/Very Good\"\nc. Significant difference between \"Acceptable\" and \"Excellent/Very Good\"\nNumber of Events Reported\na. Significant difference between \"No events reported\" and \"1 to 5 events reported\"\nb. Significant difference between \"No events reported\" and \"> 5 events reported\"\nc. Significant difference between \"1 to 5 events reported\" and \"> 5 events reported\"\n[SUBTITLE] Distribution of the Patient Safety Grade and the Number of Events Reported across Respondent and Hospital Characteristics [SUBSECTION] [SUBTITLE] Patient Safety Grade [SUBSECTION] The highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\nThe highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\n[SUBTITLE] Number of Events Reported [SUBSECTION] Respondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\nRespondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\n[SUBTITLE] Patient Safety Grade [SUBSECTION] The highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\nThe highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\n[SUBTITLE] Number of Events Reported [SUBSECTION] Respondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\nRespondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\n[SUBTITLE] Generalized Estimating Equations for the Patient Safety Composite Scores and Respondent and Hospital Characteristics against the Patient Safety Grade and the Number of Events Reported (See Table 5) [SUBSECTION] Generalized Estimating Equations\n[SUBTITLE] Patient Safety Grade [SUBSECTION] Results showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\nResults showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\n[SUBTITLE] Number of Events Reported [SUBSECTION] A one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\nA one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\nGeneralized Estimating Equations\n[SUBTITLE] Patient Safety Grade [SUBSECTION] Results showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\nResults showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\n[SUBTITLE] Number of Events Reported [SUBSECTION] A one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\nA one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\n[SUBTITLE] Linear Mixed Regression for the Patient Safety Composite Scores and Respondent and Hospital Characteristics against the Frequency of Events Reported and the Overall Perception of Safety (See Table 6) [SUBSECTION] [SUBTITLE] Frequency of Events Reported [SUBSECTION] Mixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\nMixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\n[SUBTITLE] Overall Perception of Safety [SUBSECTION] Perception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).\nPerception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).\n[SUBTITLE] Frequency of Events Reported [SUBSECTION] Mixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\nMixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\n[SUBTITLE] Overall Perception of Safety [SUBSECTION] Perception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).\nPerception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).", "Significant differences were observed between units and positions when comparing results across respondent and hospital characteristics. For both the frequency of events reported and the overall perception of safety, significantly higher means were observed for diagnostics (mean = 3.97 ± 1.01; mean = 3.96 ± 0.68) as compared to surgical (mean = 3.78 ± 1.05; mean = 3.82 ± 0.67) and medical units (mean = 3.93 ± 0.99; mean = 3.83 ± 0.68) (See Table 1).\nComparison of means for two outcome composite scores across hospital and respondent characteristics (identical letters represent significance between indicated groups)\nSignificantly higher means were observed for nurses (mean = 3.89 ± 1.00; mean = 3.80 ± 0.66) than physicians (mean = 3.78 ± 0.92; mean = 3.69 ± 0.75) on the frequency of events reported and the overall perception of safety respectively (See Table 1).\nFor frequency of events reported, higher means (and thus more events) were observed for increasing years of experience whereas the opposite trend prevailed for overall perception of safety (See Table 1). Small hospitals (mean = 3.95 ± 1.00) had a significantly higher frequency of events reported in comparison to medium (mean = 3.87 ± 1.03) and large hospitals (mean = 3.81 ± 1.08). Small hospitals (mean = 3.84 ± 0.67) also had a better overall perception of safety than large hospitals (mean = 3.76 ± 0.71). Accredited hospitals (mean = 3.91 ± 1.03) had a significantly higher frequency of events reported than non-accredited hospitals (mean = 3.84 ± 1.03). Accredited hospitals (mean = 3.84 ± 0.69) also had a better overall perception of safety than non-accredited hospitals (mean = 3.71 ± 0.68) (See Table 1).", "While significant correlations were observed among all patient safety culture composites, differences in the strength of the correlation were observed (See Table 2). The composite measuring staffing showed the weakest correlation with the outcome on frequency of event reporting (Pearson r = 0.107). The composite measuring feedback and communication about errors had the strongest correlation with this outcome variable (Pearson r = 0.378) (See Table 2).\nCorrelation between patient safety culture composites\n*Correlation is significant at the 0.01 level (2-tailed).\nThe composite with the weakest correlation with the outcome variable measuring perceptions of patient safety was non-punitive response to error (Pearson r = 0.179) while that with the strongest correlation was the composite measuring supervisor/manager expectations and actions promoting safety (Pearson r = 0.371) (See Table 2).", "Significantly different means were reported for patient safety grade with all patient safety composite scores. Respondents who gave \"Excellent/Very Good\" patient safety grades had the highest mean scores for patient safety composites (See Table 3). The number of events reported was significantly associated with the composites measuring communication openness, feedback and communication about errors, non-punitive response to error, hospital handoffs and transitions, and teamwork across hospital units. The highest means were observed when reporting more than 5 events for the composites on communication openness, feedback and communication about errors, and non-punitive response to error. The opposite was observed for the composites measuring hospital handoffs and transitions and teamwork across hospital units where the highest means were observed when no events were reported (see Table 3).\nComparison of means between patient safety grade and number of events reported with patient safety culture composite scores\nPatient Safety Grade\na. Significant difference between \"Poor or Failing\" and \"Acceptable\"\nb. Significant difference between \"Poor or Failing\" and \"Excellent/Very Good\"\nc. Significant difference between \"Acceptable\" and \"Excellent/Very Good\"\nNumber of Events Reported\na. Significant difference between \"No events reported\" and \"1 to 5 events reported\"\nb. Significant difference between \"No events reported\" and \"> 5 events reported\"\nc. Significant difference between \"1 to 5 events reported\" and \"> 5 events reported\"", "[SUBTITLE] Patient Safety Grade [SUBSECTION] The highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\nThe highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\n[SUBTITLE] Number of Events Reported [SUBSECTION] Respondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\nRespondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).", "The highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).", "Respondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).", "Generalized Estimating Equations\n[SUBTITLE] Patient Safety Grade [SUBSECTION] Results showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\nResults showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\n[SUBTITLE] Number of Events Reported [SUBSECTION] A one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\nA one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).", "Results showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.", "A one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).", "[SUBTITLE] Frequency of Events Reported [SUBSECTION] Mixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\nMixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\n[SUBTITLE] Overall Perception of Safety [SUBSECTION] Perception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).\nPerception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).", "Mixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).", "Perception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).", "The HSOPSC is one of the most common tools being used to assess the culture of safety in hospitals. Studies that utilize this tool usually report the 12 composite scores and the scores on the patient safety grade and the number of events reported. However, exploring the association between the patient safety composite scores and the hospital and respondent characteristics with the patient safety culture outcomes are not common in the literature. To our knowledge, this is one of the few studies that explore such an association. The analysis of results identified that patient safety culture predictors such as event reporting, proper communication, patient safety leadership and management, hospital size, and accreditation status are associated with the patient safety culture outcomes.\n[SUBTITLE] Event Reporting [SUBSECTION] A safety culture includes three major components a just culture, a reporting culture, and a learning culture [11]. Event reporting, an essential component for achieving a learning culture, can only happen in a non-punitive environment where events can be reported without people being blamed [12]. The non-punitive response to error composite received one of the lowest scores revealing that Lebanese hospital employees are not at ease when it comes to reporting errors.\nThe majority of physicians and nurses in our sample reported no events over the past 12 months. Training opportunities that empower physicians to improve patient safety are limited thus investing in the training of physicians is important because they can play a major role in improving patient safety initiatives primarily by improving patient safety outcomes and reducing hassles and wasted time [13].\nAccording to Leape et al. (1995), errors will be more frequent if nurses did not intercept 86% of all potential errors [14]. However, the majority of nurses in our sample reported no events over the past 12 months. Many errors in health care go unreported for many reasons including fear, humiliation, the presence of a punitive response to error, and the fact that reporting will not usually result in actual change [15]. Encouraging health professionals, specifically nurses, to report events in a non-punitive environment is crucial for improving patient safety.\nIn our study, the fewest number of respondents to report more than 5 events over the past 12 months worked in surgical units. Errors in operating rooms are not uncommon and can sometimes be catastrophic [16]. Creating a patient safety culture in surgical units by improving communication and reporting more events is a high priority for operating room staff and hospitals [16].\nEmployees who do not deal directly with the patient are more at ease when it comes to reporting errors. As mentioned in Jones et al. (2008), work in laboratory units is considered as more organized than other units since it is controlled by more professional standards and because errors investigated in these units are done as a group [17]. On the contrary, when an error is performed by a nurse, the nurse is investigated as an individual rather a member of a medical team [17].\nWork experience at the hospital also had some impact on the frequency and number of events reported. Frequency of events reported was found to increase with increasing years of experience. On the other hand, the score for perception of patient safety decreased as experience in the hospital increased. The perception of safety is defined as the extent to which procedures and systems are good at preventing errors and the lack of patient safety problems. As people become more experienced, they become more aware of the safety practices undertaken in the institutions they work in. When the perception of safety decreases with the increase in the years of experience, it means that the staff members do not agree that the patient safety practices, systems, and procedures in the hospitals act as barriers to errors and problems. A study by Bodur and Feliz (2009) showed that patient safety culture scores decreased as seniority increased [18]. The observation may be the result of an increase in medical errors done by the respondent due to frustration with hospital regulations or increasing staff awareness of safety problems and thus additional reporting.\nA safety culture includes three major components a just culture, a reporting culture, and a learning culture [11]. Event reporting, an essential component for achieving a learning culture, can only happen in a non-punitive environment where events can be reported without people being blamed [12]. The non-punitive response to error composite received one of the lowest scores revealing that Lebanese hospital employees are not at ease when it comes to reporting errors.\nThe majority of physicians and nurses in our sample reported no events over the past 12 months. Training opportunities that empower physicians to improve patient safety are limited thus investing in the training of physicians is important because they can play a major role in improving patient safety initiatives primarily by improving patient safety outcomes and reducing hassles and wasted time [13].\nAccording to Leape et al. (1995), errors will be more frequent if nurses did not intercept 86% of all potential errors [14]. However, the majority of nurses in our sample reported no events over the past 12 months. Many errors in health care go unreported for many reasons including fear, humiliation, the presence of a punitive response to error, and the fact that reporting will not usually result in actual change [15]. Encouraging health professionals, specifically nurses, to report events in a non-punitive environment is crucial for improving patient safety.\nIn our study, the fewest number of respondents to report more than 5 events over the past 12 months worked in surgical units. Errors in operating rooms are not uncommon and can sometimes be catastrophic [16]. Creating a patient safety culture in surgical units by improving communication and reporting more events is a high priority for operating room staff and hospitals [16].\nEmployees who do not deal directly with the patient are more at ease when it comes to reporting errors. As mentioned in Jones et al. (2008), work in laboratory units is considered as more organized than other units since it is controlled by more professional standards and because errors investigated in these units are done as a group [17]. On the contrary, when an error is performed by a nurse, the nurse is investigated as an individual rather a member of a medical team [17].\nWork experience at the hospital also had some impact on the frequency and number of events reported. Frequency of events reported was found to increase with increasing years of experience. On the other hand, the score for perception of patient safety decreased as experience in the hospital increased. The perception of safety is defined as the extent to which procedures and systems are good at preventing errors and the lack of patient safety problems. As people become more experienced, they become more aware of the safety practices undertaken in the institutions they work in. When the perception of safety decreases with the increase in the years of experience, it means that the staff members do not agree that the patient safety practices, systems, and procedures in the hospitals act as barriers to errors and problems. A study by Bodur and Feliz (2009) showed that patient safety culture scores decreased as seniority increased [18]. The observation may be the result of an increase in medical errors done by the respondent due to frustration with hospital regulations or increasing staff awareness of safety problems and thus additional reporting.\n[SUBTITLE] Communication [SUBSECTION] Proper communication within and across healthcare teams is essential to remove any threats to safety of patients. Communication problems have been identified as major contributing factors to adverse events [7]. An analysis of 2,455 sentinel events reported to the Joint Commission on the Accreditation of Healthcare Organizations showed that 70% of the cases were a result of failure in communication [19]. In the absence of proper communication between the different hospital units, patient safety might be jeopardized. Our results revealed that higher scores on teamwork across hospital units increase the frequency of events reported. Moreover, higher scores on hospital handoffs and transitions increased the likelihood of having a better perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\nProper communication within and across healthcare teams is essential to remove any threats to safety of patients. Communication problems have been identified as major contributing factors to adverse events [7]. An analysis of 2,455 sentinel events reported to the Joint Commission on the Accreditation of Healthcare Organizations showed that 70% of the cases were a result of failure in communication [19]. In the absence of proper communication between the different hospital units, patient safety might be jeopardized. Our results revealed that higher scores on teamwork across hospital units increase the frequency of events reported. Moreover, higher scores on hospital handoffs and transitions increased the likelihood of having a better perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\n[SUBTITLE] Patient safety leadership and management [SUBSECTION] In order for a patient safety program to be successful, strong leadership is needed. Senior leaders are the only ones who are able to create the culture and commitment needed to solve underlying system causes of medical errors and harm to patients. When leadership and management is committed to a culture of safety, the whole organization will follow and thus disclosing adverse events and finding their root causes will become an organizational process. The focus should be in emergency rooms, operating rooms, and intensive care units [20]. Our results showed that more support from hospital management for patient safety increased the frequency of events reported. It also increased the likelihood of having a better overall perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\nIn order for a patient safety program to be successful, strong leadership is needed. Senior leaders are the only ones who are able to create the culture and commitment needed to solve underlying system causes of medical errors and harm to patients. When leadership and management is committed to a culture of safety, the whole organization will follow and thus disclosing adverse events and finding their root causes will become an organizational process. The focus should be in emergency rooms, operating rooms, and intensive care units [20]. Our results showed that more support from hospital management for patient safety increased the frequency of events reported. It also increased the likelihood of having a better overall perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\n[SUBTITLE] Staffing [SUBSECTION] Staffing, a major predictor of patient safety, was one of the composites that received a very low score. Having a strong, capable, and motivated workforce is one of the biggest challenges for hospitals today [21]. Evidence has shown that a strong link exists between the availability of health care providers and population health outcomes [21]. According to Sandars & Cook (2007), major catastrophes have occurred in organizations with insufficient staffing. Medical personnel in under-staffed hospitals are overworked [7]. They also suffer from stress and sleeplessness which might lead to lapses in performance thus leading to errors affecting quality and performance [22]. Our results showed that a more positive score on staffing increased the likelihood of having a more positive perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\nStaffing, a major predictor of patient safety, was one of the composites that received a very low score. Having a strong, capable, and motivated workforce is one of the biggest challenges for hospitals today [21]. Evidence has shown that a strong link exists between the availability of health care providers and population health outcomes [21]. According to Sandars & Cook (2007), major catastrophes have occurred in organizations with insufficient staffing. Medical personnel in under-staffed hospitals are overworked [7]. They also suffer from stress and sleeplessness which might lead to lapses in performance thus leading to errors affecting quality and performance [22]. Our results showed that a more positive score on staffing increased the likelihood of having a more positive perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\n[SUBTITLE] Hospital size and accreditation [SUBSECTION] Hospital size and accreditation status were also factors affecting the culture of safety in hospitals. Small hospitals scored higher than larger hospitals on both the frequency of events reported and on the overall perception of safety. Large hospitals usually face challenges when it comes to implementing quality work especially because of bureaucracy. On the contrary, small hospitals have a more homogenous culture where staff members are more likely to share the same values [23].\nAccredited hospitals also scored higher on both frequency of events reported and on the overall perception of safety in comparison to non-accredited hospitals. Respondents working at accredited hospitals had an increased likelihood of having a more positive perception of safety. Since the development of hospital accreditation programs that made accreditation a requirement for financial reimbursement by the MOPH, quality improvement initiatives have gained increased attention in Lebanese hospitals [24]. A study by El-Jardali et al. (2008) showed that Lebanese nurses perceived that accreditation created an improvement in the quality of health care [25].\nHospital size and accreditation status were also factors affecting the culture of safety in hospitals. Small hospitals scored higher than larger hospitals on both the frequency of events reported and on the overall perception of safety. Large hospitals usually face challenges when it comes to implementing quality work especially because of bureaucracy. On the contrary, small hospitals have a more homogenous culture where staff members are more likely to share the same values [23].\nAccredited hospitals also scored higher on both frequency of events reported and on the overall perception of safety in comparison to non-accredited hospitals. Respondents working at accredited hospitals had an increased likelihood of having a more positive perception of safety. Since the development of hospital accreditation programs that made accreditation a requirement for financial reimbursement by the MOPH, quality improvement initiatives have gained increased attention in Lebanese hospitals [24]. A study by El-Jardali et al. (2008) showed that Lebanese nurses perceived that accreditation created an improvement in the quality of health care [25].\n[SUBTITLE] Limitations [SUBSECTION] A methodological limitation pertaining to this study should be acknowledged. In the literature, it is argued that social and behavioral research, particularly those that include self-reports such as surveys, is prone to Common method variance (CMV). CMV is believed to pose a threat to the validity of the data as it may either inflate or deflate the correlations among research variables. Although this issue can be attended to during survey development, evidence in the literature stipulate that although such research methods are more prone to CMV, they should not be considered weak or inappropriate if the researchers follow rigorous research conventions in research design, data collection and analysis [26]. In our study, CMV was avoided by the following [26]:\n- Assuring that the composite scores were derived based on a combination of items;\n- Counterbalancing the order of the questions;\n- Ensuring the confidentiality and anonymity of respondents;\n- Using clearly written scale items to make responses less subject to bias; and\n- Informing respondents in the consent form that there is no preferred or correct answer, and that their individual responses would not be viewed by their management and survey completion would not affect their status at their hospital so they would be encouraged to honestly assess and respond freely\nA methodological limitation pertaining to this study should be acknowledged. In the literature, it is argued that social and behavioral research, particularly those that include self-reports such as surveys, is prone to Common method variance (CMV). CMV is believed to pose a threat to the validity of the data as it may either inflate or deflate the correlations among research variables. Although this issue can be attended to during survey development, evidence in the literature stipulate that although such research methods are more prone to CMV, they should not be considered weak or inappropriate if the researchers follow rigorous research conventions in research design, data collection and analysis [26]. In our study, CMV was avoided by the following [26]:\n- Assuring that the composite scores were derived based on a combination of items;\n- Counterbalancing the order of the questions;\n- Ensuring the confidentiality and anonymity of respondents;\n- Using clearly written scale items to make responses less subject to bias; and\n- Informing respondents in the consent form that there is no preferred or correct answer, and that their individual responses would not be viewed by their management and survey completion would not affect their status at their hospital so they would be encouraged to honestly assess and respond freely", "A safety culture includes three major components a just culture, a reporting culture, and a learning culture [11]. Event reporting, an essential component for achieving a learning culture, can only happen in a non-punitive environment where events can be reported without people being blamed [12]. The non-punitive response to error composite received one of the lowest scores revealing that Lebanese hospital employees are not at ease when it comes to reporting errors.\nThe majority of physicians and nurses in our sample reported no events over the past 12 months. Training opportunities that empower physicians to improve patient safety are limited thus investing in the training of physicians is important because they can play a major role in improving patient safety initiatives primarily by improving patient safety outcomes and reducing hassles and wasted time [13].\nAccording to Leape et al. (1995), errors will be more frequent if nurses did not intercept 86% of all potential errors [14]. However, the majority of nurses in our sample reported no events over the past 12 months. Many errors in health care go unreported for many reasons including fear, humiliation, the presence of a punitive response to error, and the fact that reporting will not usually result in actual change [15]. Encouraging health professionals, specifically nurses, to report events in a non-punitive environment is crucial for improving patient safety.\nIn our study, the fewest number of respondents to report more than 5 events over the past 12 months worked in surgical units. Errors in operating rooms are not uncommon and can sometimes be catastrophic [16]. Creating a patient safety culture in surgical units by improving communication and reporting more events is a high priority for operating room staff and hospitals [16].\nEmployees who do not deal directly with the patient are more at ease when it comes to reporting errors. As mentioned in Jones et al. (2008), work in laboratory units is considered as more organized than other units since it is controlled by more professional standards and because errors investigated in these units are done as a group [17]. On the contrary, when an error is performed by a nurse, the nurse is investigated as an individual rather a member of a medical team [17].\nWork experience at the hospital also had some impact on the frequency and number of events reported. Frequency of events reported was found to increase with increasing years of experience. On the other hand, the score for perception of patient safety decreased as experience in the hospital increased. The perception of safety is defined as the extent to which procedures and systems are good at preventing errors and the lack of patient safety problems. As people become more experienced, they become more aware of the safety practices undertaken in the institutions they work in. When the perception of safety decreases with the increase in the years of experience, it means that the staff members do not agree that the patient safety practices, systems, and procedures in the hospitals act as barriers to errors and problems. A study by Bodur and Feliz (2009) showed that patient safety culture scores decreased as seniority increased [18]. The observation may be the result of an increase in medical errors done by the respondent due to frustration with hospital regulations or increasing staff awareness of safety problems and thus additional reporting.", "Proper communication within and across healthcare teams is essential to remove any threats to safety of patients. Communication problems have been identified as major contributing factors to adverse events [7]. An analysis of 2,455 sentinel events reported to the Joint Commission on the Accreditation of Healthcare Organizations showed that 70% of the cases were a result of failure in communication [19]. In the absence of proper communication between the different hospital units, patient safety might be jeopardized. Our results revealed that higher scores on teamwork across hospital units increase the frequency of events reported. Moreover, higher scores on hospital handoffs and transitions increased the likelihood of having a better perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.", "In order for a patient safety program to be successful, strong leadership is needed. Senior leaders are the only ones who are able to create the culture and commitment needed to solve underlying system causes of medical errors and harm to patients. When leadership and management is committed to a culture of safety, the whole organization will follow and thus disclosing adverse events and finding their root causes will become an organizational process. The focus should be in emergency rooms, operating rooms, and intensive care units [20]. Our results showed that more support from hospital management for patient safety increased the frequency of events reported. It also increased the likelihood of having a better overall perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.", "Staffing, a major predictor of patient safety, was one of the composites that received a very low score. Having a strong, capable, and motivated workforce is one of the biggest challenges for hospitals today [21]. Evidence has shown that a strong link exists between the availability of health care providers and population health outcomes [21]. According to Sandars & Cook (2007), major catastrophes have occurred in organizations with insufficient staffing. Medical personnel in under-staffed hospitals are overworked [7]. They also suffer from stress and sleeplessness which might lead to lapses in performance thus leading to errors affecting quality and performance [22]. Our results showed that a more positive score on staffing increased the likelihood of having a more positive perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.", "Hospital size and accreditation status were also factors affecting the culture of safety in hospitals. Small hospitals scored higher than larger hospitals on both the frequency of events reported and on the overall perception of safety. Large hospitals usually face challenges when it comes to implementing quality work especially because of bureaucracy. On the contrary, small hospitals have a more homogenous culture where staff members are more likely to share the same values [23].\nAccredited hospitals also scored higher on both frequency of events reported and on the overall perception of safety in comparison to non-accredited hospitals. Respondents working at accredited hospitals had an increased likelihood of having a more positive perception of safety. Since the development of hospital accreditation programs that made accreditation a requirement for financial reimbursement by the MOPH, quality improvement initiatives have gained increased attention in Lebanese hospitals [24]. A study by El-Jardali et al. (2008) showed that Lebanese nurses perceived that accreditation created an improvement in the quality of health care [25].", "A methodological limitation pertaining to this study should be acknowledged. In the literature, it is argued that social and behavioral research, particularly those that include self-reports such as surveys, is prone to Common method variance (CMV). CMV is believed to pose a threat to the validity of the data as it may either inflate or deflate the correlations among research variables. Although this issue can be attended to during survey development, evidence in the literature stipulate that although such research methods are more prone to CMV, they should not be considered weak or inappropriate if the researchers follow rigorous research conventions in research design, data collection and analysis [26]. In our study, CMV was avoided by the following [26]:\n- Assuring that the composite scores were derived based on a combination of items;\n- Counterbalancing the order of the questions;\n- Ensuring the confidentiality and anonymity of respondents;\n- Using clearly written scale items to make responses less subject to bias; and\n- Informing respondents in the consent form that there is no preferred or correct answer, and that their individual responses would not be viewed by their management and survey completion would not affect their status at their hospital so they would be encouraged to honestly assess and respond freely", "Investing in patient safety practices in Lebanese hospitals is crucial. While patient safety is everyone's concern, it is not easy for everyone who works in health care organizations to understand this concept. In Lebanon, most health workers have had different training, and often hold a value system that is specific to their professional group. To be truly effective, patient safety needs to be incorporated into the education of health professionals across the spectrum of health care.\nOur results demonstrate that patient safety should be a top strategic priority for the health care organizations and its leaders. There should be blame-free system for identifying threats to patient safety, sharing information and learning from events. In addition, there should be a collaborative environment so that all health workers in the healthcare organization can share and exchange information about patient safety\nTo facilitate change in cultural behaviors, hospital management should assess and redesign their current patient safety system including governance and reporting structures. In addition, they should provide their health professionals with comprehensive training on patient safety concepts, tools, and implementations.\nOur results also show that the move to prioritize patient safety by healthcare systems through accreditation is important. However, there is still a good deal skepticism on the part of health professionals who feel that by sharing highly sensitive information they may still be subject to blame and penalties. Progress in this respect will need far more enlightened policies, where health professionals are actively encouraged to report errors for the purpose of learning and improvement. In Lebanon, it is important to strengthen the new accreditation chapter on patient safety by supporting hospitals in training their staff, especially the less experienced ones, on patient safety competencies and about effective implementation of the new standards. Further research is needed to study the association between patient safety culture and clinical outcomes.\nIt is hoped that the outcome of this study will help conduct a policy dialogue meeting for policy makers and stakeholders in Lebanon to discuss the findings and make deliberation about potential next steps. Senior policy makers, managers and leaders are the only ones who are able to create the culture and commitment needed to identify and resolve underlying systemic causes related to patient safety.", "IOM: Institute of Medicine; HSOPSC: Hospital Survey on Patient Safety Culture; MOPH: Ministry of Public Health;", "The authors declare that they have no competing interests.", "FE was the principal investigator on this study and contributed to the conception, design, as well as analysis and interpretation of results. HD was involved in conceptualizing and implementing the data analysis plan as well as overseeing drafting of research results. DJ made substantial contributions to project management, including data collection and analysis as well as in drafting the manuscript. MJ was in charge of project management including data collection, she also contributed to data analysis and write-up of the final version of the paper. NH contributed to data entry and analysis and also made major contributions to the write up of the paper. All authors have read and approved the final version of the manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1472-6963/11/45/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Lebanese Context", "Methods", "Study Design, Setting, and Sample", "Survey Measures and Outcome Variables", "Bivariate Analysis", "Multivariate Regression Analysis", "Results", "Comparison of Means for the Frequency of Events Reported and the Overall Perception of Safety across Respondent and Hospital Characteristics", "Correlations between Patient Safety Culture Composites", "Comparison of Means between Outcome Variables and Patient Safety Composites", "Distribution of the Patient Safety Grade and the Number of Events Reported across Respondent and Hospital Characteristics", "Patient Safety Grade", "Number of Events Reported", "Generalized Estimating Equations for the Patient Safety Composite Scores and Respondent and Hospital Characteristics against the Patient Safety Grade and the Number of Events Reported (See Table 5)", "Patient Safety Grade", "Number of Events Reported", "Linear Mixed Regression for the Patient Safety Composite Scores and Respondent and Hospital Characteristics against the Frequency of Events Reported and the Overall Perception of Safety (See Table 6)", "Frequency of Events Reported", "Overall Perception of Safety", "Discussion", "Event Reporting", "Communication", "Patient safety leadership and management", "Staffing", "Hospital size and accreditation", "Limitations", "Conclusion", "List of abbreviations used", "Competing interests", "Authors' contributions", "Pre-publication history", "Supplementary Material" ]

[ "Developing a patient safety culture was one of the recommendations made by the Institute of Medicine to assist hospitals in improving patient safety [1,2]. Assessing the organization's existing safety culture is the first stage of developing a safety culture [3]. Patient safety culture assessments, required by international accreditation organizations, allow healthcare organizations to obtain a clear view of the patient safety aspects requiring urgent attention, identify the strengths and weaknesses of their safety culture [4], help care giving units identify their existing patient safety problems [5], and benchmark their scores with other hospitals [6].\nAccording to literature, the major predictors of a positive patient safety culture in healthcare organizations specifically hospitals include communication founded on mutual trust, good information flow, shared perception of the importance of safety, organizational learning, commitment from management and leadership, and the presence of a non-punitive approach to incident and error reporting [7]. Patient safety culture outcomes include the staff members' perception of safety, the willingness of staff members to report events, the number of events reported, and an overall patient safety grade given by staff members to their units [8].\nA multitude of evidence has been published in the area of patient safety culture in recent years. Some of the available evidence tackles patient safety culture issues that require attention, factors affecting incident reporting by hospital staff, the role of workplace environment in shaping safety, and steps that can be followed to improve safety.\nDespite the wealth of evidence on patient safety culture, limited evidence still exists about the linkage between predictors and outcomes of patient safety culture especially in countries of the Eastern Mediterranean Region. One of the first efforts to assess the culture of safety in hospitals in the region was conducted in Lebanon by El-Jardali et al. [9].\n[SUBTITLE] Lebanese Context [SUBSECTION] The study by El-Jardali et al. (2010) entitled \"The Current state of Patient Safety Culture in Lebanese Hospitals: A study at Baseline\" [9] utilized an Arabic translated version of the Hospital Survey on Patient Safety Culture (HSOPSC) [8]. It aimed at identifying the most critical issues related to patient safety culture and potential strategies to implement the patient safety accreditation standards in the light of a newly added chapter to the Lebanese handbook of hospital accreditation [10].\nThe HSOPSC measures 12 patient safety culture composites representing several patient safety culture predictors (See Box 1). The HSOPSC also requires respondents to give their work area/unit a patient safety grade and to answer a question on the number of events reported in the past 12 months [8].\nCalculating the percentage of positive responses for each composite revealed that the composites with the highest positive ratings were teamwork within units, hospital management support for patient safety, and organizational learning and continuous improvement. However, composites with the lowest ratings were teamwork across hospital units, hospital handoffs and transitions, staffing, and non-punitive response to error [9]. Approximately 60% of respondents reported not completing any event reports in the past 12 months and over 70% gave their units an \"Excellent/Very Good\" patient safety grade. Bivariate and multivariate analyses revealed significant differences across hospitals of different size and accreditation status [9].\nStudy findings outlined above represent the first component of data analysis for that stage [9]. Findings provided evidence that communication across units, staffing, event reporting, and the culture of response to error were major patient safety culture issues [9]. This paper, though, will further explore the association between patient safety culture predictors and outcomes, taking into consideration respondent and hospital characteristics. In addition, it will examine the correlation between the patient safety culture composites. Thus, the objective of this paper is to address the afore-mentioned objectives using bivariate as well as multivariate analyses.\nThe study by El-Jardali et al. (2010) entitled \"The Current state of Patient Safety Culture in Lebanese Hospitals: A study at Baseline\" [9] utilized an Arabic translated version of the Hospital Survey on Patient Safety Culture (HSOPSC) [8]. It aimed at identifying the most critical issues related to patient safety culture and potential strategies to implement the patient safety accreditation standards in the light of a newly added chapter to the Lebanese handbook of hospital accreditation [10].\nThe HSOPSC measures 12 patient safety culture composites representing several patient safety culture predictors (See Box 1). The HSOPSC also requires respondents to give their work area/unit a patient safety grade and to answer a question on the number of events reported in the past 12 months [8].\nCalculating the percentage of positive responses for each composite revealed that the composites with the highest positive ratings were teamwork within units, hospital management support for patient safety, and organizational learning and continuous improvement. However, composites with the lowest ratings were teamwork across hospital units, hospital handoffs and transitions, staffing, and non-punitive response to error [9]. Approximately 60% of respondents reported not completing any event reports in the past 12 months and over 70% gave their units an \"Excellent/Very Good\" patient safety grade. Bivariate and multivariate analyses revealed significant differences across hospitals of different size and accreditation status [9].\nStudy findings outlined above represent the first component of data analysis for that stage [9]. Findings provided evidence that communication across units, staffing, event reporting, and the culture of response to error were major patient safety culture issues [9]. This paper, though, will further explore the association between patient safety culture predictors and outcomes, taking into consideration respondent and hospital characteristics. In addition, it will examine the correlation between the patient safety culture composites. Thus, the objective of this paper is to address the afore-mentioned objectives using bivariate as well as multivariate analyses.", "The study by El-Jardali et al. (2010) entitled \"The Current state of Patient Safety Culture in Lebanese Hospitals: A study at Baseline\" [9] utilized an Arabic translated version of the Hospital Survey on Patient Safety Culture (HSOPSC) [8]. It aimed at identifying the most critical issues related to patient safety culture and potential strategies to implement the patient safety accreditation standards in the light of a newly added chapter to the Lebanese handbook of hospital accreditation [10].\nThe HSOPSC measures 12 patient safety culture composites representing several patient safety culture predictors (See Box 1). The HSOPSC also requires respondents to give their work area/unit a patient safety grade and to answer a question on the number of events reported in the past 12 months [8].\nCalculating the percentage of positive responses for each composite revealed that the composites with the highest positive ratings were teamwork within units, hospital management support for patient safety, and organizational learning and continuous improvement. However, composites with the lowest ratings were teamwork across hospital units, hospital handoffs and transitions, staffing, and non-punitive response to error [9]. Approximately 60% of respondents reported not completing any event reports in the past 12 months and over 70% gave their units an \"Excellent/Very Good\" patient safety grade. Bivariate and multivariate analyses revealed significant differences across hospitals of different size and accreditation status [9].\nStudy findings outlined above represent the first component of data analysis for that stage [9]. Findings provided evidence that communication across units, staffing, event reporting, and the culture of response to error were major patient safety culture issues [9]. This paper, though, will further explore the association between patient safety culture predictors and outcomes, taking into consideration respondent and hospital characteristics. In addition, it will examine the correlation between the patient safety culture composites. Thus, the objective of this paper is to address the afore-mentioned objectives using bivariate as well as multivariate analyses.", "[SUBTITLE] Study Design, Setting, and Sample [SUBSECTION] The study adopted a cross sectional research design utilizing a customized version of the HSOPSC. A pilot testing phase preceded data collection in order to ensure the validity and reliability of the Arabic translated version of the questionnaire. Of the 126 hospitals registered in the Lebanese Syndicate of Private Hospitals that were contacted and asked to participate, 68 consented [9]. The survey targeted hospital employees including physicians, nurses, clinical and non-clinical staff, pharmacy and laboratory staff, dietary and radiology staff, supervisors, and hospital managers. The questionnaire was distributed to 12,250 hospital employees and 6807 were returned yielding an overall response rate of 55.56%. Additional details on the study methodology can be found in El-Jardali et al. [9].\nThe study adopted a cross sectional research design utilizing a customized version of the HSOPSC. A pilot testing phase preceded data collection in order to ensure the validity and reliability of the Arabic translated version of the questionnaire. Of the 126 hospitals registered in the Lebanese Syndicate of Private Hospitals that were contacted and asked to participate, 68 consented [9]. The survey targeted hospital employees including physicians, nurses, clinical and non-clinical staff, pharmacy and laboratory staff, dietary and radiology staff, supervisors, and hospital managers. The questionnaire was distributed to 12,250 hospital employees and 6807 were returned yielding an overall response rate of 55.56%. Additional details on the study methodology can be found in El-Jardali et al. [9].\n[SUBTITLE] Survey Measures and Outcome Variables [SUBSECTION] The HSOPSC is composed of 42 items that measure 12 composites of patient safety culture (See Box1, Additional file1 ). Items were scored using a five-point scale reflecting the agreement rate on a five-point frequency scale. The percentage of positive responses for each item was calculated; negatively worded items were reversed when computing percent positive response. Composite level scores were computed by summation of the items within the composite scales and dividing by the number of items with non-missing values [9].\nTwo of the composites (frequency of events reported and overall perception of safety) are two of the four patient safety culture outcome variables [8]. The remaining two outcome variables are the patient safety grade and the number of events reported [8].\nThe HSOPSC is composed of 42 items that measure 12 composites of patient safety culture (See Box1, Additional file1 ). Items were scored using a five-point scale reflecting the agreement rate on a five-point frequency scale. The percentage of positive responses for each item was calculated; negatively worded items were reversed when computing percent positive response. Composite level scores were computed by summation of the items within the composite scales and dividing by the number of items with non-missing values [9].\nTwo of the composites (frequency of events reported and overall perception of safety) are two of the four patient safety culture outcome variables [8]. The remaining two outcome variables are the patient safety grade and the number of events reported [8].\n[SUBTITLE] Bivariate Analysis [SUBSECTION] Bivariate analyses were used to examine the associations between patient safety culture composites and differences across hospitals of different size and accreditation status.\nPearson correlations were used to examine the association between the patient safety culture composites. ANOVA f-test with multiple comparison corrected using the Bonferroni method was used to examine the association between the two outcome variables (patient safety grade and the number of events reported) with the remaining patient safety culture composites.\nStudent T-Test and ANOVA f-test with multiple comparison corrected using the Bonferroni method were then used to examine how trends in the outcome variables (frequency of events reported and overall perception of safety) differ across hospital and respondent characteristics. Finally, cross tables were constructed and chi-square tests were used to assess how trends in the outcome variables (patient safety grade and the number of events reported) differed across respondent and hospital characteristics.\nBivariate analyses were used to examine the associations between patient safety culture composites and differences across hospitals of different size and accreditation status.\nPearson correlations were used to examine the association between the patient safety culture composites. ANOVA f-test with multiple comparison corrected using the Bonferroni method was used to examine the association between the two outcome variables (patient safety grade and the number of events reported) with the remaining patient safety culture composites.\nStudent T-Test and ANOVA f-test with multiple comparison corrected using the Bonferroni method were then used to examine how trends in the outcome variables (frequency of events reported and overall perception of safety) differ across hospital and respondent characteristics. Finally, cross tables were constructed and chi-square tests were used to assess how trends in the outcome variables (patient safety grade and the number of events reported) differed across respondent and hospital characteristics.\n[SUBTITLE] Multivariate Regression Analysis [SUBSECTION] The four outcome variables were regressed against the 10 composite scores, respondent's position in the hospital, accreditation status, and hospital size. Since the data was clustered by hospital, we used appropriate statistical techniques to control for this effect. Four regression models were constructed, two adopted Generalized Estimating Equations (the two categorical outcome variables: number of events reported and patient safety grade) and the other two models followed a linear mixed regression model (the two composites for frequency of events reported and overall perception of safety).\nThe four outcome variables were regressed against the 10 composite scores, respondent's position in the hospital, accreditation status, and hospital size. Since the data was clustered by hospital, we used appropriate statistical techniques to control for this effect. Four regression models were constructed, two adopted Generalized Estimating Equations (the two categorical outcome variables: number of events reported and patient safety grade) and the other two models followed a linear mixed regression model (the two composites for frequency of events reported and overall perception of safety).", "The study adopted a cross sectional research design utilizing a customized version of the HSOPSC. A pilot testing phase preceded data collection in order to ensure the validity and reliability of the Arabic translated version of the questionnaire. Of the 126 hospitals registered in the Lebanese Syndicate of Private Hospitals that were contacted and asked to participate, 68 consented [9]. The survey targeted hospital employees including physicians, nurses, clinical and non-clinical staff, pharmacy and laboratory staff, dietary and radiology staff, supervisors, and hospital managers. The questionnaire was distributed to 12,250 hospital employees and 6807 were returned yielding an overall response rate of 55.56%. Additional details on the study methodology can be found in El-Jardali et al. [9].", "The HSOPSC is composed of 42 items that measure 12 composites of patient safety culture (See Box1, Additional file1 ). Items were scored using a five-point scale reflecting the agreement rate on a five-point frequency scale. The percentage of positive responses for each item was calculated; negatively worded items were reversed when computing percent positive response. Composite level scores were computed by summation of the items within the composite scales and dividing by the number of items with non-missing values [9].\nTwo of the composites (frequency of events reported and overall perception of safety) are two of the four patient safety culture outcome variables [8]. The remaining two outcome variables are the patient safety grade and the number of events reported [8].", "Bivariate analyses were used to examine the associations between patient safety culture composites and differences across hospitals of different size and accreditation status.\nPearson correlations were used to examine the association between the patient safety culture composites. ANOVA f-test with multiple comparison corrected using the Bonferroni method was used to examine the association between the two outcome variables (patient safety grade and the number of events reported) with the remaining patient safety culture composites.\nStudent T-Test and ANOVA f-test with multiple comparison corrected using the Bonferroni method were then used to examine how trends in the outcome variables (frequency of events reported and overall perception of safety) differ across hospital and respondent characteristics. Finally, cross tables were constructed and chi-square tests were used to assess how trends in the outcome variables (patient safety grade and the number of events reported) differed across respondent and hospital characteristics.", "The four outcome variables were regressed against the 10 composite scores, respondent's position in the hospital, accreditation status, and hospital size. Since the data was clustered by hospital, we used appropriate statistical techniques to control for this effect. Four regression models were constructed, two adopted Generalized Estimating Equations (the two categorical outcome variables: number of events reported and patient safety grade) and the other two models followed a linear mixed regression model (the two composites for frequency of events reported and overall perception of safety).", "[SUBTITLE] Comparison of Means for the Frequency of Events Reported and the Overall Perception of Safety across Respondent and Hospital Characteristics [SUBSECTION] Significant differences were observed between units and positions when comparing results across respondent and hospital characteristics. For both the frequency of events reported and the overall perception of safety, significantly higher means were observed for diagnostics (mean = 3.97 ± 1.01; mean = 3.96 ± 0.68) as compared to surgical (mean = 3.78 ± 1.05; mean = 3.82 ± 0.67) and medical units (mean = 3.93 ± 0.99; mean = 3.83 ± 0.68) (See Table 1).\nComparison of means for two outcome composite scores across hospital and respondent characteristics (identical letters represent significance between indicated groups)\nSignificantly higher means were observed for nurses (mean = 3.89 ± 1.00; mean = 3.80 ± 0.66) than physicians (mean = 3.78 ± 0.92; mean = 3.69 ± 0.75) on the frequency of events reported and the overall perception of safety respectively (See Table 1).\nFor frequency of events reported, higher means (and thus more events) were observed for increasing years of experience whereas the opposite trend prevailed for overall perception of safety (See Table 1). Small hospitals (mean = 3.95 ± 1.00) had a significantly higher frequency of events reported in comparison to medium (mean = 3.87 ± 1.03) and large hospitals (mean = 3.81 ± 1.08). Small hospitals (mean = 3.84 ± 0.67) also had a better overall perception of safety than large hospitals (mean = 3.76 ± 0.71). Accredited hospitals (mean = 3.91 ± 1.03) had a significantly higher frequency of events reported than non-accredited hospitals (mean = 3.84 ± 1.03). Accredited hospitals (mean = 3.84 ± 0.69) also had a better overall perception of safety than non-accredited hospitals (mean = 3.71 ± 0.68) (See Table 1).\nSignificant differences were observed between units and positions when comparing results across respondent and hospital characteristics. For both the frequency of events reported and the overall perception of safety, significantly higher means were observed for diagnostics (mean = 3.97 ± 1.01; mean = 3.96 ± 0.68) as compared to surgical (mean = 3.78 ± 1.05; mean = 3.82 ± 0.67) and medical units (mean = 3.93 ± 0.99; mean = 3.83 ± 0.68) (See Table 1).\nComparison of means for two outcome composite scores across hospital and respondent characteristics (identical letters represent significance between indicated groups)\nSignificantly higher means were observed for nurses (mean = 3.89 ± 1.00; mean = 3.80 ± 0.66) than physicians (mean = 3.78 ± 0.92; mean = 3.69 ± 0.75) on the frequency of events reported and the overall perception of safety respectively (See Table 1).\nFor frequency of events reported, higher means (and thus more events) were observed for increasing years of experience whereas the opposite trend prevailed for overall perception of safety (See Table 1). Small hospitals (mean = 3.95 ± 1.00) had a significantly higher frequency of events reported in comparison to medium (mean = 3.87 ± 1.03) and large hospitals (mean = 3.81 ± 1.08). Small hospitals (mean = 3.84 ± 0.67) also had a better overall perception of safety than large hospitals (mean = 3.76 ± 0.71). Accredited hospitals (mean = 3.91 ± 1.03) had a significantly higher frequency of events reported than non-accredited hospitals (mean = 3.84 ± 1.03). Accredited hospitals (mean = 3.84 ± 0.69) also had a better overall perception of safety than non-accredited hospitals (mean = 3.71 ± 0.68) (See Table 1).\n[SUBTITLE] Correlations between Patient Safety Culture Composites [SUBSECTION] While significant correlations were observed among all patient safety culture composites, differences in the strength of the correlation were observed (See Table 2). The composite measuring staffing showed the weakest correlation with the outcome on frequency of event reporting (Pearson r = 0.107). The composite measuring feedback and communication about errors had the strongest correlation with this outcome variable (Pearson r = 0.378) (See Table 2).\nCorrelation between patient safety culture composites\n*Correlation is significant at the 0.01 level (2-tailed).\nThe composite with the weakest correlation with the outcome variable measuring perceptions of patient safety was non-punitive response to error (Pearson r = 0.179) while that with the strongest correlation was the composite measuring supervisor/manager expectations and actions promoting safety (Pearson r = 0.371) (See Table 2).\nWhile significant correlations were observed among all patient safety culture composites, differences in the strength of the correlation were observed (See Table 2). The composite measuring staffing showed the weakest correlation with the outcome on frequency of event reporting (Pearson r = 0.107). The composite measuring feedback and communication about errors had the strongest correlation with this outcome variable (Pearson r = 0.378) (See Table 2).\nCorrelation between patient safety culture composites\n*Correlation is significant at the 0.01 level (2-tailed).\nThe composite with the weakest correlation with the outcome variable measuring perceptions of patient safety was non-punitive response to error (Pearson r = 0.179) while that with the strongest correlation was the composite measuring supervisor/manager expectations and actions promoting safety (Pearson r = 0.371) (See Table 2).\n[SUBTITLE] Comparison of Means between Outcome Variables and Patient Safety Composites [SUBSECTION] Significantly different means were reported for patient safety grade with all patient safety composite scores. Respondents who gave \"Excellent/Very Good\" patient safety grades had the highest mean scores for patient safety composites (See Table 3). The number of events reported was significantly associated with the composites measuring communication openness, feedback and communication about errors, non-punitive response to error, hospital handoffs and transitions, and teamwork across hospital units. The highest means were observed when reporting more than 5 events for the composites on communication openness, feedback and communication about errors, and non-punitive response to error. The opposite was observed for the composites measuring hospital handoffs and transitions and teamwork across hospital units where the highest means were observed when no events were reported (see Table 3).\nComparison of means between patient safety grade and number of events reported with patient safety culture composite scores\nPatient Safety Grade\na. Significant difference between \"Poor or Failing\" and \"Acceptable\"\nb. Significant difference between \"Poor or Failing\" and \"Excellent/Very Good\"\nc. Significant difference between \"Acceptable\" and \"Excellent/Very Good\"\nNumber of Events Reported\na. Significant difference between \"No events reported\" and \"1 to 5 events reported\"\nb. Significant difference between \"No events reported\" and \"> 5 events reported\"\nc. Significant difference between \"1 to 5 events reported\" and \"> 5 events reported\"\nSignificantly different means were reported for patient safety grade with all patient safety composite scores. Respondents who gave \"Excellent/Very Good\" patient safety grades had the highest mean scores for patient safety composites (See Table 3). The number of events reported was significantly associated with the composites measuring communication openness, feedback and communication about errors, non-punitive response to error, hospital handoffs and transitions, and teamwork across hospital units. The highest means were observed when reporting more than 5 events for the composites on communication openness, feedback and communication about errors, and non-punitive response to error. The opposite was observed for the composites measuring hospital handoffs and transitions and teamwork across hospital units where the highest means were observed when no events were reported (see Table 3).\nComparison of means between patient safety grade and number of events reported with patient safety culture composite scores\nPatient Safety Grade\na. Significant difference between \"Poor or Failing\" and \"Acceptable\"\nb. Significant difference between \"Poor or Failing\" and \"Excellent/Very Good\"\nc. Significant difference between \"Acceptable\" and \"Excellent/Very Good\"\nNumber of Events Reported\na. Significant difference between \"No events reported\" and \"1 to 5 events reported\"\nb. Significant difference between \"No events reported\" and \"> 5 events reported\"\nc. Significant difference between \"1 to 5 events reported\" and \"> 5 events reported\"\n[SUBTITLE] Distribution of the Patient Safety Grade and the Number of Events Reported across Respondent and Hospital Characteristics [SUBSECTION] [SUBTITLE] Patient Safety Grade [SUBSECTION] The highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\nThe highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\n[SUBTITLE] Number of Events Reported [SUBSECTION] Respondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\nRespondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\n[SUBTITLE] Patient Safety Grade [SUBSECTION] The highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\nThe highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\n[SUBTITLE] Number of Events Reported [SUBSECTION] Respondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\nRespondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\n[SUBTITLE] Generalized Estimating Equations for the Patient Safety Composite Scores and Respondent and Hospital Characteristics against the Patient Safety Grade and the Number of Events Reported (See Table 5) [SUBSECTION] Generalized Estimating Equations\n[SUBTITLE] Patient Safety Grade [SUBSECTION] Results showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\nResults showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\n[SUBTITLE] Number of Events Reported [SUBSECTION] A one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\nA one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\nGeneralized Estimating Equations\n[SUBTITLE] Patient Safety Grade [SUBSECTION] Results showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\nResults showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\n[SUBTITLE] Number of Events Reported [SUBSECTION] A one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\nA one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\n[SUBTITLE] Linear Mixed Regression for the Patient Safety Composite Scores and Respondent and Hospital Characteristics against the Frequency of Events Reported and the Overall Perception of Safety (See Table 6) [SUBSECTION] [SUBTITLE] Frequency of Events Reported [SUBSECTION] Mixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\nMixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\n[SUBTITLE] Overall Perception of Safety [SUBSECTION] Perception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).\nPerception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).\n[SUBTITLE] Frequency of Events Reported [SUBSECTION] Mixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\nMixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\n[SUBTITLE] Overall Perception of Safety [SUBSECTION] Perception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).\nPerception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).", "Significant differences were observed between units and positions when comparing results across respondent and hospital characteristics. For both the frequency of events reported and the overall perception of safety, significantly higher means were observed for diagnostics (mean = 3.97 ± 1.01; mean = 3.96 ± 0.68) as compared to surgical (mean = 3.78 ± 1.05; mean = 3.82 ± 0.67) and medical units (mean = 3.93 ± 0.99; mean = 3.83 ± 0.68) (See Table 1).\nComparison of means for two outcome composite scores across hospital and respondent characteristics (identical letters represent significance between indicated groups)\nSignificantly higher means were observed for nurses (mean = 3.89 ± 1.00; mean = 3.80 ± 0.66) than physicians (mean = 3.78 ± 0.92; mean = 3.69 ± 0.75) on the frequency of events reported and the overall perception of safety respectively (See Table 1).\nFor frequency of events reported, higher means (and thus more events) were observed for increasing years of experience whereas the opposite trend prevailed for overall perception of safety (See Table 1). Small hospitals (mean = 3.95 ± 1.00) had a significantly higher frequency of events reported in comparison to medium (mean = 3.87 ± 1.03) and large hospitals (mean = 3.81 ± 1.08). Small hospitals (mean = 3.84 ± 0.67) also had a better overall perception of safety than large hospitals (mean = 3.76 ± 0.71). Accredited hospitals (mean = 3.91 ± 1.03) had a significantly higher frequency of events reported than non-accredited hospitals (mean = 3.84 ± 1.03). Accredited hospitals (mean = 3.84 ± 0.69) also had a better overall perception of safety than non-accredited hospitals (mean = 3.71 ± 0.68) (See Table 1).", "While significant correlations were observed among all patient safety culture composites, differences in the strength of the correlation were observed (See Table 2). The composite measuring staffing showed the weakest correlation with the outcome on frequency of event reporting (Pearson r = 0.107). The composite measuring feedback and communication about errors had the strongest correlation with this outcome variable (Pearson r = 0.378) (See Table 2).\nCorrelation between patient safety culture composites\n*Correlation is significant at the 0.01 level (2-tailed).\nThe composite with the weakest correlation with the outcome variable measuring perceptions of patient safety was non-punitive response to error (Pearson r = 0.179) while that with the strongest correlation was the composite measuring supervisor/manager expectations and actions promoting safety (Pearson r = 0.371) (See Table 2).", "Significantly different means were reported for patient safety grade with all patient safety composite scores. Respondents who gave \"Excellent/Very Good\" patient safety grades had the highest mean scores for patient safety composites (See Table 3). The number of events reported was significantly associated with the composites measuring communication openness, feedback and communication about errors, non-punitive response to error, hospital handoffs and transitions, and teamwork across hospital units. The highest means were observed when reporting more than 5 events for the composites on communication openness, feedback and communication about errors, and non-punitive response to error. The opposite was observed for the composites measuring hospital handoffs and transitions and teamwork across hospital units where the highest means were observed when no events were reported (see Table 3).\nComparison of means between patient safety grade and number of events reported with patient safety culture composite scores\nPatient Safety Grade\na. Significant difference between \"Poor or Failing\" and \"Acceptable\"\nb. Significant difference between \"Poor or Failing\" and \"Excellent/Very Good\"\nc. Significant difference between \"Acceptable\" and \"Excellent/Very Good\"\nNumber of Events Reported\na. Significant difference between \"No events reported\" and \"1 to 5 events reported\"\nb. Significant difference between \"No events reported\" and \"> 5 events reported\"\nc. Significant difference between \"1 to 5 events reported\" and \"> 5 events reported\"", "[SUBTITLE] Patient Safety Grade [SUBSECTION] The highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\nThe highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).\n[SUBTITLE] Number of Events Reported [SUBSECTION] Respondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).\nRespondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).", "The highest percentage of respondents reporting \"Excellent/Very Good\" patient safety grades were those working in diagnostics (83.6%). The highest percentage of respondents reporting a \"Poor/Failing\" patient safety grade (4.2%) worked across many different hospital units. Although the majority of respondents gave their units an \"Excellent/Very Good\" patient safety grade, physicians (63.3%) and administrators (63.3%) were the least to report an \"Excellent/Very Good\" patient safety grade (See Table 4).\nDistribution of two outcome variables across hospital and respondent characteristics\nThe association between years of experience and patient safety grade is similar to an inverted J-shaped association. Respondents with less than 1 year of experience were less likely to give \"Excellent/Very Good\" patient safety grade but this increased as years of experience gradually increased. However, after experience exceeded 16 years, the percentage of respondents giving \"Excellent/Very Good\" patient safety grade started to decrease gradually (See Table 4).\nAs detailed in El-Jardali et al. (2010), respondents working at accredited hospitals were more likely to give their units an \"Excellent/Very Good\" patient safety grade than respondents working at non-accredited hospitals (74.2% vs. 68.8%) (See Table 4).", "Respondents working in administrative units were most likely to report more than 5 events over the past year (12.9%) while respondents working in surgical units were the least likely to report more than 5 events over the past year (5.7%) (See Table 4). Quality and safety officers had the highest percent reporting of more than 5 events (29.6%) followed by pharmacists (26.2%). It is worth noting that 57.1% of physicians and 57.2% of nurses reported no events over the past year (See Table 4).\nRespondents with less than 1 year of experience were the largest group of respondents to report no events over the past 12 months (68.7%). This percentage dropped as years of experience increased and then rose again after experience exceeded 10 years (See Table 4).\nInteraction with patients was significantly associated with number of events reported where respondents having no direct contact with patients were more likely to report more than 5 events over the past 12 months (8.1%).\nRespondents working at accredited hospitals were more likely to report more than 5 events (7.7%) over the past 12 months as compared to respondents working at non-accredited hospitals (5.9%) (See Table 4).", "Generalized Estimating Equations\n[SUBTITLE] Patient Safety Grade [SUBSECTION] Results showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\nResults showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.\n[SUBTITLE] Number of Events Reported [SUBSECTION] A one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).\nA one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).", "Results showed that a one unit increase in the composite score for supervisor/manager expectations and actions to promote patient safety increased the odds of reporting a better patient safety grade by 1.23 (95%CI = 1.03-1.47; P-Value = 0.024). Similarly, the odds of reporting a higher patient safety grade increased by 1.30 (95% CI = 1.03-1.65; P-Value = 0.029) for a one unit increase in the composite score for organizational learning and continuous improvement. Moreover, a one unit increase in the composite score on teamwork within hospital units increased the odds of reporting a higher patient safety grade by 1.64 (95%CI = 1.40-1.93; P-Value < 0.001). The odds of reporting a higher patient safety grade increased by 1.54 (95% CI = 1.33-1.78; P-Value < 0.001) for a one unit increase in the composite score for feedback and communication about errors. The odds of reporting a higher patient safety grade increased by 1.31 (95%CI = 1.16-1.48; P-Value < 0.001) for a one unit increase in the composite score for staffing. A one unit increase in the composites score for hospital management support for patient safety and hospital handoffs and transitions increased the odds of reporting a higher patient safety grade by 1.85 (95% CI = 1.53-2.28; P-Value < 0.001) and 1.36 (95%CI = 1.15-1. 61; P-Value < 0.001) respectively (See Table 5).\nRespondents who held positions as unit assistants and clerks had 1.79 higher odds (95%CI = 1.16-2.74; P-Value = 0.008) of reporting a higher patient safety grade as compared to the nurses while those working in the administration had 0.39 lower odds (95%CI = 0.27-0.57; P-Value < 0.001) of reporting a higher patient safety grade. Respondents working in \"other\" hospital departments had 1.66 higher odds (95% CI = 1.12-2.48, P-Value = 0.012) of reporting a higher patient safety grade.", "A one unit increase in the composite score for feedback and communication about errors increased the odds of reporting a higher number of events by 1.17 (95%CI = 1.03-1.32; P-Value = 0.013). Similarly, the odds of reporting a higher number of events decreased by 0.79 (95% CI = 0.68-0.91; P-Value = 0.003) for a one unit increase in the composite score for hospital handoffs and transition (See Table 5).\nRespondents who reported working in administration had 3.57 higher odds (95%CI = 1.99-6.40; P-Value < 0.001) of reporting a higher number of events compared to nurses. Moreover, respondents working in Quality and Safety had 3.14 higher odds (95%CI = 1.24-7.91) of reporting a higher number of events compared to nurses (See Table 5).", "[SUBTITLE] Frequency of Events Reported [SUBSECTION] Mixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\nMixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).\n[SUBTITLE] Overall Perception of Safety [SUBSECTION] Perception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).\nPerception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).", "Mixed linear regression analysis showed that a one unit increase in the score on organizational learning and continuous improvement increased the frequency of events reported by 0.159 (P-Value < 0.001). An increase of 0.044 (P-Value = 0.05) in the composite measuring communication and openness was observed for one unit increase in frequency of events reported. A one unit increase in the score on feedback and communication about errors increased the frequency of events reported by 0.366 (P-Value < 0.001). An increase of 0.040 (P-Value = 0.06) in the frequency of events reported was observed for a one unit increase in the score on non-punitive response to errors. Furthermore, an increase of 0.050 (P-Value = 0.05) in the frequency of events reported was observed for a one unit increase in the score on hospital management support for patient safety. A one unit increase in the score on teamwork across hospital units was found to increase the frequency of events reported by 0.100 (P-Value = 0.001) (See Table 6).\nLinear Mixed Model Regression\nRespondents working in the quality and safety departments had lower frequency of events reported (beta = -0.364, P-Value = 0.020) as compared to nurses (See Table 6).", "Perception of patient safety improved by 0.094 (P-Value < 0.001) for a one unit increase in the score on supervisor/manager expectations and actions promoting safety, by 0.129 (P-Value < 0.001) a one unit increase in the score on organizational learning and continuous improvement, and by 0.111 (P-Value < 0.001) for a one unit increase in the score on teamwork within hospital units. Similarly, perception of patient safety also improved by 0.031 for a one unit increase in the score on non-punitive response to error (P-Value = 0.01), 0.120 for staffing (p-Value < 0.001), 0.099 for hospital management support for patient safety (p-Value < 0.001) and 0.163 for hospital handoffs and transitions (p-Value < 0.001). However, perception of patient safety decreased by -0.045 (P-Value < 0.001) for a one unit increase in the score on teamwork across hospital units (See Table 6).\nRespondents working as unit assistant/clerk/secretary/technicians had 0.054 (P-Value = 0.08) better perception of patient safety as compared to nurses (See Table 6).\nAccredited hospitals were found to have better perception of patient safety as compared to non-accredited hospitals (beta = 0.113, p-value = 0.01) (See Table 6).", "The HSOPSC is one of the most common tools being used to assess the culture of safety in hospitals. Studies that utilize this tool usually report the 12 composite scores and the scores on the patient safety grade and the number of events reported. However, exploring the association between the patient safety composite scores and the hospital and respondent characteristics with the patient safety culture outcomes are not common in the literature. To our knowledge, this is one of the few studies that explore such an association. The analysis of results identified that patient safety culture predictors such as event reporting, proper communication, patient safety leadership and management, hospital size, and accreditation status are associated with the patient safety culture outcomes.\n[SUBTITLE] Event Reporting [SUBSECTION] A safety culture includes three major components a just culture, a reporting culture, and a learning culture [11]. Event reporting, an essential component for achieving a learning culture, can only happen in a non-punitive environment where events can be reported without people being blamed [12]. The non-punitive response to error composite received one of the lowest scores revealing that Lebanese hospital employees are not at ease when it comes to reporting errors.\nThe majority of physicians and nurses in our sample reported no events over the past 12 months. Training opportunities that empower physicians to improve patient safety are limited thus investing in the training of physicians is important because they can play a major role in improving patient safety initiatives primarily by improving patient safety outcomes and reducing hassles and wasted time [13].\nAccording to Leape et al. (1995), errors will be more frequent if nurses did not intercept 86% of all potential errors [14]. However, the majority of nurses in our sample reported no events over the past 12 months. Many errors in health care go unreported for many reasons including fear, humiliation, the presence of a punitive response to error, and the fact that reporting will not usually result in actual change [15]. Encouraging health professionals, specifically nurses, to report events in a non-punitive environment is crucial for improving patient safety.\nIn our study, the fewest number of respondents to report more than 5 events over the past 12 months worked in surgical units. Errors in operating rooms are not uncommon and can sometimes be catastrophic [16]. Creating a patient safety culture in surgical units by improving communication and reporting more events is a high priority for operating room staff and hospitals [16].\nEmployees who do not deal directly with the patient are more at ease when it comes to reporting errors. As mentioned in Jones et al. (2008), work in laboratory units is considered as more organized than other units since it is controlled by more professional standards and because errors investigated in these units are done as a group [17]. On the contrary, when an error is performed by a nurse, the nurse is investigated as an individual rather a member of a medical team [17].\nWork experience at the hospital also had some impact on the frequency and number of events reported. Frequency of events reported was found to increase with increasing years of experience. On the other hand, the score for perception of patient safety decreased as experience in the hospital increased. The perception of safety is defined as the extent to which procedures and systems are good at preventing errors and the lack of patient safety problems. As people become more experienced, they become more aware of the safety practices undertaken in the institutions they work in. When the perception of safety decreases with the increase in the years of experience, it means that the staff members do not agree that the patient safety practices, systems, and procedures in the hospitals act as barriers to errors and problems. A study by Bodur and Feliz (2009) showed that patient safety culture scores decreased as seniority increased [18]. The observation may be the result of an increase in medical errors done by the respondent due to frustration with hospital regulations or increasing staff awareness of safety problems and thus additional reporting.\nA safety culture includes three major components a just culture, a reporting culture, and a learning culture [11]. Event reporting, an essential component for achieving a learning culture, can only happen in a non-punitive environment where events can be reported without people being blamed [12]. The non-punitive response to error composite received one of the lowest scores revealing that Lebanese hospital employees are not at ease when it comes to reporting errors.\nThe majority of physicians and nurses in our sample reported no events over the past 12 months. Training opportunities that empower physicians to improve patient safety are limited thus investing in the training of physicians is important because they can play a major role in improving patient safety initiatives primarily by improving patient safety outcomes and reducing hassles and wasted time [13].\nAccording to Leape et al. (1995), errors will be more frequent if nurses did not intercept 86% of all potential errors [14]. However, the majority of nurses in our sample reported no events over the past 12 months. Many errors in health care go unreported for many reasons including fear, humiliation, the presence of a punitive response to error, and the fact that reporting will not usually result in actual change [15]. Encouraging health professionals, specifically nurses, to report events in a non-punitive environment is crucial for improving patient safety.\nIn our study, the fewest number of respondents to report more than 5 events over the past 12 months worked in surgical units. Errors in operating rooms are not uncommon and can sometimes be catastrophic [16]. Creating a patient safety culture in surgical units by improving communication and reporting more events is a high priority for operating room staff and hospitals [16].\nEmployees who do not deal directly with the patient are more at ease when it comes to reporting errors. As mentioned in Jones et al. (2008), work in laboratory units is considered as more organized than other units since it is controlled by more professional standards and because errors investigated in these units are done as a group [17]. On the contrary, when an error is performed by a nurse, the nurse is investigated as an individual rather a member of a medical team [17].\nWork experience at the hospital also had some impact on the frequency and number of events reported. Frequency of events reported was found to increase with increasing years of experience. On the other hand, the score for perception of patient safety decreased as experience in the hospital increased. The perception of safety is defined as the extent to which procedures and systems are good at preventing errors and the lack of patient safety problems. As people become more experienced, they become more aware of the safety practices undertaken in the institutions they work in. When the perception of safety decreases with the increase in the years of experience, it means that the staff members do not agree that the patient safety practices, systems, and procedures in the hospitals act as barriers to errors and problems. A study by Bodur and Feliz (2009) showed that patient safety culture scores decreased as seniority increased [18]. The observation may be the result of an increase in medical errors done by the respondent due to frustration with hospital regulations or increasing staff awareness of safety problems and thus additional reporting.\n[SUBTITLE] Communication [SUBSECTION] Proper communication within and across healthcare teams is essential to remove any threats to safety of patients. Communication problems have been identified as major contributing factors to adverse events [7]. An analysis of 2,455 sentinel events reported to the Joint Commission on the Accreditation of Healthcare Organizations showed that 70% of the cases were a result of failure in communication [19]. In the absence of proper communication between the different hospital units, patient safety might be jeopardized. Our results revealed that higher scores on teamwork across hospital units increase the frequency of events reported. Moreover, higher scores on hospital handoffs and transitions increased the likelihood of having a better perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\nProper communication within and across healthcare teams is essential to remove any threats to safety of patients. Communication problems have been identified as major contributing factors to adverse events [7]. An analysis of 2,455 sentinel events reported to the Joint Commission on the Accreditation of Healthcare Organizations showed that 70% of the cases were a result of failure in communication [19]. In the absence of proper communication between the different hospital units, patient safety might be jeopardized. Our results revealed that higher scores on teamwork across hospital units increase the frequency of events reported. Moreover, higher scores on hospital handoffs and transitions increased the likelihood of having a better perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\n[SUBTITLE] Patient safety leadership and management [SUBSECTION] In order for a patient safety program to be successful, strong leadership is needed. Senior leaders are the only ones who are able to create the culture and commitment needed to solve underlying system causes of medical errors and harm to patients. When leadership and management is committed to a culture of safety, the whole organization will follow and thus disclosing adverse events and finding their root causes will become an organizational process. The focus should be in emergency rooms, operating rooms, and intensive care units [20]. Our results showed that more support from hospital management for patient safety increased the frequency of events reported. It also increased the likelihood of having a better overall perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\nIn order for a patient safety program to be successful, strong leadership is needed. Senior leaders are the only ones who are able to create the culture and commitment needed to solve underlying system causes of medical errors and harm to patients. When leadership and management is committed to a culture of safety, the whole organization will follow and thus disclosing adverse events and finding their root causes will become an organizational process. The focus should be in emergency rooms, operating rooms, and intensive care units [20]. Our results showed that more support from hospital management for patient safety increased the frequency of events reported. It also increased the likelihood of having a better overall perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\n[SUBTITLE] Staffing [SUBSECTION] Staffing, a major predictor of patient safety, was one of the composites that received a very low score. Having a strong, capable, and motivated workforce is one of the biggest challenges for hospitals today [21]. Evidence has shown that a strong link exists between the availability of health care providers and population health outcomes [21]. According to Sandars & Cook (2007), major catastrophes have occurred in organizations with insufficient staffing. Medical personnel in under-staffed hospitals are overworked [7]. They also suffer from stress and sleeplessness which might lead to lapses in performance thus leading to errors affecting quality and performance [22]. Our results showed that a more positive score on staffing increased the likelihood of having a more positive perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\nStaffing, a major predictor of patient safety, was one of the composites that received a very low score. Having a strong, capable, and motivated workforce is one of the biggest challenges for hospitals today [21]. Evidence has shown that a strong link exists between the availability of health care providers and population health outcomes [21]. According to Sandars & Cook (2007), major catastrophes have occurred in organizations with insufficient staffing. Medical personnel in under-staffed hospitals are overworked [7]. They also suffer from stress and sleeplessness which might lead to lapses in performance thus leading to errors affecting quality and performance [22]. Our results showed that a more positive score on staffing increased the likelihood of having a more positive perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.\n[SUBTITLE] Hospital size and accreditation [SUBSECTION] Hospital size and accreditation status were also factors affecting the culture of safety in hospitals. Small hospitals scored higher than larger hospitals on both the frequency of events reported and on the overall perception of safety. Large hospitals usually face challenges when it comes to implementing quality work especially because of bureaucracy. On the contrary, small hospitals have a more homogenous culture where staff members are more likely to share the same values [23].\nAccredited hospitals also scored higher on both frequency of events reported and on the overall perception of safety in comparison to non-accredited hospitals. Respondents working at accredited hospitals had an increased likelihood of having a more positive perception of safety. Since the development of hospital accreditation programs that made accreditation a requirement for financial reimbursement by the MOPH, quality improvement initiatives have gained increased attention in Lebanese hospitals [24]. A study by El-Jardali et al. (2008) showed that Lebanese nurses perceived that accreditation created an improvement in the quality of health care [25].\nHospital size and accreditation status were also factors affecting the culture of safety in hospitals. Small hospitals scored higher than larger hospitals on both the frequency of events reported and on the overall perception of safety. Large hospitals usually face challenges when it comes to implementing quality work especially because of bureaucracy. On the contrary, small hospitals have a more homogenous culture where staff members are more likely to share the same values [23].\nAccredited hospitals also scored higher on both frequency of events reported and on the overall perception of safety in comparison to non-accredited hospitals. Respondents working at accredited hospitals had an increased likelihood of having a more positive perception of safety. Since the development of hospital accreditation programs that made accreditation a requirement for financial reimbursement by the MOPH, quality improvement initiatives have gained increased attention in Lebanese hospitals [24]. A study by El-Jardali et al. (2008) showed that Lebanese nurses perceived that accreditation created an improvement in the quality of health care [25].\n[SUBTITLE] Limitations [SUBSECTION] A methodological limitation pertaining to this study should be acknowledged. In the literature, it is argued that social and behavioral research, particularly those that include self-reports such as surveys, is prone to Common method variance (CMV). CMV is believed to pose a threat to the validity of the data as it may either inflate or deflate the correlations among research variables. Although this issue can be attended to during survey development, evidence in the literature stipulate that although such research methods are more prone to CMV, they should not be considered weak or inappropriate if the researchers follow rigorous research conventions in research design, data collection and analysis [26]. In our study, CMV was avoided by the following [26]:\n- Assuring that the composite scores were derived based on a combination of items;\n- Counterbalancing the order of the questions;\n- Ensuring the confidentiality and anonymity of respondents;\n- Using clearly written scale items to make responses less subject to bias; and\n- Informing respondents in the consent form that there is no preferred or correct answer, and that their individual responses would not be viewed by their management and survey completion would not affect their status at their hospital so they would be encouraged to honestly assess and respond freely\nA methodological limitation pertaining to this study should be acknowledged. In the literature, it is argued that social and behavioral research, particularly those that include self-reports such as surveys, is prone to Common method variance (CMV). CMV is believed to pose a threat to the validity of the data as it may either inflate or deflate the correlations among research variables. Although this issue can be attended to during survey development, evidence in the literature stipulate that although such research methods are more prone to CMV, they should not be considered weak or inappropriate if the researchers follow rigorous research conventions in research design, data collection and analysis [26]. In our study, CMV was avoided by the following [26]:\n- Assuring that the composite scores were derived based on a combination of items;\n- Counterbalancing the order of the questions;\n- Ensuring the confidentiality and anonymity of respondents;\n- Using clearly written scale items to make responses less subject to bias; and\n- Informing respondents in the consent form that there is no preferred or correct answer, and that their individual responses would not be viewed by their management and survey completion would not affect their status at their hospital so they would be encouraged to honestly assess and respond freely", "A safety culture includes three major components a just culture, a reporting culture, and a learning culture [11]. Event reporting, an essential component for achieving a learning culture, can only happen in a non-punitive environment where events can be reported without people being blamed [12]. The non-punitive response to error composite received one of the lowest scores revealing that Lebanese hospital employees are not at ease when it comes to reporting errors.\nThe majority of physicians and nurses in our sample reported no events over the past 12 months. Training opportunities that empower physicians to improve patient safety are limited thus investing in the training of physicians is important because they can play a major role in improving patient safety initiatives primarily by improving patient safety outcomes and reducing hassles and wasted time [13].\nAccording to Leape et al. (1995), errors will be more frequent if nurses did not intercept 86% of all potential errors [14]. However, the majority of nurses in our sample reported no events over the past 12 months. Many errors in health care go unreported for many reasons including fear, humiliation, the presence of a punitive response to error, and the fact that reporting will not usually result in actual change [15]. Encouraging health professionals, specifically nurses, to report events in a non-punitive environment is crucial for improving patient safety.\nIn our study, the fewest number of respondents to report more than 5 events over the past 12 months worked in surgical units. Errors in operating rooms are not uncommon and can sometimes be catastrophic [16]. Creating a patient safety culture in surgical units by improving communication and reporting more events is a high priority for operating room staff and hospitals [16].\nEmployees who do not deal directly with the patient are more at ease when it comes to reporting errors. As mentioned in Jones et al. (2008), work in laboratory units is considered as more organized than other units since it is controlled by more professional standards and because errors investigated in these units are done as a group [17]. On the contrary, when an error is performed by a nurse, the nurse is investigated as an individual rather a member of a medical team [17].\nWork experience at the hospital also had some impact on the frequency and number of events reported. Frequency of events reported was found to increase with increasing years of experience. On the other hand, the score for perception of patient safety decreased as experience in the hospital increased. The perception of safety is defined as the extent to which procedures and systems are good at preventing errors and the lack of patient safety problems. As people become more experienced, they become more aware of the safety practices undertaken in the institutions they work in. When the perception of safety decreases with the increase in the years of experience, it means that the staff members do not agree that the patient safety practices, systems, and procedures in the hospitals act as barriers to errors and problems. A study by Bodur and Feliz (2009) showed that patient safety culture scores decreased as seniority increased [18]. The observation may be the result of an increase in medical errors done by the respondent due to frustration with hospital regulations or increasing staff awareness of safety problems and thus additional reporting.", "Proper communication within and across healthcare teams is essential to remove any threats to safety of patients. Communication problems have been identified as major contributing factors to adverse events [7]. An analysis of 2,455 sentinel events reported to the Joint Commission on the Accreditation of Healthcare Organizations showed that 70% of the cases were a result of failure in communication [19]. In the absence of proper communication between the different hospital units, patient safety might be jeopardized. Our results revealed that higher scores on teamwork across hospital units increase the frequency of events reported. Moreover, higher scores on hospital handoffs and transitions increased the likelihood of having a better perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.", "In order for a patient safety program to be successful, strong leadership is needed. Senior leaders are the only ones who are able to create the culture and commitment needed to solve underlying system causes of medical errors and harm to patients. When leadership and management is committed to a culture of safety, the whole organization will follow and thus disclosing adverse events and finding their root causes will become an organizational process. The focus should be in emergency rooms, operating rooms, and intensive care units [20]. Our results showed that more support from hospital management for patient safety increased the frequency of events reported. It also increased the likelihood of having a better overall perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.", "Staffing, a major predictor of patient safety, was one of the composites that received a very low score. Having a strong, capable, and motivated workforce is one of the biggest challenges for hospitals today [21]. Evidence has shown that a strong link exists between the availability of health care providers and population health outcomes [21]. According to Sandars & Cook (2007), major catastrophes have occurred in organizations with insufficient staffing. Medical personnel in under-staffed hospitals are overworked [7]. They also suffer from stress and sleeplessness which might lead to lapses in performance thus leading to errors affecting quality and performance [22]. Our results showed that a more positive score on staffing increased the likelihood of having a more positive perception of safety among respondents and the likelihood of respondents to report a higher patient safety grade.", "Hospital size and accreditation status were also factors affecting the culture of safety in hospitals. Small hospitals scored higher than larger hospitals on both the frequency of events reported and on the overall perception of safety. Large hospitals usually face challenges when it comes to implementing quality work especially because of bureaucracy. On the contrary, small hospitals have a more homogenous culture where staff members are more likely to share the same values [23].\nAccredited hospitals also scored higher on both frequency of events reported and on the overall perception of safety in comparison to non-accredited hospitals. Respondents working at accredited hospitals had an increased likelihood of having a more positive perception of safety. Since the development of hospital accreditation programs that made accreditation a requirement for financial reimbursement by the MOPH, quality improvement initiatives have gained increased attention in Lebanese hospitals [24]. A study by El-Jardali et al. (2008) showed that Lebanese nurses perceived that accreditation created an improvement in the quality of health care [25].", "A methodological limitation pertaining to this study should be acknowledged. In the literature, it is argued that social and behavioral research, particularly those that include self-reports such as surveys, is prone to Common method variance (CMV). CMV is believed to pose a threat to the validity of the data as it may either inflate or deflate the correlations among research variables. Although this issue can be attended to during survey development, evidence in the literature stipulate that although such research methods are more prone to CMV, they should not be considered weak or inappropriate if the researchers follow rigorous research conventions in research design, data collection and analysis [26]. In our study, CMV was avoided by the following [26]:\n- Assuring that the composite scores were derived based on a combination of items;\n- Counterbalancing the order of the questions;\n- Ensuring the confidentiality and anonymity of respondents;\n- Using clearly written scale items to make responses less subject to bias; and\n- Informing respondents in the consent form that there is no preferred or correct answer, and that their individual responses would not be viewed by their management and survey completion would not affect their status at their hospital so they would be encouraged to honestly assess and respond freely", "Investing in patient safety practices in Lebanese hospitals is crucial. While patient safety is everyone's concern, it is not easy for everyone who works in health care organizations to understand this concept. In Lebanon, most health workers have had different training, and often hold a value system that is specific to their professional group. To be truly effective, patient safety needs to be incorporated into the education of health professionals across the spectrum of health care.\nOur results demonstrate that patient safety should be a top strategic priority for the health care organizations and its leaders. There should be blame-free system for identifying threats to patient safety, sharing information and learning from events. In addition, there should be a collaborative environment so that all health workers in the healthcare organization can share and exchange information about patient safety\nTo facilitate change in cultural behaviors, hospital management should assess and redesign their current patient safety system including governance and reporting structures. In addition, they should provide their health professionals with comprehensive training on patient safety concepts, tools, and implementations.\nOur results also show that the move to prioritize patient safety by healthcare systems through accreditation is important. However, there is still a good deal skepticism on the part of health professionals who feel that by sharing highly sensitive information they may still be subject to blame and penalties. Progress in this respect will need far more enlightened policies, where health professionals are actively encouraged to report errors for the purpose of learning and improvement. In Lebanon, it is important to strengthen the new accreditation chapter on patient safety by supporting hospitals in training their staff, especially the less experienced ones, on patient safety competencies and about effective implementation of the new standards. Further research is needed to study the association between patient safety culture and clinical outcomes.\nIt is hoped that the outcome of this study will help conduct a policy dialogue meeting for policy makers and stakeholders in Lebanon to discuss the findings and make deliberation about potential next steps. Senior policy makers, managers and leaders are the only ones who are able to create the culture and commitment needed to identify and resolve underlying systemic causes related to patient safety.", "IOM: Institute of Medicine; HSOPSC: Hospital Survey on Patient Safety Culture; MOPH: Ministry of Public Health;", "The authors declare that they have no competing interests.", "FE was the principal investigator on this study and contributed to the conception, design, as well as analysis and interpretation of results. HD was involved in conceptualizing and implementing the data analysis plan as well as overseeing drafting of research results. DJ made substantial contributions to project management, including data collection and analysis as well as in drafting the manuscript. MJ was in charge of project management including data collection, she also contributed to data analysis and write-up of the final version of the paper. NH contributed to data entry and analysis and also made major contributions to the write up of the paper. All authors have read and approved the final version of the manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1472-6963/11/45/prepub\n", "Description of Patient Safety Culture Composites and Cronbach's Alpha. File contains a definition of each of the Patient Safety Culture Composites in addition to Cronach's Alpha for each of the composites.\nClick here for file" ]

[ null, null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Animals", "Blotting, Western", "Capsid Proteins", "Cell Nucleus", "Chromatography, Affinity", "Ducks", "Enzyme-Linked Immunosorbent Assay", "Escherichia coli", "Fibroblasts", "Gene Expression", "Mardivirus", "Molecular Weight", "Rabbits", "Recombinant Proteins" ]

[ "Background", "Results", "Expression and purification of recombinant DEV VP19c and VP19c(N)", "Antigenicities analysis of rVP19c and rVP19c(N)", "Characterization of the polyclonal antibodies against rVP19c and rVP19c(N) proteins", "Discussion", "Conclusions", "Construction of plasmid", "Expression", "Purification", "Western blot", "Indirect ELISA", "Production of polyclonal antibody", "Indirect immunofluorescent assay", "Competing interests", "Authors' contributions" ]

[ "Duck viral enteritis (DVE) is an acute, contagious, and lethal disease of waterfowl of the family Anatidae worldwide [1]. The causative agent, duck enteritis virus (DEV), is a member of the family Herpesviridae, in which herpes simplex virus type 1 (HSV-1) is studied most completely. While research on molecular epidemiology of DEV has advanced over the years [2], relatively little is known concerning the structural, functional and immunogenic role of the structural proteins. The DEV virion is enveloped and the genome consists of double-stranded DNA segments packaged in an icosahedral capsid of several structural proteins [3]. The genetic information of viruses is enclosed in a capsid shell, a protein coat whose function is to protect the nucleic acid and to aid in the infectious process. In the HSV-1, capsid is an icosahedral shell, three of whose primary structural components are a major capsid protein (VP5; coded by the UL19 gene) and two minor capsid proteins, VP19c (UL38 gene) and VP23 (UL18 gene). VP19c and VP23 make up the triplex, which plays an essential role in capsid assembly and architecture [4]. Cell localization studies have also demonstrated the requirement of VP19c for the nuclear localization of VP23 [5]. Interestingly, the HSV-1 UL38 is regulated with late kinetics [6], whereas the bovine herpesvirus type 1 (BHV-1) and pseudorabies virus (PRV) UL38 transcript belong to the early kinetic class [7,8]. Most of the information of DEV UL38 gene currently is from bioinformatic approaches. Lacking an antibody against DEV VP19c, studies on biofunctions related to it are limited.\nComputational predictions of the VP19c amino acid sequence revealed that epitopes were more abundant on the N-terminal half of the VP19c protein than the C-terminal half of it [9]. Hence, in the present study, partial and full-length coding open reading frame (ORF) of UL38 gene were cloned, for the first time, into pET-32a(+) expression vector to obtain abundant recombinant proteins in E. coli. Moreover, their antigenic properties were characterized by western blot analysis and ELISA. Subsequently, two polyclonal antibodies were raised against the purified recombinant proteins in rabbits, and the titer and specificity of the polyclonal antibodies were characterized further by ELISA and immunofluorescent assays.", "[SUBTITLE] Expression and purification of recombinant DEV VP19c and VP19c(N) [SUBSECTION] The cloning strategy for constructing the recombinant plasmids is shown in Figure 1. The N-terminally His-tagged rVP19c and rVP19c(N) were produced in E. coli strain BL21 (DE3) with molecular weight of about 66 and 46 kDa (Figure 2). The optimal temperatures for rVP19c and rVP19c(N) expression were 30°C and 37°C (Figure 3) respectively, and optimal induction times for them were both about 4h (Figure 4), while concentrations of IPTG (data not shown) had a little influence on their expression. With the analysis of software BandScan 5.0, the induced rVP19c and rVP19c(N) reached approximately 40% and 70% in the total bacteria protein of each 1 ml culture, respectively.\nStrategy of vector construction. (A) Scheme of the arrangement of UL38 gene in DEV genome and the PCR amplification strategy. (B) Map of plasmid pET32a-VP19c and pET32a-VP19c(N).\nSDS-PAGE analysis of recombinant proteins. M, wide molecular weight protein marker; 1, control E. coli cell transformed without the insert; 2, the whole bacterium of pET32a-VP19c; 3, the whole bacterium of pET32a-VP19c(N). All of strains were grown in the presence (+) or in the absence (-) of IPTG.\nTemperature optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, and 3, the bacterial cultures of pET32a-VP19c were induced in 25°C, 30°C, 37°C, respectively; 4, 5, and 6, the bacterial cultures of pET32a-VP19c(N) were induced in 37°C, 30°C, 25°C, respectively.\nTime optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, 3, and 4, the bacterial cultures of pET32a-VP19c were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively; 5, 6, 7, and 8, the bacterial cultures of pET32a-VP19c(N) were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively.\nThe rVP19c and rVP19c(N) (Figure 5) were successfully expressed as inclusion bodies in E. coli and isolated roughly. The final yields of denatured soluble inclusion bodies were estimated to be approximately 100 mg/L (rVP19c) and 150 mg/L (rVP19c(N)) of initial bacterial culture. Then the proteins were purified by IMAC under denaturing conditions to reach the level of homogeneity (Figure 5).\nAnalysis of solubility and purification of recombinant protein by SDS-PAGE. M, wide molecular weight protein marker; 1, the supernatant of induced pET32a-VP19c after sonication; 2, the pellet of induced pET32a-VP19c after sonication; 3, the inclusion bodies of rVP19c by rough extraction; 4, the purified rVP19c by Ni-NTA affinity column; 5, the supernatant of induced pET32a-VP19c(N) after sonication; 6, the pellet of induced pET32a-VP19c(N) after sonication; 7, the inclusion bodies of rVP19c(N) by rough extraction; 8, the purified rVP19c(N) by Ni-NTA affinity column.\nThe cloning strategy for constructing the recombinant plasmids is shown in Figure 1. The N-terminally His-tagged rVP19c and rVP19c(N) were produced in E. coli strain BL21 (DE3) with molecular weight of about 66 and 46 kDa (Figure 2). The optimal temperatures for rVP19c and rVP19c(N) expression were 30°C and 37°C (Figure 3) respectively, and optimal induction times for them were both about 4h (Figure 4), while concentrations of IPTG (data not shown) had a little influence on their expression. With the analysis of software BandScan 5.0, the induced rVP19c and rVP19c(N) reached approximately 40% and 70% in the total bacteria protein of each 1 ml culture, respectively.\nStrategy of vector construction. (A) Scheme of the arrangement of UL38 gene in DEV genome and the PCR amplification strategy. (B) Map of plasmid pET32a-VP19c and pET32a-VP19c(N).\nSDS-PAGE analysis of recombinant proteins. M, wide molecular weight protein marker; 1, control E. coli cell transformed without the insert; 2, the whole bacterium of pET32a-VP19c; 3, the whole bacterium of pET32a-VP19c(N). All of strains were grown in the presence (+) or in the absence (-) of IPTG.\nTemperature optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, and 3, the bacterial cultures of pET32a-VP19c were induced in 25°C, 30°C, 37°C, respectively; 4, 5, and 6, the bacterial cultures of pET32a-VP19c(N) were induced in 37°C, 30°C, 25°C, respectively.\nTime optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, 3, and 4, the bacterial cultures of pET32a-VP19c were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively; 5, 6, 7, and 8, the bacterial cultures of pET32a-VP19c(N) were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively.\nThe rVP19c and rVP19c(N) (Figure 5) were successfully expressed as inclusion bodies in E. coli and isolated roughly. The final yields of denatured soluble inclusion bodies were estimated to be approximately 100 mg/L (rVP19c) and 150 mg/L (rVP19c(N)) of initial bacterial culture. Then the proteins were purified by IMAC under denaturing conditions to reach the level of homogeneity (Figure 5).\nAnalysis of solubility and purification of recombinant protein by SDS-PAGE. M, wide molecular weight protein marker; 1, the supernatant of induced pET32a-VP19c after sonication; 2, the pellet of induced pET32a-VP19c after sonication; 3, the inclusion bodies of rVP19c by rough extraction; 4, the purified rVP19c by Ni-NTA affinity column; 5, the supernatant of induced pET32a-VP19c(N) after sonication; 6, the pellet of induced pET32a-VP19c(N) after sonication; 7, the inclusion bodies of rVP19c(N) by rough extraction; 8, the purified rVP19c(N) by Ni-NTA affinity column.\n[SUBTITLE] Antigenicities analysis of rVP19c and rVP19c(N) [SUBSECTION] As shown in Figure 6, polyclonal antibody from rabbit immunized against purified DEV reacted with rVP19c and rVP19c(N). The results showed a similar immune-detection pattern as the native VP19c protein presented in the virus and the VP19c(N) was involved in antigenic recognition of lineal-epitopes recognized by PAb under denaturing conditions. In ELISA assay, the positive serum was able to recognize the rVP19c(N) protein at a similar level as rVP19c protein(Figure 7), which indicated that the linear epitopes within VP19c(N) recognized by positive serum were the same in the VP19c protein.\nWestern blot analysis of recombinant proteins with rabbit polyclonal antibody to DEV. 1, the rVP19c(N) protein incubated with rabbit pre-serum; 2, the rVP19c protein incubated with rabbit pre-serum; 3, the rVP19c(N) protein incubated with polyclonal anti-DEV antibodies; 4, the rVP19c protein incubated with polyclonal anti-DEV antibodies; M, prestained protein ladder.\nAntigenic properties analysis. ELISA analysis of recombinant proteins with duck anti-DEV positive serum. Duck negative serum was used to be a control.\nAs shown in Figure 6, polyclonal antibody from rabbit immunized against purified DEV reacted with rVP19c and rVP19c(N). The results showed a similar immune-detection pattern as the native VP19c protein presented in the virus and the VP19c(N) was involved in antigenic recognition of lineal-epitopes recognized by PAb under denaturing conditions. In ELISA assay, the positive serum was able to recognize the rVP19c(N) protein at a similar level as rVP19c protein(Figure 7), which indicated that the linear epitopes within VP19c(N) recognized by positive serum were the same in the VP19c protein.\nWestern blot analysis of recombinant proteins with rabbit polyclonal antibody to DEV. 1, the rVP19c(N) protein incubated with rabbit pre-serum; 2, the rVP19c protein incubated with rabbit pre-serum; 3, the rVP19c(N) protein incubated with polyclonal anti-DEV antibodies; 4, the rVP19c protein incubated with polyclonal anti-DEV antibodies; M, prestained protein ladder.\nAntigenic properties analysis. ELISA analysis of recombinant proteins with duck anti-DEV positive serum. Duck negative serum was used to be a control.\n[SUBTITLE] Characterization of the polyclonal antibodies against rVP19c and rVP19c(N) proteins [SUBSECTION] The antibodies against rVP19c and rVP19c(N) proteins at different dilutions were reacted with fusion proteins, pre-immunized rabbit serum served as the negative control. The two antibodies titer were found to be approximately 1:12 800. At the same time, negative control did not result in a detectable signal (data not shown).\nAs shown in Figure 8G, anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells and the VP19c protein targeted to the nucleus as distinct punctate speckles. In contrast, there was no staining in mock infected cells (Figure 8A) or DEV infected cells detected with the preimmune serum (Figure 8D). Similarly, the antibody against rVP19c(N) was able to recognize the VP19c protein in the cells infected with DEV (data not shown).\nIndirect immunofluorescent assay of the VP19c protein in DEV infected DEF cells. Mock infected DEF cells probed with the VP19c antibody (A). DEV infected DEF cells probed with the preimmune sera (D) or VP19c antibody (G). Cells were labeled with FITC-conjugated goat anti-rabbit immunoglobulin G and counterstained with DAPI to visualize the nuclei (B, E, H). Then FITC and DAPI were merged (C, F, I). (Images were acquired by using 60 × objective).\nThe antibodies against rVP19c and rVP19c(N) proteins at different dilutions were reacted with fusion proteins, pre-immunized rabbit serum served as the negative control. The two antibodies titer were found to be approximately 1:12 800. At the same time, negative control did not result in a detectable signal (data not shown).\nAs shown in Figure 8G, anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells and the VP19c protein targeted to the nucleus as distinct punctate speckles. In contrast, there was no staining in mock infected cells (Figure 8A) or DEV infected cells detected with the preimmune serum (Figure 8D). Similarly, the antibody against rVP19c(N) was able to recognize the VP19c protein in the cells infected with DEV (data not shown).\nIndirect immunofluorescent assay of the VP19c protein in DEV infected DEF cells. Mock infected DEF cells probed with the VP19c antibody (A). DEV infected DEF cells probed with the preimmune sera (D) or VP19c antibody (G). Cells were labeled with FITC-conjugated goat anti-rabbit immunoglobulin G and counterstained with DAPI to visualize the nuclei (B, E, H). Then FITC and DAPI were merged (C, F, I). (Images were acquired by using 60 × objective).", "The cloning strategy for constructing the recombinant plasmids is shown in Figure 1. The N-terminally His-tagged rVP19c and rVP19c(N) were produced in E. coli strain BL21 (DE3) with molecular weight of about 66 and 46 kDa (Figure 2). The optimal temperatures for rVP19c and rVP19c(N) expression were 30°C and 37°C (Figure 3) respectively, and optimal induction times for them were both about 4h (Figure 4), while concentrations of IPTG (data not shown) had a little influence on their expression. With the analysis of software BandScan 5.0, the induced rVP19c and rVP19c(N) reached approximately 40% and 70% in the total bacteria protein of each 1 ml culture, respectively.\nStrategy of vector construction. (A) Scheme of the arrangement of UL38 gene in DEV genome and the PCR amplification strategy. (B) Map of plasmid pET32a-VP19c and pET32a-VP19c(N).\nSDS-PAGE analysis of recombinant proteins. M, wide molecular weight protein marker; 1, control E. coli cell transformed without the insert; 2, the whole bacterium of pET32a-VP19c; 3, the whole bacterium of pET32a-VP19c(N). All of strains were grown in the presence (+) or in the absence (-) of IPTG.\nTemperature optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, and 3, the bacterial cultures of pET32a-VP19c were induced in 25°C, 30°C, 37°C, respectively; 4, 5, and 6, the bacterial cultures of pET32a-VP19c(N) were induced in 37°C, 30°C, 25°C, respectively.\nTime optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, 3, and 4, the bacterial cultures of pET32a-VP19c were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively; 5, 6, 7, and 8, the bacterial cultures of pET32a-VP19c(N) were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively.\nThe rVP19c and rVP19c(N) (Figure 5) were successfully expressed as inclusion bodies in E. coli and isolated roughly. The final yields of denatured soluble inclusion bodies were estimated to be approximately 100 mg/L (rVP19c) and 150 mg/L (rVP19c(N)) of initial bacterial culture. Then the proteins were purified by IMAC under denaturing conditions to reach the level of homogeneity (Figure 5).\nAnalysis of solubility and purification of recombinant protein by SDS-PAGE. M, wide molecular weight protein marker; 1, the supernatant of induced pET32a-VP19c after sonication; 2, the pellet of induced pET32a-VP19c after sonication; 3, the inclusion bodies of rVP19c by rough extraction; 4, the purified rVP19c by Ni-NTA affinity column; 5, the supernatant of induced pET32a-VP19c(N) after sonication; 6, the pellet of induced pET32a-VP19c(N) after sonication; 7, the inclusion bodies of rVP19c(N) by rough extraction; 8, the purified rVP19c(N) by Ni-NTA affinity column.", "As shown in Figure 6, polyclonal antibody from rabbit immunized against purified DEV reacted with rVP19c and rVP19c(N). The results showed a similar immune-detection pattern as the native VP19c protein presented in the virus and the VP19c(N) was involved in antigenic recognition of lineal-epitopes recognized by PAb under denaturing conditions. In ELISA assay, the positive serum was able to recognize the rVP19c(N) protein at a similar level as rVP19c protein(Figure 7), which indicated that the linear epitopes within VP19c(N) recognized by positive serum were the same in the VP19c protein.\nWestern blot analysis of recombinant proteins with rabbit polyclonal antibody to DEV. 1, the rVP19c(N) protein incubated with rabbit pre-serum; 2, the rVP19c protein incubated with rabbit pre-serum; 3, the rVP19c(N) protein incubated with polyclonal anti-DEV antibodies; 4, the rVP19c protein incubated with polyclonal anti-DEV antibodies; M, prestained protein ladder.\nAntigenic properties analysis. ELISA analysis of recombinant proteins with duck anti-DEV positive serum. Duck negative serum was used to be a control.", "The antibodies against rVP19c and rVP19c(N) proteins at different dilutions were reacted with fusion proteins, pre-immunized rabbit serum served as the negative control. The two antibodies titer were found to be approximately 1:12 800. At the same time, negative control did not result in a detectable signal (data not shown).\nAs shown in Figure 8G, anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells and the VP19c protein targeted to the nucleus as distinct punctate speckles. In contrast, there was no staining in mock infected cells (Figure 8A) or DEV infected cells detected with the preimmune serum (Figure 8D). Similarly, the antibody against rVP19c(N) was able to recognize the VP19c protein in the cells infected with DEV (data not shown).\nIndirect immunofluorescent assay of the VP19c protein in DEV infected DEF cells. Mock infected DEF cells probed with the VP19c antibody (A). DEV infected DEF cells probed with the preimmune sera (D) or VP19c antibody (G). Cells were labeled with FITC-conjugated goat anti-rabbit immunoglobulin G and counterstained with DAPI to visualize the nuclei (B, E, H). Then FITC and DAPI were merged (C, F, I). (Images were acquired by using 60 × objective).", "Currently, morphological features [3,10] and complete genome sequence [11] of DEV have been defined, but studies on proteins of DEV virion, especially structural and functional characteristics of viral structural proteins, have been very few. Generally, the structural viral proteins have various functions during the replicative cycle, which might be closely related to their structure [12]. Therefore, the understanding of the function of each structural protein depends on the study at a molecular level.\nComputational predictions of the VP19c amino acid sequence suggested that it did not require any post-translational modifications and lacked transmembrane regions and signal peptides, so it is more suitable for prokaryotic expression. And predictions presented the N-terminus of VP19c protein may contain most antigenic linear-epitopes [9].\nTo obtain high-level expression, several expression conditions, such as induction time, induction temperature, and IPTG concentration were optimized. It could be seen that temperature of induction is critical. After induction conditions optimized, the expected high-level expression result was obtained.\nWestern blot confirmed that the expressed recombinant proteins were specific to DEV and had good antigenicity. These results strongly suggested that the purified rVP19c protein seemed to be structurally and antigenically very similar to native VP19c protein and the antigenicity of VP19c was also mainly determined by linear antigenic determinant of it. Inclusion bodies (IBs) are denatured proteins and lack conformational epitopes. Thus it can be seen that though the IBs are inactivated form of protein, they are sufficient for animal immunity and harvest of specific sera [13]. Moreover, the rVP19c(N) was deleted C-terminus, but still remained good antigenicity. The results of ELISA assays showed that the antibody in the duck sera raised against the complete VP19c protein were primarily against the N-terminal region of the VP19c protein, which indicated that the N-terminal portion of the VP19c protein may harbor most of the linear epitopes. Furthermore, anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells, which indicated that this specific polyclonal antibody provide a good tool for further studying structural and functional characterization of DEV VP19c.\nCurrently, little is known about DEV VP19c function or regulation of expression, as well as other features. Homology analysis with other herpesviruses suggested that the DEV VP19c protein could be a crucial capsid protein [14]. Conservation of the protein suggests that it may play a role in the viral life cycle. Because of the localization of VP19c to viral factories, we investigated the possibility that this protein might be incorporated into DEV virions.", "To our knowledge, this was the first report on basic properties of DEV VP19c protein. In the present study, we obtained a high-level expression of the recombinant VP19c protein as well as high titers of rabbit polyclonal antibody against to VP19c protein. The anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells and the VP19c protein targeted to the nucleus as distinct punctate speckles. This specific polyclonal antibody provides a good tool for further studying structural and functional characterization of DEV VP19c.", "The viral DNA was extracted from partially purified DEV and used as template for polymerase chain reaction (PCR) [15]. A fragment (1515 bp) containing the complete ORF of DEV UL38 gene (1398 bp) (Genbank accession no. EU071041) was amplified using PCR with the following primers: the upstream primer (P1) 5'-GGATCCACGATGAAAGTACCAAATG-3' containing the BamHI site and the downstream primer (P2)5'-CTCGAGAGCCAGATACGAC AAGAAG-3' containing the XhoI site. The PCR parameters were 10 min at 95°C and 30 cycles of 1 min at 94°C, 1 min at 58.5°C, 1 min at 72°C, and a final extension time of 10 min at 72°C. A fragment (723 bp) containing the partial ORF (711 bp) encoding the N-terminal of VP19c (VP19c(N)) protein was amplified with the following primers: the upstream primer (P1) 5'-GGATCCACGATGAAAGTACCAAATG-3' containing the BamHI siteand the downstream primer (P3) 5' -CTCGAGGTGTACGTGACATCTAACCA TAG-3' containing the XhoI site. The PCR assay and programme were carried out following the protocol described above.\nThe PCR products and plasmid pET-32a(+) (Qiagen GmbH, Hilden, Germany) were both digested with BamH I and Xho I, and then ligated with T4 DNA ligase to yield the constructs. The constructs were transformed into E.coli, and the selected bacterial transformants were verified by colony PCR, restriction enzyme analysis and sequencing.", "The positive individual clone was cultured in 5 mL Luria bertani (LB) medium containing 100 ug/mL amp and then induced at 37°C by adding isopropyl-β-D-thiogalactoside (IPTG) at a final concentration of 1 mM for 4 h. For IPTG dose optimization, the bacterial culture was induced with different concentrations of IPTG [0.2, 0.3, 0.4, 0.5, 0.6,0.7,0.8,0.9,1.0 (mM)] and allowed to grow for 4h at 37°C. For temperature optimization, the bacterial culture was induced with IPTG [1.0 (mM)] and allowed to grow for 4 h at three different temperatures (25, 30 and 37°C). For time optimization, the bacterial culture was induced with IPTG [1.0 (mM)] and allowed to grow for 3 h, 4 h, 5 h and overnight (~16 h) at 37°C. Total cell proteins from each optimization experiment were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Then, small-scale expression was done by optimized conditions as described above to prepare for purification [16]. The protein amount was determined with reference to standard bovine serum albumin (BSA) in the Bradford assay [17].", "500 ml induced bacterial culture was harvested after 4 h, centrifuged at 6000 × g for 10 min and the cell pellet was suspended in 20 mM Tris buffer (pH = 8.0). The cells were later lysed by using lysozyme (0.1 mg/mL) at 4°C for 1 h and sonicated on ice for 5 min at an amplitude of 30% with a 30s pulse frequency. The lysate was centrifuged at 10,000 × g for 20 min at 4°C and the supernatant was collected as soluble fraction. The resulting pellet was washed twice with 10 mL 2M urea containing 50 mM Tris buffer (pH = 8.0), 1 mM EDTA, 150 mM NaCl and 0.1% Triton X-100.\nThe suspension was centrifuged at 10,000 × g for 20 min at 4°C and then the resulting subsidence was resuspended in regeneration buffer containing 6 M urea, 0.5 M NaCl, 20 mM Tris-HCl (pH 7.9) and incubated at room temperature for 30 min. The incubated mixtures were then centrifuged at 10,000 × g for 20 min, and the supernatant was submitted to further purification. The supernatant was then poured on to a purification column and allowed to bind for 1 h with gentle shaking. The recombinant His-tagged proteins were purified from the supernatant obtained above by immobilized metal affinity chromatography (IMAC) on Ni-NTA affinity resin (Bio-Rad) following the conventional protocol [18]. Finally, the proteins were collected and analyzed by SDS-PAGE to assess the level of homogeneity.", "The recombinant proteins were electrophoresed with SDS-PAGE using 12% polyacrylamide gel and then electroblotted onto polyvinylidene fluoride (PVDF) membrane. The PVDF membrane was blocked with 3% bovine serum albumin (BSA) in Tris-Buffered Saline Tween-20 (TBST) Buffer containing 20 mM Tris-HCl,150 mM NaCl and 0.05% Tween 20 overnight at 4°C. The membrane was washed three times with TBST and incubated for 1 h at 37°C with rabbit anti-DEV polyclonal antibody and rabbit pre-serum at 1:100 of dilution in TBST buffer containing 0.5% BSA. After washing, the membrane was incubated for 1 h at 37°C with anti-rabbit-HRPO conjugated IgG(GE Healthcare Limited, Buckinghamshire, United Kingdom) at 1:5000 of dilution in TBST buffer containing 0.5% BSA. Finally, the membrane was washed with TBST and placed in diamino benzidine (DAB) solution as a chromogen to visualize the binding.", "Flat bottomed 96 well plate (Corning, Corning Costar Corp., MA, USA) was coated overnight at 4°C with the recombinant proteins at 7 ug/ml in carbonate bicarbonate buffer, pH 9.6. After blocking with 1% BSA in PBS, the wells were washed with PBS containing 0.05% (v/v) Tween-20 (PBST) and later incubated at 37°C for 1 h with various dilutions (1:5120-1:10) of duck anti-DEV strain Chv positive serum and negative serum. After washing with PBST, the plates were incubated at 37°C for 1 h with anti-duck-HRPO conjugated IgG(GE Healthcare Limited, Buckinghamshire, United Kingdom). Finally, the plate was washed and the peroxidase reaction was visualized by using tetrame-thylbenzidine(TMB) (Sigma) as substrate after incubation for 10 min at room temperature. The reaction was stopped by adding 2 M H2SO4 and absorbance was read at 450 nm by a microplate autoreader (Bio-Rad).", "Preimmune serum was collected prior to immunization. New Zealand white rabbits were immunized firstly intradermally with a mixture of 1 mg purified recombinant rVP19c or rVP19c(N) mixed with an equal volume of complete Freund's adjuvant (Sigma). Two weeks later, the rabbits were boosted twice subcutaneously with the same amount of recombinant proteins mixed with an equal volume of incomplete Freund's adjuvant at a one-week interval. Two weeks after the last immunization, the two antiserums were harvested from the carotid artery. Then the polyclonal antibody was purified by protein A affinity IgG purification kit according the user's guide. The titer of the specific antibodies was determined by ELISA.", "For immunofluorescence assays, monolayers of duck embryo fibroblast (DEF) cells were infected (MOI 5) with the DEV strain Chv and then incubated for 30 h at 37°C. Cells were fixed with 4% paraformaldehyde for 30 min, washed with PBS and permeabilized with 0.5% Triton X-100 for 10 min. After washing, the cells were blocked with PBS containing 5% BSA for 1.5 h at 37°C. Subsequently, the cells were incubated with rVP19c-specific polyclonal antibody diluted in PBS containing 0.5% BSA for 1 h at 37°C. Finally, fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit immunoglobulin G (Sigma) was added at a dilution of 1:200 and incubated at 4°C for 16 h. After each incubation step, the cells were washed extensively with PBS. The cell nuclei were visualized by DAPI counterstaining [19-21]. The images captured with fluorescence microscopy (Nikon, Japan).", "The authors declare that they have no competing interests.", "JX and SCZ carried out most of the experiments and drafted the manuscript. ACC and MSW have critically revised the manuscript and the experimental design. HC, CJS, DKZ, RYJ, QHL, ZLC and XYC helped in experiments. All authors have read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Results", "Expression and purification of recombinant DEV VP19c and VP19c(N)", "Antigenicities analysis of rVP19c and rVP19c(N)", "Characterization of the polyclonal antibodies against rVP19c and rVP19c(N) proteins", "Discussion", "Conclusions", "Methods", "Construction of plasmid", "Expression", "Purification", "Western blot", "Indirect ELISA", "Production of polyclonal antibody", "Indirect immunofluorescent assay", "Competing interests", "Authors' contributions" ]

[ "Duck viral enteritis (DVE) is an acute, contagious, and lethal disease of waterfowl of the family Anatidae worldwide [1]. The causative agent, duck enteritis virus (DEV), is a member of the family Herpesviridae, in which herpes simplex virus type 1 (HSV-1) is studied most completely. While research on molecular epidemiology of DEV has advanced over the years [2], relatively little is known concerning the structural, functional and immunogenic role of the structural proteins. The DEV virion is enveloped and the genome consists of double-stranded DNA segments packaged in an icosahedral capsid of several structural proteins [3]. The genetic information of viruses is enclosed in a capsid shell, a protein coat whose function is to protect the nucleic acid and to aid in the infectious process. In the HSV-1, capsid is an icosahedral shell, three of whose primary structural components are a major capsid protein (VP5; coded by the UL19 gene) and two minor capsid proteins, VP19c (UL38 gene) and VP23 (UL18 gene). VP19c and VP23 make up the triplex, which plays an essential role in capsid assembly and architecture [4]. Cell localization studies have also demonstrated the requirement of VP19c for the nuclear localization of VP23 [5]. Interestingly, the HSV-1 UL38 is regulated with late kinetics [6], whereas the bovine herpesvirus type 1 (BHV-1) and pseudorabies virus (PRV) UL38 transcript belong to the early kinetic class [7,8]. Most of the information of DEV UL38 gene currently is from bioinformatic approaches. Lacking an antibody against DEV VP19c, studies on biofunctions related to it are limited.\nComputational predictions of the VP19c amino acid sequence revealed that epitopes were more abundant on the N-terminal half of the VP19c protein than the C-terminal half of it [9]. Hence, in the present study, partial and full-length coding open reading frame (ORF) of UL38 gene were cloned, for the first time, into pET-32a(+) expression vector to obtain abundant recombinant proteins in E. coli. Moreover, their antigenic properties were characterized by western blot analysis and ELISA. Subsequently, two polyclonal antibodies were raised against the purified recombinant proteins in rabbits, and the titer and specificity of the polyclonal antibodies were characterized further by ELISA and immunofluorescent assays.", "[SUBTITLE] Expression and purification of recombinant DEV VP19c and VP19c(N) [SUBSECTION] The cloning strategy for constructing the recombinant plasmids is shown in Figure 1. The N-terminally His-tagged rVP19c and rVP19c(N) were produced in E. coli strain BL21 (DE3) with molecular weight of about 66 and 46 kDa (Figure 2). The optimal temperatures for rVP19c and rVP19c(N) expression were 30°C and 37°C (Figure 3) respectively, and optimal induction times for them were both about 4h (Figure 4), while concentrations of IPTG (data not shown) had a little influence on their expression. With the analysis of software BandScan 5.0, the induced rVP19c and rVP19c(N) reached approximately 40% and 70% in the total bacteria protein of each 1 ml culture, respectively.\nStrategy of vector construction. (A) Scheme of the arrangement of UL38 gene in DEV genome and the PCR amplification strategy. (B) Map of plasmid pET32a-VP19c and pET32a-VP19c(N).\nSDS-PAGE analysis of recombinant proteins. M, wide molecular weight protein marker; 1, control E. coli cell transformed without the insert; 2, the whole bacterium of pET32a-VP19c; 3, the whole bacterium of pET32a-VP19c(N). All of strains were grown in the presence (+) or in the absence (-) of IPTG.\nTemperature optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, and 3, the bacterial cultures of pET32a-VP19c were induced in 25°C, 30°C, 37°C, respectively; 4, 5, and 6, the bacterial cultures of pET32a-VP19c(N) were induced in 37°C, 30°C, 25°C, respectively.\nTime optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, 3, and 4, the bacterial cultures of pET32a-VP19c were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively; 5, 6, 7, and 8, the bacterial cultures of pET32a-VP19c(N) were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively.\nThe rVP19c and rVP19c(N) (Figure 5) were successfully expressed as inclusion bodies in E. coli and isolated roughly. The final yields of denatured soluble inclusion bodies were estimated to be approximately 100 mg/L (rVP19c) and 150 mg/L (rVP19c(N)) of initial bacterial culture. Then the proteins were purified by IMAC under denaturing conditions to reach the level of homogeneity (Figure 5).\nAnalysis of solubility and purification of recombinant protein by SDS-PAGE. M, wide molecular weight protein marker; 1, the supernatant of induced pET32a-VP19c after sonication; 2, the pellet of induced pET32a-VP19c after sonication; 3, the inclusion bodies of rVP19c by rough extraction; 4, the purified rVP19c by Ni-NTA affinity column; 5, the supernatant of induced pET32a-VP19c(N) after sonication; 6, the pellet of induced pET32a-VP19c(N) after sonication; 7, the inclusion bodies of rVP19c(N) by rough extraction; 8, the purified rVP19c(N) by Ni-NTA affinity column.\nThe cloning strategy for constructing the recombinant plasmids is shown in Figure 1. The N-terminally His-tagged rVP19c and rVP19c(N) were produced in E. coli strain BL21 (DE3) with molecular weight of about 66 and 46 kDa (Figure 2). The optimal temperatures for rVP19c and rVP19c(N) expression were 30°C and 37°C (Figure 3) respectively, and optimal induction times for them were both about 4h (Figure 4), while concentrations of IPTG (data not shown) had a little influence on their expression. With the analysis of software BandScan 5.0, the induced rVP19c and rVP19c(N) reached approximately 40% and 70% in the total bacteria protein of each 1 ml culture, respectively.\nStrategy of vector construction. (A) Scheme of the arrangement of UL38 gene in DEV genome and the PCR amplification strategy. (B) Map of plasmid pET32a-VP19c and pET32a-VP19c(N).\nSDS-PAGE analysis of recombinant proteins. M, wide molecular weight protein marker; 1, control E. coli cell transformed without the insert; 2, the whole bacterium of pET32a-VP19c; 3, the whole bacterium of pET32a-VP19c(N). All of strains were grown in the presence (+) or in the absence (-) of IPTG.\nTemperature optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, and 3, the bacterial cultures of pET32a-VP19c were induced in 25°C, 30°C, 37°C, respectively; 4, 5, and 6, the bacterial cultures of pET32a-VP19c(N) were induced in 37°C, 30°C, 25°C, respectively.\nTime optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, 3, and 4, the bacterial cultures of pET32a-VP19c were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively; 5, 6, 7, and 8, the bacterial cultures of pET32a-VP19c(N) were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively.\nThe rVP19c and rVP19c(N) (Figure 5) were successfully expressed as inclusion bodies in E. coli and isolated roughly. The final yields of denatured soluble inclusion bodies were estimated to be approximately 100 mg/L (rVP19c) and 150 mg/L (rVP19c(N)) of initial bacterial culture. Then the proteins were purified by IMAC under denaturing conditions to reach the level of homogeneity (Figure 5).\nAnalysis of solubility and purification of recombinant protein by SDS-PAGE. M, wide molecular weight protein marker; 1, the supernatant of induced pET32a-VP19c after sonication; 2, the pellet of induced pET32a-VP19c after sonication; 3, the inclusion bodies of rVP19c by rough extraction; 4, the purified rVP19c by Ni-NTA affinity column; 5, the supernatant of induced pET32a-VP19c(N) after sonication; 6, the pellet of induced pET32a-VP19c(N) after sonication; 7, the inclusion bodies of rVP19c(N) by rough extraction; 8, the purified rVP19c(N) by Ni-NTA affinity column.\n[SUBTITLE] Antigenicities analysis of rVP19c and rVP19c(N) [SUBSECTION] As shown in Figure 6, polyclonal antibody from rabbit immunized against purified DEV reacted with rVP19c and rVP19c(N). The results showed a similar immune-detection pattern as the native VP19c protein presented in the virus and the VP19c(N) was involved in antigenic recognition of lineal-epitopes recognized by PAb under denaturing conditions. In ELISA assay, the positive serum was able to recognize the rVP19c(N) protein at a similar level as rVP19c protein(Figure 7), which indicated that the linear epitopes within VP19c(N) recognized by positive serum were the same in the VP19c protein.\nWestern blot analysis of recombinant proteins with rabbit polyclonal antibody to DEV. 1, the rVP19c(N) protein incubated with rabbit pre-serum; 2, the rVP19c protein incubated with rabbit pre-serum; 3, the rVP19c(N) protein incubated with polyclonal anti-DEV antibodies; 4, the rVP19c protein incubated with polyclonal anti-DEV antibodies; M, prestained protein ladder.\nAntigenic properties analysis. ELISA analysis of recombinant proteins with duck anti-DEV positive serum. Duck negative serum was used to be a control.\nAs shown in Figure 6, polyclonal antibody from rabbit immunized against purified DEV reacted with rVP19c and rVP19c(N). The results showed a similar immune-detection pattern as the native VP19c protein presented in the virus and the VP19c(N) was involved in antigenic recognition of lineal-epitopes recognized by PAb under denaturing conditions. In ELISA assay, the positive serum was able to recognize the rVP19c(N) protein at a similar level as rVP19c protein(Figure 7), which indicated that the linear epitopes within VP19c(N) recognized by positive serum were the same in the VP19c protein.\nWestern blot analysis of recombinant proteins with rabbit polyclonal antibody to DEV. 1, the rVP19c(N) protein incubated with rabbit pre-serum; 2, the rVP19c protein incubated with rabbit pre-serum; 3, the rVP19c(N) protein incubated with polyclonal anti-DEV antibodies; 4, the rVP19c protein incubated with polyclonal anti-DEV antibodies; M, prestained protein ladder.\nAntigenic properties analysis. ELISA analysis of recombinant proteins with duck anti-DEV positive serum. Duck negative serum was used to be a control.\n[SUBTITLE] Characterization of the polyclonal antibodies against rVP19c and rVP19c(N) proteins [SUBSECTION] The antibodies against rVP19c and rVP19c(N) proteins at different dilutions were reacted with fusion proteins, pre-immunized rabbit serum served as the negative control. The two antibodies titer were found to be approximately 1:12 800. At the same time, negative control did not result in a detectable signal (data not shown).\nAs shown in Figure 8G, anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells and the VP19c protein targeted to the nucleus as distinct punctate speckles. In contrast, there was no staining in mock infected cells (Figure 8A) or DEV infected cells detected with the preimmune serum (Figure 8D). Similarly, the antibody against rVP19c(N) was able to recognize the VP19c protein in the cells infected with DEV (data not shown).\nIndirect immunofluorescent assay of the VP19c protein in DEV infected DEF cells. Mock infected DEF cells probed with the VP19c antibody (A). DEV infected DEF cells probed with the preimmune sera (D) or VP19c antibody (G). Cells were labeled with FITC-conjugated goat anti-rabbit immunoglobulin G and counterstained with DAPI to visualize the nuclei (B, E, H). Then FITC and DAPI were merged (C, F, I). (Images were acquired by using 60 × objective).\nThe antibodies against rVP19c and rVP19c(N) proteins at different dilutions were reacted with fusion proteins, pre-immunized rabbit serum served as the negative control. The two antibodies titer were found to be approximately 1:12 800. At the same time, negative control did not result in a detectable signal (data not shown).\nAs shown in Figure 8G, anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells and the VP19c protein targeted to the nucleus as distinct punctate speckles. In contrast, there was no staining in mock infected cells (Figure 8A) or DEV infected cells detected with the preimmune serum (Figure 8D). Similarly, the antibody against rVP19c(N) was able to recognize the VP19c protein in the cells infected with DEV (data not shown).\nIndirect immunofluorescent assay of the VP19c protein in DEV infected DEF cells. Mock infected DEF cells probed with the VP19c antibody (A). DEV infected DEF cells probed with the preimmune sera (D) or VP19c antibody (G). Cells were labeled with FITC-conjugated goat anti-rabbit immunoglobulin G and counterstained with DAPI to visualize the nuclei (B, E, H). Then FITC and DAPI were merged (C, F, I). (Images were acquired by using 60 × objective).", "The cloning strategy for constructing the recombinant plasmids is shown in Figure 1. The N-terminally His-tagged rVP19c and rVP19c(N) were produced in E. coli strain BL21 (DE3) with molecular weight of about 66 and 46 kDa (Figure 2). The optimal temperatures for rVP19c and rVP19c(N) expression were 30°C and 37°C (Figure 3) respectively, and optimal induction times for them were both about 4h (Figure 4), while concentrations of IPTG (data not shown) had a little influence on their expression. With the analysis of software BandScan 5.0, the induced rVP19c and rVP19c(N) reached approximately 40% and 70% in the total bacteria protein of each 1 ml culture, respectively.\nStrategy of vector construction. (A) Scheme of the arrangement of UL38 gene in DEV genome and the PCR amplification strategy. (B) Map of plasmid pET32a-VP19c and pET32a-VP19c(N).\nSDS-PAGE analysis of recombinant proteins. M, wide molecular weight protein marker; 1, control E. coli cell transformed without the insert; 2, the whole bacterium of pET32a-VP19c; 3, the whole bacterium of pET32a-VP19c(N). All of strains were grown in the presence (+) or in the absence (-) of IPTG.\nTemperature optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, and 3, the bacterial cultures of pET32a-VP19c were induced in 25°C, 30°C, 37°C, respectively; 4, 5, and 6, the bacterial cultures of pET32a-VP19c(N) were induced in 37°C, 30°C, 25°C, respectively.\nTime optimization analysis by SDS-PAGE. M, low molecular protein marker; 1, 2, 3, and 4, the bacterial cultures of pET32a-VP19c were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively; 5, 6, 7, and 8, the bacterial cultures of pET32a-VP19c(N) were induced for 3 h, 4 h, 5 h, overnight (16 h), respectively.\nThe rVP19c and rVP19c(N) (Figure 5) were successfully expressed as inclusion bodies in E. coli and isolated roughly. The final yields of denatured soluble inclusion bodies were estimated to be approximately 100 mg/L (rVP19c) and 150 mg/L (rVP19c(N)) of initial bacterial culture. Then the proteins were purified by IMAC under denaturing conditions to reach the level of homogeneity (Figure 5).\nAnalysis of solubility and purification of recombinant protein by SDS-PAGE. M, wide molecular weight protein marker; 1, the supernatant of induced pET32a-VP19c after sonication; 2, the pellet of induced pET32a-VP19c after sonication; 3, the inclusion bodies of rVP19c by rough extraction; 4, the purified rVP19c by Ni-NTA affinity column; 5, the supernatant of induced pET32a-VP19c(N) after sonication; 6, the pellet of induced pET32a-VP19c(N) after sonication; 7, the inclusion bodies of rVP19c(N) by rough extraction; 8, the purified rVP19c(N) by Ni-NTA affinity column.", "As shown in Figure 6, polyclonal antibody from rabbit immunized against purified DEV reacted with rVP19c and rVP19c(N). The results showed a similar immune-detection pattern as the native VP19c protein presented in the virus and the VP19c(N) was involved in antigenic recognition of lineal-epitopes recognized by PAb under denaturing conditions. In ELISA assay, the positive serum was able to recognize the rVP19c(N) protein at a similar level as rVP19c protein(Figure 7), which indicated that the linear epitopes within VP19c(N) recognized by positive serum were the same in the VP19c protein.\nWestern blot analysis of recombinant proteins with rabbit polyclonal antibody to DEV. 1, the rVP19c(N) protein incubated with rabbit pre-serum; 2, the rVP19c protein incubated with rabbit pre-serum; 3, the rVP19c(N) protein incubated with polyclonal anti-DEV antibodies; 4, the rVP19c protein incubated with polyclonal anti-DEV antibodies; M, prestained protein ladder.\nAntigenic properties analysis. ELISA analysis of recombinant proteins with duck anti-DEV positive serum. Duck negative serum was used to be a control.", "The antibodies against rVP19c and rVP19c(N) proteins at different dilutions were reacted with fusion proteins, pre-immunized rabbit serum served as the negative control. The two antibodies titer were found to be approximately 1:12 800. At the same time, negative control did not result in a detectable signal (data not shown).\nAs shown in Figure 8G, anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells and the VP19c protein targeted to the nucleus as distinct punctate speckles. In contrast, there was no staining in mock infected cells (Figure 8A) or DEV infected cells detected with the preimmune serum (Figure 8D). Similarly, the antibody against rVP19c(N) was able to recognize the VP19c protein in the cells infected with DEV (data not shown).\nIndirect immunofluorescent assay of the VP19c protein in DEV infected DEF cells. Mock infected DEF cells probed with the VP19c antibody (A). DEV infected DEF cells probed with the preimmune sera (D) or VP19c antibody (G). Cells were labeled with FITC-conjugated goat anti-rabbit immunoglobulin G and counterstained with DAPI to visualize the nuclei (B, E, H). Then FITC and DAPI were merged (C, F, I). (Images were acquired by using 60 × objective).", "Currently, morphological features [3,10] and complete genome sequence [11] of DEV have been defined, but studies on proteins of DEV virion, especially structural and functional characteristics of viral structural proteins, have been very few. Generally, the structural viral proteins have various functions during the replicative cycle, which might be closely related to their structure [12]. Therefore, the understanding of the function of each structural protein depends on the study at a molecular level.\nComputational predictions of the VP19c amino acid sequence suggested that it did not require any post-translational modifications and lacked transmembrane regions and signal peptides, so it is more suitable for prokaryotic expression. And predictions presented the N-terminus of VP19c protein may contain most antigenic linear-epitopes [9].\nTo obtain high-level expression, several expression conditions, such as induction time, induction temperature, and IPTG concentration were optimized. It could be seen that temperature of induction is critical. After induction conditions optimized, the expected high-level expression result was obtained.\nWestern blot confirmed that the expressed recombinant proteins were specific to DEV and had good antigenicity. These results strongly suggested that the purified rVP19c protein seemed to be structurally and antigenically very similar to native VP19c protein and the antigenicity of VP19c was also mainly determined by linear antigenic determinant of it. Inclusion bodies (IBs) are denatured proteins and lack conformational epitopes. Thus it can be seen that though the IBs are inactivated form of protein, they are sufficient for animal immunity and harvest of specific sera [13]. Moreover, the rVP19c(N) was deleted C-terminus, but still remained good antigenicity. The results of ELISA assays showed that the antibody in the duck sera raised against the complete VP19c protein were primarily against the N-terminal region of the VP19c protein, which indicated that the N-terminal portion of the VP19c protein may harbor most of the linear epitopes. Furthermore, anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells, which indicated that this specific polyclonal antibody provide a good tool for further studying structural and functional characterization of DEV VP19c.\nCurrently, little is known about DEV VP19c function or regulation of expression, as well as other features. Homology analysis with other herpesviruses suggested that the DEV VP19c protein could be a crucial capsid protein [14]. Conservation of the protein suggests that it may play a role in the viral life cycle. Because of the localization of VP19c to viral factories, we investigated the possibility that this protein might be incorporated into DEV virions.", "To our knowledge, this was the first report on basic properties of DEV VP19c protein. In the present study, we obtained a high-level expression of the recombinant VP19c protein as well as high titers of rabbit polyclonal antibody against to VP19c protein. The anti-rVP19c serum was able to detect the VP19c protein in DEV infected cells and the VP19c protein targeted to the nucleus as distinct punctate speckles. This specific polyclonal antibody provides a good tool for further studying structural and functional characterization of DEV VP19c.", "[SUBTITLE] Construction of plasmid [SUBSECTION] The viral DNA was extracted from partially purified DEV and used as template for polymerase chain reaction (PCR) [15]. A fragment (1515 bp) containing the complete ORF of DEV UL38 gene (1398 bp) (Genbank accession no. EU071041) was amplified using PCR with the following primers: the upstream primer (P1) 5'-GGATCCACGATGAAAGTACCAAATG-3' containing the BamHI site and the downstream primer (P2)5'-CTCGAGAGCCAGATACGAC AAGAAG-3' containing the XhoI site. The PCR parameters were 10 min at 95°C and 30 cycles of 1 min at 94°C, 1 min at 58.5°C, 1 min at 72°C, and a final extension time of 10 min at 72°C. A fragment (723 bp) containing the partial ORF (711 bp) encoding the N-terminal of VP19c (VP19c(N)) protein was amplified with the following primers: the upstream primer (P1) 5'-GGATCCACGATGAAAGTACCAAATG-3' containing the BamHI siteand the downstream primer (P3) 5' -CTCGAGGTGTACGTGACATCTAACCA TAG-3' containing the XhoI site. The PCR assay and programme were carried out following the protocol described above.\nThe PCR products and plasmid pET-32a(+) (Qiagen GmbH, Hilden, Germany) were both digested with BamH I and Xho I, and then ligated with T4 DNA ligase to yield the constructs. The constructs were transformed into E.coli, and the selected bacterial transformants were verified by colony PCR, restriction enzyme analysis and sequencing.\nThe viral DNA was extracted from partially purified DEV and used as template for polymerase chain reaction (PCR) [15]. A fragment (1515 bp) containing the complete ORF of DEV UL38 gene (1398 bp) (Genbank accession no. EU071041) was amplified using PCR with the following primers: the upstream primer (P1) 5'-GGATCCACGATGAAAGTACCAAATG-3' containing the BamHI site and the downstream primer (P2)5'-CTCGAGAGCCAGATACGAC AAGAAG-3' containing the XhoI site. The PCR parameters were 10 min at 95°C and 30 cycles of 1 min at 94°C, 1 min at 58.5°C, 1 min at 72°C, and a final extension time of 10 min at 72°C. A fragment (723 bp) containing the partial ORF (711 bp) encoding the N-terminal of VP19c (VP19c(N)) protein was amplified with the following primers: the upstream primer (P1) 5'-GGATCCACGATGAAAGTACCAAATG-3' containing the BamHI siteand the downstream primer (P3) 5' -CTCGAGGTGTACGTGACATCTAACCA TAG-3' containing the XhoI site. The PCR assay and programme were carried out following the protocol described above.\nThe PCR products and plasmid pET-32a(+) (Qiagen GmbH, Hilden, Germany) were both digested with BamH I and Xho I, and then ligated with T4 DNA ligase to yield the constructs. The constructs were transformed into E.coli, and the selected bacterial transformants were verified by colony PCR, restriction enzyme analysis and sequencing.\n[SUBTITLE] Expression [SUBSECTION] The positive individual clone was cultured in 5 mL Luria bertani (LB) medium containing 100 ug/mL amp and then induced at 37°C by adding isopropyl-β-D-thiogalactoside (IPTG) at a final concentration of 1 mM for 4 h. For IPTG dose optimization, the bacterial culture was induced with different concentrations of IPTG [0.2, 0.3, 0.4, 0.5, 0.6,0.7,0.8,0.9,1.0 (mM)] and allowed to grow for 4h at 37°C. For temperature optimization, the bacterial culture was induced with IPTG [1.0 (mM)] and allowed to grow for 4 h at three different temperatures (25, 30 and 37°C). For time optimization, the bacterial culture was induced with IPTG [1.0 (mM)] and allowed to grow for 3 h, 4 h, 5 h and overnight (~16 h) at 37°C. Total cell proteins from each optimization experiment were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Then, small-scale expression was done by optimized conditions as described above to prepare for purification [16]. The protein amount was determined with reference to standard bovine serum albumin (BSA) in the Bradford assay [17].\nThe positive individual clone was cultured in 5 mL Luria bertani (LB) medium containing 100 ug/mL amp and then induced at 37°C by adding isopropyl-β-D-thiogalactoside (IPTG) at a final concentration of 1 mM for 4 h. For IPTG dose optimization, the bacterial culture was induced with different concentrations of IPTG [0.2, 0.3, 0.4, 0.5, 0.6,0.7,0.8,0.9,1.0 (mM)] and allowed to grow for 4h at 37°C. For temperature optimization, the bacterial culture was induced with IPTG [1.0 (mM)] and allowed to grow for 4 h at three different temperatures (25, 30 and 37°C). For time optimization, the bacterial culture was induced with IPTG [1.0 (mM)] and allowed to grow for 3 h, 4 h, 5 h and overnight (~16 h) at 37°C. Total cell proteins from each optimization experiment were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Then, small-scale expression was done by optimized conditions as described above to prepare for purification [16]. The protein amount was determined with reference to standard bovine serum albumin (BSA) in the Bradford assay [17].\n[SUBTITLE] Purification [SUBSECTION] 500 ml induced bacterial culture was harvested after 4 h, centrifuged at 6000 × g for 10 min and the cell pellet was suspended in 20 mM Tris buffer (pH = 8.0). The cells were later lysed by using lysozyme (0.1 mg/mL) at 4°C for 1 h and sonicated on ice for 5 min at an amplitude of 30% with a 30s pulse frequency. The lysate was centrifuged at 10,000 × g for 20 min at 4°C and the supernatant was collected as soluble fraction. The resulting pellet was washed twice with 10 mL 2M urea containing 50 mM Tris buffer (pH = 8.0), 1 mM EDTA, 150 mM NaCl and 0.1% Triton X-100.\nThe suspension was centrifuged at 10,000 × g for 20 min at 4°C and then the resulting subsidence was resuspended in regeneration buffer containing 6 M urea, 0.5 M NaCl, 20 mM Tris-HCl (pH 7.9) and incubated at room temperature for 30 min. The incubated mixtures were then centrifuged at 10,000 × g for 20 min, and the supernatant was submitted to further purification. The supernatant was then poured on to a purification column and allowed to bind for 1 h with gentle shaking. The recombinant His-tagged proteins were purified from the supernatant obtained above by immobilized metal affinity chromatography (IMAC) on Ni-NTA affinity resin (Bio-Rad) following the conventional protocol [18]. Finally, the proteins were collected and analyzed by SDS-PAGE to assess the level of homogeneity.\n500 ml induced bacterial culture was harvested after 4 h, centrifuged at 6000 × g for 10 min and the cell pellet was suspended in 20 mM Tris buffer (pH = 8.0). The cells were later lysed by using lysozyme (0.1 mg/mL) at 4°C for 1 h and sonicated on ice for 5 min at an amplitude of 30% with a 30s pulse frequency. The lysate was centrifuged at 10,000 × g for 20 min at 4°C and the supernatant was collected as soluble fraction. The resulting pellet was washed twice with 10 mL 2M urea containing 50 mM Tris buffer (pH = 8.0), 1 mM EDTA, 150 mM NaCl and 0.1% Triton X-100.\nThe suspension was centrifuged at 10,000 × g for 20 min at 4°C and then the resulting subsidence was resuspended in regeneration buffer containing 6 M urea, 0.5 M NaCl, 20 mM Tris-HCl (pH 7.9) and incubated at room temperature for 30 min. The incubated mixtures were then centrifuged at 10,000 × g for 20 min, and the supernatant was submitted to further purification. The supernatant was then poured on to a purification column and allowed to bind for 1 h with gentle shaking. The recombinant His-tagged proteins were purified from the supernatant obtained above by immobilized metal affinity chromatography (IMAC) on Ni-NTA affinity resin (Bio-Rad) following the conventional protocol [18]. Finally, the proteins were collected and analyzed by SDS-PAGE to assess the level of homogeneity.\n[SUBTITLE] Western blot [SUBSECTION] The recombinant proteins were electrophoresed with SDS-PAGE using 12% polyacrylamide gel and then electroblotted onto polyvinylidene fluoride (PVDF) membrane. The PVDF membrane was blocked with 3% bovine serum albumin (BSA) in Tris-Buffered Saline Tween-20 (TBST) Buffer containing 20 mM Tris-HCl,150 mM NaCl and 0.05% Tween 20 overnight at 4°C. The membrane was washed three times with TBST and incubated for 1 h at 37°C with rabbit anti-DEV polyclonal antibody and rabbit pre-serum at 1:100 of dilution in TBST buffer containing 0.5% BSA. After washing, the membrane was incubated for 1 h at 37°C with anti-rabbit-HRPO conjugated IgG(GE Healthcare Limited, Buckinghamshire, United Kingdom) at 1:5000 of dilution in TBST buffer containing 0.5% BSA. Finally, the membrane was washed with TBST and placed in diamino benzidine (DAB) solution as a chromogen to visualize the binding.\nThe recombinant proteins were electrophoresed with SDS-PAGE using 12% polyacrylamide gel and then electroblotted onto polyvinylidene fluoride (PVDF) membrane. The PVDF membrane was blocked with 3% bovine serum albumin (BSA) in Tris-Buffered Saline Tween-20 (TBST) Buffer containing 20 mM Tris-HCl,150 mM NaCl and 0.05% Tween 20 overnight at 4°C. The membrane was washed three times with TBST and incubated for 1 h at 37°C with rabbit anti-DEV polyclonal antibody and rabbit pre-serum at 1:100 of dilution in TBST buffer containing 0.5% BSA. After washing, the membrane was incubated for 1 h at 37°C with anti-rabbit-HRPO conjugated IgG(GE Healthcare Limited, Buckinghamshire, United Kingdom) at 1:5000 of dilution in TBST buffer containing 0.5% BSA. Finally, the membrane was washed with TBST and placed in diamino benzidine (DAB) solution as a chromogen to visualize the binding.\n[SUBTITLE] Indirect ELISA [SUBSECTION] Flat bottomed 96 well plate (Corning, Corning Costar Corp., MA, USA) was coated overnight at 4°C with the recombinant proteins at 7 ug/ml in carbonate bicarbonate buffer, pH 9.6. After blocking with 1% BSA in PBS, the wells were washed with PBS containing 0.05% (v/v) Tween-20 (PBST) and later incubated at 37°C for 1 h with various dilutions (1:5120-1:10) of duck anti-DEV strain Chv positive serum and negative serum. After washing with PBST, the plates were incubated at 37°C for 1 h with anti-duck-HRPO conjugated IgG(GE Healthcare Limited, Buckinghamshire, United Kingdom). Finally, the plate was washed and the peroxidase reaction was visualized by using tetrame-thylbenzidine(TMB) (Sigma) as substrate after incubation for 10 min at room temperature. The reaction was stopped by adding 2 M H2SO4 and absorbance was read at 450 nm by a microplate autoreader (Bio-Rad).\nFlat bottomed 96 well plate (Corning, Corning Costar Corp., MA, USA) was coated overnight at 4°C with the recombinant proteins at 7 ug/ml in carbonate bicarbonate buffer, pH 9.6. After blocking with 1% BSA in PBS, the wells were washed with PBS containing 0.05% (v/v) Tween-20 (PBST) and later incubated at 37°C for 1 h with various dilutions (1:5120-1:10) of duck anti-DEV strain Chv positive serum and negative serum. After washing with PBST, the plates were incubated at 37°C for 1 h with anti-duck-HRPO conjugated IgG(GE Healthcare Limited, Buckinghamshire, United Kingdom). Finally, the plate was washed and the peroxidase reaction was visualized by using tetrame-thylbenzidine(TMB) (Sigma) as substrate after incubation for 10 min at room temperature. The reaction was stopped by adding 2 M H2SO4 and absorbance was read at 450 nm by a microplate autoreader (Bio-Rad).\n[SUBTITLE] Production of polyclonal antibody [SUBSECTION] Preimmune serum was collected prior to immunization. New Zealand white rabbits were immunized firstly intradermally with a mixture of 1 mg purified recombinant rVP19c or rVP19c(N) mixed with an equal volume of complete Freund's adjuvant (Sigma). Two weeks later, the rabbits were boosted twice subcutaneously with the same amount of recombinant proteins mixed with an equal volume of incomplete Freund's adjuvant at a one-week interval. Two weeks after the last immunization, the two antiserums were harvested from the carotid artery. Then the polyclonal antibody was purified by protein A affinity IgG purification kit according the user's guide. The titer of the specific antibodies was determined by ELISA.\nPreimmune serum was collected prior to immunization. New Zealand white rabbits were immunized firstly intradermally with a mixture of 1 mg purified recombinant rVP19c or rVP19c(N) mixed with an equal volume of complete Freund's adjuvant (Sigma). Two weeks later, the rabbits were boosted twice subcutaneously with the same amount of recombinant proteins mixed with an equal volume of incomplete Freund's adjuvant at a one-week interval. Two weeks after the last immunization, the two antiserums were harvested from the carotid artery. Then the polyclonal antibody was purified by protein A affinity IgG purification kit according the user's guide. The titer of the specific antibodies was determined by ELISA.\n[SUBTITLE] Indirect immunofluorescent assay [SUBSECTION] For immunofluorescence assays, monolayers of duck embryo fibroblast (DEF) cells were infected (MOI 5) with the DEV strain Chv and then incubated for 30 h at 37°C. Cells were fixed with 4% paraformaldehyde for 30 min, washed with PBS and permeabilized with 0.5% Triton X-100 for 10 min. After washing, the cells were blocked with PBS containing 5% BSA for 1.5 h at 37°C. Subsequently, the cells were incubated with rVP19c-specific polyclonal antibody diluted in PBS containing 0.5% BSA for 1 h at 37°C. Finally, fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit immunoglobulin G (Sigma) was added at a dilution of 1:200 and incubated at 4°C for 16 h. After each incubation step, the cells were washed extensively with PBS. The cell nuclei were visualized by DAPI counterstaining [19-21]. The images captured with fluorescence microscopy (Nikon, Japan).\nFor immunofluorescence assays, monolayers of duck embryo fibroblast (DEF) cells were infected (MOI 5) with the DEV strain Chv and then incubated for 30 h at 37°C. Cells were fixed with 4% paraformaldehyde for 30 min, washed with PBS and permeabilized with 0.5% Triton X-100 for 10 min. After washing, the cells were blocked with PBS containing 5% BSA for 1.5 h at 37°C. Subsequently, the cells were incubated with rVP19c-specific polyclonal antibody diluted in PBS containing 0.5% BSA for 1 h at 37°C. Finally, fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit immunoglobulin G (Sigma) was added at a dilution of 1:200 and incubated at 4°C for 16 h. After each incubation step, the cells were washed extensively with PBS. The cell nuclei were visualized by DAPI counterstaining [19-21]. The images captured with fluorescence microscopy (Nikon, Japan).", "The viral DNA was extracted from partially purified DEV and used as template for polymerase chain reaction (PCR) [15]. A fragment (1515 bp) containing the complete ORF of DEV UL38 gene (1398 bp) (Genbank accession no. EU071041) was amplified using PCR with the following primers: the upstream primer (P1) 5'-GGATCCACGATGAAAGTACCAAATG-3' containing the BamHI site and the downstream primer (P2)5'-CTCGAGAGCCAGATACGAC AAGAAG-3' containing the XhoI site. The PCR parameters were 10 min at 95°C and 30 cycles of 1 min at 94°C, 1 min at 58.5°C, 1 min at 72°C, and a final extension time of 10 min at 72°C. A fragment (723 bp) containing the partial ORF (711 bp) encoding the N-terminal of VP19c (VP19c(N)) protein was amplified with the following primers: the upstream primer (P1) 5'-GGATCCACGATGAAAGTACCAAATG-3' containing the BamHI siteand the downstream primer (P3) 5' -CTCGAGGTGTACGTGACATCTAACCA TAG-3' containing the XhoI site. The PCR assay and programme were carried out following the protocol described above.\nThe PCR products and plasmid pET-32a(+) (Qiagen GmbH, Hilden, Germany) were both digested with BamH I and Xho I, and then ligated with T4 DNA ligase to yield the constructs. The constructs were transformed into E.coli, and the selected bacterial transformants were verified by colony PCR, restriction enzyme analysis and sequencing.", "The positive individual clone was cultured in 5 mL Luria bertani (LB) medium containing 100 ug/mL amp and then induced at 37°C by adding isopropyl-β-D-thiogalactoside (IPTG) at a final concentration of 1 mM for 4 h. For IPTG dose optimization, the bacterial culture was induced with different concentrations of IPTG [0.2, 0.3, 0.4, 0.5, 0.6,0.7,0.8,0.9,1.0 (mM)] and allowed to grow for 4h at 37°C. For temperature optimization, the bacterial culture was induced with IPTG [1.0 (mM)] and allowed to grow for 4 h at three different temperatures (25, 30 and 37°C). For time optimization, the bacterial culture was induced with IPTG [1.0 (mM)] and allowed to grow for 3 h, 4 h, 5 h and overnight (~16 h) at 37°C. Total cell proteins from each optimization experiment were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Then, small-scale expression was done by optimized conditions as described above to prepare for purification [16]. The protein amount was determined with reference to standard bovine serum albumin (BSA) in the Bradford assay [17].", "500 ml induced bacterial culture was harvested after 4 h, centrifuged at 6000 × g for 10 min and the cell pellet was suspended in 20 mM Tris buffer (pH = 8.0). The cells were later lysed by using lysozyme (0.1 mg/mL) at 4°C for 1 h and sonicated on ice for 5 min at an amplitude of 30% with a 30s pulse frequency. The lysate was centrifuged at 10,000 × g for 20 min at 4°C and the supernatant was collected as soluble fraction. The resulting pellet was washed twice with 10 mL 2M urea containing 50 mM Tris buffer (pH = 8.0), 1 mM EDTA, 150 mM NaCl and 0.1% Triton X-100.\nThe suspension was centrifuged at 10,000 × g for 20 min at 4°C and then the resulting subsidence was resuspended in regeneration buffer containing 6 M urea, 0.5 M NaCl, 20 mM Tris-HCl (pH 7.9) and incubated at room temperature for 30 min. The incubated mixtures were then centrifuged at 10,000 × g for 20 min, and the supernatant was submitted to further purification. The supernatant was then poured on to a purification column and allowed to bind for 1 h with gentle shaking. The recombinant His-tagged proteins were purified from the supernatant obtained above by immobilized metal affinity chromatography (IMAC) on Ni-NTA affinity resin (Bio-Rad) following the conventional protocol [18]. Finally, the proteins were collected and analyzed by SDS-PAGE to assess the level of homogeneity.", "The recombinant proteins were electrophoresed with SDS-PAGE using 12% polyacrylamide gel and then electroblotted onto polyvinylidene fluoride (PVDF) membrane. The PVDF membrane was blocked with 3% bovine serum albumin (BSA) in Tris-Buffered Saline Tween-20 (TBST) Buffer containing 20 mM Tris-HCl,150 mM NaCl and 0.05% Tween 20 overnight at 4°C. The membrane was washed three times with TBST and incubated for 1 h at 37°C with rabbit anti-DEV polyclonal antibody and rabbit pre-serum at 1:100 of dilution in TBST buffer containing 0.5% BSA. After washing, the membrane was incubated for 1 h at 37°C with anti-rabbit-HRPO conjugated IgG(GE Healthcare Limited, Buckinghamshire, United Kingdom) at 1:5000 of dilution in TBST buffer containing 0.5% BSA. Finally, the membrane was washed with TBST and placed in diamino benzidine (DAB) solution as a chromogen to visualize the binding.", "Flat bottomed 96 well plate (Corning, Corning Costar Corp., MA, USA) was coated overnight at 4°C with the recombinant proteins at 7 ug/ml in carbonate bicarbonate buffer, pH 9.6. After blocking with 1% BSA in PBS, the wells were washed with PBS containing 0.05% (v/v) Tween-20 (PBST) and later incubated at 37°C for 1 h with various dilutions (1:5120-1:10) of duck anti-DEV strain Chv positive serum and negative serum. After washing with PBST, the plates were incubated at 37°C for 1 h with anti-duck-HRPO conjugated IgG(GE Healthcare Limited, Buckinghamshire, United Kingdom). Finally, the plate was washed and the peroxidase reaction was visualized by using tetrame-thylbenzidine(TMB) (Sigma) as substrate after incubation for 10 min at room temperature. The reaction was stopped by adding 2 M H2SO4 and absorbance was read at 450 nm by a microplate autoreader (Bio-Rad).", "Preimmune serum was collected prior to immunization. New Zealand white rabbits were immunized firstly intradermally with a mixture of 1 mg purified recombinant rVP19c or rVP19c(N) mixed with an equal volume of complete Freund's adjuvant (Sigma). Two weeks later, the rabbits were boosted twice subcutaneously with the same amount of recombinant proteins mixed with an equal volume of incomplete Freund's adjuvant at a one-week interval. Two weeks after the last immunization, the two antiserums were harvested from the carotid artery. Then the polyclonal antibody was purified by protein A affinity IgG purification kit according the user's guide. The titer of the specific antibodies was determined by ELISA.", "For immunofluorescence assays, monolayers of duck embryo fibroblast (DEF) cells were infected (MOI 5) with the DEV strain Chv and then incubated for 30 h at 37°C. Cells were fixed with 4% paraformaldehyde for 30 min, washed with PBS and permeabilized with 0.5% Triton X-100 for 10 min. After washing, the cells were blocked with PBS containing 5% BSA for 1.5 h at 37°C. Subsequently, the cells were incubated with rVP19c-specific polyclonal antibody diluted in PBS containing 0.5% BSA for 1 h at 37°C. Finally, fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit immunoglobulin G (Sigma) was added at a dilution of 1:200 and incubated at 4°C for 16 h. After each incubation step, the cells were washed extensively with PBS. The cell nuclei were visualized by DAPI counterstaining [19-21]. The images captured with fluorescence microscopy (Nikon, Japan).", "The authors declare that they have no competing interests.", "JX and SCZ carried out most of the experiments and drafted the manuscript. ACC and MSW have critically revised the manuscript and the experimental design. HC, CJS, DKZ, RYJ, QHL, ZLC and XYC helped in experiments. All authors have read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, "methods", null, null, null, null, null, null, null, null, null ]

[]

[ "Adult", "Brain", "Cluster Headache", "Electric Stimulation Therapy", "Electrodes, Implanted", "Female", "Fluorodeoxyglucose F18", "Glucose", "Humans", "Male", "Middle Aged", "Pain", "Pain Management", "Positron-Emission Tomography", "Spinal Nerves" ]

[ "Background", "Subjects", "Surgical and stimulation procedure", "18-FDG-PET study design", "Study groups", "Data acquisition", "Statistical analysis (Cyclotron Research Centre, Liège)", "Results", "Clinical outcome", "PET results", "Discussion", "PET findings - interpretation", "Metabolic pattern in drCCH compared to HV", "Short-term changes associated with ONS", "Long-term changes associated with ONS", "Perigenual ACC (PACC) activation in responders", "Study limitations", "Conclusions", "Abbreviations used in the text", "Competing interests", "Authors' contributions", "Author's information", "Pre-publication history" ]

[ "Cluster headache (CH) is one of the most painful primary headaches and is characterized by attacks of severe unilateral periorbital pain associated with ipsilateral autonomic features [1]. About 10% of patients have, or develop over time, a chronic form (CCH) [2] characterised by recurrent attacks for at least 1 year without remissions or with remissions of less than 1 month [1]. About 1% of CCH patients become drug-resistant (drCCH) to most prophylactic drug treatments and fulfil published criteria for intractable headaches [3].\nCH is the most prevalent member of the so-called trigeminal autonomic cephalalgias (TACs), which include paroxysmal hemicrania, SUNCT (Short-lasting Unilateral Neuralgiform headache with Conjunctival injection and Tearing) and probably hemicrania continua [4]. Neuroimaging studies have provided new insight into the pathophysiology of these disorders. Besides non-specific changes in activity of brain areas belonging to the pain matrix like the anterior cingular cortex (ACC), insula(e), and thalamus, TACs are associated with ictal activation of ipsilateral posterior hypothalamus (CH, SUNCT) or dorsal pons (hemicrania continua) or contralateral posterior hypothalamus (paroxysmal hemicrania) which may be more specific and disease-related [5].\nAs a consequence, deep brain stimulation (DBS) targeting the posterior hypothalamus was proposed for drCCH and was found to be more effective than any previously used invasive therapy [6,7]. However hypothalamic DBS is not a riskless procedure [7] and less invasive methods were thus explored. Among them, occipital nerve stimulation (ONS) had comparable efficacy to hypothalamic DBS, except for slower onset of action [8,9].\nThe mechanisms by which ONS improves drCCH remain unclear. In a study of ONS in drCCH we found no significant change in pain thresholds, which argues against a diffuse analgesic effect [8]. It was speculated that ONS might exert its action by decreasing excitability of second order nociceptors in trigeminal nucleus caudalis on which converge cervical, somatic trigeminal and visceral trigeminovascular afferents [10,11]. Yet, the nociception-specific blink reflex, mediated by spinal trigeminal nucleus interneurons, was increased rather than decreased in our study of ONS in drCCH [8] and it remained unchanged in healthy subjects after short low frequency transcutaneous stimulation of the greater occipital nerve [12]. A more likely explanation for the therapeutic effect of ONS in headache including drCCH is the induction of slow neuromodulatory changes in brain regions belonging to the pain matrix or in centres more specifically involved in CCH pathophysiology. Hence, chronic migraine patients treated with ONS [13] show significant blood flow increases on H215O-PET in dorsal rostral pons, anterior cingulate cortex and cuneus, directly correlated to pain scores, and in left pulvinar, inversely correlated to such scores. Dorsal pons activation persisted after ONS, supporting the role of this structure in migraine pathophysiology [13].\nSo far, functional imaging studies have not been performed in ONS-treated drCCH patients. We performed such a study focusing on the pain matrix, but also on hypothalamus and brainstem that seem more specifically involved in the pathophysiology of TACs. We enrolled patients from the published cohort [8] and newly implanted ones. We used 18-fluorodeoxyglucose-positron emission tomography (18-FDG-PET) in order to detect long term activity modulation.", "We studied 10 patients with drCCH (9 males and 1 female, mean age at implantation 44.2 ± 9.9 years SD). Inclusion criteria were: CCH for at least 2 years, daily attacks by history, side-locked attacks from the beginning, resistance to drug treatment according to expert consensus guidelines [3] and absence of associated disabling organic or psychiatric disorder. Five patients had left-sided, 5 right-sided attacks. At the time of the study, all patients were taking one or several of the following preventive drugs: verapamil (n = 9), lithium carbonate (n = 6), methylprednisolone (n = 2), methysergide (n = 2), melatonine (n = 1), gabapentine (n = 1). None took analgesics, in particular opioids.\nPatients were recruited in two phases (1st and 2nd group), with written informed consent. Approval of the local Ethics Committee for ONS in drCCH was first obtained for 5 patients (EUDRACT-2004-004551-19). Because of the favourable results in ONS-treated patients after a 16 months follow-up, we requested Ethics Committee approval for a protocol amendment allowing us to recruit 6 supplementary patients who were implanted and agreed to undergo PET before and after surgery. At the same time, patients of the 1st group were also asked to participate in the PET study and 4 of them accepted (the last patient of group 1 had been explanted [8]).", "Surgical procedure and stimulation protocols have been described previously [8]. We used unilateral subcutaneous implantation of paddle style stimulating leads with 4 electrode plots (Medtronic 3587A Resume II®; Medtronic Inc., Minneapolis, USA) via a retromastoid C1-2-3 approach [14] and Medtronic Itrel III® or Synergy® stimulators.\nStimulation protocols were adapted using a programming matrix [8] such as to induce paraesthesias in the largest possible occipital territory. The clinical evolution of patients was monitored with cluster headache paper diaries.", "[SUBTITLE] Study groups [SUBSECTION] In the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.\nIn the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.\n[SUBTITLE] Data acquisition [SUBSECTION] PET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.\nPET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.\n[SUBTITLE] Statistical analysis (Cyclotron Research Centre, Liège) [SUBSECTION] Because we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).\nBecause we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).", "In the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.", "PET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.", "Because we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).", "[SUBTITLE] Clinical outcome [SUBSECTION] Clinical data of patients and changes in their attack frequency after various durations of ONS are summarized in table 1. All patients but one in the 1st group (N = 4, 24-30 months follow-up) improved after ONS: one patient was pain free; 2 patients had a 90% and 93% reduction in attack frequency; the \"non-responder\" patient had a 25% improvement. In the 2nd group (N = 6), at 6 months follow-up, 3 patients were pain free and one was improved by 90%; these 4 patients already reported significant improvement after 1 month of ONS. The 5th patient only had a 33% reduction in attack frequency while in the 6th patient, attack frequency was slightly increased. According to the 50% cut-off criterion, 7 patients were thus considered responders and 3 non-responders to ONS for the binary analysis. During the 3-day period of ONS interruption, only one responder had recurrence of attacks. In the 2nd group, all patients kept the same preventive drug treatment during the follow-up scans except for one responder (8) who was able stop all medications after 6 months.\nPatients characteristics and clinical outcome\nONS: occipital nerve stimulation, R: right, L: left, Y: yes, N: no.\nClinical data of patients and changes in their attack frequency after various durations of ONS are summarized in table 1. All patients but one in the 1st group (N = 4, 24-30 months follow-up) improved after ONS: one patient was pain free; 2 patients had a 90% and 93% reduction in attack frequency; the \"non-responder\" patient had a 25% improvement. In the 2nd group (N = 6), at 6 months follow-up, 3 patients were pain free and one was improved by 90%; these 4 patients already reported significant improvement after 1 month of ONS. The 5th patient only had a 33% reduction in attack frequency while in the 6th patient, attack frequency was slightly increased. According to the 50% cut-off criterion, 7 patients were thus considered responders and 3 non-responders to ONS for the binary analysis. During the 3-day period of ONS interruption, only one responder had recurrence of attacks. In the 2nd group, all patients kept the same preventive drug treatment during the follow-up scans except for one responder (8) who was able stop all medications after 6 months.\nPatients characteristics and clinical outcome\nONS: occipital nerve stimulation, R: right, L: left, Y: yes, N: no.\n[SUBTITLE] PET results [SUBSECTION] The main areas of peak voxels where a metabolic change was found for the various comparisons are shown in table 2.\nMain statistical results and localization of peak voxels where cerebral metabolism was activated (>) or deactivated (<)\nCoordinates are in the standardized stereotactic Montreal Neurological Institute space (mm). FDR = False discovery rate corrected. drCCH: drug-resistant chronic cluster headache; HV: healthy volunteers; ACC: anterior cingulate cortex, R: right, L: left. * FDR corrected in region of interest identified in whole group analysis\nWe first pooled all scans performed in drCCH patients and compared them with those of the HV. In comparison to HV, a significant hypermetabolism was found in anterior cingulate cortex (ACC), left hypothalamus, left pulvinar, left visual cortex, cerebellum and brain stem (left lower pons and midbrain) (figure 1). By contrast, a significant hypometabolism appeared in both sensori-motor areas.\nHypermetabolic areas in drCCH patients (all conditions: baseline - 1 month, 6 months, 24/30 months) compared with HV (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (through midline and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, MB: midbrain, LP: lower pons, VC: visual cortex, P: pulvinar, H: hypothalamus.\nThere was no significant difference between scans performed with stimulator ON or OFF, regardless of ONS duration (1, 6 or 24-30 months). For scans obtained in the late phase (≥ 6 months) we observed a hypermetabolism in the left frontal lobe (BA 10, uncorrected p = 0.03, x = -12, y = 46, z = 4) and the left lower brainstem (uncorrected p = 0.014, x = -6, y = -30, z = -36) when the stimulator was turned ON, but these results, reported for the sake of completeness, did not survive correction for multiple comparisons.\nOver time, ONS changed glucose uptake in several brain areas (independent of the stimulator settings ON or OFF). The anterior cingulate, mid cingulate, left pulvinar, midbrain, lower pons, visual cortex and cerebellum had decreased metabolism over time, i.e. they became less hypermetabolic when comparing baseline to the early phase (1 month) or the late phase (≥ 6 months). By contrast, metabolism increased over time in sensorimotor cortices, i.e. they became less hypometabolic, and it was not modified in the left hypothalamus (figure 2).\nAreas progressively deactivated by ONS over time (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (right and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, B: brainstem, VC: visual cortex, P: pulvinar. White circle highlights hypothalamic area, which is not modified by the stimulation.\nWhen finally comparing responders and non-responders in the late phase, there was a significant hypermetabolism in previously identified perigenual ACC ipsilateral to the pain and stimulation side in responders (figure 3). No hypometabolic area differentiated responders from non-responders.\nActivation of perigenual cortex in ONS responders vs. non responders after 6 to 30 months stimulation (p < 0.05 FDR corrected). Result is displayed on a left sagittal section of a normalized MRI template.\nThe main areas of peak voxels where a metabolic change was found for the various comparisons are shown in table 2.\nMain statistical results and localization of peak voxels where cerebral metabolism was activated (>) or deactivated (<)\nCoordinates are in the standardized stereotactic Montreal Neurological Institute space (mm). FDR = False discovery rate corrected. drCCH: drug-resistant chronic cluster headache; HV: healthy volunteers; ACC: anterior cingulate cortex, R: right, L: left. * FDR corrected in region of interest identified in whole group analysis\nWe first pooled all scans performed in drCCH patients and compared them with those of the HV. In comparison to HV, a significant hypermetabolism was found in anterior cingulate cortex (ACC), left hypothalamus, left pulvinar, left visual cortex, cerebellum and brain stem (left lower pons and midbrain) (figure 1). By contrast, a significant hypometabolism appeared in both sensori-motor areas.\nHypermetabolic areas in drCCH patients (all conditions: baseline - 1 month, 6 months, 24/30 months) compared with HV (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (through midline and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, MB: midbrain, LP: lower pons, VC: visual cortex, P: pulvinar, H: hypothalamus.\nThere was no significant difference between scans performed with stimulator ON or OFF, regardless of ONS duration (1, 6 or 24-30 months). For scans obtained in the late phase (≥ 6 months) we observed a hypermetabolism in the left frontal lobe (BA 10, uncorrected p = 0.03, x = -12, y = 46, z = 4) and the left lower brainstem (uncorrected p = 0.014, x = -6, y = -30, z = -36) when the stimulator was turned ON, but these results, reported for the sake of completeness, did not survive correction for multiple comparisons.\nOver time, ONS changed glucose uptake in several brain areas (independent of the stimulator settings ON or OFF). The anterior cingulate, mid cingulate, left pulvinar, midbrain, lower pons, visual cortex and cerebellum had decreased metabolism over time, i.e. they became less hypermetabolic when comparing baseline to the early phase (1 month) or the late phase (≥ 6 months). By contrast, metabolism increased over time in sensorimotor cortices, i.e. they became less hypometabolic, and it was not modified in the left hypothalamus (figure 2).\nAreas progressively deactivated by ONS over time (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (right and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, B: brainstem, VC: visual cortex, P: pulvinar. White circle highlights hypothalamic area, which is not modified by the stimulation.\nWhen finally comparing responders and non-responders in the late phase, there was a significant hypermetabolism in previously identified perigenual ACC ipsilateral to the pain and stimulation side in responders (figure 3). No hypometabolic area differentiated responders from non-responders.\nActivation of perigenual cortex in ONS responders vs. non responders after 6 to 30 months stimulation (p < 0.05 FDR corrected). Result is displayed on a left sagittal section of a normalized MRI template.", "Clinical data of patients and changes in their attack frequency after various durations of ONS are summarized in table 1. All patients but one in the 1st group (N = 4, 24-30 months follow-up) improved after ONS: one patient was pain free; 2 patients had a 90% and 93% reduction in attack frequency; the \"non-responder\" patient had a 25% improvement. In the 2nd group (N = 6), at 6 months follow-up, 3 patients were pain free and one was improved by 90%; these 4 patients already reported significant improvement after 1 month of ONS. The 5th patient only had a 33% reduction in attack frequency while in the 6th patient, attack frequency was slightly increased. According to the 50% cut-off criterion, 7 patients were thus considered responders and 3 non-responders to ONS for the binary analysis. During the 3-day period of ONS interruption, only one responder had recurrence of attacks. In the 2nd group, all patients kept the same preventive drug treatment during the follow-up scans except for one responder (8) who was able stop all medications after 6 months.\nPatients characteristics and clinical outcome\nONS: occipital nerve stimulation, R: right, L: left, Y: yes, N: no.", "The main areas of peak voxels where a metabolic change was found for the various comparisons are shown in table 2.\nMain statistical results and localization of peak voxels where cerebral metabolism was activated (>) or deactivated (<)\nCoordinates are in the standardized stereotactic Montreal Neurological Institute space (mm). FDR = False discovery rate corrected. drCCH: drug-resistant chronic cluster headache; HV: healthy volunteers; ACC: anterior cingulate cortex, R: right, L: left. * FDR corrected in region of interest identified in whole group analysis\nWe first pooled all scans performed in drCCH patients and compared them with those of the HV. In comparison to HV, a significant hypermetabolism was found in anterior cingulate cortex (ACC), left hypothalamus, left pulvinar, left visual cortex, cerebellum and brain stem (left lower pons and midbrain) (figure 1). By contrast, a significant hypometabolism appeared in both sensori-motor areas.\nHypermetabolic areas in drCCH patients (all conditions: baseline - 1 month, 6 months, 24/30 months) compared with HV (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (through midline and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, MB: midbrain, LP: lower pons, VC: visual cortex, P: pulvinar, H: hypothalamus.\nThere was no significant difference between scans performed with stimulator ON or OFF, regardless of ONS duration (1, 6 or 24-30 months). For scans obtained in the late phase (≥ 6 months) we observed a hypermetabolism in the left frontal lobe (BA 10, uncorrected p = 0.03, x = -12, y = 46, z = 4) and the left lower brainstem (uncorrected p = 0.014, x = -6, y = -30, z = -36) when the stimulator was turned ON, but these results, reported for the sake of completeness, did not survive correction for multiple comparisons.\nOver time, ONS changed glucose uptake in several brain areas (independent of the stimulator settings ON or OFF). The anterior cingulate, mid cingulate, left pulvinar, midbrain, lower pons, visual cortex and cerebellum had decreased metabolism over time, i.e. they became less hypermetabolic when comparing baseline to the early phase (1 month) or the late phase (≥ 6 months). By contrast, metabolism increased over time in sensorimotor cortices, i.e. they became less hypometabolic, and it was not modified in the left hypothalamus (figure 2).\nAreas progressively deactivated by ONS over time (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (right and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, B: brainstem, VC: visual cortex, P: pulvinar. White circle highlights hypothalamic area, which is not modified by the stimulation.\nWhen finally comparing responders and non-responders in the late phase, there was a significant hypermetabolism in previously identified perigenual ACC ipsilateral to the pain and stimulation side in responders (figure 3). No hypometabolic area differentiated responders from non-responders.\nActivation of perigenual cortex in ONS responders vs. non responders after 6 to 30 months stimulation (p < 0.05 FDR corrected). Result is displayed on a left sagittal section of a normalized MRI template.", "The therapeutic outcome after ONS in 1st and 2nd groups was overall similar, with good response as defined above in nearly 70% of patients [8,17]. A noticeable difference between the 2 groups was the latency to significant efficacy, which was on average shortened to 1 month in the 2nd group compared to 3 months in the 1st one. This remains compatible with a slow modulatory effect on the central nervous system. We ascribe faster efficacy in group 2 to a learning effect for investigators and to quicker optimisation of stimulator settings [8].\nWe will focus our discussion on the FDG-PET findings.\n[SUBTITLE] PET findings - interpretation [SUBSECTION] [SUBTITLE] Metabolic pattern in drCCH compared to HV [SUBSECTION] The enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\nThe enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\n[SUBTITLE] Short-term changes associated with ONS [SUBSECTION] We found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\nWe found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\n[SUBTITLE] Long-term changes associated with ONS [SUBSECTION] Hypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\nHypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\n[SUBTITLE] Perigenual ACC (PACC) activation in responders [SUBSECTION] Comparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\nComparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\n[SUBTITLE] Study limitations [SUBSECTION] We are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.\nWe are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.\n[SUBTITLE] Metabolic pattern in drCCH compared to HV [SUBSECTION] The enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\nThe enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\n[SUBTITLE] Short-term changes associated with ONS [SUBSECTION] We found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\nWe found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\n[SUBTITLE] Long-term changes associated with ONS [SUBSECTION] Hypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\nHypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\n[SUBTITLE] Perigenual ACC (PACC) activation in responders [SUBSECTION] Comparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\nComparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\n[SUBTITLE] Study limitations [SUBSECTION] We are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.\nWe are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.", "[SUBTITLE] Metabolic pattern in drCCH compared to HV [SUBSECTION] The enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\nThe enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\n[SUBTITLE] Short-term changes associated with ONS [SUBSECTION] We found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\nWe found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\n[SUBTITLE] Long-term changes associated with ONS [SUBSECTION] Hypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\nHypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\n[SUBTITLE] Perigenual ACC (PACC) activation in responders [SUBSECTION] Comparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\nComparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\n[SUBTITLE] Study limitations [SUBSECTION] We are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.\nWe are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.", "The enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).", "We found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].", "Hypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.", "Comparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.", "We are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.", "We confirm that ONS is effective and safe in drCCH, reducing attack frequency by ≥ 50% in more than 60% of patients, which is similar to the results obtained with hypothalamic DBS.\nThe FDG-PET results in our small sample appear consistent with the clinical impression that ONS exerts its beneficial effects via slow neuromodulatory processes in the central pain matrix. The finding of a possible selective perigenual ACC activation in responders compared to non-responders might advocate that ONS activates descending pain control systems in a top-down manner and restores an equilibrium in anti-nociceptive opioidergic pathways. We suggest for the first time that metabolic activity could be increased in ipsilateral posterior hypothalamus in chronic cluster headache patients outside of an attack. That ONS, as suspected on clinical grounds, does not cure drCCH, but merely acts as a symptomatic treatment is underlined by its inability to reduce this ipsilateral hypothalamic hyperactivity which is typically found during attacks in episodic cluster headache. Persistent hypothalamic activation might also explain why ONS-treated pain-free drCCH patients may still have autonomic attacks and why attacks rapidly recur after interruption of ONS.", "18-FDG-PET: 18-Fluorodeoxyglucose-positron emission tomography; ACC: anterior cingular cortex; BA: Brodmann area; DBS: deep brain stimulation; drCCH: drug-resistant chronic cluster headache; HV: healthy volunteers; OFF: stimulator switched off; ON: stimulator switched on; ONS: occipital nerve stimulation; PACC: perigenual anterior cingular cortex; PAG: periaqueductal grey; TACs: trigeminal autonomic cephalalgias; TENS: transcutaneous electrical nerve stimulation", "The authors declare that they have no competing interests.", "All authors read and approved the final manuscript. DM participated to the clinical follow-up of patients, interpreted the PET results and drafted the manuscript. MAB analyzed the PET data and wrote the methodology. AF designed the PET study protocol. PYG participated to the clinical follow-up of patients. RH is head of the University Nuclear Medicine Department where patients underwent the PET scans. SL analyzed the PET data and did the matrix design. JS participated to the clinical follow-up of patients, interpreted the PET results and drafted the manuscript with DM.", "DM, MD, is clinical chief associate at the University Neurology Department, Liège, Belgium.\nMAB, MSc, is research fellow at the National Fund for Scientific Research (FNRS), Belgium.\nAF, MD, PhD, is clinical chief at the University Neurology Department, Liège, Belgium.\nPYG, MD, is research fellow at the University Neurology Department, Liège, Belgium.\nRH, MD, PhD, is head of the University Nuclear Medicine Department, Liège, Belgium.\nSL, MD, PhD, is senior research associate at the FNRS, Belgium.\nJS, MD, PhD, is head of the Headache Research Unit at the University of Liège, Belgium and Professor of Neuroanatomy.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2377/11/25/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Subjects", "Surgical and stimulation procedure", "18-FDG-PET study design", "Study groups", "Data acquisition", "Statistical analysis (Cyclotron Research Centre, Liège)", "Results", "Clinical outcome", "PET results", "Discussion", "PET findings - interpretation", "Metabolic pattern in drCCH compared to HV", "Short-term changes associated with ONS", "Long-term changes associated with ONS", "Perigenual ACC (PACC) activation in responders", "Study limitations", "Conclusions", "Abbreviations used in the text", "Competing interests", "Authors' contributions", "Author's information", "Pre-publication history" ]

[ "Cluster headache (CH) is one of the most painful primary headaches and is characterized by attacks of severe unilateral periorbital pain associated with ipsilateral autonomic features [1]. About 10% of patients have, or develop over time, a chronic form (CCH) [2] characterised by recurrent attacks for at least 1 year without remissions or with remissions of less than 1 month [1]. About 1% of CCH patients become drug-resistant (drCCH) to most prophylactic drug treatments and fulfil published criteria for intractable headaches [3].\nCH is the most prevalent member of the so-called trigeminal autonomic cephalalgias (TACs), which include paroxysmal hemicrania, SUNCT (Short-lasting Unilateral Neuralgiform headache with Conjunctival injection and Tearing) and probably hemicrania continua [4]. Neuroimaging studies have provided new insight into the pathophysiology of these disorders. Besides non-specific changes in activity of brain areas belonging to the pain matrix like the anterior cingular cortex (ACC), insula(e), and thalamus, TACs are associated with ictal activation of ipsilateral posterior hypothalamus (CH, SUNCT) or dorsal pons (hemicrania continua) or contralateral posterior hypothalamus (paroxysmal hemicrania) which may be more specific and disease-related [5].\nAs a consequence, deep brain stimulation (DBS) targeting the posterior hypothalamus was proposed for drCCH and was found to be more effective than any previously used invasive therapy [6,7]. However hypothalamic DBS is not a riskless procedure [7] and less invasive methods were thus explored. Among them, occipital nerve stimulation (ONS) had comparable efficacy to hypothalamic DBS, except for slower onset of action [8,9].\nThe mechanisms by which ONS improves drCCH remain unclear. In a study of ONS in drCCH we found no significant change in pain thresholds, which argues against a diffuse analgesic effect [8]. It was speculated that ONS might exert its action by decreasing excitability of second order nociceptors in trigeminal nucleus caudalis on which converge cervical, somatic trigeminal and visceral trigeminovascular afferents [10,11]. Yet, the nociception-specific blink reflex, mediated by spinal trigeminal nucleus interneurons, was increased rather than decreased in our study of ONS in drCCH [8] and it remained unchanged in healthy subjects after short low frequency transcutaneous stimulation of the greater occipital nerve [12]. A more likely explanation for the therapeutic effect of ONS in headache including drCCH is the induction of slow neuromodulatory changes in brain regions belonging to the pain matrix or in centres more specifically involved in CCH pathophysiology. Hence, chronic migraine patients treated with ONS [13] show significant blood flow increases on H215O-PET in dorsal rostral pons, anterior cingulate cortex and cuneus, directly correlated to pain scores, and in left pulvinar, inversely correlated to such scores. Dorsal pons activation persisted after ONS, supporting the role of this structure in migraine pathophysiology [13].\nSo far, functional imaging studies have not been performed in ONS-treated drCCH patients. We performed such a study focusing on the pain matrix, but also on hypothalamus and brainstem that seem more specifically involved in the pathophysiology of TACs. We enrolled patients from the published cohort [8] and newly implanted ones. We used 18-fluorodeoxyglucose-positron emission tomography (18-FDG-PET) in order to detect long term activity modulation.", "[SUBTITLE] Subjects [SUBSECTION] We studied 10 patients with drCCH (9 males and 1 female, mean age at implantation 44.2 ± 9.9 years SD). Inclusion criteria were: CCH for at least 2 years, daily attacks by history, side-locked attacks from the beginning, resistance to drug treatment according to expert consensus guidelines [3] and absence of associated disabling organic or psychiatric disorder. Five patients had left-sided, 5 right-sided attacks. At the time of the study, all patients were taking one or several of the following preventive drugs: verapamil (n = 9), lithium carbonate (n = 6), methylprednisolone (n = 2), methysergide (n = 2), melatonine (n = 1), gabapentine (n = 1). None took analgesics, in particular opioids.\nPatients were recruited in two phases (1st and 2nd group), with written informed consent. Approval of the local Ethics Committee for ONS in drCCH was first obtained for 5 patients (EUDRACT-2004-004551-19). Because of the favourable results in ONS-treated patients after a 16 months follow-up, we requested Ethics Committee approval for a protocol amendment allowing us to recruit 6 supplementary patients who were implanted and agreed to undergo PET before and after surgery. At the same time, patients of the 1st group were also asked to participate in the PET study and 4 of them accepted (the last patient of group 1 had been explanted [8]).\nWe studied 10 patients with drCCH (9 males and 1 female, mean age at implantation 44.2 ± 9.9 years SD). Inclusion criteria were: CCH for at least 2 years, daily attacks by history, side-locked attacks from the beginning, resistance to drug treatment according to expert consensus guidelines [3] and absence of associated disabling organic or psychiatric disorder. Five patients had left-sided, 5 right-sided attacks. At the time of the study, all patients were taking one or several of the following preventive drugs: verapamil (n = 9), lithium carbonate (n = 6), methylprednisolone (n = 2), methysergide (n = 2), melatonine (n = 1), gabapentine (n = 1). None took analgesics, in particular opioids.\nPatients were recruited in two phases (1st and 2nd group), with written informed consent. Approval of the local Ethics Committee for ONS in drCCH was first obtained for 5 patients (EUDRACT-2004-004551-19). Because of the favourable results in ONS-treated patients after a 16 months follow-up, we requested Ethics Committee approval for a protocol amendment allowing us to recruit 6 supplementary patients who were implanted and agreed to undergo PET before and after surgery. At the same time, patients of the 1st group were also asked to participate in the PET study and 4 of them accepted (the last patient of group 1 had been explanted [8]).\n[SUBTITLE] Surgical and stimulation procedure [SUBSECTION] Surgical procedure and stimulation protocols have been described previously [8]. We used unilateral subcutaneous implantation of paddle style stimulating leads with 4 electrode plots (Medtronic 3587A Resume II®; Medtronic Inc., Minneapolis, USA) via a retromastoid C1-2-3 approach [14] and Medtronic Itrel III® or Synergy® stimulators.\nStimulation protocols were adapted using a programming matrix [8] such as to induce paraesthesias in the largest possible occipital territory. The clinical evolution of patients was monitored with cluster headache paper diaries.\nSurgical procedure and stimulation protocols have been described previously [8]. We used unilateral subcutaneous implantation of paddle style stimulating leads with 4 electrode plots (Medtronic 3587A Resume II®; Medtronic Inc., Minneapolis, USA) via a retromastoid C1-2-3 approach [14] and Medtronic Itrel III® or Synergy® stimulators.\nStimulation protocols were adapted using a programming matrix [8] such as to induce paraesthesias in the largest possible occipital territory. The clinical evolution of patients was monitored with cluster headache paper diaries.\n[SUBTITLE] 18-FDG-PET study design [SUBSECTION] [SUBTITLE] Study groups [SUBSECTION] In the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.\nIn the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.\n[SUBTITLE] Data acquisition [SUBSECTION] PET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.\nPET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.\n[SUBTITLE] Statistical analysis (Cyclotron Research Centre, Liège) [SUBSECTION] Because we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).\nBecause we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).\n[SUBTITLE] Study groups [SUBSECTION] In the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.\nIn the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.\n[SUBTITLE] Data acquisition [SUBSECTION] PET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.\nPET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.\n[SUBTITLE] Statistical analysis (Cyclotron Research Centre, Liège) [SUBSECTION] Because we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).\nBecause we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).", "We studied 10 patients with drCCH (9 males and 1 female, mean age at implantation 44.2 ± 9.9 years SD). Inclusion criteria were: CCH for at least 2 years, daily attacks by history, side-locked attacks from the beginning, resistance to drug treatment according to expert consensus guidelines [3] and absence of associated disabling organic or psychiatric disorder. Five patients had left-sided, 5 right-sided attacks. At the time of the study, all patients were taking one or several of the following preventive drugs: verapamil (n = 9), lithium carbonate (n = 6), methylprednisolone (n = 2), methysergide (n = 2), melatonine (n = 1), gabapentine (n = 1). None took analgesics, in particular opioids.\nPatients were recruited in two phases (1st and 2nd group), with written informed consent. Approval of the local Ethics Committee for ONS in drCCH was first obtained for 5 patients (EUDRACT-2004-004551-19). Because of the favourable results in ONS-treated patients after a 16 months follow-up, we requested Ethics Committee approval for a protocol amendment allowing us to recruit 6 supplementary patients who were implanted and agreed to undergo PET before and after surgery. At the same time, patients of the 1st group were also asked to participate in the PET study and 4 of them accepted (the last patient of group 1 had been explanted [8]).", "Surgical procedure and stimulation protocols have been described previously [8]. We used unilateral subcutaneous implantation of paddle style stimulating leads with 4 electrode plots (Medtronic 3587A Resume II®; Medtronic Inc., Minneapolis, USA) via a retromastoid C1-2-3 approach [14] and Medtronic Itrel III® or Synergy® stimulators.\nStimulation protocols were adapted using a programming matrix [8] such as to induce paraesthesias in the largest possible occipital territory. The clinical evolution of patients was monitored with cluster headache paper diaries.", "[SUBTITLE] Study groups [SUBSECTION] In the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.\nIn the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.\n[SUBTITLE] Data acquisition [SUBSECTION] PET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.\nPET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.\n[SUBTITLE] Statistical analysis (Cyclotron Research Centre, Liège) [SUBSECTION] Because we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).\nBecause we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).", "In the 1st group, patients (n = 4), underwent a PET session after 24 to 30 months of ONS. All patients but one belonging to the 2nd group (n = 6) were scanned before implantation (baseline). In both groups, for each session after implantation, patients were scanned with the stimulator switched on (ON) and off for 3 days (OFF). This period was arbitrarily fixed in line with our previous observation of headache recurrence within 2 days on average after switching off the stimulator [8]. They underwent 2 more pairs of scans after 1 and 6 months of ONS. None of the patients had a cluster attack during PET, nor within the 12 hours preceding or following the procedure.\nData collected in patients were compared to a pool of 39 drug-free healthy volunteers (HV) (18 males, 21 females, mean age 45 ± 16 years SD) without headache history.", "PET data were obtained on a Siemens CTI 951 16/32® scanner (Siemens, Erlangen). Data were corrected for attenuation and background activity. Resting cerebral metabolism was studied after intravenous injection of 5-10 mCi (185-370 MBq) [18F]fluorodeoxyglucose (FDG). Subjects were scanned in a dark room, with minimal environmental noise.", "Because we selected CH patients with side-locked unilateral attacks and as there was no side shift during the observational period, we flipped the PET scans of the patients with right-sided symptoms in the axial plane so that we could analyze all subjects together (all \"left-sided\"). We used a binary classification of patients responding or not to ONS with an arbitrary, but clinically relevant, cut-off point for responders set at 50% decrease in attack frequency. Given the data from previous functional imaging studies, we conducted our analysis with an a priori hypothesis towards regions known to be involved in CH and other TACs [5], areas which are modulated by ONS in chronic migraine [13] and areas belonging to the pain matrix.\nData were analysed using statistical parametric mapping (SPM8 version; Wellcome Department of Cognitive Neurology Institute of Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) implemented in MATLAB (version 7.1, MathworksInc., Sherborn, MA). Images were spatially normalized into a standard stereotactic space using a symmetrical MNI (Montreal Neurological Institute) PET template [15] and smoothed using a 14 mm full-width-half-maximum (FWHM) isotropic kernel [16]. T-test was used with a significance level set at p < 0.001 uncorrected (p < 0.05 FDR).\nThe first design matrix included the scans of the 39 HV, of the 4 patients of the 1st group performed 24 to 30 months post-ONS and of the 6 patients of the 2nd group, scanned before implantation (baseline), 1 month and 6 months post-ONS. Our first analysis identified brain regions with a significant hyper- or hypometabolism in drCCH patients as compared to HV independent on time of scanning or ONS stimulation. Then, we looked for brain regions showing short term ONS-induced (ON versus OFF) increase or decrease in metabolism independent on delay since ONS implantation, in the early post-ONS phase (scans obtained after 1 month) and in the late post-ONS phase (scans obtained after 6 and 24-30 months). Finally, we compared metabolic activity measured during baseline, early phase (1 month) and late phase (≥ 6 months) post-ONS (independent of ONS stimulator settings) searching for progressive increases and decreases in metabolism over time.\nIn a second design matrix we searched for differences between the subgroups of 7 responders and 3 non-responders. Here, we included only the PET data obtained in the late phase (≥ 6 months; both with stimulator ON and OFF) and the HV scans, searching for regions with metabolic differences between the two groups (i.e., increased metabolism - as compared to HV - present in responders but not in non-responders).\nFor all analyses, the resulting set of voxel values for each contrast, constituting an SPM of the t-statistics (SPM{t}), was transformed to the unit normal distribution (SPM{Z}) and thresholded at p = 0.001. All group results were thresholded at false discovery-corrected p < 0.05, corrected for the whole brain volume. For differences between responder and non-responder subgroups, results were corrected for multiple comparisons within the regions of interest identified during the previous whole group analyses by employing a small volume correction (10 mm radius sphere).", "[SUBTITLE] Clinical outcome [SUBSECTION] Clinical data of patients and changes in their attack frequency after various durations of ONS are summarized in table 1. All patients but one in the 1st group (N = 4, 24-30 months follow-up) improved after ONS: one patient was pain free; 2 patients had a 90% and 93% reduction in attack frequency; the \"non-responder\" patient had a 25% improvement. In the 2nd group (N = 6), at 6 months follow-up, 3 patients were pain free and one was improved by 90%; these 4 patients already reported significant improvement after 1 month of ONS. The 5th patient only had a 33% reduction in attack frequency while in the 6th patient, attack frequency was slightly increased. According to the 50% cut-off criterion, 7 patients were thus considered responders and 3 non-responders to ONS for the binary analysis. During the 3-day period of ONS interruption, only one responder had recurrence of attacks. In the 2nd group, all patients kept the same preventive drug treatment during the follow-up scans except for one responder (8) who was able stop all medications after 6 months.\nPatients characteristics and clinical outcome\nONS: occipital nerve stimulation, R: right, L: left, Y: yes, N: no.\nClinical data of patients and changes in their attack frequency after various durations of ONS are summarized in table 1. All patients but one in the 1st group (N = 4, 24-30 months follow-up) improved after ONS: one patient was pain free; 2 patients had a 90% and 93% reduction in attack frequency; the \"non-responder\" patient had a 25% improvement. In the 2nd group (N = 6), at 6 months follow-up, 3 patients were pain free and one was improved by 90%; these 4 patients already reported significant improvement after 1 month of ONS. The 5th patient only had a 33% reduction in attack frequency while in the 6th patient, attack frequency was slightly increased. According to the 50% cut-off criterion, 7 patients were thus considered responders and 3 non-responders to ONS for the binary analysis. During the 3-day period of ONS interruption, only one responder had recurrence of attacks. In the 2nd group, all patients kept the same preventive drug treatment during the follow-up scans except for one responder (8) who was able stop all medications after 6 months.\nPatients characteristics and clinical outcome\nONS: occipital nerve stimulation, R: right, L: left, Y: yes, N: no.\n[SUBTITLE] PET results [SUBSECTION] The main areas of peak voxels where a metabolic change was found for the various comparisons are shown in table 2.\nMain statistical results and localization of peak voxels where cerebral metabolism was activated (>) or deactivated (<)\nCoordinates are in the standardized stereotactic Montreal Neurological Institute space (mm). FDR = False discovery rate corrected. drCCH: drug-resistant chronic cluster headache; HV: healthy volunteers; ACC: anterior cingulate cortex, R: right, L: left. * FDR corrected in region of interest identified in whole group analysis\nWe first pooled all scans performed in drCCH patients and compared them with those of the HV. In comparison to HV, a significant hypermetabolism was found in anterior cingulate cortex (ACC), left hypothalamus, left pulvinar, left visual cortex, cerebellum and brain stem (left lower pons and midbrain) (figure 1). By contrast, a significant hypometabolism appeared in both sensori-motor areas.\nHypermetabolic areas in drCCH patients (all conditions: baseline - 1 month, 6 months, 24/30 months) compared with HV (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (through midline and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, MB: midbrain, LP: lower pons, VC: visual cortex, P: pulvinar, H: hypothalamus.\nThere was no significant difference between scans performed with stimulator ON or OFF, regardless of ONS duration (1, 6 or 24-30 months). For scans obtained in the late phase (≥ 6 months) we observed a hypermetabolism in the left frontal lobe (BA 10, uncorrected p = 0.03, x = -12, y = 46, z = 4) and the left lower brainstem (uncorrected p = 0.014, x = -6, y = -30, z = -36) when the stimulator was turned ON, but these results, reported for the sake of completeness, did not survive correction for multiple comparisons.\nOver time, ONS changed glucose uptake in several brain areas (independent of the stimulator settings ON or OFF). The anterior cingulate, mid cingulate, left pulvinar, midbrain, lower pons, visual cortex and cerebellum had decreased metabolism over time, i.e. they became less hypermetabolic when comparing baseline to the early phase (1 month) or the late phase (≥ 6 months). By contrast, metabolism increased over time in sensorimotor cortices, i.e. they became less hypometabolic, and it was not modified in the left hypothalamus (figure 2).\nAreas progressively deactivated by ONS over time (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (right and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, B: brainstem, VC: visual cortex, P: pulvinar. White circle highlights hypothalamic area, which is not modified by the stimulation.\nWhen finally comparing responders and non-responders in the late phase, there was a significant hypermetabolism in previously identified perigenual ACC ipsilateral to the pain and stimulation side in responders (figure 3). No hypometabolic area differentiated responders from non-responders.\nActivation of perigenual cortex in ONS responders vs. non responders after 6 to 30 months stimulation (p < 0.05 FDR corrected). Result is displayed on a left sagittal section of a normalized MRI template.\nThe main areas of peak voxels where a metabolic change was found for the various comparisons are shown in table 2.\nMain statistical results and localization of peak voxels where cerebral metabolism was activated (>) or deactivated (<)\nCoordinates are in the standardized stereotactic Montreal Neurological Institute space (mm). FDR = False discovery rate corrected. drCCH: drug-resistant chronic cluster headache; HV: healthy volunteers; ACC: anterior cingulate cortex, R: right, L: left. * FDR corrected in region of interest identified in whole group analysis\nWe first pooled all scans performed in drCCH patients and compared them with those of the HV. In comparison to HV, a significant hypermetabolism was found in anterior cingulate cortex (ACC), left hypothalamus, left pulvinar, left visual cortex, cerebellum and brain stem (left lower pons and midbrain) (figure 1). By contrast, a significant hypometabolism appeared in both sensori-motor areas.\nHypermetabolic areas in drCCH patients (all conditions: baseline - 1 month, 6 months, 24/30 months) compared with HV (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (through midline and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, MB: midbrain, LP: lower pons, VC: visual cortex, P: pulvinar, H: hypothalamus.\nThere was no significant difference between scans performed with stimulator ON or OFF, regardless of ONS duration (1, 6 or 24-30 months). For scans obtained in the late phase (≥ 6 months) we observed a hypermetabolism in the left frontal lobe (BA 10, uncorrected p = 0.03, x = -12, y = 46, z = 4) and the left lower brainstem (uncorrected p = 0.014, x = -6, y = -30, z = -36) when the stimulator was turned ON, but these results, reported for the sake of completeness, did not survive correction for multiple comparisons.\nOver time, ONS changed glucose uptake in several brain areas (independent of the stimulator settings ON or OFF). The anterior cingulate, mid cingulate, left pulvinar, midbrain, lower pons, visual cortex and cerebellum had decreased metabolism over time, i.e. they became less hypermetabolic when comparing baseline to the early phase (1 month) or the late phase (≥ 6 months). By contrast, metabolism increased over time in sensorimotor cortices, i.e. they became less hypometabolic, and it was not modified in the left hypothalamus (figure 2).\nAreas progressively deactivated by ONS over time (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (right and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, B: brainstem, VC: visual cortex, P: pulvinar. White circle highlights hypothalamic area, which is not modified by the stimulation.\nWhen finally comparing responders and non-responders in the late phase, there was a significant hypermetabolism in previously identified perigenual ACC ipsilateral to the pain and stimulation side in responders (figure 3). No hypometabolic area differentiated responders from non-responders.\nActivation of perigenual cortex in ONS responders vs. non responders after 6 to 30 months stimulation (p < 0.05 FDR corrected). Result is displayed on a left sagittal section of a normalized MRI template.", "Clinical data of patients and changes in their attack frequency after various durations of ONS are summarized in table 1. All patients but one in the 1st group (N = 4, 24-30 months follow-up) improved after ONS: one patient was pain free; 2 patients had a 90% and 93% reduction in attack frequency; the \"non-responder\" patient had a 25% improvement. In the 2nd group (N = 6), at 6 months follow-up, 3 patients were pain free and one was improved by 90%; these 4 patients already reported significant improvement after 1 month of ONS. The 5th patient only had a 33% reduction in attack frequency while in the 6th patient, attack frequency was slightly increased. According to the 50% cut-off criterion, 7 patients were thus considered responders and 3 non-responders to ONS for the binary analysis. During the 3-day period of ONS interruption, only one responder had recurrence of attacks. In the 2nd group, all patients kept the same preventive drug treatment during the follow-up scans except for one responder (8) who was able stop all medications after 6 months.\nPatients characteristics and clinical outcome\nONS: occipital nerve stimulation, R: right, L: left, Y: yes, N: no.", "The main areas of peak voxels where a metabolic change was found for the various comparisons are shown in table 2.\nMain statistical results and localization of peak voxels where cerebral metabolism was activated (>) or deactivated (<)\nCoordinates are in the standardized stereotactic Montreal Neurological Institute space (mm). FDR = False discovery rate corrected. drCCH: drug-resistant chronic cluster headache; HV: healthy volunteers; ACC: anterior cingulate cortex, R: right, L: left. * FDR corrected in region of interest identified in whole group analysis\nWe first pooled all scans performed in drCCH patients and compared them with those of the HV. In comparison to HV, a significant hypermetabolism was found in anterior cingulate cortex (ACC), left hypothalamus, left pulvinar, left visual cortex, cerebellum and brain stem (left lower pons and midbrain) (figure 1). By contrast, a significant hypometabolism appeared in both sensori-motor areas.\nHypermetabolic areas in drCCH patients (all conditions: baseline - 1 month, 6 months, 24/30 months) compared with HV (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (through midline and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, MB: midbrain, LP: lower pons, VC: visual cortex, P: pulvinar, H: hypothalamus.\nThere was no significant difference between scans performed with stimulator ON or OFF, regardless of ONS duration (1, 6 or 24-30 months). For scans obtained in the late phase (≥ 6 months) we observed a hypermetabolism in the left frontal lobe (BA 10, uncorrected p = 0.03, x = -12, y = 46, z = 4) and the left lower brainstem (uncorrected p = 0.014, x = -6, y = -30, z = -36) when the stimulator was turned ON, but these results, reported for the sake of completeness, did not survive correction for multiple comparisons.\nOver time, ONS changed glucose uptake in several brain areas (independent of the stimulator settings ON or OFF). The anterior cingulate, mid cingulate, left pulvinar, midbrain, lower pons, visual cortex and cerebellum had decreased metabolism over time, i.e. they became less hypermetabolic when comparing baseline to the early phase (1 month) or the late phase (≥ 6 months). By contrast, metabolism increased over time in sensorimotor cortices, i.e. they became less hypometabolic, and it was not modified in the left hypothalamus (figure 2).\nAreas progressively deactivated by ONS over time (p < 0.05 FDR corrected). Results are displayed on 2 sagittal sections of a normalized MRI template (right and left hemisphere). ACC: anterior cingulate cortex, C: cerebellum, B: brainstem, VC: visual cortex, P: pulvinar. White circle highlights hypothalamic area, which is not modified by the stimulation.\nWhen finally comparing responders and non-responders in the late phase, there was a significant hypermetabolism in previously identified perigenual ACC ipsilateral to the pain and stimulation side in responders (figure 3). No hypometabolic area differentiated responders from non-responders.\nActivation of perigenual cortex in ONS responders vs. non responders after 6 to 30 months stimulation (p < 0.05 FDR corrected). Result is displayed on a left sagittal section of a normalized MRI template.", "The therapeutic outcome after ONS in 1st and 2nd groups was overall similar, with good response as defined above in nearly 70% of patients [8,17]. A noticeable difference between the 2 groups was the latency to significant efficacy, which was on average shortened to 1 month in the 2nd group compared to 3 months in the 1st one. This remains compatible with a slow modulatory effect on the central nervous system. We ascribe faster efficacy in group 2 to a learning effect for investigators and to quicker optimisation of stimulator settings [8].\nWe will focus our discussion on the FDG-PET findings.\n[SUBTITLE] PET findings - interpretation [SUBSECTION] [SUBTITLE] Metabolic pattern in drCCH compared to HV [SUBSECTION] The enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\nThe enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\n[SUBTITLE] Short-term changes associated with ONS [SUBSECTION] We found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\nWe found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\n[SUBTITLE] Long-term changes associated with ONS [SUBSECTION] Hypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\nHypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\n[SUBTITLE] Perigenual ACC (PACC) activation in responders [SUBSECTION] Comparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\nComparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\n[SUBTITLE] Study limitations [SUBSECTION] We are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.\nWe are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.\n[SUBTITLE] Metabolic pattern in drCCH compared to HV [SUBSECTION] The enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\nThe enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\n[SUBTITLE] Short-term changes associated with ONS [SUBSECTION] We found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\nWe found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\n[SUBTITLE] Long-term changes associated with ONS [SUBSECTION] Hypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\nHypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\n[SUBTITLE] Perigenual ACC (PACC) activation in responders [SUBSECTION] Comparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\nComparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\n[SUBTITLE] Study limitations [SUBSECTION] We are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.\nWe are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.", "[SUBTITLE] Metabolic pattern in drCCH compared to HV [SUBSECTION] The enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\nThe enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).\n[SUBTITLE] Short-term changes associated with ONS [SUBSECTION] We found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\nWe found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].\n[SUBTITLE] Long-term changes associated with ONS [SUBSECTION] Hypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\nHypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.\n[SUBTITLE] Perigenual ACC (PACC) activation in responders [SUBSECTION] Comparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\nComparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.\n[SUBTITLE] Study limitations [SUBSECTION] We are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.\nWe are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.", "The enhanced FDG uptake found ipsilaterally in hypothalamus of drCCH patients respective to HV is in line with the reports showing increased activity in this area with H215O-PET or fMRI during attacks [18-20]. However in our study all patients were scanned between attacks, i.e. at a time point when hypothalamic activation has not been reported yet. Another FDG-PET study of episodic CH comparing patients during and between bouts revealed no change in hypothalamic glucose uptake [21].\nAreas belonging to the pain matrix like the cingulate gyrus or midbrain (periaqueductal grey - PAG) are classically activated in various pain states including headaches [5]. We also found activation in the cerebellum, in line with previous imaging studies showing consistent cerebellar activation across the spectrum of pain from visceral to somatic, acute to chronic [22]. Activation of cerebellar vermis and anterior lobe in drCCH may be particularly strong due to dense somatotopically arranged trigemino-cerebellar connexions [22], but also because there are direct connections between the ventro-posterior hypothalamic area and the cerebellum as shown by MRI tractography in a drCCH patient treated with hypothalamic DBS [23].\nThe perigenual ACC (PACC) is of particular interest. Contrary to our results, it was found hypometabolic compared to HV in episodic CH [21]. However, when the authors compared the same patients during and between bouts, the PACC was hypermetabolic during the bout despite the absence of an attack at the time of scanning.\nLower pons activation has been described during attacks of hemicrania continua, but not of CH [5]. In hemicrania continua, dorsal pontine activation is ipsilateral, like in our study, while posterior hypothalamic activation is observed on the opposite side, unlike in CH.\nThe pulvinar where we found ipsilateral activation in drCCH patients has not been a region of interest in CH before. Pulvinotomy and electrical stimulation of the pulvinar have been used successfully in the treatment of chronic pain in humans [24]. In functional neuroimaging studies, pulvinar hypermetabolism is either associated with pain state [25] or with pain relief after various procedures [24], including ONS in chronic migraine [13].\nWe have no straightforward explanation for the increased metabolism of the ipsilateral visual cortex in our patients. Such activation has not been reported hitherto. Photophobia ipsilateral to the pain is a frequent attack-associated symptom in various TACs, in particular CH [26]. Like in migraineurs, the visual cortex of CH patients might thus be more sensitive to light stimuli. This hypothesis seems unlikely, however, as all subjects were scanned in a dark room.\nTo summarize, the interictal FDG-PET hyperactive pattern of drCCH patients comprises areas reported to be hypermetabolic mainly during TAC attacks (ipsilateral hypothalamus and pons), but also during a bout of the disorder outside of an attack (perigenual ACC).", "We found no significant differences between PET recordings performed with stimulator ON or OFF within a 72 hour delay. This finding is similar to Matharu et al.'s [13] observations in chronic migraine, except that they were not able to scan the patients OFF and pain-free because of an almost immediate recurrence of pain after interrupting the stimulation. They concluded that central structures were not modulated in chronic migraine by ONS beyond the stimulation period.\nHere, lack of short-term metabolic modification favours a slow neuromodulatory effect of ONS, as we suspected before [8]. Interestingly, in CCH patients treated with hypothalamic DBS, May et al. [27] found rapid metabolic changes with H215O-PET in various brain structures involved in cluster headache and more generally in the pain matrix within only 10 minutes of switching the stimulator ON or OFF, but there were no clinical correlates suggesting that the therapeutic effect of DBS is also due to slow CNS changes.\nOur subanalysis with a smaller significance level revealed a change in left lower brainstem metabolism. It might indicate a short term ONS effect in the trigemino-cervical complex. As expected from the neuro-anatomical connexions, the trigemino-cervical complex could relay stimulation to more rostral structures allowing for neuroplastic modulation of their activity. In comparison, a recent study of transcutaneous electrical nerve stimulation (TENS) in arthritic rats suggests that its antinociceptive effect is mediated by ascending activation of the opioid system originating in the ventrolateral PAG and projecting via the rostral ventro-medial medulla to the spinal cord [28].", "Hypermetabolism of most overactive areas in drCCH patients compared to HV decreased after ONS. This was particularly obvious for the anterior cingulate cortex, left pulvinar, left visual cortex, left lower pons, cerebellum and midbrain. Conversely, baseline hypometabolism of sensori-motor cortices increased after ONS. The noteworthy exception to these post-ONS metabolic changes is the ipsilateral hypothalamus. This is precisely the region activated during CH attacks on the side of the pain [18] and where increased gray matter density is found on voxel-based MRI between attacks [29]. Our findings in drCCH contrast with those by Sprenger et al [21] in episodic CH where no significant metabolic change was found in the hypothalamus, either outside or during the bout. If replicated, our data suggest that persistent hypothalamic activation is a hallmark of chronic CH. They may explain why attacks are non-remittent but also why some ONS-treated, pain-free patients still have autonomic attacks [8]. A similar conclusion was drawn for the dorsal rostral pons in chronic migraine following the finding of persistent activation in this area despite ONS-induced pain relief [13]. The persistence of an ipsilateral hypothalamic activation despite reduced attack frequency also confirms that ONS is no more than a symptomatic therapy, as already suggested by the recurrence of attacks after interruption of the stimulation [8].\nThe metabolic changes observed after ONS could be due to the stimulation itself or to the reduction of attack frequency. Our protocol did not allow to favour either mechanism. However, despite the small number of non-responders in our study, some insight might be gained from the comparison of patients who clearly responded to ONS and those who did not.", "Comparison of ONS responders and non-responders revealed increased FDG uptake in the PACC of the former. This area is of interest for several reasons. First, PACC plays a major role in central opioidergic pain control system. It is selectively activated during analgesia induced by the μ-receptor agonists fentanyl and remifentanyl compared with placebo [30], providing evidence that opioidergic analgesia is mediated by activation of descending antinociceptive pathways. Second, PACC was found hypometabolic respective to HV in episodic CH, but its activity increased significantly during the bout [21]. Knowing the pivotal role of PACC in descending pain control, these authors hypothesized that deficient endogenous antinociceptive mechanisms between bouts might predispose CH patients to the disorder and to its recurrence. Concordantly, Sprenger et al [31], using PET with the opioidergic ligand [11C]diprenorphine, demonstrated an inverse linear relationship between the duration of CH and opioid receptor availability in the rostral ACC (and ipsilateral hypothalamus). A recent case report by the same group of a drCCH patient in whom low dose levomethadone induced complete remission of attacks favours this hypothesis [32].\nCompared to TENS that was shown to induce analgesia through activation of a PAG-RVM-spinal cord pathway [28], ONS could activate this descending pain-control pathway even further up-stream at the level of PACC. Its therapeutic effect in drCCH patients could thus be due to progressive restoration of activity in deficient opioidergic antinociceptive pathways.", "We are well aware of some methodological flaws that may limit the strength of our findings.\nFirst, the number of patients included is rather small. This is the case in most similar studies as drCCH patients are rare and ONS is an emerging treatment modality for which only 38 cases, including ours, have been published. In a much commoner condition like chronic migraine, metabolic imaging studies have been limited to less than 10 patients [13].\nA second shortcoming is the dichotomy of the PET design. PET studies were not planned in our initial pilot study of ONS in 5 patients as the outcome was uncertain and the sample considered too small. As clinical efficacy was encouraging, we were allowed to recruit 6 additional patients, for whom imaging studies were planned prospectively. For greater sample size, patients from the 1st group were also proposed to undergo PET. This explains why the latter had neither baseline nor 1 month scans and why long-term treatment periods vary between 6 and 30 months. We know since that there is no further clinical improvement after 6 months of ONS and that in most patients attacks recur after stimulation interruption whatever the duration of ONS [8]. This is why we decide to merge scans obtained between 6 and 30 months of ONS, albeit statistically questionable.\nFinally, one may argue that the prophylactic drugs taken by the patients may have influenced the PET results. This cannot be ruled out, as HV did not take any medication. However, pharmacotherapy remained stable in all patients of the 2nd group except one and was similar in ONS responders and non-responders.", "We confirm that ONS is effective and safe in drCCH, reducing attack frequency by ≥ 50% in more than 60% of patients, which is similar to the results obtained with hypothalamic DBS.\nThe FDG-PET results in our small sample appear consistent with the clinical impression that ONS exerts its beneficial effects via slow neuromodulatory processes in the central pain matrix. The finding of a possible selective perigenual ACC activation in responders compared to non-responders might advocate that ONS activates descending pain control systems in a top-down manner and restores an equilibrium in anti-nociceptive opioidergic pathways. We suggest for the first time that metabolic activity could be increased in ipsilateral posterior hypothalamus in chronic cluster headache patients outside of an attack. That ONS, as suspected on clinical grounds, does not cure drCCH, but merely acts as a symptomatic treatment is underlined by its inability to reduce this ipsilateral hypothalamic hyperactivity which is typically found during attacks in episodic cluster headache. Persistent hypothalamic activation might also explain why ONS-treated pain-free drCCH patients may still have autonomic attacks and why attacks rapidly recur after interruption of ONS.", "18-FDG-PET: 18-Fluorodeoxyglucose-positron emission tomography; ACC: anterior cingular cortex; BA: Brodmann area; DBS: deep brain stimulation; drCCH: drug-resistant chronic cluster headache; HV: healthy volunteers; OFF: stimulator switched off; ON: stimulator switched on; ONS: occipital nerve stimulation; PACC: perigenual anterior cingular cortex; PAG: periaqueductal grey; TACs: trigeminal autonomic cephalalgias; TENS: transcutaneous electrical nerve stimulation", "The authors declare that they have no competing interests.", "All authors read and approved the final manuscript. DM participated to the clinical follow-up of patients, interpreted the PET results and drafted the manuscript. MAB analyzed the PET data and wrote the methodology. AF designed the PET study protocol. PYG participated to the clinical follow-up of patients. RH is head of the University Nuclear Medicine Department where patients underwent the PET scans. SL analyzed the PET data and did the matrix design. JS participated to the clinical follow-up of patients, interpreted the PET results and drafted the manuscript with DM.", "DM, MD, is clinical chief associate at the University Neurology Department, Liège, Belgium.\nMAB, MSc, is research fellow at the National Fund for Scientific Research (FNRS), Belgium.\nAF, MD, PhD, is clinical chief at the University Neurology Department, Liège, Belgium.\nPYG, MD, is research fellow at the University Neurology Department, Liège, Belgium.\nRH, MD, PhD, is head of the University Nuclear Medicine Department, Liège, Belgium.\nSL, MD, PhD, is senior research associate at the FNRS, Belgium.\nJS, MD, PhD, is head of the Headache Research Unit at the University of Liège, Belgium and Professor of Neuroanatomy.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2377/11/25/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adolescent", "Amino Acid Substitution", "Base Sequence", "Brain", "Child", "Child, Preschool", "Consanguinity", "DNA Mutational Analysis", "Extracellular Matrix Proteins", "Female", "Humans", "Intellectual Disability", "Lipoid Proteinosis of Urbach and Wiethe", "Male", "Mutation", "Mutation, Missense", "Pedigree", "Saudi Arabia", "Sequence Deletion", "Young Adult" ]

[ "Background", "Results", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Lipoid proteinosis (LP; MIM 247100) is a rare autosomal recessive disease characterized by cutaneous and mucosal lesions and hoarseness appearing in early childhood[1] that is caused by homozygous or compound heterozygous mutations in the ECM1 gene located on chromosome 1q21[2]. This gene encodes a protein that is an important structural component of basement membrane and extracellular matrix[3,4], and loss of protein-protein interactions due to ECM1 gene mutations is the likely cause of dermatologic abnormalities including warty skin, scarring, and mucosal thickening[5]. These changes also affect the nasopharynx, tongue, and vocal cords, resulting in severe fibrosis and the hoarseness characteristic of the disorder.\nApproximately one third of affected individuals are reported to have mild mental retardation[6], andneuropsychological problems may be more common[7-9]. Other reported neurologic abnormalities include complex partial seizures, memory loss, and emotional difficulties, often beginning in teenage years and progressing from that time onward[6,8]. Calcification of the mesial temporal lobes may be observedon brain imaging and is progressive with age. Pathologic material is limited, and the mechanism of brain injury remains uncertain.\nWe had the opportunity to study seven individuals from three unrelated consanguineous families who had the typical spectrum of skin, voice, and neurologiccharacteristics of LP with mutations in the ECM1 gene.", "Direct sequencing of the ECM1 gene inthe two patients (II-1 and II-6) from Family 1 (Figure 1a) identified a novel homozygous two base deletion near the end of exon 8 (c.1300-1301delAA). The deletion occurred at codon 434 and was expected to create a premature stop codon two codons later, likely to result in low levels of allECM1 transcriptsand in severely truncated proteins[3].\nPedigrees and Genetic Results. (a)Family 1 pedigree and ECM1 chromatogram showing a homozygous two base deletion detected in affected family members. b) Family 2 pedigree and ECM1 chromatogram showing a homozygous base substitution in both affected individuals (II-3 and II-6) and heterozygous base substitution in their mother. Similar heterozygous results were also found in father (II-1) and an unaffected sibling (II-4) but are not shown. c) Family 3 pedigree with photograph of a 1.5% agarose gel showing failed amplification of exons 9 and 10 in all affected individuals but not in their mother or unaffected brother. Forward primer was designed in intron 8 and the reverse primer was designed in intron 10.\nFamily 2 (Figure 1b) had a novel one-base substitution mutation c.806G > Ain exon 7, which resulted in the replacement of cystine with tyrosine at codon 269 (p.Cys269Tyr). This mutation was detected in homozygous status in both patients (II-3 and II-6) and in heterozygous status in both parents (I-1 and I-2) and one unaffected siblings (II-4). An assessment of the pathologic effect of this mutation utilizing PolyPhen, a tool predicting the possible impact of an amino acid substitution on the structure and function of a human protein using physical and comparative considerations http://genetics.bwh.harvard.edu/pph/[10], predicted this mutation to be probably damaging (i.e., very likely to affect protein structure and/or function). Exon 7 is represented in transcripts ECM1a and ECM1c[11]but not in the ECM1b transcript, and therefore, presumably the ECM1b transcript would be normal in this family.\nFamily 3 (Figure 1c) had a 1163 bp deletion starting 34 bp into intron 8 and encompassing all of exon-9, intron-9, and exon 10, including the termination codon and part of the 3'-UTR. This mutation was detected in homozygous status in all affected individuals in this family (II-1, II-2, II-3), but was not present in an unaffected sibling (II-4). The mother was a carrier of the mutation; the father was deceased but was also expected to be a carrier. This mutation was previously reported in an affected Saudi family from the same tribe thatmay be related by founder effect to the current family[2]. It results in complete loss of exons 9 and 10 from transcripts ECM1a/c and ECM1b and is predicted to have a deleterious effect on protein structure and function[2].\nThe two siblings in Family 1 were the only affected individuals in this consanguineous family with seven children (Figure 1A). Patient II-1 was born at eight months gestation and stayed in a neonatal intensive care unit for six weeks, while Patient II-6 was the product of a normal pregnancy and delivery. They each had mild motor delay, walking at age 18 months and talking at approximately age three years. They were modestly mentally retarded with intelligence quotients of approximately 60 and a requirement for special education. Neither had developed seizures or obvious emotional problems when last examined at ages 25 (Patient II-1) and 14 (Patient II-6)years.\nBoth children had normal skin appearance at birth, but each had an abnormal cry and developed hoarseness by age six months. Patient II-1 had the most severe presentation during early childhood in this group with irregular, indurated, partially coalescent yellowish papules with hypo- and hyperpigmented macular plaques in the lower back and gluteal regions. He developed skin-colored, verrucous papules and nodules on the elbows, knees, and sites of friction and movement including the hands. His tongue was enlarged and woody, hard on palpation, with inability to protrude tongue. The frenulum and oral mucosa were infiltrated, but dentition was normal. Laryngoscopy revealed hypertrophic vocal cords with normal movement. Patient II-6 had a milder presentation in early childhood with mildly itchy, erythematous maculopapular skin eruptions over both extremities that developed gradually into hyperpigmented, verrucous areas over elbows, knees, and knuckles. Both had smallhands and digits. They had mild microcephaly with a saddle nose and eyelid thickening (moniliform blepharosis). Muscle mass and strength were normal, but both had mild lower extremity spasticity with modest asymmetric hyperreflexia but without dystonic changes. Gait was grossly normal.\nSkin biopsies revealed extensive deposition of amorphous eosinophilic, hyaline material characteristic of LP consisting of thick, homogenous bundles extending perpendicular to the skin surface. These bundles surrounded blood vessels in the thickened papillary dermis, and this material was PAS positive and diastase resistant (Figure 2). There were no ischemic changes around blood vessels.\nPapillary dermis. (a) Deposition of PAS positive and diastase-resistant material from Patient II-1 of Family 1. Rare inflammatory cells are seen in the upper dermis (original magnification X 200). (b) PAS positive material deposited around blood vessels which have thickened and hyalinized walls. (arrows, original magnification X 400). There was no evidence of infarction.\nBrain computed tomography (CT) of Patient II-6 at age 10 years showed small calcifications in the amygdala bilaterally (Figure 3a). CT and magnetic resonance imaging (MRI) scans were otherwise normal with no evidence of atrophy or ischemia (Figure 3b). CT of Patient II-1revealed no brain calcification; however, both CT and MRI showed a watershed ischemic injury involving the right cerebral hemisphere with periventricular white matter loss (Figure 3c) and thickening of the overlying skull implying chronicity. He also had atrophy of the right cerebellar hemisphere and middle cerebellar peduncle (Figure 3d).\nBrain CT and MRI of Patients 1 and 2. (a) Brain CT image of Patient II-6 of Family 1 at age of 10 years showing a small calcification in the left amygdale. Calcification in the right amygdale was seen on another section (not shown). (b) Brain axial T2-weighted MR image of the same patientat the level of temporal lobes showing a tiny hypointensity (arrow) representing amygdala calcification. The mesial temporal lobes were otherwise normal. (c) Brain axial T2-weighted MR images of Patient II-1 of Family 1 showing enlarged sulci and small gyri in the watershed zones of the right frontal lobe (open arrows) associated with loss of white matter due to old ischemia. (d) The right cerebellar hemisphere of the same patientwas small with widened folia, white matter gliosis (arrow), and a small right middle cerebellar peduncle (asterisk). Remodeling of the posterior fossa confirmed the long-standing nature of this abnormality. CT and MRI scans on patients from Families 2 and 3 were normal.\nIn general, patients from Family 2 (Figure 1b) were slightly more affected than patients from Family 3 (Figure 1c), although both families were less affected than Family 1. Rough, yellowish-white papular deposits in the skin and the oral mucosa were presentwithin the first year of life. They all developed multiple, small, severe blisters over the fingers of both hands leading to excoriations and post inflammatory hyperpigmentation mainly over joints with mild facial skin changes but no striking scarring. They all had beaded papules on the upper and lower labial mucosa and tongue together with multiple, small, waxy, yellow beaded papules along the eyelids (moniliform blepharosis). Hoarseness and pharyngeal involvement were present but also less evident.\nIndividuals from Families 2 and 3 had no developmental delay, and head circumference was normal in each. They were not mentally retarded, and each attended age-appropriate grades in normal schools, except for the youngest, who was not yet in school, and the girl from Family 3 (II-3), who dropped out of school, possibly for emotional reasons. Basic neurologic examinations were normal with no reflex or gait abnormality. CT and MRI scans (including Susceptibility-Weighted sequences, which are particularly sensitive to calcification and iron deposition) in each affected individual, except for Patient II-6 of Family 2, showed no temporal lobe calcification and no other brain abnormalities.", "These seven patients alldeveloped characteristic skin lesions and hoarseness in infancy and had unequivocal LP by skin biopsy. Two individuals from Family 1 had modest mental retardation, and one had small amygdala calcifications, but the other affected patients had no definitive neurologic involvement when last examined between ages five and 19 years. All seven individuals had homozygous mutations in the ECM1 gene[3,12,13];Families 1 and 2 had novel ECM1 mutations, while Family 3 had a previously reported large deletion[2].\nInnate severity of LP phenotypic expression is difficult to gauge for a number of reasons, even in carefully evaluated patients. Skin manifestations are recognized to be variable within families and even from time to time within the same individual, generally increasing with age and skin exposure, which is often determined by environment and employment. Pharyngeal involvement is usually obvious because of hoarseness and fibrosis involving the tongue and pharyngeal tissue, but these changes are difficult to quantify. Overt mental retardation may be straightforward to recognize, as in two of our patients, but more subtle intelligence changesmay not be easy to detect and may not be reported by families wishing to minimize their own impression of the impact of this genetic disease. Detecting subtle emotional or personality changes requires formal, quantitative neuropsychological testing standardized to the appropriate language and ethnic group. Such testing is not often available, as in the case of the families reported here. These families also demonstrate that current quality brain CT and MRI may be entirely normal in the temporal lobes and elsewhere even in teenage years and even with some indication of emotional difficulties (e.g., Patient II-3of Family 3).\nLP has been reported to cause a variety of neurologic problems, including psychosis[9], partial complex seizures[6], emotional incontinence including rage and panic attacks[14], and progressive memory loss[7], most commonly of memories with emotional content[8]. Mental retardation has been reported in approximately one third of individuals with LP [6,9,15]and was documented in two of these seven individuals. The relative infrequency of other obviousneurologic abnormalities may be due tochance in a small series of patients or to the subtlety of changes in cognitive function or personality. Alternatively, patients with advanced or obvious neurologic abnormalities may have been over-represented in classic reports, or neuropsychological changes may be more frequent in certain well-studied populations (e.g., South African) than in others[16].\nThe etiology of mental retardation in LP remains uncertain because currently there is no information about ECM1 expression in human brain outside of blood vessels[12]. The ECM1 gene is an integral part of the extracellular matrix and basement membrane structure. These roles likely lead directly to the classic skin and pharyngeal damage when ECM1 is mutated [3], but the relevance to brain is less obvious. ECM1 interacts with other proteins making up the extracellular matrix [17,18], often regulating bioactivity of the binding partner[11], includingmatrix metalloproteinase 9 [19], which has an important role in temporal lobe synaptic physiology [20]. ECM1 mutations, therefore, might affect brain function, compromise homeostasis, or impair repair mechanisms independent of ischemia [19,21].\nThe brain pathophysiology of ECM1 has been hypothesized to result from a defect in cerebrovascular basement membrane that causes fibrinoid changes and calcification of vessel walls [8,15,22]. However, available neuropathology is limited, is from older individuals, and leaves open the question of whether the vascular pathology described is primary or secondary to some other process. Careful review of CT and MRI images yielded no hint of brain damage, including in the area immediately contiguous to temporal lobe calcification[14]in previously reported patients [4,8,14,23] or in Patient II-6 of Family 1. Therefore, a cerebral vasculopathic mechanism is not supported by the absence of ischemic changes on CT or MRI neuroimaging or in skin biopsies[6]. Even if ischemia were the cause of temporal lobe calcification, it is still unclear why this process should affect a particular area of the mesial temporal lobe and, when advanced, cause a fairly consistent comma-shaped area of calcification[24].\nPatients from Family 1 were more affected in terms of skin, pharyngeal, and neurologic involvement than patients from Families 2 and 3, suggesting that severity ofintegumentinvolvement may roughly correspond to severity of brain involvement in certain LP individuals and families. Hamada and colleagues noted that individuals with mutations outside of exon 7 seemed more affected than individuals with mutations involving exon 7 of the ECM1 gene[3]. Indeed, Family 1, with a mutation in exon 8, was more affected than Family 2, with a mutation in exon 7. However, Family 1 had a two nucleotide deletion at the end of exon 8, leading almost immediately to a premature stop codon and effectively truncating all transcripts at that point. Family 3 had a large deletion removing exons 9 and 10 and accomplishing a comparable genetic defect. Nevertheless, skin, pharynx, and nervous systemof Family 1 seemedmore affected than in Family 3.\nMost genetic reports regarding LP have observed that phenotype-genotype correlations are not obvious[3,25], and this seems also to be true in our three families. The ECM1 gene has several transcripts, each of which may be differentially active in various tissues, including various parts of the brain, at different times during an individual's lifetime, offering the possibility of substantial variability in the effect of mutations in particular individuals. In addition, novel transcripts may provide partial rescue through in-frame skipping of mutated exons in certain circumstances[3]. This genetic complexity is currently difficult to either predict or assess in individual patients. Similarly, our current ability to completely quantify clinical, including neurologic, involvement is limited. Taken together, these factors mean that the lack of phenotype-genotype correlation is not surprising. Therefore, a suspected diagnosis of LP should be confirmed by sequencing ECM1 exons 6 and 7 first because this is where most mutations occur[3], but homozygous or compound heterozygous mutations of any type may occur in any exon.", "These individuals illustrate the neurologic spectrum of LP, including variable mental retardation, personality changes, and mesial temporal calcification but imply that significant neurologic involvement may be somewhat less common that previously thought. The cause of neurologic abnormalities was not clear from either neuroimaging or from what is known about ECM1 function. The severity of dermatologic abnormalities and hoarseness generally correlated with neurologic abnormalities, with Family 1 being somewhat more affected in all spheres than the other two families. Nevertheless, phenotype-genotype correlation was not obvious, possibly because of difficulty quantifying the neurologic phenotype and because of genetic complexity.", "CT: Computed tomography; ECM1: Extracellular matrix protein 1; LP: Lipoid proteinosis; MRI: magnetic resonance imaging; UTR: Untranslated Region", "The authors declare that they have no competing interests.", "MASevaluated clinical and neurologic features of the patients; IAO was in charge of the radiological evaluation; TMB evaluatedneurologic and neuro-ophthalmologic features; DTO evaluated ocular motility;LL, JAM, and LVM evaluated additional dermatologic and clinical features; and YHF and AHS evaluated ophthalmological features. KKA was in charge of analysis of genetic data and writing the genetic portion of the manuscript. KKA and TMB were responsible for study design and manuscript completion. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2350/12/31/prepub\n" ]

[ null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Results", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Lipoid proteinosis (LP; MIM 247100) is a rare autosomal recessive disease characterized by cutaneous and mucosal lesions and hoarseness appearing in early childhood[1] that is caused by homozygous or compound heterozygous mutations in the ECM1 gene located on chromosome 1q21[2]. This gene encodes a protein that is an important structural component of basement membrane and extracellular matrix[3,4], and loss of protein-protein interactions due to ECM1 gene mutations is the likely cause of dermatologic abnormalities including warty skin, scarring, and mucosal thickening[5]. These changes also affect the nasopharynx, tongue, and vocal cords, resulting in severe fibrosis and the hoarseness characteristic of the disorder.\nApproximately one third of affected individuals are reported to have mild mental retardation[6], andneuropsychological problems may be more common[7-9]. Other reported neurologic abnormalities include complex partial seizures, memory loss, and emotional difficulties, often beginning in teenage years and progressing from that time onward[6,8]. Calcification of the mesial temporal lobes may be observedon brain imaging and is progressive with age. Pathologic material is limited, and the mechanism of brain injury remains uncertain.\nWe had the opportunity to study seven individuals from three unrelated consanguineous families who had the typical spectrum of skin, voice, and neurologiccharacteristics of LP with mutations in the ECM1 gene.", "Seven participants from three unrelated Saudi Arabian (Table 1) wereenrolled and all signed informed consent. This study was approved by the institute IRB and adhered to the tenets of the Declaration of Helsinki. Blood samples were collected from all participants and DNA was extracted and stored at -20°C until needed for genetic testing.\nDemographics of three families\nIntelligence testing was performed by routine assessment on neurologic examination and by using the Stanford Binet Test (Patient II-1, Family 1) and the Wechsler Bellevue Intelligence scale (Patient II-6, Family 1). Patients from each family had skin punch biopsies including epidermis, dermis, and subcutaneous tissue that were fixed in formalin and stained with PAS and PAS/diastase. All affected individualsexcept Patient II-6 of Family 2had brain computed tomography (CT) and either 1.5 or 3.0 Tesla magnetic resonance imaging(MRI). MRI for patients from Family 3 included Susceptibility-Weighted sequences.\nThe ECM1 gene was directly sequenced as described previously in affected individuals, in available first degree relatives, and in 100 chromosomes from healthy individuals of matching ethnicity[2,3]. Nucleotide numbering for mutations reported in all three families was referenced to the NCBI sequence (NG_012062).", "Direct sequencing of the ECM1 gene inthe two patients (II-1 and II-6) from Family 1 (Figure 1a) identified a novel homozygous two base deletion near the end of exon 8 (c.1300-1301delAA). The deletion occurred at codon 434 and was expected to create a premature stop codon two codons later, likely to result in low levels of allECM1 transcriptsand in severely truncated proteins[3].\nPedigrees and Genetic Results. (a)Family 1 pedigree and ECM1 chromatogram showing a homozygous two base deletion detected in affected family members. b) Family 2 pedigree and ECM1 chromatogram showing a homozygous base substitution in both affected individuals (II-3 and II-6) and heterozygous base substitution in their mother. Similar heterozygous results were also found in father (II-1) and an unaffected sibling (II-4) but are not shown. c) Family 3 pedigree with photograph of a 1.5% agarose gel showing failed amplification of exons 9 and 10 in all affected individuals but not in their mother or unaffected brother. Forward primer was designed in intron 8 and the reverse primer was designed in intron 10.\nFamily 2 (Figure 1b) had a novel one-base substitution mutation c.806G > Ain exon 7, which resulted in the replacement of cystine with tyrosine at codon 269 (p.Cys269Tyr). This mutation was detected in homozygous status in both patients (II-3 and II-6) and in heterozygous status in both parents (I-1 and I-2) and one unaffected siblings (II-4). An assessment of the pathologic effect of this mutation utilizing PolyPhen, a tool predicting the possible impact of an amino acid substitution on the structure and function of a human protein using physical and comparative considerations http://genetics.bwh.harvard.edu/pph/[10], predicted this mutation to be probably damaging (i.e., very likely to affect protein structure and/or function). Exon 7 is represented in transcripts ECM1a and ECM1c[11]but not in the ECM1b transcript, and therefore, presumably the ECM1b transcript would be normal in this family.\nFamily 3 (Figure 1c) had a 1163 bp deletion starting 34 bp into intron 8 and encompassing all of exon-9, intron-9, and exon 10, including the termination codon and part of the 3'-UTR. This mutation was detected in homozygous status in all affected individuals in this family (II-1, II-2, II-3), but was not present in an unaffected sibling (II-4). The mother was a carrier of the mutation; the father was deceased but was also expected to be a carrier. This mutation was previously reported in an affected Saudi family from the same tribe thatmay be related by founder effect to the current family[2]. It results in complete loss of exons 9 and 10 from transcripts ECM1a/c and ECM1b and is predicted to have a deleterious effect on protein structure and function[2].\nThe two siblings in Family 1 were the only affected individuals in this consanguineous family with seven children (Figure 1A). Patient II-1 was born at eight months gestation and stayed in a neonatal intensive care unit for six weeks, while Patient II-6 was the product of a normal pregnancy and delivery. They each had mild motor delay, walking at age 18 months and talking at approximately age three years. They were modestly mentally retarded with intelligence quotients of approximately 60 and a requirement for special education. Neither had developed seizures or obvious emotional problems when last examined at ages 25 (Patient II-1) and 14 (Patient II-6)years.\nBoth children had normal skin appearance at birth, but each had an abnormal cry and developed hoarseness by age six months. Patient II-1 had the most severe presentation during early childhood in this group with irregular, indurated, partially coalescent yellowish papules with hypo- and hyperpigmented macular plaques in the lower back and gluteal regions. He developed skin-colored, verrucous papules and nodules on the elbows, knees, and sites of friction and movement including the hands. His tongue was enlarged and woody, hard on palpation, with inability to protrude tongue. The frenulum and oral mucosa were infiltrated, but dentition was normal. Laryngoscopy revealed hypertrophic vocal cords with normal movement. Patient II-6 had a milder presentation in early childhood with mildly itchy, erythematous maculopapular skin eruptions over both extremities that developed gradually into hyperpigmented, verrucous areas over elbows, knees, and knuckles. Both had smallhands and digits. They had mild microcephaly with a saddle nose and eyelid thickening (moniliform blepharosis). Muscle mass and strength were normal, but both had mild lower extremity spasticity with modest asymmetric hyperreflexia but without dystonic changes. Gait was grossly normal.\nSkin biopsies revealed extensive deposition of amorphous eosinophilic, hyaline material characteristic of LP consisting of thick, homogenous bundles extending perpendicular to the skin surface. These bundles surrounded blood vessels in the thickened papillary dermis, and this material was PAS positive and diastase resistant (Figure 2). There were no ischemic changes around blood vessels.\nPapillary dermis. (a) Deposition of PAS positive and diastase-resistant material from Patient II-1 of Family 1. Rare inflammatory cells are seen in the upper dermis (original magnification X 200). (b) PAS positive material deposited around blood vessels which have thickened and hyalinized walls. (arrows, original magnification X 400). There was no evidence of infarction.\nBrain computed tomography (CT) of Patient II-6 at age 10 years showed small calcifications in the amygdala bilaterally (Figure 3a). CT and magnetic resonance imaging (MRI) scans were otherwise normal with no evidence of atrophy or ischemia (Figure 3b). CT of Patient II-1revealed no brain calcification; however, both CT and MRI showed a watershed ischemic injury involving the right cerebral hemisphere with periventricular white matter loss (Figure 3c) and thickening of the overlying skull implying chronicity. He also had atrophy of the right cerebellar hemisphere and middle cerebellar peduncle (Figure 3d).\nBrain CT and MRI of Patients 1 and 2. (a) Brain CT image of Patient II-6 of Family 1 at age of 10 years showing a small calcification in the left amygdale. Calcification in the right amygdale was seen on another section (not shown). (b) Brain axial T2-weighted MR image of the same patientat the level of temporal lobes showing a tiny hypointensity (arrow) representing amygdala calcification. The mesial temporal lobes were otherwise normal. (c) Brain axial T2-weighted MR images of Patient II-1 of Family 1 showing enlarged sulci and small gyri in the watershed zones of the right frontal lobe (open arrows) associated with loss of white matter due to old ischemia. (d) The right cerebellar hemisphere of the same patientwas small with widened folia, white matter gliosis (arrow), and a small right middle cerebellar peduncle (asterisk). Remodeling of the posterior fossa confirmed the long-standing nature of this abnormality. CT and MRI scans on patients from Families 2 and 3 were normal.\nIn general, patients from Family 2 (Figure 1b) were slightly more affected than patients from Family 3 (Figure 1c), although both families were less affected than Family 1. Rough, yellowish-white papular deposits in the skin and the oral mucosa were presentwithin the first year of life. They all developed multiple, small, severe blisters over the fingers of both hands leading to excoriations and post inflammatory hyperpigmentation mainly over joints with mild facial skin changes but no striking scarring. They all had beaded papules on the upper and lower labial mucosa and tongue together with multiple, small, waxy, yellow beaded papules along the eyelids (moniliform blepharosis). Hoarseness and pharyngeal involvement were present but also less evident.\nIndividuals from Families 2 and 3 had no developmental delay, and head circumference was normal in each. They were not mentally retarded, and each attended age-appropriate grades in normal schools, except for the youngest, who was not yet in school, and the girl from Family 3 (II-3), who dropped out of school, possibly for emotional reasons. Basic neurologic examinations were normal with no reflex or gait abnormality. CT and MRI scans (including Susceptibility-Weighted sequences, which are particularly sensitive to calcification and iron deposition) in each affected individual, except for Patient II-6 of Family 2, showed no temporal lobe calcification and no other brain abnormalities.", "These seven patients alldeveloped characteristic skin lesions and hoarseness in infancy and had unequivocal LP by skin biopsy. Two individuals from Family 1 had modest mental retardation, and one had small amygdala calcifications, but the other affected patients had no definitive neurologic involvement when last examined between ages five and 19 years. All seven individuals had homozygous mutations in the ECM1 gene[3,12,13];Families 1 and 2 had novel ECM1 mutations, while Family 3 had a previously reported large deletion[2].\nInnate severity of LP phenotypic expression is difficult to gauge for a number of reasons, even in carefully evaluated patients. Skin manifestations are recognized to be variable within families and even from time to time within the same individual, generally increasing with age and skin exposure, which is often determined by environment and employment. Pharyngeal involvement is usually obvious because of hoarseness and fibrosis involving the tongue and pharyngeal tissue, but these changes are difficult to quantify. Overt mental retardation may be straightforward to recognize, as in two of our patients, but more subtle intelligence changesmay not be easy to detect and may not be reported by families wishing to minimize their own impression of the impact of this genetic disease. Detecting subtle emotional or personality changes requires formal, quantitative neuropsychological testing standardized to the appropriate language and ethnic group. Such testing is not often available, as in the case of the families reported here. These families also demonstrate that current quality brain CT and MRI may be entirely normal in the temporal lobes and elsewhere even in teenage years and even with some indication of emotional difficulties (e.g., Patient II-3of Family 3).\nLP has been reported to cause a variety of neurologic problems, including psychosis[9], partial complex seizures[6], emotional incontinence including rage and panic attacks[14], and progressive memory loss[7], most commonly of memories with emotional content[8]. Mental retardation has been reported in approximately one third of individuals with LP [6,9,15]and was documented in two of these seven individuals. The relative infrequency of other obviousneurologic abnormalities may be due tochance in a small series of patients or to the subtlety of changes in cognitive function or personality. Alternatively, patients with advanced or obvious neurologic abnormalities may have been over-represented in classic reports, or neuropsychological changes may be more frequent in certain well-studied populations (e.g., South African) than in others[16].\nThe etiology of mental retardation in LP remains uncertain because currently there is no information about ECM1 expression in human brain outside of blood vessels[12]. The ECM1 gene is an integral part of the extracellular matrix and basement membrane structure. These roles likely lead directly to the classic skin and pharyngeal damage when ECM1 is mutated [3], but the relevance to brain is less obvious. ECM1 interacts with other proteins making up the extracellular matrix [17,18], often regulating bioactivity of the binding partner[11], includingmatrix metalloproteinase 9 [19], which has an important role in temporal lobe synaptic physiology [20]. ECM1 mutations, therefore, might affect brain function, compromise homeostasis, or impair repair mechanisms independent of ischemia [19,21].\nThe brain pathophysiology of ECM1 has been hypothesized to result from a defect in cerebrovascular basement membrane that causes fibrinoid changes and calcification of vessel walls [8,15,22]. However, available neuropathology is limited, is from older individuals, and leaves open the question of whether the vascular pathology described is primary or secondary to some other process. Careful review of CT and MRI images yielded no hint of brain damage, including in the area immediately contiguous to temporal lobe calcification[14]in previously reported patients [4,8,14,23] or in Patient II-6 of Family 1. Therefore, a cerebral vasculopathic mechanism is not supported by the absence of ischemic changes on CT or MRI neuroimaging or in skin biopsies[6]. Even if ischemia were the cause of temporal lobe calcification, it is still unclear why this process should affect a particular area of the mesial temporal lobe and, when advanced, cause a fairly consistent comma-shaped area of calcification[24].\nPatients from Family 1 were more affected in terms of skin, pharyngeal, and neurologic involvement than patients from Families 2 and 3, suggesting that severity ofintegumentinvolvement may roughly correspond to severity of brain involvement in certain LP individuals and families. Hamada and colleagues noted that individuals with mutations outside of exon 7 seemed more affected than individuals with mutations involving exon 7 of the ECM1 gene[3]. Indeed, Family 1, with a mutation in exon 8, was more affected than Family 2, with a mutation in exon 7. However, Family 1 had a two nucleotide deletion at the end of exon 8, leading almost immediately to a premature stop codon and effectively truncating all transcripts at that point. Family 3 had a large deletion removing exons 9 and 10 and accomplishing a comparable genetic defect. Nevertheless, skin, pharynx, and nervous systemof Family 1 seemedmore affected than in Family 3.\nMost genetic reports regarding LP have observed that phenotype-genotype correlations are not obvious[3,25], and this seems also to be true in our three families. The ECM1 gene has several transcripts, each of which may be differentially active in various tissues, including various parts of the brain, at different times during an individual's lifetime, offering the possibility of substantial variability in the effect of mutations in particular individuals. In addition, novel transcripts may provide partial rescue through in-frame skipping of mutated exons in certain circumstances[3]. This genetic complexity is currently difficult to either predict or assess in individual patients. Similarly, our current ability to completely quantify clinical, including neurologic, involvement is limited. Taken together, these factors mean that the lack of phenotype-genotype correlation is not surprising. Therefore, a suspected diagnosis of LP should be confirmed by sequencing ECM1 exons 6 and 7 first because this is where most mutations occur[3], but homozygous or compound heterozygous mutations of any type may occur in any exon.", "These individuals illustrate the neurologic spectrum of LP, including variable mental retardation, personality changes, and mesial temporal calcification but imply that significant neurologic involvement may be somewhat less common that previously thought. The cause of neurologic abnormalities was not clear from either neuroimaging or from what is known about ECM1 function. The severity of dermatologic abnormalities and hoarseness generally correlated with neurologic abnormalities, with Family 1 being somewhat more affected in all spheres than the other two families. Nevertheless, phenotype-genotype correlation was not obvious, possibly because of difficulty quantifying the neurologic phenotype and because of genetic complexity.", "CT: Computed tomography; ECM1: Extracellular matrix protein 1; LP: Lipoid proteinosis; MRI: magnetic resonance imaging; UTR: Untranslated Region", "The authors declare that they have no competing interests.", "MASevaluated clinical and neurologic features of the patients; IAO was in charge of the radiological evaluation; TMB evaluatedneurologic and neuro-ophthalmologic features; DTO evaluated ocular motility;LL, JAM, and LVM evaluated additional dermatologic and clinical features; and YHF and AHS evaluated ophthalmological features. KKA was in charge of analysis of genetic data and writing the genetic portion of the manuscript. KKA and TMB were responsible for study design and manuscript completion. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2350/12/31/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null ]

[]

[ "Cross-Sectional Studies", "Data Interpretation, Statistical", "Databases, Factual", "Humans", "Meta-Analysis as Topic", "Models, Statistical", "Research Design" ]

[ "Background", "Results", "Choice of model in relation to degree of heterogeneity", "Significant effects in relation to choice of model", "Cautions about the heterogeneity", "Discussion", "Limitations", "Other studies of heterogeneity", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Variability among individual study results in systematic reviews virtually always occurs. This is caused partly by random error (chance) and partly by systematic differences between the trials. The variation in the true effects is called heterogeneity. Its impact on meta-analyses can be assessed by I2 that describes the percentage of the variability that is due to heterogeneity [1,2]. Values greater than 50% are - rather arbitrarily - considered substantial heterogeneity [1].\nStrategies for addressing heterogeneity in systematic reviews include checking that the data extracted from the trial reports are correct, which may often not be the case [3]; omitting meta-analysis; conducting subgroup analysis or meta-regression; choosing a fixed effect or a random effects model [2]; changing the statistical metric, e.g. from a risk difference to a relative risk [4,5]; and excluding studies.\nThe fixed effect model assumes that all the included studies are estimating the same true effect. The variation in findings among studies is therefore due to chance [2]. Each study will be assigned a weight depending on the study's precision (within-trial variance) and an overall estimate can be calculated. Small studies will contribute relatively little to the outcome because they have less precision [6].\nThe random effects model assumes that the effects being estimated in the different studies are not identical, but follow a distribution. The confidence interval takes account of the additional uncertainty in the location of the mean of the systematically different effects in the different studies (this between-trial variance is added to the within-trial variance). Small studies will therefore contribute more to the average than in a fixed effect analysis, which is reasonable because the studies represent different true effects. Thus, when heterogeneity is present, the confidence interval around a random effects pooled estimate is wider than a confidence interval around a fixed effect pooled estimate [6].\nDealing with heterogeneity is often tricky, and there is only limited advice for authors on what to do, e.g. on when a particular model should be chosen for the other [7], or when the heterogeneity becomes too large for a meaningful meta-analysis.\nAn additional complexity is that the test for detecting heterogeneity has low power when the sample sizes are small or when few trials are included. For example, 11 trials give just 10 degrees of freedom, like a t-test on two groups of 6 people each does. There is also variation in practice as to which P-value demonstrates significant heterogeneity [2], but as the power of the test is so low, it is common to choose P = 0.10. It is important to be aware, however, that the choice of statistical model should not be based on the outcome of a test of heterogeneity [2].\nThe aim of our study was to investigate how authors address different degrees of substantial heterogeneity in meta-analyses in Cochrane reviews.", "[SUBTITLE] Choice of model in relation to degree of heterogeneity [SUBSECTION] Several of the 60 selected reviews had been published more than once. The oldest was most recently updated in 1996 and the newest in 2007 (median 2005). A fixed effect model was used in 33 reviews (55%), and a random effects model in 27 reviews (the Cochrane software does not allow mixed models [8]). There was no correlation between degree of heterogeneity and choice of model (Figure 2, P = 0.79, Mann-Whitney test for trend, with correction for ties), in fact the average I2 was 71%, both for reviews using a fixed effect model and for those using a random effects model.\nChoice of model, reviews grouped by I2.\nThe authors selected the random effects model more often in the newest half of the reviews (updated later than 1 June 2005, Table 1), than in the oldest half. The same pattern was evident for the subgroup of reviews with marginally statistically significant heterogeneity (P-value between 0.05 and 0.10, P = 0.007, Fisher's exact test).\nChoice of model in relation to the P-value for the heterogeneity test.\nNewer reviews are those updated after 1 June 2005 (n = 31). P = 0.007 for those reviews where the heterogeneity test yielded a P-value between 0.05 and 0.10\nSeveral of the 60 selected reviews had been published more than once. The oldest was most recently updated in 1996 and the newest in 2007 (median 2005). A fixed effect model was used in 33 reviews (55%), and a random effects model in 27 reviews (the Cochrane software does not allow mixed models [8]). There was no correlation between degree of heterogeneity and choice of model (Figure 2, P = 0.79, Mann-Whitney test for trend, with correction for ties), in fact the average I2 was 71%, both for reviews using a fixed effect model and for those using a random effects model.\nChoice of model, reviews grouped by I2.\nThe authors selected the random effects model more often in the newest half of the reviews (updated later than 1 June 2005, Table 1), than in the oldest half. The same pattern was evident for the subgroup of reviews with marginally statistically significant heterogeneity (P-value between 0.05 and 0.10, P = 0.007, Fisher's exact test).\nChoice of model in relation to the P-value for the heterogeneity test.\nNewer reviews are those updated after 1 June 2005 (n = 31). P = 0.007 for those reviews where the heterogeneity test yielded a P-value between 0.05 and 0.10\n[SUBTITLE] Significant effects in relation to choice of model [SUBSECTION] A significant effect estimate (P < 0.05) was presented in 34 reviews. For 6 of the 60 reviews, a significant result changed to a non-significant result when we applied the alternative model (discussed further below). For 5 reviews, a non-significant result became significant when we used the alternative model (Table 2). These 5 reviews had all used a random effects model and addressed heterogeneity this way; they were free of major methodological problems and will therefore not be discussed further.\nResults using the authors' model and the alternative model we applied\nFor 2 of the 6 reviews [9,10] where a significant result changed to a non-significant when we used a random effects model, the authors were cautious about their heterogeneous result and didn't base their conclusion on the significant finding they had obtained with a fixed effect model, which we consider a correct approach. One review was explicit about this: \"Substantial heterogeneity was also detected (p = 0.03, I2 = 79%). Because of this, the result of this analysis should be interpreted with caution and not be considered a definitive statement\" [10].\nThe authors of the other 4 reviews were less cautious. One review [11] calculated mean differences instead of standardized mean differences, although the outcomes were measured on very different scales. Because of this error, both the means and the standard deviations differed by a factor of 10. This resulted in extreme heterogeneity (I2 = 93%, P = 0.0002) despite very low power, as only two studies were included. In the methods section, the authors promised to use a random effects model in case of heterogeneity, but this was not done (and would not have solved the other problem).\nIn another review [12], the authors calculated the standardized mean difference both with a fixed effect model (1.07, 95% confidence interval 0.43 to 1.70) and a random effects model (1.74, -0.71 to 4.19). In the methods section, they stated they would use a random effects model if heterogeneity was present, which it was (I2 = 89%, P = 0.002). With a random effects model, the result was not significant. They wrote that no definite conclusion could be made but added that there was reasonable evidence that cognitive therapy was beneficial in treating depression. We find this conclusion doubtful, given the data and their declared methods. In this example, the effect estimate calculated by the two models differs substantially due to a one small outlying study. Hence, the choice of model should have been considered and explained in detail.\nAnother review [13] used Peto's odds ratio (0.28, 0.11 to 0.73). Significant heterogeneity was present (I2 = 59%, P = 0.05), and when using the ordinary odds ratio and a random effects model, the result became 0.28 (0.05 to 1.55). The authors concluded in the abstract that albumin showed a clear benefit at preventing severe ovarian hyperstimulation syndrome, although they were much more cautious in the main text.\nThe authors of the last review [14] reported that the P-value for heterogeneity was insignificant even though it was 0.07 and the power of the heterogeneity test was very low, as there were only 5 studies. They reported less mortality in the intervention group, relative risk 0.86 (0.74 to 1.00). With a random effects model the relative risk became 0.82 (0.57 to 1.18), with P = 0.30. The authors mentioned that heterogeneity was present and noted that one outlying trial had a very low mortality in the control group. The meta-analysis was driven by a big trial, which comprised 69% of the deaths and showed the same result as the pooled result, 0.86 (0.75 to 1.00). Even so, we find it pretty bold that the authors believed in a result with borderline significance (P = 0.05), and only when using the fixed effect model, with so much heterogeneity, and with unexplained discrepancies between the results of the trials.\nA significant effect estimate (P < 0.05) was presented in 34 reviews. For 6 of the 60 reviews, a significant result changed to a non-significant result when we applied the alternative model (discussed further below). For 5 reviews, a non-significant result became significant when we used the alternative model (Table 2). These 5 reviews had all used a random effects model and addressed heterogeneity this way; they were free of major methodological problems and will therefore not be discussed further.\nResults using the authors' model and the alternative model we applied\nFor 2 of the 6 reviews [9,10] where a significant result changed to a non-significant when we used a random effects model, the authors were cautious about their heterogeneous result and didn't base their conclusion on the significant finding they had obtained with a fixed effect model, which we consider a correct approach. One review was explicit about this: \"Substantial heterogeneity was also detected (p = 0.03, I2 = 79%). Because of this, the result of this analysis should be interpreted with caution and not be considered a definitive statement\" [10].\nThe authors of the other 4 reviews were less cautious. One review [11] calculated mean differences instead of standardized mean differences, although the outcomes were measured on very different scales. Because of this error, both the means and the standard deviations differed by a factor of 10. This resulted in extreme heterogeneity (I2 = 93%, P = 0.0002) despite very low power, as only two studies were included. In the methods section, the authors promised to use a random effects model in case of heterogeneity, but this was not done (and would not have solved the other problem).\nIn another review [12], the authors calculated the standardized mean difference both with a fixed effect model (1.07, 95% confidence interval 0.43 to 1.70) and a random effects model (1.74, -0.71 to 4.19). In the methods section, they stated they would use a random effects model if heterogeneity was present, which it was (I2 = 89%, P = 0.002). With a random effects model, the result was not significant. They wrote that no definite conclusion could be made but added that there was reasonable evidence that cognitive therapy was beneficial in treating depression. We find this conclusion doubtful, given the data and their declared methods. In this example, the effect estimate calculated by the two models differs substantially due to a one small outlying study. Hence, the choice of model should have been considered and explained in detail.\nAnother review [13] used Peto's odds ratio (0.28, 0.11 to 0.73). Significant heterogeneity was present (I2 = 59%, P = 0.05), and when using the ordinary odds ratio and a random effects model, the result became 0.28 (0.05 to 1.55). The authors concluded in the abstract that albumin showed a clear benefit at preventing severe ovarian hyperstimulation syndrome, although they were much more cautious in the main text.\nThe authors of the last review [14] reported that the P-value for heterogeneity was insignificant even though it was 0.07 and the power of the heterogeneity test was very low, as there were only 5 studies. They reported less mortality in the intervention group, relative risk 0.86 (0.74 to 1.00). With a random effects model the relative risk became 0.82 (0.57 to 1.18), with P = 0.30. The authors mentioned that heterogeneity was present and noted that one outlying trial had a very low mortality in the control group. The meta-analysis was driven by a big trial, which comprised 69% of the deaths and showed the same result as the pooled result, 0.86 (0.75 to 1.00). Even so, we find it pretty bold that the authors believed in a result with borderline significance (P = 0.05), and only when using the fixed effect model, with so much heterogeneity, and with unexplained discrepancies between the results of the trials.\n[SUBTITLE] Cautions about the heterogeneity [SUBSECTION] In our judgment, 40 of the 60 reviews were devoid of major problems in relation to their handling of heterogeneity (Table 3). However, only 27 reviews (45%) gave a rationale for choice of statistical model. Our overall judgment of methodological quality was not related to I2 (P = 0.26, Mann-Whitney test for trend corrected for ties, grouping I2 into intervals of 10%). However, the unproblematic reviews contained more studies than problematic reviews, 8.0 versus 4.3, on average (P = 0.024, student t-test, two-tailed).\nOur assessments of the 60 reviews in relation to their handling of heterogeneity\nSixteen of the 20 problematic reviews had chosen a fixed effect model. There can be plausible reasons for this, even in cases with substantial heterogeneity, but authors should then explain what they are. Eight of these 20 reviews had no explanations or reservations and did not address the heterogeneity statistically [15-22]. One review [21] had only included one study, but the patients were split into two subgroups, according to whether they had a rash. Both results were significant, but one showed harm and the other benefit (Figure 3). Combining such results is inappropriate and doesn't represent today's standards of Cochrane reviews (the review was last updated in 1996).\nExtremely diverging results. Both results originate from the same study [21].\nFive other reviews used Peto's odds ratio [23,24,13,18,19], and three reviews didn't follow the analysis plan that was set out in the methods section [12,25,26], which included using a random effects model or omitting meta-analysis in case of heterogeneity, and there was no explanation why. Three other reviews paid no attention to the heterogeneity and didn't discuss it, even though the P-value was between 0.05 and 0.10 [27,14,28]. An additional review described the heterogeneity (I2 = 71%, P = 0.06) but ignored it due to \"lack of stability of the known tests\" [29], which is not a valid reason for ignoring heterogeneity. In another review, the authors divided the analysis into subgroups because they had found heterogeneity, but although the consequence was that the chi-square test for heterogeneity was no longer significant due to loss of power, the I2 actually increased, which the authors failed to comment on [30]. In yet another review, which was discussed above, the authors pooled two risk scores measured on different scales that varied by a factor of ten [11]. The complete list of included reviews can be found on the web http://sites.google.com/site/dealingwithheterogeneity/.\nIn our judgment, 40 of the 60 reviews were devoid of major problems in relation to their handling of heterogeneity (Table 3). However, only 27 reviews (45%) gave a rationale for choice of statistical model. Our overall judgment of methodological quality was not related to I2 (P = 0.26, Mann-Whitney test for trend corrected for ties, grouping I2 into intervals of 10%). However, the unproblematic reviews contained more studies than problematic reviews, 8.0 versus 4.3, on average (P = 0.024, student t-test, two-tailed).\nOur assessments of the 60 reviews in relation to their handling of heterogeneity\nSixteen of the 20 problematic reviews had chosen a fixed effect model. There can be plausible reasons for this, even in cases with substantial heterogeneity, but authors should then explain what they are. Eight of these 20 reviews had no explanations or reservations and did not address the heterogeneity statistically [15-22]. One review [21] had only included one study, but the patients were split into two subgroups, according to whether they had a rash. Both results were significant, but one showed harm and the other benefit (Figure 3). Combining such results is inappropriate and doesn't represent today's standards of Cochrane reviews (the review was last updated in 1996).\nExtremely diverging results. Both results originate from the same study [21].\nFive other reviews used Peto's odds ratio [23,24,13,18,19], and three reviews didn't follow the analysis plan that was set out in the methods section [12,25,26], which included using a random effects model or omitting meta-analysis in case of heterogeneity, and there was no explanation why. Three other reviews paid no attention to the heterogeneity and didn't discuss it, even though the P-value was between 0.05 and 0.10 [27,14,28]. An additional review described the heterogeneity (I2 = 71%, P = 0.06) but ignored it due to \"lack of stability of the known tests\" [29], which is not a valid reason for ignoring heterogeneity. In another review, the authors divided the analysis into subgroups because they had found heterogeneity, but although the consequence was that the chi-square test for heterogeneity was no longer significant due to loss of power, the I2 actually increased, which the authors failed to comment on [30]. In yet another review, which was discussed above, the authors pooled two risk scores measured on different scales that varied by a factor of ten [11]. The complete list of included reviews can be found on the web http://sites.google.com/site/dealingwithheterogeneity/.", "Several of the 60 selected reviews had been published more than once. The oldest was most recently updated in 1996 and the newest in 2007 (median 2005). A fixed effect model was used in 33 reviews (55%), and a random effects model in 27 reviews (the Cochrane software does not allow mixed models [8]). There was no correlation between degree of heterogeneity and choice of model (Figure 2, P = 0.79, Mann-Whitney test for trend, with correction for ties), in fact the average I2 was 71%, both for reviews using a fixed effect model and for those using a random effects model.\nChoice of model, reviews grouped by I2.\nThe authors selected the random effects model more often in the newest half of the reviews (updated later than 1 June 2005, Table 1), than in the oldest half. The same pattern was evident for the subgroup of reviews with marginally statistically significant heterogeneity (P-value between 0.05 and 0.10, P = 0.007, Fisher's exact test).\nChoice of model in relation to the P-value for the heterogeneity test.\nNewer reviews are those updated after 1 June 2005 (n = 31). P = 0.007 for those reviews where the heterogeneity test yielded a P-value between 0.05 and 0.10", "A significant effect estimate (P < 0.05) was presented in 34 reviews. For 6 of the 60 reviews, a significant result changed to a non-significant result when we applied the alternative model (discussed further below). For 5 reviews, a non-significant result became significant when we used the alternative model (Table 2). These 5 reviews had all used a random effects model and addressed heterogeneity this way; they were free of major methodological problems and will therefore not be discussed further.\nResults using the authors' model and the alternative model we applied\nFor 2 of the 6 reviews [9,10] where a significant result changed to a non-significant when we used a random effects model, the authors were cautious about their heterogeneous result and didn't base their conclusion on the significant finding they had obtained with a fixed effect model, which we consider a correct approach. One review was explicit about this: \"Substantial heterogeneity was also detected (p = 0.03, I2 = 79%). Because of this, the result of this analysis should be interpreted with caution and not be considered a definitive statement\" [10].\nThe authors of the other 4 reviews were less cautious. One review [11] calculated mean differences instead of standardized mean differences, although the outcomes were measured on very different scales. Because of this error, both the means and the standard deviations differed by a factor of 10. This resulted in extreme heterogeneity (I2 = 93%, P = 0.0002) despite very low power, as only two studies were included. In the methods section, the authors promised to use a random effects model in case of heterogeneity, but this was not done (and would not have solved the other problem).\nIn another review [12], the authors calculated the standardized mean difference both with a fixed effect model (1.07, 95% confidence interval 0.43 to 1.70) and a random effects model (1.74, -0.71 to 4.19). In the methods section, they stated they would use a random effects model if heterogeneity was present, which it was (I2 = 89%, P = 0.002). With a random effects model, the result was not significant. They wrote that no definite conclusion could be made but added that there was reasonable evidence that cognitive therapy was beneficial in treating depression. We find this conclusion doubtful, given the data and their declared methods. In this example, the effect estimate calculated by the two models differs substantially due to a one small outlying study. Hence, the choice of model should have been considered and explained in detail.\nAnother review [13] used Peto's odds ratio (0.28, 0.11 to 0.73). Significant heterogeneity was present (I2 = 59%, P = 0.05), and when using the ordinary odds ratio and a random effects model, the result became 0.28 (0.05 to 1.55). The authors concluded in the abstract that albumin showed a clear benefit at preventing severe ovarian hyperstimulation syndrome, although they were much more cautious in the main text.\nThe authors of the last review [14] reported that the P-value for heterogeneity was insignificant even though it was 0.07 and the power of the heterogeneity test was very low, as there were only 5 studies. They reported less mortality in the intervention group, relative risk 0.86 (0.74 to 1.00). With a random effects model the relative risk became 0.82 (0.57 to 1.18), with P = 0.30. The authors mentioned that heterogeneity was present and noted that one outlying trial had a very low mortality in the control group. The meta-analysis was driven by a big trial, which comprised 69% of the deaths and showed the same result as the pooled result, 0.86 (0.75 to 1.00). Even so, we find it pretty bold that the authors believed in a result with borderline significance (P = 0.05), and only when using the fixed effect model, with so much heterogeneity, and with unexplained discrepancies between the results of the trials.", "In our judgment, 40 of the 60 reviews were devoid of major problems in relation to their handling of heterogeneity (Table 3). However, only 27 reviews (45%) gave a rationale for choice of statistical model. Our overall judgment of methodological quality was not related to I2 (P = 0.26, Mann-Whitney test for trend corrected for ties, grouping I2 into intervals of 10%). However, the unproblematic reviews contained more studies than problematic reviews, 8.0 versus 4.3, on average (P = 0.024, student t-test, two-tailed).\nOur assessments of the 60 reviews in relation to their handling of heterogeneity\nSixteen of the 20 problematic reviews had chosen a fixed effect model. There can be plausible reasons for this, even in cases with substantial heterogeneity, but authors should then explain what they are. Eight of these 20 reviews had no explanations or reservations and did not address the heterogeneity statistically [15-22]. One review [21] had only included one study, but the patients were split into two subgroups, according to whether they had a rash. Both results were significant, but one showed harm and the other benefit (Figure 3). Combining such results is inappropriate and doesn't represent today's standards of Cochrane reviews (the review was last updated in 1996).\nExtremely diverging results. Both results originate from the same study [21].\nFive other reviews used Peto's odds ratio [23,24,13,18,19], and three reviews didn't follow the analysis plan that was set out in the methods section [12,25,26], which included using a random effects model or omitting meta-analysis in case of heterogeneity, and there was no explanation why. Three other reviews paid no attention to the heterogeneity and didn't discuss it, even though the P-value was between 0.05 and 0.10 [27,14,28]. An additional review described the heterogeneity (I2 = 71%, P = 0.06) but ignored it due to \"lack of stability of the known tests\" [29], which is not a valid reason for ignoring heterogeneity. In another review, the authors divided the analysis into subgroups because they had found heterogeneity, but although the consequence was that the chi-square test for heterogeneity was no longer significant due to loss of power, the I2 actually increased, which the authors failed to comment on [30]. In yet another review, which was discussed above, the authors pooled two risk scores measured on different scales that varied by a factor of ten [11]. The complete list of included reviews can be found on the web http://sites.google.com/site/dealingwithheterogeneity/.", "It can be challenging to choose the most appropriate model for meta-analysis, as there are pros and cons with both models [2]. It is pretty clear that the larger the heterogeneity, the harder it is to defend choosing a fixed effect model, as different studies cannot be assumed to provide estimates of a common, true effect. However, one also needs to consider that a random effects model may apply too much weight to small studies, which are often poorly done and biased. An example is shown in Figure 4, where it appeared reasonable that the authors used a fixed effect model despite pronounced heterogeneity, as it gives more weight to the only large study, which, moreover, had a result that was closer to no effect than most other studies. It is also possible that the studies were so different that they should not have been combined in a meta-analysis. Only a closer look at the review, and possibly also at the individual studies, can elucidate this, and even then, researchers may disagree what would be the most appropriate approach.\nResults analysed using a fixed effect model, which gives more weight to the only large study [22].\nIt is also important to consider that the fixed effect model only allows an inference about the studies included in the meta-analysis, whereas the random effects model allows an inference about the mean effect in a hypothetical population of studies if we can assume that the studies included in the meta-analysis constitute a random selection of studies from this hypothetical population.\nThe random effects model is more conservative than the fixed effect model in the sense that the confidence interval is broader, but sometimes the point estimate is farther from the null and the P value for the pooled effect smaller than with a fixed effect model [31].\nWhen using a random effects model, the between-study variance needs to be calculated, but if there are few studies, this cannot be calculated with any precision, and a fixed effect model is therefore sometimes used in this situation [6].\nIt was surprising that we did not find a relation between the degree of heterogeneity and the choice of model. Some Cochrane groups instruct their authors to routinely use a fixed effect model, although few statisticians would find such blanket recommendations reasonable. Furthermore, in all types of research, authors should change their planned analysis and explain why if it would not be sensible.\nReaders might be more willing to accept the results if they are robust to both types of analyses, but we found only one example of this approach [32]. Authors should also consider the possibility of abstaining from meta-analysis and explore the reasons for the heterogeneity instead, and we identified several reviews where this might have been better. For example, one review with extreme heterogeneity (I2 = 98%, P < 0.0001) pooled three trials with a random effects model although none of them had overlapping confidence intervals (Figure 5)[20].\nExample of extreme heterogeneity [20].\nAlthough our sample consisted only of reviews with substantial heterogeneity, about a third of the authors had not paid any attention to it. This omission was quite uniform over the spectrum of I2 values, and it might therefore partly reflect the well-known lack of statistical skills among authors of medical research papers [33-35]. However, as authors are recommended to routinely assess whether the results are consistent across studies [2], and what the likely causes are if they are not, they could do better even without having access to statistical expertise. Cochrane review groups could also do better, as they are required to have access to statistical expertise [2]. Recently, summary of findings tables were introduced in Cochrane reviews as part of GRADEprofiler, where the authors are asked to assess the quality of the body of evidence. This includes assessing the likelihood that the pooled estimate for each outcome is free from bias [2], and a judgment related to the degree of heterogeneity.\nReviews that were devoid of major problems had included more trials than those with problems. The likely reason for this is that authors are usually too influenced by whether or not a P-value is significant and often do not take into account, or do not know, that P-values depend on the number of trials. When fewer trials are included, it is harder to identify heterogeneity using a chi-square test. This test is therefore not the recommended way to investigate heterogeneity [1]. I2 is more sensitive but with few included trials there is a small risk of false positives.\n[SUBTITLE] Limitations [SUBSECTION] Our sampling method precludes us from drawing general conclusions about the quality of Cochrane reviews in relation to heterogeneity. As we sampled meta-analyses, we did not assess how often the authors had abstained from pooling the results because of heterogeneity, which would have been an arduous task, given our total sample of 3,385 reviews.\nThe most important assessment - whether a review was devoid of major problems related to heterogeneity - was not as thoroughly specified in our protocol as we would have wished. It would not have been possible to specify in advance rigid rules because of the great diversity in handling and reporting heterogeneity. We have compensated for this limitation by describing the problematic reviews we encountered. More strict criteria could be used in future studies based on our findings.\nIn a few reviews, our outcome was not a primary one, which could be the reason that the heterogeneity was not addressed. On the other hand, these reviews tended to not address heterogeneity at all, for any outcomes.\nWe specified in our protocol that we wanted to investigate to which extent the point estimates and the confidence interval varied when a different model was chosen, but decided to focus on reviews where the result changed from significant to nonsignificant and vice versa.\nSome of our analyses were exploratory. During data extraction, we decided to investigate if there was a relation between the choice of model and the P-value for heterogeneity, and we couldn't help noticing that the reviews we judged to be most problematic also tended to be those that had included fewest trials.\nIt is known that I2 increases when the sizes of the included studies increase and alternative measures of heterogeneity have been suggested [36]. However, the problematic reviews identified in our study included very few trials and relatively few participants. When there are only few included trials there is a small risk of I2 above 50% even though no heterogeneity is present.\nOur sampling method precludes us from drawing general conclusions about the quality of Cochrane reviews in relation to heterogeneity. As we sampled meta-analyses, we did not assess how often the authors had abstained from pooling the results because of heterogeneity, which would have been an arduous task, given our total sample of 3,385 reviews.\nThe most important assessment - whether a review was devoid of major problems related to heterogeneity - was not as thoroughly specified in our protocol as we would have wished. It would not have been possible to specify in advance rigid rules because of the great diversity in handling and reporting heterogeneity. We have compensated for this limitation by describing the problematic reviews we encountered. More strict criteria could be used in future studies based on our findings.\nIn a few reviews, our outcome was not a primary one, which could be the reason that the heterogeneity was not addressed. On the other hand, these reviews tended to not address heterogeneity at all, for any outcomes.\nWe specified in our protocol that we wanted to investigate to which extent the point estimates and the confidence interval varied when a different model was chosen, but decided to focus on reviews where the result changed from significant to nonsignificant and vice versa.\nSome of our analyses were exploratory. During data extraction, we decided to investigate if there was a relation between the choice of model and the P-value for heterogeneity, and we couldn't help noticing that the reviews we judged to be most problematic also tended to be those that had included fewest trials.\nIt is known that I2 increases when the sizes of the included studies increase and alternative measures of heterogeneity have been suggested [36]. However, the problematic reviews identified in our study included very few trials and relatively few participants. When there are only few included trials there is a small risk of I2 above 50% even though no heterogeneity is present.\n[SUBTITLE] Other studies of heterogeneity [SUBSECTION] In the early years of the Cochrane Collaboration, randomly selected Cochrane reviews were assessed by two different observers, and 29% were judged to have major problems [37], but these concerned other issues than heterogeneity. In another study of Cochrane reviews, heterogeneity, defined as P < 0.10, was identified in 34 out of 86 meta-analyses, and in 12 of the 34 meta-analyses, heterogeneity was not addressed [38]. In 2002, Higgins et al. [7] investigated the newest Cochrane reviews and tested if heterogeneity was present, and collected information about choice of model and subgroup analyses. The study compared the protocol to the review and identified problems concerning choice of statistical model and problems with conducting subgroup analyses, as there were often too few included trials.\nIn the early years of the Cochrane Collaboration, randomly selected Cochrane reviews were assessed by two different observers, and 29% were judged to have major problems [37], but these concerned other issues than heterogeneity. In another study of Cochrane reviews, heterogeneity, defined as P < 0.10, was identified in 34 out of 86 meta-analyses, and in 12 of the 34 meta-analyses, heterogeneity was not addressed [38]. In 2002, Higgins et al. [7] investigated the newest Cochrane reviews and tested if heterogeneity was present, and collected information about choice of model and subgroup analyses. The study compared the protocol to the review and identified problems concerning choice of statistical model and problems with conducting subgroup analyses, as there were often too few included trials.", "Our sampling method precludes us from drawing general conclusions about the quality of Cochrane reviews in relation to heterogeneity. As we sampled meta-analyses, we did not assess how often the authors had abstained from pooling the results because of heterogeneity, which would have been an arduous task, given our total sample of 3,385 reviews.\nThe most important assessment - whether a review was devoid of major problems related to heterogeneity - was not as thoroughly specified in our protocol as we would have wished. It would not have been possible to specify in advance rigid rules because of the great diversity in handling and reporting heterogeneity. We have compensated for this limitation by describing the problematic reviews we encountered. More strict criteria could be used in future studies based on our findings.\nIn a few reviews, our outcome was not a primary one, which could be the reason that the heterogeneity was not addressed. On the other hand, these reviews tended to not address heterogeneity at all, for any outcomes.\nWe specified in our protocol that we wanted to investigate to which extent the point estimates and the confidence interval varied when a different model was chosen, but decided to focus on reviews where the result changed from significant to nonsignificant and vice versa.\nSome of our analyses were exploratory. During data extraction, we decided to investigate if there was a relation between the choice of model and the P-value for heterogeneity, and we couldn't help noticing that the reviews we judged to be most problematic also tended to be those that had included fewest trials.\nIt is known that I2 increases when the sizes of the included studies increase and alternative measures of heterogeneity have been suggested [36]. However, the problematic reviews identified in our study included very few trials and relatively few participants. When there are only few included trials there is a small risk of I2 above 50% even though no heterogeneity is present.", "In the early years of the Cochrane Collaboration, randomly selected Cochrane reviews were assessed by two different observers, and 29% were judged to have major problems [37], but these concerned other issues than heterogeneity. In another study of Cochrane reviews, heterogeneity, defined as P < 0.10, was identified in 34 out of 86 meta-analyses, and in 12 of the 34 meta-analyses, heterogeneity was not addressed [38]. In 2002, Higgins et al. [7] investigated the newest Cochrane reviews and tested if heterogeneity was present, and collected information about choice of model and subgroup analyses. The study compared the protocol to the review and identified problems concerning choice of statistical model and problems with conducting subgroup analyses, as there were often too few included trials.", "One-third of Cochrane reviews with substantial heterogeneity in the first reported outcome had major problems in relation to their handling of heterogeneity. These consisted mainly of the use of a fixed effect model without an explicit rationale for choice of that model, and lack of reservations and explanations of the likely causes of the heterogeneity. These problems became less pronounced with time, as those reviews that were most recently updated much more often used a random effects model. More attention is needed to this issue, as the problems we identified can be essential for the conclusions of the reviews.", "We all work at a Cochrane Centre.", "JBS and PCG designed the study, carried out the statistical analysis and analyzed the data. RM extracted the list of meta-analysis and revised the manuscript. JS extracted data from each meta-analysis later verified by PCG. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2288/11/22/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Results", "Choice of model in relation to degree of heterogeneity", "Significant effects in relation to choice of model", "Cautions about the heterogeneity", "Discussion", "Limitations", "Other studies of heterogeneity", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Variability among individual study results in systematic reviews virtually always occurs. This is caused partly by random error (chance) and partly by systematic differences between the trials. The variation in the true effects is called heterogeneity. Its impact on meta-analyses can be assessed by I2 that describes the percentage of the variability that is due to heterogeneity [1,2]. Values greater than 50% are - rather arbitrarily - considered substantial heterogeneity [1].\nStrategies for addressing heterogeneity in systematic reviews include checking that the data extracted from the trial reports are correct, which may often not be the case [3]; omitting meta-analysis; conducting subgroup analysis or meta-regression; choosing a fixed effect or a random effects model [2]; changing the statistical metric, e.g. from a risk difference to a relative risk [4,5]; and excluding studies.\nThe fixed effect model assumes that all the included studies are estimating the same true effect. The variation in findings among studies is therefore due to chance [2]. Each study will be assigned a weight depending on the study's precision (within-trial variance) and an overall estimate can be calculated. Small studies will contribute relatively little to the outcome because they have less precision [6].\nThe random effects model assumes that the effects being estimated in the different studies are not identical, but follow a distribution. The confidence interval takes account of the additional uncertainty in the location of the mean of the systematically different effects in the different studies (this between-trial variance is added to the within-trial variance). Small studies will therefore contribute more to the average than in a fixed effect analysis, which is reasonable because the studies represent different true effects. Thus, when heterogeneity is present, the confidence interval around a random effects pooled estimate is wider than a confidence interval around a fixed effect pooled estimate [6].\nDealing with heterogeneity is often tricky, and there is only limited advice for authors on what to do, e.g. on when a particular model should be chosen for the other [7], or when the heterogeneity becomes too large for a meaningful meta-analysis.\nAn additional complexity is that the test for detecting heterogeneity has low power when the sample sizes are small or when few trials are included. For example, 11 trials give just 10 degrees of freedom, like a t-test on two groups of 6 people each does. There is also variation in practice as to which P-value demonstrates significant heterogeneity [2], but as the power of the test is so low, it is common to choose P = 0.10. It is important to be aware, however, that the choice of statistical model should not be based on the outcome of a test of heterogeneity [2].\nThe aim of our study was to investigate how authors address different degrees of substantial heterogeneity in meta-analyses in Cochrane reviews.", "We listed all Cochrane reviews from the Cochrane Database of Systematic Reviews 2008, Issue 1, which had at least one meta-analysis and where the first outcome in the first comparison involved all studies ('total'), and not only subgroups of studies ('subtotals'). We assumed that in most cases the first outcome in the first comparison would be the primary outcome and that in the remainder, it would still be important for the review.\nThere were 3,385 Cochrane reviews, and of these, 2,354 (70%) had a least one meta-analysis, and 1,366 (40%) had a 'total' for the first outcome. Figure 1 shows that the distribution of I2 for the 1,366 meta-analyses was rather uniform, apart from a decline in numbers of reviews with the most extreme degrees of heterogeneity and a large number of reviews in the group with I2 of 0%. The latter result is expected since the calculation of I2 gives many negative values, in which case I2 is by definition set to zero [1].\nFrequency distribution of Cochrane reviews in relation to I2 for the first outcome in the first comparison.\nBecause of the relatively smooth distribution, we randomly selected 60 reviews with an I2 of more than 50% for our study, using the random numbers generator in Excel. After having assessed the 60 reviews, it was clear that we had enough information to elucidate how authors address different degrees of substantial heterogeneity.\nFor every review, one observer (JBS) copied the relevant data into an Excel spreadsheet and a second observer (PCG) checked the data. Disagreements were few and were resolved by discussion. The extracted data were: i) The selected statistical model (random or fixed); ii) Any rationale for choosing the model; iii) The critical value for considering heterogeneity statistically significant; iv) Reservations about the results in relation to choice of model and comments on the heterogeneity; v) Attempts at explaining the heterogeneity narratively, e.g. different doses, populations, length of follow-up or quality of the included studies; vi) Attempts at addressing the heterogeneity statistically, e.g. by division of studies in subgroups, test for interaction, sensitivity analysis with omission of some studies, or meta-regression; this information was extracted from the Results section and in some cases directly from the graphs; vii) Point estimate and its P-value; the point estimate was also calculated with the alternative effect model, using the built-in facility for this in The Cochrane Library; viii) The P-value for the chi-square test for heterogeneity.\nWe assessed the overall methodological quality of the review based on whether the above points were addressed at all and focusing on if there were major problems in handling and interpretation of heterogeneity. We decided a priori that using a random effects model was a reasonable way of addressing substantial heterogeneity (unless there were special circumstances, as discussed below), and our assessments therefore focused mostly on those reviews where the authors had used a fixed effect model or where only one of the two models yielded a statistically significant estimate. We strived to be conservative in our judgments. If, for example, the authors had used a fixed effect model and gave the result of the heterogeneity test in the Results section, we interpreted this as a reservation about the result in relation to choice of model even if the authors provided no comments. Similarly, when a random effects model was chosen we interpreted this as a reservation.\nWe investigated whether the choice of model depended on the degree of heterogeneity, or on the P-value for the heterogeneity test. We did this because the reviews were produced at different times. Before the I2 was developed, authors often relied on the P-value to identify heterogeneity [1].", "[SUBTITLE] Choice of model in relation to degree of heterogeneity [SUBSECTION] Several of the 60 selected reviews had been published more than once. The oldest was most recently updated in 1996 and the newest in 2007 (median 2005). A fixed effect model was used in 33 reviews (55%), and a random effects model in 27 reviews (the Cochrane software does not allow mixed models [8]). There was no correlation between degree of heterogeneity and choice of model (Figure 2, P = 0.79, Mann-Whitney test for trend, with correction for ties), in fact the average I2 was 71%, both for reviews using a fixed effect model and for those using a random effects model.\nChoice of model, reviews grouped by I2.\nThe authors selected the random effects model more often in the newest half of the reviews (updated later than 1 June 2005, Table 1), than in the oldest half. The same pattern was evident for the subgroup of reviews with marginally statistically significant heterogeneity (P-value between 0.05 and 0.10, P = 0.007, Fisher's exact test).\nChoice of model in relation to the P-value for the heterogeneity test.\nNewer reviews are those updated after 1 June 2005 (n = 31). P = 0.007 for those reviews where the heterogeneity test yielded a P-value between 0.05 and 0.10\nSeveral of the 60 selected reviews had been published more than once. The oldest was most recently updated in 1996 and the newest in 2007 (median 2005). A fixed effect model was used in 33 reviews (55%), and a random effects model in 27 reviews (the Cochrane software does not allow mixed models [8]). There was no correlation between degree of heterogeneity and choice of model (Figure 2, P = 0.79, Mann-Whitney test for trend, with correction for ties), in fact the average I2 was 71%, both for reviews using a fixed effect model and for those using a random effects model.\nChoice of model, reviews grouped by I2.\nThe authors selected the random effects model more often in the newest half of the reviews (updated later than 1 June 2005, Table 1), than in the oldest half. The same pattern was evident for the subgroup of reviews with marginally statistically significant heterogeneity (P-value between 0.05 and 0.10, P = 0.007, Fisher's exact test).\nChoice of model in relation to the P-value for the heterogeneity test.\nNewer reviews are those updated after 1 June 2005 (n = 31). P = 0.007 for those reviews where the heterogeneity test yielded a P-value between 0.05 and 0.10\n[SUBTITLE] Significant effects in relation to choice of model [SUBSECTION] A significant effect estimate (P < 0.05) was presented in 34 reviews. For 6 of the 60 reviews, a significant result changed to a non-significant result when we applied the alternative model (discussed further below). For 5 reviews, a non-significant result became significant when we used the alternative model (Table 2). These 5 reviews had all used a random effects model and addressed heterogeneity this way; they were free of major methodological problems and will therefore not be discussed further.\nResults using the authors' model and the alternative model we applied\nFor 2 of the 6 reviews [9,10] where a significant result changed to a non-significant when we used a random effects model, the authors were cautious about their heterogeneous result and didn't base their conclusion on the significant finding they had obtained with a fixed effect model, which we consider a correct approach. One review was explicit about this: \"Substantial heterogeneity was also detected (p = 0.03, I2 = 79%). Because of this, the result of this analysis should be interpreted with caution and not be considered a definitive statement\" [10].\nThe authors of the other 4 reviews were less cautious. One review [11] calculated mean differences instead of standardized mean differences, although the outcomes were measured on very different scales. Because of this error, both the means and the standard deviations differed by a factor of 10. This resulted in extreme heterogeneity (I2 = 93%, P = 0.0002) despite very low power, as only two studies were included. In the methods section, the authors promised to use a random effects model in case of heterogeneity, but this was not done (and would not have solved the other problem).\nIn another review [12], the authors calculated the standardized mean difference both with a fixed effect model (1.07, 95% confidence interval 0.43 to 1.70) and a random effects model (1.74, -0.71 to 4.19). In the methods section, they stated they would use a random effects model if heterogeneity was present, which it was (I2 = 89%, P = 0.002). With a random effects model, the result was not significant. They wrote that no definite conclusion could be made but added that there was reasonable evidence that cognitive therapy was beneficial in treating depression. We find this conclusion doubtful, given the data and their declared methods. In this example, the effect estimate calculated by the two models differs substantially due to a one small outlying study. Hence, the choice of model should have been considered and explained in detail.\nAnother review [13] used Peto's odds ratio (0.28, 0.11 to 0.73). Significant heterogeneity was present (I2 = 59%, P = 0.05), and when using the ordinary odds ratio and a random effects model, the result became 0.28 (0.05 to 1.55). The authors concluded in the abstract that albumin showed a clear benefit at preventing severe ovarian hyperstimulation syndrome, although they were much more cautious in the main text.\nThe authors of the last review [14] reported that the P-value for heterogeneity was insignificant even though it was 0.07 and the power of the heterogeneity test was very low, as there were only 5 studies. They reported less mortality in the intervention group, relative risk 0.86 (0.74 to 1.00). With a random effects model the relative risk became 0.82 (0.57 to 1.18), with P = 0.30. The authors mentioned that heterogeneity was present and noted that one outlying trial had a very low mortality in the control group. The meta-analysis was driven by a big trial, which comprised 69% of the deaths and showed the same result as the pooled result, 0.86 (0.75 to 1.00). Even so, we find it pretty bold that the authors believed in a result with borderline significance (P = 0.05), and only when using the fixed effect model, with so much heterogeneity, and with unexplained discrepancies between the results of the trials.\nA significant effect estimate (P < 0.05) was presented in 34 reviews. For 6 of the 60 reviews, a significant result changed to a non-significant result when we applied the alternative model (discussed further below). For 5 reviews, a non-significant result became significant when we used the alternative model (Table 2). These 5 reviews had all used a random effects model and addressed heterogeneity this way; they were free of major methodological problems and will therefore not be discussed further.\nResults using the authors' model and the alternative model we applied\nFor 2 of the 6 reviews [9,10] where a significant result changed to a non-significant when we used a random effects model, the authors were cautious about their heterogeneous result and didn't base their conclusion on the significant finding they had obtained with a fixed effect model, which we consider a correct approach. One review was explicit about this: \"Substantial heterogeneity was also detected (p = 0.03, I2 = 79%). Because of this, the result of this analysis should be interpreted with caution and not be considered a definitive statement\" [10].\nThe authors of the other 4 reviews were less cautious. One review [11] calculated mean differences instead of standardized mean differences, although the outcomes were measured on very different scales. Because of this error, both the means and the standard deviations differed by a factor of 10. This resulted in extreme heterogeneity (I2 = 93%, P = 0.0002) despite very low power, as only two studies were included. In the methods section, the authors promised to use a random effects model in case of heterogeneity, but this was not done (and would not have solved the other problem).\nIn another review [12], the authors calculated the standardized mean difference both with a fixed effect model (1.07, 95% confidence interval 0.43 to 1.70) and a random effects model (1.74, -0.71 to 4.19). In the methods section, they stated they would use a random effects model if heterogeneity was present, which it was (I2 = 89%, P = 0.002). With a random effects model, the result was not significant. They wrote that no definite conclusion could be made but added that there was reasonable evidence that cognitive therapy was beneficial in treating depression. We find this conclusion doubtful, given the data and their declared methods. In this example, the effect estimate calculated by the two models differs substantially due to a one small outlying study. Hence, the choice of model should have been considered and explained in detail.\nAnother review [13] used Peto's odds ratio (0.28, 0.11 to 0.73). Significant heterogeneity was present (I2 = 59%, P = 0.05), and when using the ordinary odds ratio and a random effects model, the result became 0.28 (0.05 to 1.55). The authors concluded in the abstract that albumin showed a clear benefit at preventing severe ovarian hyperstimulation syndrome, although they were much more cautious in the main text.\nThe authors of the last review [14] reported that the P-value for heterogeneity was insignificant even though it was 0.07 and the power of the heterogeneity test was very low, as there were only 5 studies. They reported less mortality in the intervention group, relative risk 0.86 (0.74 to 1.00). With a random effects model the relative risk became 0.82 (0.57 to 1.18), with P = 0.30. The authors mentioned that heterogeneity was present and noted that one outlying trial had a very low mortality in the control group. The meta-analysis was driven by a big trial, which comprised 69% of the deaths and showed the same result as the pooled result, 0.86 (0.75 to 1.00). Even so, we find it pretty bold that the authors believed in a result with borderline significance (P = 0.05), and only when using the fixed effect model, with so much heterogeneity, and with unexplained discrepancies between the results of the trials.\n[SUBTITLE] Cautions about the heterogeneity [SUBSECTION] In our judgment, 40 of the 60 reviews were devoid of major problems in relation to their handling of heterogeneity (Table 3). However, only 27 reviews (45%) gave a rationale for choice of statistical model. Our overall judgment of methodological quality was not related to I2 (P = 0.26, Mann-Whitney test for trend corrected for ties, grouping I2 into intervals of 10%). However, the unproblematic reviews contained more studies than problematic reviews, 8.0 versus 4.3, on average (P = 0.024, student t-test, two-tailed).\nOur assessments of the 60 reviews in relation to their handling of heterogeneity\nSixteen of the 20 problematic reviews had chosen a fixed effect model. There can be plausible reasons for this, even in cases with substantial heterogeneity, but authors should then explain what they are. Eight of these 20 reviews had no explanations or reservations and did not address the heterogeneity statistically [15-22]. One review [21] had only included one study, but the patients were split into two subgroups, according to whether they had a rash. Both results were significant, but one showed harm and the other benefit (Figure 3). Combining such results is inappropriate and doesn't represent today's standards of Cochrane reviews (the review was last updated in 1996).\nExtremely diverging results. Both results originate from the same study [21].\nFive other reviews used Peto's odds ratio [23,24,13,18,19], and three reviews didn't follow the analysis plan that was set out in the methods section [12,25,26], which included using a random effects model or omitting meta-analysis in case of heterogeneity, and there was no explanation why. Three other reviews paid no attention to the heterogeneity and didn't discuss it, even though the P-value was between 0.05 and 0.10 [27,14,28]. An additional review described the heterogeneity (I2 = 71%, P = 0.06) but ignored it due to \"lack of stability of the known tests\" [29], which is not a valid reason for ignoring heterogeneity. In another review, the authors divided the analysis into subgroups because they had found heterogeneity, but although the consequence was that the chi-square test for heterogeneity was no longer significant due to loss of power, the I2 actually increased, which the authors failed to comment on [30]. In yet another review, which was discussed above, the authors pooled two risk scores measured on different scales that varied by a factor of ten [11]. The complete list of included reviews can be found on the web http://sites.google.com/site/dealingwithheterogeneity/.\nIn our judgment, 40 of the 60 reviews were devoid of major problems in relation to their handling of heterogeneity (Table 3). However, only 27 reviews (45%) gave a rationale for choice of statistical model. Our overall judgment of methodological quality was not related to I2 (P = 0.26, Mann-Whitney test for trend corrected for ties, grouping I2 into intervals of 10%). However, the unproblematic reviews contained more studies than problematic reviews, 8.0 versus 4.3, on average (P = 0.024, student t-test, two-tailed).\nOur assessments of the 60 reviews in relation to their handling of heterogeneity\nSixteen of the 20 problematic reviews had chosen a fixed effect model. There can be plausible reasons for this, even in cases with substantial heterogeneity, but authors should then explain what they are. Eight of these 20 reviews had no explanations or reservations and did not address the heterogeneity statistically [15-22]. One review [21] had only included one study, but the patients were split into two subgroups, according to whether they had a rash. Both results were significant, but one showed harm and the other benefit (Figure 3). Combining such results is inappropriate and doesn't represent today's standards of Cochrane reviews (the review was last updated in 1996).\nExtremely diverging results. Both results originate from the same study [21].\nFive other reviews used Peto's odds ratio [23,24,13,18,19], and three reviews didn't follow the analysis plan that was set out in the methods section [12,25,26], which included using a random effects model or omitting meta-analysis in case of heterogeneity, and there was no explanation why. Three other reviews paid no attention to the heterogeneity and didn't discuss it, even though the P-value was between 0.05 and 0.10 [27,14,28]. An additional review described the heterogeneity (I2 = 71%, P = 0.06) but ignored it due to \"lack of stability of the known tests\" [29], which is not a valid reason for ignoring heterogeneity. In another review, the authors divided the analysis into subgroups because they had found heterogeneity, but although the consequence was that the chi-square test for heterogeneity was no longer significant due to loss of power, the I2 actually increased, which the authors failed to comment on [30]. In yet another review, which was discussed above, the authors pooled two risk scores measured on different scales that varied by a factor of ten [11]. The complete list of included reviews can be found on the web http://sites.google.com/site/dealingwithheterogeneity/.", "Several of the 60 selected reviews had been published more than once. The oldest was most recently updated in 1996 and the newest in 2007 (median 2005). A fixed effect model was used in 33 reviews (55%), and a random effects model in 27 reviews (the Cochrane software does not allow mixed models [8]). There was no correlation between degree of heterogeneity and choice of model (Figure 2, P = 0.79, Mann-Whitney test for trend, with correction for ties), in fact the average I2 was 71%, both for reviews using a fixed effect model and for those using a random effects model.\nChoice of model, reviews grouped by I2.\nThe authors selected the random effects model more often in the newest half of the reviews (updated later than 1 June 2005, Table 1), than in the oldest half. The same pattern was evident for the subgroup of reviews with marginally statistically significant heterogeneity (P-value between 0.05 and 0.10, P = 0.007, Fisher's exact test).\nChoice of model in relation to the P-value for the heterogeneity test.\nNewer reviews are those updated after 1 June 2005 (n = 31). P = 0.007 for those reviews where the heterogeneity test yielded a P-value between 0.05 and 0.10", "A significant effect estimate (P < 0.05) was presented in 34 reviews. For 6 of the 60 reviews, a significant result changed to a non-significant result when we applied the alternative model (discussed further below). For 5 reviews, a non-significant result became significant when we used the alternative model (Table 2). These 5 reviews had all used a random effects model and addressed heterogeneity this way; they were free of major methodological problems and will therefore not be discussed further.\nResults using the authors' model and the alternative model we applied\nFor 2 of the 6 reviews [9,10] where a significant result changed to a non-significant when we used a random effects model, the authors were cautious about their heterogeneous result and didn't base their conclusion on the significant finding they had obtained with a fixed effect model, which we consider a correct approach. One review was explicit about this: \"Substantial heterogeneity was also detected (p = 0.03, I2 = 79%). Because of this, the result of this analysis should be interpreted with caution and not be considered a definitive statement\" [10].\nThe authors of the other 4 reviews were less cautious. One review [11] calculated mean differences instead of standardized mean differences, although the outcomes were measured on very different scales. Because of this error, both the means and the standard deviations differed by a factor of 10. This resulted in extreme heterogeneity (I2 = 93%, P = 0.0002) despite very low power, as only two studies were included. In the methods section, the authors promised to use a random effects model in case of heterogeneity, but this was not done (and would not have solved the other problem).\nIn another review [12], the authors calculated the standardized mean difference both with a fixed effect model (1.07, 95% confidence interval 0.43 to 1.70) and a random effects model (1.74, -0.71 to 4.19). In the methods section, they stated they would use a random effects model if heterogeneity was present, which it was (I2 = 89%, P = 0.002). With a random effects model, the result was not significant. They wrote that no definite conclusion could be made but added that there was reasonable evidence that cognitive therapy was beneficial in treating depression. We find this conclusion doubtful, given the data and their declared methods. In this example, the effect estimate calculated by the two models differs substantially due to a one small outlying study. Hence, the choice of model should have been considered and explained in detail.\nAnother review [13] used Peto's odds ratio (0.28, 0.11 to 0.73). Significant heterogeneity was present (I2 = 59%, P = 0.05), and when using the ordinary odds ratio and a random effects model, the result became 0.28 (0.05 to 1.55). The authors concluded in the abstract that albumin showed a clear benefit at preventing severe ovarian hyperstimulation syndrome, although they were much more cautious in the main text.\nThe authors of the last review [14] reported that the P-value for heterogeneity was insignificant even though it was 0.07 and the power of the heterogeneity test was very low, as there were only 5 studies. They reported less mortality in the intervention group, relative risk 0.86 (0.74 to 1.00). With a random effects model the relative risk became 0.82 (0.57 to 1.18), with P = 0.30. The authors mentioned that heterogeneity was present and noted that one outlying trial had a very low mortality in the control group. The meta-analysis was driven by a big trial, which comprised 69% of the deaths and showed the same result as the pooled result, 0.86 (0.75 to 1.00). Even so, we find it pretty bold that the authors believed in a result with borderline significance (P = 0.05), and only when using the fixed effect model, with so much heterogeneity, and with unexplained discrepancies between the results of the trials.", "In our judgment, 40 of the 60 reviews were devoid of major problems in relation to their handling of heterogeneity (Table 3). However, only 27 reviews (45%) gave a rationale for choice of statistical model. Our overall judgment of methodological quality was not related to I2 (P = 0.26, Mann-Whitney test for trend corrected for ties, grouping I2 into intervals of 10%). However, the unproblematic reviews contained more studies than problematic reviews, 8.0 versus 4.3, on average (P = 0.024, student t-test, two-tailed).\nOur assessments of the 60 reviews in relation to their handling of heterogeneity\nSixteen of the 20 problematic reviews had chosen a fixed effect model. There can be plausible reasons for this, even in cases with substantial heterogeneity, but authors should then explain what they are. Eight of these 20 reviews had no explanations or reservations and did not address the heterogeneity statistically [15-22]. One review [21] had only included one study, but the patients were split into two subgroups, according to whether they had a rash. Both results were significant, but one showed harm and the other benefit (Figure 3). Combining such results is inappropriate and doesn't represent today's standards of Cochrane reviews (the review was last updated in 1996).\nExtremely diverging results. Both results originate from the same study [21].\nFive other reviews used Peto's odds ratio [23,24,13,18,19], and three reviews didn't follow the analysis plan that was set out in the methods section [12,25,26], which included using a random effects model or omitting meta-analysis in case of heterogeneity, and there was no explanation why. Three other reviews paid no attention to the heterogeneity and didn't discuss it, even though the P-value was between 0.05 and 0.10 [27,14,28]. An additional review described the heterogeneity (I2 = 71%, P = 0.06) but ignored it due to \"lack of stability of the known tests\" [29], which is not a valid reason for ignoring heterogeneity. In another review, the authors divided the analysis into subgroups because they had found heterogeneity, but although the consequence was that the chi-square test for heterogeneity was no longer significant due to loss of power, the I2 actually increased, which the authors failed to comment on [30]. In yet another review, which was discussed above, the authors pooled two risk scores measured on different scales that varied by a factor of ten [11]. The complete list of included reviews can be found on the web http://sites.google.com/site/dealingwithheterogeneity/.", "It can be challenging to choose the most appropriate model for meta-analysis, as there are pros and cons with both models [2]. It is pretty clear that the larger the heterogeneity, the harder it is to defend choosing a fixed effect model, as different studies cannot be assumed to provide estimates of a common, true effect. However, one also needs to consider that a random effects model may apply too much weight to small studies, which are often poorly done and biased. An example is shown in Figure 4, where it appeared reasonable that the authors used a fixed effect model despite pronounced heterogeneity, as it gives more weight to the only large study, which, moreover, had a result that was closer to no effect than most other studies. It is also possible that the studies were so different that they should not have been combined in a meta-analysis. Only a closer look at the review, and possibly also at the individual studies, can elucidate this, and even then, researchers may disagree what would be the most appropriate approach.\nResults analysed using a fixed effect model, which gives more weight to the only large study [22].\nIt is also important to consider that the fixed effect model only allows an inference about the studies included in the meta-analysis, whereas the random effects model allows an inference about the mean effect in a hypothetical population of studies if we can assume that the studies included in the meta-analysis constitute a random selection of studies from this hypothetical population.\nThe random effects model is more conservative than the fixed effect model in the sense that the confidence interval is broader, but sometimes the point estimate is farther from the null and the P value for the pooled effect smaller than with a fixed effect model [31].\nWhen using a random effects model, the between-study variance needs to be calculated, but if there are few studies, this cannot be calculated with any precision, and a fixed effect model is therefore sometimes used in this situation [6].\nIt was surprising that we did not find a relation between the degree of heterogeneity and the choice of model. Some Cochrane groups instruct their authors to routinely use a fixed effect model, although few statisticians would find such blanket recommendations reasonable. Furthermore, in all types of research, authors should change their planned analysis and explain why if it would not be sensible.\nReaders might be more willing to accept the results if they are robust to both types of analyses, but we found only one example of this approach [32]. Authors should also consider the possibility of abstaining from meta-analysis and explore the reasons for the heterogeneity instead, and we identified several reviews where this might have been better. For example, one review with extreme heterogeneity (I2 = 98%, P < 0.0001) pooled three trials with a random effects model although none of them had overlapping confidence intervals (Figure 5)[20].\nExample of extreme heterogeneity [20].\nAlthough our sample consisted only of reviews with substantial heterogeneity, about a third of the authors had not paid any attention to it. This omission was quite uniform over the spectrum of I2 values, and it might therefore partly reflect the well-known lack of statistical skills among authors of medical research papers [33-35]. However, as authors are recommended to routinely assess whether the results are consistent across studies [2], and what the likely causes are if they are not, they could do better even without having access to statistical expertise. Cochrane review groups could also do better, as they are required to have access to statistical expertise [2]. Recently, summary of findings tables were introduced in Cochrane reviews as part of GRADEprofiler, where the authors are asked to assess the quality of the body of evidence. This includes assessing the likelihood that the pooled estimate for each outcome is free from bias [2], and a judgment related to the degree of heterogeneity.\nReviews that were devoid of major problems had included more trials than those with problems. The likely reason for this is that authors are usually too influenced by whether or not a P-value is significant and often do not take into account, or do not know, that P-values depend on the number of trials. When fewer trials are included, it is harder to identify heterogeneity using a chi-square test. This test is therefore not the recommended way to investigate heterogeneity [1]. I2 is more sensitive but with few included trials there is a small risk of false positives.\n[SUBTITLE] Limitations [SUBSECTION] Our sampling method precludes us from drawing general conclusions about the quality of Cochrane reviews in relation to heterogeneity. As we sampled meta-analyses, we did not assess how often the authors had abstained from pooling the results because of heterogeneity, which would have been an arduous task, given our total sample of 3,385 reviews.\nThe most important assessment - whether a review was devoid of major problems related to heterogeneity - was not as thoroughly specified in our protocol as we would have wished. It would not have been possible to specify in advance rigid rules because of the great diversity in handling and reporting heterogeneity. We have compensated for this limitation by describing the problematic reviews we encountered. More strict criteria could be used in future studies based on our findings.\nIn a few reviews, our outcome was not a primary one, which could be the reason that the heterogeneity was not addressed. On the other hand, these reviews tended to not address heterogeneity at all, for any outcomes.\nWe specified in our protocol that we wanted to investigate to which extent the point estimates and the confidence interval varied when a different model was chosen, but decided to focus on reviews where the result changed from significant to nonsignificant and vice versa.\nSome of our analyses were exploratory. During data extraction, we decided to investigate if there was a relation between the choice of model and the P-value for heterogeneity, and we couldn't help noticing that the reviews we judged to be most problematic also tended to be those that had included fewest trials.\nIt is known that I2 increases when the sizes of the included studies increase and alternative measures of heterogeneity have been suggested [36]. However, the problematic reviews identified in our study included very few trials and relatively few participants. When there are only few included trials there is a small risk of I2 above 50% even though no heterogeneity is present.\nOur sampling method precludes us from drawing general conclusions about the quality of Cochrane reviews in relation to heterogeneity. As we sampled meta-analyses, we did not assess how often the authors had abstained from pooling the results because of heterogeneity, which would have been an arduous task, given our total sample of 3,385 reviews.\nThe most important assessment - whether a review was devoid of major problems related to heterogeneity - was not as thoroughly specified in our protocol as we would have wished. It would not have been possible to specify in advance rigid rules because of the great diversity in handling and reporting heterogeneity. We have compensated for this limitation by describing the problematic reviews we encountered. More strict criteria could be used in future studies based on our findings.\nIn a few reviews, our outcome was not a primary one, which could be the reason that the heterogeneity was not addressed. On the other hand, these reviews tended to not address heterogeneity at all, for any outcomes.\nWe specified in our protocol that we wanted to investigate to which extent the point estimates and the confidence interval varied when a different model was chosen, but decided to focus on reviews where the result changed from significant to nonsignificant and vice versa.\nSome of our analyses were exploratory. During data extraction, we decided to investigate if there was a relation between the choice of model and the P-value for heterogeneity, and we couldn't help noticing that the reviews we judged to be most problematic also tended to be those that had included fewest trials.\nIt is known that I2 increases when the sizes of the included studies increase and alternative measures of heterogeneity have been suggested [36]. However, the problematic reviews identified in our study included very few trials and relatively few participants. When there are only few included trials there is a small risk of I2 above 50% even though no heterogeneity is present.\n[SUBTITLE] Other studies of heterogeneity [SUBSECTION] In the early years of the Cochrane Collaboration, randomly selected Cochrane reviews were assessed by two different observers, and 29% were judged to have major problems [37], but these concerned other issues than heterogeneity. In another study of Cochrane reviews, heterogeneity, defined as P < 0.10, was identified in 34 out of 86 meta-analyses, and in 12 of the 34 meta-analyses, heterogeneity was not addressed [38]. In 2002, Higgins et al. [7] investigated the newest Cochrane reviews and tested if heterogeneity was present, and collected information about choice of model and subgroup analyses. The study compared the protocol to the review and identified problems concerning choice of statistical model and problems with conducting subgroup analyses, as there were often too few included trials.\nIn the early years of the Cochrane Collaboration, randomly selected Cochrane reviews were assessed by two different observers, and 29% were judged to have major problems [37], but these concerned other issues than heterogeneity. In another study of Cochrane reviews, heterogeneity, defined as P < 0.10, was identified in 34 out of 86 meta-analyses, and in 12 of the 34 meta-analyses, heterogeneity was not addressed [38]. In 2002, Higgins et al. [7] investigated the newest Cochrane reviews and tested if heterogeneity was present, and collected information about choice of model and subgroup analyses. The study compared the protocol to the review and identified problems concerning choice of statistical model and problems with conducting subgroup analyses, as there were often too few included trials.", "Our sampling method precludes us from drawing general conclusions about the quality of Cochrane reviews in relation to heterogeneity. As we sampled meta-analyses, we did not assess how often the authors had abstained from pooling the results because of heterogeneity, which would have been an arduous task, given our total sample of 3,385 reviews.\nThe most important assessment - whether a review was devoid of major problems related to heterogeneity - was not as thoroughly specified in our protocol as we would have wished. It would not have been possible to specify in advance rigid rules because of the great diversity in handling and reporting heterogeneity. We have compensated for this limitation by describing the problematic reviews we encountered. More strict criteria could be used in future studies based on our findings.\nIn a few reviews, our outcome was not a primary one, which could be the reason that the heterogeneity was not addressed. On the other hand, these reviews tended to not address heterogeneity at all, for any outcomes.\nWe specified in our protocol that we wanted to investigate to which extent the point estimates and the confidence interval varied when a different model was chosen, but decided to focus on reviews where the result changed from significant to nonsignificant and vice versa.\nSome of our analyses were exploratory. During data extraction, we decided to investigate if there was a relation between the choice of model and the P-value for heterogeneity, and we couldn't help noticing that the reviews we judged to be most problematic also tended to be those that had included fewest trials.\nIt is known that I2 increases when the sizes of the included studies increase and alternative measures of heterogeneity have been suggested [36]. However, the problematic reviews identified in our study included very few trials and relatively few participants. When there are only few included trials there is a small risk of I2 above 50% even though no heterogeneity is present.", "In the early years of the Cochrane Collaboration, randomly selected Cochrane reviews were assessed by two different observers, and 29% were judged to have major problems [37], but these concerned other issues than heterogeneity. In another study of Cochrane reviews, heterogeneity, defined as P < 0.10, was identified in 34 out of 86 meta-analyses, and in 12 of the 34 meta-analyses, heterogeneity was not addressed [38]. In 2002, Higgins et al. [7] investigated the newest Cochrane reviews and tested if heterogeneity was present, and collected information about choice of model and subgroup analyses. The study compared the protocol to the review and identified problems concerning choice of statistical model and problems with conducting subgroup analyses, as there were often too few included trials.", "One-third of Cochrane reviews with substantial heterogeneity in the first reported outcome had major problems in relation to their handling of heterogeneity. These consisted mainly of the use of a fixed effect model without an explicit rationale for choice of that model, and lack of reservations and explanations of the likely causes of the heterogeneity. These problems became less pronounced with time, as those reviews that were most recently updated much more often used a random effects model. More attention is needed to this issue, as the problems we identified can be essential for the conclusions of the reviews.", "We all work at a Cochrane Centre.", "JBS and PCG designed the study, carried out the statistical analysis and analyzed the data. RM extracted the list of meta-analysis and revised the manuscript. JS extracted data from each meta-analysis later verified by PCG. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2288/11/22/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adult", "Aged", "Aged, 80 and over", "Asian People", "Case-Control Studies", "China", "Female", "Humans", "Male", "Melanoma", "Middle Aged", "Prognosis", "Skin Neoplasms", "Survival Analysis", "Young Adult" ]

[ "Background", "Database Design", "Patients and Staging Evaluation", "Data Analysis", "Results", "Treatment", "Surgery", "Adjuvant Therapy", "Treatment for Stage IV Disease", "Clinical staging", "Overall Survival", "Disease-free Survival and Prognostic-free Survival", "Prognostic Factors", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Malignant melanoma demonstrates a clear demographic and ethnic disparity, and is a common malignancy in Western countries among people with light-colored skin. It has the highest incidence in Queensland, Australia [1], and is the 5th most commonly diagnosed cancer in the United States [2]. However, the malignancy is relatively rare among Africans, Hispanics, and Asians. In addition to the variations demonstrated in incidences, the Surveillance, Epidemiology, and End Results (SEER) data from the United States showed that the clinical characteristics such as pathology, anatomical origin, and patients' prognoses differ significantly among different ethnic groups [3]. Furthermore, although darker-pigmented populations including Asians may benefit from protective effect of melanin to ultraviolet radiation, worse prognoses for non-Caucasian melanoma patients including a significantly shorter survival time have been reported [4-6].\nBased on the published literatures and reported disparities, it is reasonable to postulate that malignant melanoma diagnosed in non-Caucasian population differs from those diagnosed in endemic regions. However, despite the extensively published results from Western countries, knowledge of melanoma in Asian patients is scant. Current available literatures for the disease in Asia are limited to a survey, a small retrospective series, as well as an epidemiology report from U.S. that included Asian patients as a minority group [3,7-9]. Clinical evidence for Asian patients especially of large-scale does not exist due to, at least in part, the rarity of the disease in this region. As such, the etiology, characteristics, biological behavior, as well as outcome after treatment are largely unknown for melanoma in Asian patients. Clearly, additional knowledge on the characteristics of the disease as well as the outcome after treatment is needed to permit a better understanding of this highly aggressive and racial specific malignancy.\nThe aim of this analysis is to bolster the existing but highly limited literatures on malignant melanoma of Asian people by documenting the clinical presentation as well as outcome after active treatment of a relatively large group of Chinese patients with pathologically confirmed malignant melanoma recently treated in our tertiary referral center specialized for the management of this disease.", "A database was prospectively designed for the current analyses after the approval by the institutional review board (IRB) of the Beijing Cancer Hospital (BCH) (Beijing, China) prior to the retrieval and reviewing of medical records. Elements of the database were based on published prognostic factors in addition to basic demographic data, and included characteristics of the patients (age at diagnosis, gender, and performance status), disease (anatomic location, histological subtype, presence of ulceration, Breslow thickness, and stage), treatment (modality of treatment, type of surgery and systemic treatment agents used), and follow-up (time of disease recurrence/progression, patients' death, interval of local control and disease-free survival).\nAfter the designing of the database, medical records of all patients presented to the Department of Renal Cancer and Melanoma of BCH with pathologically confirmed malignant melanoma were retrieved, reviewed, and accrued into the database.", "According to the institutional protocol, pretreatment evaluation of all patients consisted of a complete history and physical examination, biopsy of the primary or secondary lesion with pathology study, complete blood count, serum chemistry, metabolic and liver panels, thoracic CT scan, and abdominal CT or ultrasound. Local excision is performed in patients with clinical stage I cutaneous melanoma without biopsy. Additional evaluation tests were required for patients who were accrued into our institutional prospective trials.\nThe American Joint Committee on Cancer (AJCC) staging system (6th edition) was used for either clinical or pathological staging [10]. For externally referred patients without pathological evaluation of the Breslow thickness, clinical stages were determined based on the available T-category, regional lymph node involvement, as well as status of distant metastasis.", "The duration of time to locoregional failure or distant metastasis was measured from the end of treatment for non-metastatic diseases until documented treatment failure. The duration of progression-free survival (PFS) in stage IV patients were from the initiation of any treatment (either systemic or local palliation) until documented disease progression at any site. The duration of overall survival (OS) was calculated from the pathologic diagnosis of melanoma until death or until the date of the last follow-up visit for patients still alive.\nThe actuarial local control, disease-free survival (DFS), PFS, and overall survival rates were calculated by Kaplan-Meier method [11]. Mantel-Cox log-rank test stratified by every factor was applied to compare the Kaplan-Meier curves for survival. Multivariate analyses of prognosticators for OS, DFS, and PFS were performed using Cox proportional hazard model. Variables with a P < 0.10 in univariate analyses were included in multivariate analysis.", "Between January 2006 and March 2010, a total of 522 consecutive and non-selected cases of malignant melanoma with pathologic confirmation were identified and reviewed. No patient was excluded in this analysis. The median follow-up time for the entire group of patients was 16 months (range 1-87 months).\nAmong the 357 cases of cutaneous melanoma including acral lentiginous melanoma (ALM), Superficial Spreading Melanoma (SSM), nodular melanoma (NM), and Lentigo Maligna Melanoma (LMM), 234 (65.5%) patients had ulceration in their primary lesion. Characteristics of the patients and their diseases are detailed in Table 1 and Figure 1.\nCharacteristics of the patients (at diagnosis) and their diseases\n* Abbreviation: LMM, lentigo maligna melanoma; ALM, acral lentiginous melanoma; MCM, Mucosal melanoma; NM, nodular melanoma; SSM, superficial spreading melanoma.\n† Only includes cutaneous melanoma. MCM and unclassifiable type were excluded.\n‡ Includes 4 mm but excludes 1 mm.\nPatient distribution by (A) anatomic location (H-N, head and neck) and (B) histological type (Un, unclassifiable).\n[SUBTITLE] Treatment [SUBSECTION] [SUBTITLE] Surgery [SUBSECTION] All patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\nAll patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\n[SUBTITLE] Adjuvant Therapy [SUBSECTION] Among patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\nAmong patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\n[SUBTITLE] Treatment for Stage IV Disease [SUBSECTION] All patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\nAll patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\n[SUBTITLE] Surgery [SUBSECTION] All patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\nAll patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\n[SUBTITLE] Adjuvant Therapy [SUBSECTION] Among patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\nAmong patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\n[SUBTITLE] Treatment for Stage IV Disease [SUBSECTION] All patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\nAll patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\n[SUBTITLE] Clinical staging [SUBSECTION] Since Breslow thickness was available in 180 of the 455 patients underwent surgery, clinical staging was based on the available knowledge of T-category, regional nodal status, and distant metastasis.\nThirty-two patients presented with limited primary disease (≤ 1.0 mm or ≤ 2.0 mm without ulceration based on pathology study or clinical evaluation) were staged as stage I disease. A total of 292 patients with more advanced T-disease on pathology or physical examination but no evidence of regional or distant metastasis were classified as stage II melanoma. And 131 patients with regional lymph adenopathy without distant metastasis, and 67 patients with distant metastasis were staged as stage III and stage IV diseases, respectively, according to the AJCC clinical classification for melanoma.\nSince close to 40% of all patients had stage III or IV diseases, and Breslow thickness were recorded in less than 70% of patients with stage I and II diseases, pathological TNM staging was not established in our database.\nSince Breslow thickness was available in 180 of the 455 patients underwent surgery, clinical staging was based on the available knowledge of T-category, regional nodal status, and distant metastasis.\nThirty-two patients presented with limited primary disease (≤ 1.0 mm or ≤ 2.0 mm without ulceration based on pathology study or clinical evaluation) were staged as stage I disease. A total of 292 patients with more advanced T-disease on pathology or physical examination but no evidence of regional or distant metastasis were classified as stage II melanoma. And 131 patients with regional lymph adenopathy without distant metastasis, and 67 patients with distant metastasis were staged as stage III and stage IV diseases, respectively, according to the AJCC clinical classification for melanoma.\nSince close to 40% of all patients had stage III or IV diseases, and Breslow thickness were recorded in less than 70% of patients with stage I and II diseases, pathological TNM staging was not established in our database.\n[SUBTITLE] Overall Survival [SUBSECTION] The 5-year overall survival rate of all 522 patients was 41.6%. The median survival time (MST) was 3.92 years (95% CI, 3.282 to 4.558) (Figure 2A).\nKaplan-Meier analyses of overall survival (OS) for the entire group of patients according to different stratums by prognostic factors (overall comparison was administered by Mantel-Cox log-rank test). Overall survival (A), OS based on stage (P < .001) (B), histology (P < .001) (C), MCM and non-MCM patients (P = .036) (D), Breslow thickness (P = .29) (E), and Ulceration status (P = .08) (F).\nKaplan-Meier analyses of disease-free survival (DFS) of 450 patients with stage I-III melanoma received definitive therapy according to different stratums by prognostic factors (overall comparison was performed by Mantel-Cox log-rank test): (A) Male vs. Female (P = .035), (B) DFS estimates according to stage (P < .001), (C) DFS of patients receiving different excision (EE for extended excision, LE for local excision, LND for lymph node dissection) of primary tumor (P < .001), (D) DFS of patients receiving different adjuvant therapy (P < .001).\nThe overall survival rates at 5-years stratified by stages at diagnosis were 94.1%, 44.0%, 38.4%, and 4.6%, respectively for stages I-IV diseases (Figure 2B) (P < 0.001); The median survival for stage I patients was not reached after a median follow-up of 12 months (range: 3-48 months). The median survival time were 4.25 years (95% CI, 3.557-4.943), 2.83 years (95% CI, 0.595-5.065) and 1.42 years (95% CI, 1.166-1.674), respectively for patients with stage II, III, and IV diseases.\nThe median survival time was 3.58 versus 4.67 years (P = 0.036) for patients with mucosal melanoma (MCM) versus non-MCM (Figure 2D). And the 5-year survival rates for MCM and ALM were 26.8% versus 53.9%, respectively (P = 0.003)\nNo significant difference in OS was observed among patients with different Breslow thickness of the primary tumor, although the 5-year OS of patients with disease ≤ 1 mm (92.3%) was higher than the other 2 groups (48.9% and 50.1% for disease of 1-4 mm and > 4 mm in thickness) (Figure 2E).\nUnivariate analyses revealed that histological type, clinical stage, as well as origin of the disease (mucosa vs. non-mucosa) were significant prognosticators for overall survival (Table 2).\nUnivariate analyses of prognostic factors for OS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\nThe 5-year overall survival rate of all 522 patients was 41.6%. The median survival time (MST) was 3.92 years (95% CI, 3.282 to 4.558) (Figure 2A).\nKaplan-Meier analyses of overall survival (OS) for the entire group of patients according to different stratums by prognostic factors (overall comparison was administered by Mantel-Cox log-rank test). Overall survival (A), OS based on stage (P < .001) (B), histology (P < .001) (C), MCM and non-MCM patients (P = .036) (D), Breslow thickness (P = .29) (E), and Ulceration status (P = .08) (F).\nKaplan-Meier analyses of disease-free survival (DFS) of 450 patients with stage I-III melanoma received definitive therapy according to different stratums by prognostic factors (overall comparison was performed by Mantel-Cox log-rank test): (A) Male vs. Female (P = .035), (B) DFS estimates according to stage (P < .001), (C) DFS of patients receiving different excision (EE for extended excision, LE for local excision, LND for lymph node dissection) of primary tumor (P < .001), (D) DFS of patients receiving different adjuvant therapy (P < .001).\nThe overall survival rates at 5-years stratified by stages at diagnosis were 94.1%, 44.0%, 38.4%, and 4.6%, respectively for stages I-IV diseases (Figure 2B) (P < 0.001); The median survival for stage I patients was not reached after a median follow-up of 12 months (range: 3-48 months). The median survival time were 4.25 years (95% CI, 3.557-4.943), 2.83 years (95% CI, 0.595-5.065) and 1.42 years (95% CI, 1.166-1.674), respectively for patients with stage II, III, and IV diseases.\nThe median survival time was 3.58 versus 4.67 years (P = 0.036) for patients with mucosal melanoma (MCM) versus non-MCM (Figure 2D). And the 5-year survival rates for MCM and ALM were 26.8% versus 53.9%, respectively (P = 0.003)\nNo significant difference in OS was observed among patients with different Breslow thickness of the primary tumor, although the 5-year OS of patients with disease ≤ 1 mm (92.3%) was higher than the other 2 groups (48.9% and 50.1% for disease of 1-4 mm and > 4 mm in thickness) (Figure 2E).\nUnivariate analyses revealed that histological type, clinical stage, as well as origin of the disease (mucosa vs. non-mucosa) were significant prognosticators for overall survival (Table 2).\nUnivariate analyses of prognostic factors for OS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\n[SUBTITLE] Disease-free Survival and Prognostic-free Survival [SUBSECTION] The 5-year disease-free survival (DFS) rate and the median survival time of 455 patients with stage I-III diseases were 12.3% (95% CI, 7.4%-17.2%) and 20 months, respectively.\nThe histology subtypes, Breslow thickness, presence of ulceration, location of the primary disease, and age had no significant value in predicting DFS in univariate analyses (Table 3).\nUnivariate analyses of prognostic factors for DFS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\nUnivariate analysis of all factors for progression-free survival (PFS) among patients with stage IV diseases was not significant.\nThe 5-year disease-free survival (DFS) rate and the median survival time of 455 patients with stage I-III diseases were 12.3% (95% CI, 7.4%-17.2%) and 20 months, respectively.\nThe histology subtypes, Breslow thickness, presence of ulceration, location of the primary disease, and age had no significant value in predicting DFS in univariate analyses (Table 3).\nUnivariate analyses of prognostic factors for DFS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\nUnivariate analysis of all factors for progression-free survival (PFS) among patients with stage IV diseases was not significant.\n[SUBTITLE] Prognostic Factors [SUBSECTION] Potential factors for overall survival (OS) of all patients, disease-free survival (DFS) for patients with stage I-III diseases indicated by univariate analyses described above that demonstrated significance or a trend (P < 0.10) were further analyzed in multivariate analysis to identify the independent prognostic factors.\nThe results of multivariate analysis indicated that stage at diagnosis and the presence of ulceration were two significant predictive factors, whereas anatomic origin of the disease and histological subtype provided no significant prognostic value for OS. Surgical modality (extended surgery vs. local excision) and the use of adjuvant therapy were significant prognosticators for DFS of patients with stage I-III malignant melanoma. Stage of the disease demonstrated a trend in predicting DFS in non-metastatic melanoma patients (Table 4).\nMultivariate analyses of prognostic factors for OS and DFS\n* DFS in non-metastatic patients who received definitive surgery.\nPotential factors for overall survival (OS) of all patients, disease-free survival (DFS) for patients with stage I-III diseases indicated by univariate analyses described above that demonstrated significance or a trend (P < 0.10) were further analyzed in multivariate analysis to identify the independent prognostic factors.\nThe results of multivariate analysis indicated that stage at diagnosis and the presence of ulceration were two significant predictive factors, whereas anatomic origin of the disease and histological subtype provided no significant prognostic value for OS. Surgical modality (extended surgery vs. local excision) and the use of adjuvant therapy were significant prognosticators for DFS of patients with stage I-III malignant melanoma. Stage of the disease demonstrated a trend in predicting DFS in non-metastatic melanoma patients (Table 4).\nMultivariate analyses of prognostic factors for OS and DFS\n* DFS in non-metastatic patients who received definitive surgery.", "[SUBTITLE] Surgery [SUBSECTION] All patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\nAll patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\n[SUBTITLE] Adjuvant Therapy [SUBSECTION] Among patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\nAmong patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\n[SUBTITLE] Treatment for Stage IV Disease [SUBSECTION] All patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\nAll patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.", "All patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.", "Among patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.", "All patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.", "Since Breslow thickness was available in 180 of the 455 patients underwent surgery, clinical staging was based on the available knowledge of T-category, regional nodal status, and distant metastasis.\nThirty-two patients presented with limited primary disease (≤ 1.0 mm or ≤ 2.0 mm without ulceration based on pathology study or clinical evaluation) were staged as stage I disease. A total of 292 patients with more advanced T-disease on pathology or physical examination but no evidence of regional or distant metastasis were classified as stage II melanoma. And 131 patients with regional lymph adenopathy without distant metastasis, and 67 patients with distant metastasis were staged as stage III and stage IV diseases, respectively, according to the AJCC clinical classification for melanoma.\nSince close to 40% of all patients had stage III or IV diseases, and Breslow thickness were recorded in less than 70% of patients with stage I and II diseases, pathological TNM staging was not established in our database.", "The 5-year overall survival rate of all 522 patients was 41.6%. The median survival time (MST) was 3.92 years (95% CI, 3.282 to 4.558) (Figure 2A).\nKaplan-Meier analyses of overall survival (OS) for the entire group of patients according to different stratums by prognostic factors (overall comparison was administered by Mantel-Cox log-rank test). Overall survival (A), OS based on stage (P < .001) (B), histology (P < .001) (C), MCM and non-MCM patients (P = .036) (D), Breslow thickness (P = .29) (E), and Ulceration status (P = .08) (F).\nKaplan-Meier analyses of disease-free survival (DFS) of 450 patients with stage I-III melanoma received definitive therapy according to different stratums by prognostic factors (overall comparison was performed by Mantel-Cox log-rank test): (A) Male vs. Female (P = .035), (B) DFS estimates according to stage (P < .001), (C) DFS of patients receiving different excision (EE for extended excision, LE for local excision, LND for lymph node dissection) of primary tumor (P < .001), (D) DFS of patients receiving different adjuvant therapy (P < .001).\nThe overall survival rates at 5-years stratified by stages at diagnosis were 94.1%, 44.0%, 38.4%, and 4.6%, respectively for stages I-IV diseases (Figure 2B) (P < 0.001); The median survival for stage I patients was not reached after a median follow-up of 12 months (range: 3-48 months). The median survival time were 4.25 years (95% CI, 3.557-4.943), 2.83 years (95% CI, 0.595-5.065) and 1.42 years (95% CI, 1.166-1.674), respectively for patients with stage II, III, and IV diseases.\nThe median survival time was 3.58 versus 4.67 years (P = 0.036) for patients with mucosal melanoma (MCM) versus non-MCM (Figure 2D). And the 5-year survival rates for MCM and ALM were 26.8% versus 53.9%, respectively (P = 0.003)\nNo significant difference in OS was observed among patients with different Breslow thickness of the primary tumor, although the 5-year OS of patients with disease ≤ 1 mm (92.3%) was higher than the other 2 groups (48.9% and 50.1% for disease of 1-4 mm and > 4 mm in thickness) (Figure 2E).\nUnivariate analyses revealed that histological type, clinical stage, as well as origin of the disease (mucosa vs. non-mucosa) were significant prognosticators for overall survival (Table 2).\nUnivariate analyses of prognostic factors for OS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness", "The 5-year disease-free survival (DFS) rate and the median survival time of 455 patients with stage I-III diseases were 12.3% (95% CI, 7.4%-17.2%) and 20 months, respectively.\nThe histology subtypes, Breslow thickness, presence of ulceration, location of the primary disease, and age had no significant value in predicting DFS in univariate analyses (Table 3).\nUnivariate analyses of prognostic factors for DFS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\nUnivariate analysis of all factors for progression-free survival (PFS) among patients with stage IV diseases was not significant.", "Potential factors for overall survival (OS) of all patients, disease-free survival (DFS) for patients with stage I-III diseases indicated by univariate analyses described above that demonstrated significance or a trend (P < 0.10) were further analyzed in multivariate analysis to identify the independent prognostic factors.\nThe results of multivariate analysis indicated that stage at diagnosis and the presence of ulceration were two significant predictive factors, whereas anatomic origin of the disease and histological subtype provided no significant prognostic value for OS. Surgical modality (extended surgery vs. local excision) and the use of adjuvant therapy were significant prognosticators for DFS of patients with stage I-III malignant melanoma. Stage of the disease demonstrated a trend in predicting DFS in non-metastatic melanoma patients (Table 4).\nMultivariate analyses of prognostic factors for OS and DFS\n* DFS in non-metastatic patients who received definitive surgery.", "In the current analyses of 522 Asian patients diagnosed with malignant melanoma, we discovered that histology subtypes of melanoma diagnosed in Asian patients differ substantially from those reported in Western populations. Specifically, the two most commonly diagnosed subtypes were acral lentiginous melanoma (ALM) and mucosal melanoma (MCM), which accounted for close to 65% of all patients collectively. The overall survival of Asian patients with melanoma was clearly suboptimal: The 5-year overall and disease-free survival (DFS) rates and median survival time were 41.6% and 43 months, and 12.3% and 20 months, respectively. These results were substantially worse than those observed in the SEER data from the United States (5-year survival rate of 91.4%) [12]. However, they were more comparable to those patients treated for high-risk disease at 37%/1.7 years and 26%/1.98 years, respectively for overall survival and DFS [13].\nA number of prognostic factors were analyzed, and demonstrated that in addition to the two most commonly observed prognostic factors emphasized by the current version of AJCC staging system, i.e., stage of the disease and the presence of ulceration, other factors including the pathology subtypes and the origin of the primary disease were not significant for predicting overall survival (OS) in multivariate analyses. Nevertheless, the use of adjuvant therapy and extent of surgery were found to be significant factors, and stage at diagnosis showed a clear trend, in predicting the disease-free survival (DFS) after treatment for patients with non-metastatic melanoma.\nStage and the extent of the primary disease (i.e., Breslow thickness) have been repeatedly confirmed to be the most important prognostic indicators for melanoma [3,14,15]. However, most evidences presented in the literatures were originated in endemic regions particularly Western countries, and the most commonly diagnosed subtype of malignant melanoma is superficial spreading melanoma (SMM). Due to the high prevalence of the disease, knowledge on diagnosis and screening is readily available and diagnoses are usually made in relatively earlier stages. The applicability of the AJCC/UICC TNM staging system in melanoma patients from non-endemic regions particularly for histological subtypes rarely observed in endemic regions has not been adequately addressed. Although malignant melanoma is a relatively rare disease in Asia, the incidence of melanoma is increasing. In a nationwide survey of 4495 cases of melanoma from 1992 to 1998 in Japan, the incidence of the disease increased 5-folds [16]. A similar pattern was observed in the metropolitan areas in China, although systemically established tumor registry is lacking.\nThe lack of knowledge of the disease in Asian patients clearly hampers the understanding thus the development of proper treatment strategy in this particular group of patients, as well as further research for more effective diagnosis and treatment. Therefore, we consider our finding important as the results confirmed the applicability of the updated AJCC staging to Asian patients with different subtypes of malignant melanoma. Both univariate and multivariate analyses of our data showed that clinical staging as well as presence of ulceration were two significant prognosticator for overall survival, in consistent with those reported in endemic regions including United States and Australia, regardless of histological subtypes [10-18]. In addition, a trend was demonstrated in our series for clinical staging in predicting DFS for patients with non-metastatic disease.\nAmong the five histological types in our series, ALM is the most common type in China accounting for nearly half of all patients, while the sum of NM, SMM and LMM is less than half of ALM. These results are consistent with the data from other Asian countries [7-9], but differed from those reported in the endemic regions where SSM, NM, LMM, and ALM account for > 70%, 15%, 13%, and 2-3%, respectively [3,15,18]. It is suggested that racial status play a key role in worldwide proportion of different histological types. In 1976, RJ Reed first described ALM and noted that this type of melanoma was the most common expression of melanoma in blacks [20]. Lately, ALM was found to be more commonly diagnosed in Asians and African Americans. In addition, African-Americans were found to have significantly shorter survival time as compared to their Caucasian counterparts in the United States [5]. Results from the series from Taiwan also suggested that ALM possess worse prognosis as compared to other types of melanoma such as SSM [9,19,21]. In the current series, we found that histological subtype was a significant prognosticator for overall survival in univariate analysis. However, when stage and presence of ulceration was added, multivariate analyses revealed that histological subtypes per se was not a significant predictor for OS. One of the potential reasons for this discovery was that ALM might have higher incidence of non-cutaneous (e.g., mucosal lesions) thereby a later stage presentation at diagnosis as compared to cutaneous SSM in endemic regions. Whereas in China, stage III and IV diseases were the majority thus the confounding effect of late diagnosis of ALM does not exist. Since the distribution of various subtypes of melanoma in our series was uneven, directly comparison for clinical presentation and outcome between ALM versus SSM, as well as other pathologies were not feasible.\nAs far as we know, this is the first large-scale observational study for malignant melanoma in Asian patients. Few reports have been published to address the epidemiology in Asia. In an updated survey from Japan published by Ishihara et al, malignant melanoma demonstrated a clear increasing incidence [22]. Furthermore, ALM and nodular melanoma (NM) were the cost common types and collectively accounted for close to 2/3 of all melanoma diagnosed in Japan. These data collide well with our results and those observed in Taiwan [9]. However, evaluation of the effects of certain treatment modality in the Japanese survey such as prophylactic lymph node dissection and adjuvant systemic therapy were inconclusive, and detailed knowledge on treatment as well as outcome is lacking due to the nature of a survey. On the other hand, our data demonstrated a clear benefit of adjuvant therapy, and suggested an efficacy of aggressive treatment in melanoma of Asian patients.\nOne of the prognostic factors that have not been addressed previously in literature is the mucosal origin of the primary. Mendenhall et al. reported the head and neck mucosal melanoma to be a rare entity comprising less than 1% for all Western melanomas. The likelihood of local recurrence after resection is approximately 50% with 5-year survival rates ranging from 20% to 50% [23]. Since mucosal melanoma occurred in 22.6% of all cases in our series, it was analyzed against non-MCM as a prognostic factor for OS and was found to be significant (P = 0.036). The primary lesions of MCM varied from nasal cavity, paranasal sinuses, oral cavity, choroids, to cervix and even vagina (not shown in this article). The features of occult onset, early recurrence after resection, and present at an advanced stage even with metastasis contribute to the poor survival of MCM.\nAlthough our study represented the largest clinical series and analyses of malignant melanoma in Asia, a number of limitations need to be addressed. Firstly, the database for the current series was prospectively designed and prepared based on the characteristics and factors published in the literature; however, the observational nature of the analysis precluded exhaustive record keeping for all the factors in all cases. As a cancer center specialized in the management of malignant melanoma, referral after surgical resection of the disease to our center is not uncommon. Malignant melanoma is among one of the rarest cancers in China, thus knowledge on proper diagnosis and treatment may not be readily available among the primary care physicians. As such, a number of parameters of the disease including Breslow thickness and the presence of ulceration were not recorded exhaustively. Thickness is the primary determinant of T-category in AJCC staging system of melanoma and in the 6th edition of AJCC in 2002, thickness thresholds were revised to 1.0, 2.0 and 4.0 mm due to a large statistic result of 17,600 patients [10,18]. Although our data revealed the 5-year OS of patients with thickness ≤ 1 mm (92.3%) was substantially higher than those of 1-4 mm (48.9%) and > 4 mm (50.1%), the incomplete data especially the extent of the primary disease may be the key reason that Breslow thickness was insignificant in predicting the overall survival in multivariate analyses. Another potential reason might be the advanced stage at presentation, i.e., the majority of patients had lymph node metastasis and distant metastasis.\nSince malignant melanoma is a relatively rare disease in Asia, the majority of patients initiated their treatment in our center were accrued into clinical trials, especially those with more advanced, i.e., stage II-IV, diseases. The effect of the heterogeneous treatment modalities used in clinical trials may also affect the treatment outcome. The results of a recently published prospective randomized trial showed that adjuvant radiation therapy to the regional lymph nodes significantly improved DFS in patients with high-risk melanoma after Lymphadenectomy [24]. In addition, adjuvant systemic treatment with interferon in patients with advanced disease is controversial, based on results of systemic reviews and meta-analyses [25,26]. Furthermore, most of these series were published in endemic regions and ALM and MCM melanoma are usually not a substantial component of the samples studied. Our analyses showed that adjuvant treatments and types of surgery in patients with non-metastatic disease might be a significant prognosticator for DFS. These findings appeared similar to those demonstrated in the literature based on other subtypes of melanoma. However, the selection of type of surgery and adjuvant treatment largely depended on the extent of disease, i.e. clinical staging in our series. The observation of an improved DFS but not OS by more extended surgical subtypes questioned whether improved local control could translate to survival in the more commonly diagnosed melanoma subtypes in Asia. Similar questions also need to be answered for the use of aggressive local and/or systemic adjuvant therapy. Clearly, more effective treatment is clearly needed. Unfortunately, the heterogeneity of our adjuvant treatment strategy precluded detailed and meaningful analyses of the efficacy of individual types of therapy.\nPatients with stage IV malignant melanoma are usually treated with systemic biological and/or chemotherapy. However, treatment for metastatic melanoma has always been a challenge, and aggressive treatment including combining chemotherapy and immunotherapy failed to show additive efficacy [27]. Furthermore, targeted therapy such as sorafenib or bevacizumab has been the focus of study for metastatic disease; however, most of the clinical trials revealed moderate or limited efficacy [28-30]. Currently, only one targeted agent demonstrated significant efficacy when used alone [31]. In our institute, stage IV melanoma patients are encouraged to participate clinical trial(s), and most of the 61 patients with stage IV disease were accrued in prospective studies including a combined molecular targeted therapy regimen. The outcomes of these clinical trials are pending.\nAlthough the current series presented the largest clinical data in Asia, epidemiology and etiology of malignant melanoma diagnosed in China were not the focus of the study. One of the interesting issues that await further research is the influence of sunlight to the incidence of cutaneous melanoma (not including ALM) in China. Currently there is no epidemiologic research reported from Asia that has addressed the cause-effect association between sun exposure and cutaneous melanoma. Although the protective effect of melanin in Asian population is well known, the anecdotal observation of the increasing incidence in China may be associated with the lifestyle changes. In addition, as a tertiary cancer institute specialized in melanoma treatment, the number of more advanced disease seen in our center may be overestimated. As such, a regional statistical investigation is needed and is currently under active planning to address the epidemiology of malignant in melanoma more effectively.", "Most malignant melanoma patients in China were diagnosed with locally advanced disease (i.e., stage II or above), and their prognoses were suboptimal. ALM and MCM are the two most commonly diagnosed pathological subtypes of malignant melanoma in China. Clinical staging and presence of ulceration was significantly associated with clinical outcome in terms of OS, while treatment strategy including extent of surgery and use of adjuvant therapy were significant predictors of DFS.", "The authors declare that they have no competing interests.", "ZC, SL, and JG participated in the study design. ZC, SL, XS, LS, CC, and MH participated in data collection and analysis. All authors participated in the interpretation and manuscript writing. SL, ZC, and JG participated in editing and proof reading. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2407/11/85/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Database Design", "Patients and Staging Evaluation", "Data Analysis", "Results", "Treatment", "Surgery", "Adjuvant Therapy", "Treatment for Stage IV Disease", "Clinical staging", "Overall Survival", "Disease-free Survival and Prognostic-free Survival", "Prognostic Factors", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Malignant melanoma demonstrates a clear demographic and ethnic disparity, and is a common malignancy in Western countries among people with light-colored skin. It has the highest incidence in Queensland, Australia [1], and is the 5th most commonly diagnosed cancer in the United States [2]. However, the malignancy is relatively rare among Africans, Hispanics, and Asians. In addition to the variations demonstrated in incidences, the Surveillance, Epidemiology, and End Results (SEER) data from the United States showed that the clinical characteristics such as pathology, anatomical origin, and patients' prognoses differ significantly among different ethnic groups [3]. Furthermore, although darker-pigmented populations including Asians may benefit from protective effect of melanin to ultraviolet radiation, worse prognoses for non-Caucasian melanoma patients including a significantly shorter survival time have been reported [4-6].\nBased on the published literatures and reported disparities, it is reasonable to postulate that malignant melanoma diagnosed in non-Caucasian population differs from those diagnosed in endemic regions. However, despite the extensively published results from Western countries, knowledge of melanoma in Asian patients is scant. Current available literatures for the disease in Asia are limited to a survey, a small retrospective series, as well as an epidemiology report from U.S. that included Asian patients as a minority group [3,7-9]. Clinical evidence for Asian patients especially of large-scale does not exist due to, at least in part, the rarity of the disease in this region. As such, the etiology, characteristics, biological behavior, as well as outcome after treatment are largely unknown for melanoma in Asian patients. Clearly, additional knowledge on the characteristics of the disease as well as the outcome after treatment is needed to permit a better understanding of this highly aggressive and racial specific malignancy.\nThe aim of this analysis is to bolster the existing but highly limited literatures on malignant melanoma of Asian people by documenting the clinical presentation as well as outcome after active treatment of a relatively large group of Chinese patients with pathologically confirmed malignant melanoma recently treated in our tertiary referral center specialized for the management of this disease.", "[SUBTITLE] Database Design [SUBSECTION] A database was prospectively designed for the current analyses after the approval by the institutional review board (IRB) of the Beijing Cancer Hospital (BCH) (Beijing, China) prior to the retrieval and reviewing of medical records. Elements of the database were based on published prognostic factors in addition to basic demographic data, and included characteristics of the patients (age at diagnosis, gender, and performance status), disease (anatomic location, histological subtype, presence of ulceration, Breslow thickness, and stage), treatment (modality of treatment, type of surgery and systemic treatment agents used), and follow-up (time of disease recurrence/progression, patients' death, interval of local control and disease-free survival).\nAfter the designing of the database, medical records of all patients presented to the Department of Renal Cancer and Melanoma of BCH with pathologically confirmed malignant melanoma were retrieved, reviewed, and accrued into the database.\nA database was prospectively designed for the current analyses after the approval by the institutional review board (IRB) of the Beijing Cancer Hospital (BCH) (Beijing, China) prior to the retrieval and reviewing of medical records. Elements of the database were based on published prognostic factors in addition to basic demographic data, and included characteristics of the patients (age at diagnosis, gender, and performance status), disease (anatomic location, histological subtype, presence of ulceration, Breslow thickness, and stage), treatment (modality of treatment, type of surgery and systemic treatment agents used), and follow-up (time of disease recurrence/progression, patients' death, interval of local control and disease-free survival).\nAfter the designing of the database, medical records of all patients presented to the Department of Renal Cancer and Melanoma of BCH with pathologically confirmed malignant melanoma were retrieved, reviewed, and accrued into the database.\n[SUBTITLE] Patients and Staging Evaluation [SUBSECTION] According to the institutional protocol, pretreatment evaluation of all patients consisted of a complete history and physical examination, biopsy of the primary or secondary lesion with pathology study, complete blood count, serum chemistry, metabolic and liver panels, thoracic CT scan, and abdominal CT or ultrasound. Local excision is performed in patients with clinical stage I cutaneous melanoma without biopsy. Additional evaluation tests were required for patients who were accrued into our institutional prospective trials.\nThe American Joint Committee on Cancer (AJCC) staging system (6th edition) was used for either clinical or pathological staging [10]. For externally referred patients without pathological evaluation of the Breslow thickness, clinical stages were determined based on the available T-category, regional lymph node involvement, as well as status of distant metastasis.\nAccording to the institutional protocol, pretreatment evaluation of all patients consisted of a complete history and physical examination, biopsy of the primary or secondary lesion with pathology study, complete blood count, serum chemistry, metabolic and liver panels, thoracic CT scan, and abdominal CT or ultrasound. Local excision is performed in patients with clinical stage I cutaneous melanoma without biopsy. Additional evaluation tests were required for patients who were accrued into our institutional prospective trials.\nThe American Joint Committee on Cancer (AJCC) staging system (6th edition) was used for either clinical or pathological staging [10]. For externally referred patients without pathological evaluation of the Breslow thickness, clinical stages were determined based on the available T-category, regional lymph node involvement, as well as status of distant metastasis.\n[SUBTITLE] Data Analysis [SUBSECTION] The duration of time to locoregional failure or distant metastasis was measured from the end of treatment for non-metastatic diseases until documented treatment failure. The duration of progression-free survival (PFS) in stage IV patients were from the initiation of any treatment (either systemic or local palliation) until documented disease progression at any site. The duration of overall survival (OS) was calculated from the pathologic diagnosis of melanoma until death or until the date of the last follow-up visit for patients still alive.\nThe actuarial local control, disease-free survival (DFS), PFS, and overall survival rates were calculated by Kaplan-Meier method [11]. Mantel-Cox log-rank test stratified by every factor was applied to compare the Kaplan-Meier curves for survival. Multivariate analyses of prognosticators for OS, DFS, and PFS were performed using Cox proportional hazard model. Variables with a P < 0.10 in univariate analyses were included in multivariate analysis.\nThe duration of time to locoregional failure or distant metastasis was measured from the end of treatment for non-metastatic diseases until documented treatment failure. The duration of progression-free survival (PFS) in stage IV patients were from the initiation of any treatment (either systemic or local palliation) until documented disease progression at any site. The duration of overall survival (OS) was calculated from the pathologic diagnosis of melanoma until death or until the date of the last follow-up visit for patients still alive.\nThe actuarial local control, disease-free survival (DFS), PFS, and overall survival rates were calculated by Kaplan-Meier method [11]. Mantel-Cox log-rank test stratified by every factor was applied to compare the Kaplan-Meier curves for survival. Multivariate analyses of prognosticators for OS, DFS, and PFS were performed using Cox proportional hazard model. Variables with a P < 0.10 in univariate analyses were included in multivariate analysis.", "A database was prospectively designed for the current analyses after the approval by the institutional review board (IRB) of the Beijing Cancer Hospital (BCH) (Beijing, China) prior to the retrieval and reviewing of medical records. Elements of the database were based on published prognostic factors in addition to basic demographic data, and included characteristics of the patients (age at diagnosis, gender, and performance status), disease (anatomic location, histological subtype, presence of ulceration, Breslow thickness, and stage), treatment (modality of treatment, type of surgery and systemic treatment agents used), and follow-up (time of disease recurrence/progression, patients' death, interval of local control and disease-free survival).\nAfter the designing of the database, medical records of all patients presented to the Department of Renal Cancer and Melanoma of BCH with pathologically confirmed malignant melanoma were retrieved, reviewed, and accrued into the database.", "According to the institutional protocol, pretreatment evaluation of all patients consisted of a complete history and physical examination, biopsy of the primary or secondary lesion with pathology study, complete blood count, serum chemistry, metabolic and liver panels, thoracic CT scan, and abdominal CT or ultrasound. Local excision is performed in patients with clinical stage I cutaneous melanoma without biopsy. Additional evaluation tests were required for patients who were accrued into our institutional prospective trials.\nThe American Joint Committee on Cancer (AJCC) staging system (6th edition) was used for either clinical or pathological staging [10]. For externally referred patients without pathological evaluation of the Breslow thickness, clinical stages were determined based on the available T-category, regional lymph node involvement, as well as status of distant metastasis.", "The duration of time to locoregional failure or distant metastasis was measured from the end of treatment for non-metastatic diseases until documented treatment failure. The duration of progression-free survival (PFS) in stage IV patients were from the initiation of any treatment (either systemic or local palliation) until documented disease progression at any site. The duration of overall survival (OS) was calculated from the pathologic diagnosis of melanoma until death or until the date of the last follow-up visit for patients still alive.\nThe actuarial local control, disease-free survival (DFS), PFS, and overall survival rates were calculated by Kaplan-Meier method [11]. Mantel-Cox log-rank test stratified by every factor was applied to compare the Kaplan-Meier curves for survival. Multivariate analyses of prognosticators for OS, DFS, and PFS were performed using Cox proportional hazard model. Variables with a P < 0.10 in univariate analyses were included in multivariate analysis.", "Between January 2006 and March 2010, a total of 522 consecutive and non-selected cases of malignant melanoma with pathologic confirmation were identified and reviewed. No patient was excluded in this analysis. The median follow-up time for the entire group of patients was 16 months (range 1-87 months).\nAmong the 357 cases of cutaneous melanoma including acral lentiginous melanoma (ALM), Superficial Spreading Melanoma (SSM), nodular melanoma (NM), and Lentigo Maligna Melanoma (LMM), 234 (65.5%) patients had ulceration in their primary lesion. Characteristics of the patients and their diseases are detailed in Table 1 and Figure 1.\nCharacteristics of the patients (at diagnosis) and their diseases\n* Abbreviation: LMM, lentigo maligna melanoma; ALM, acral lentiginous melanoma; MCM, Mucosal melanoma; NM, nodular melanoma; SSM, superficial spreading melanoma.\n† Only includes cutaneous melanoma. MCM and unclassifiable type were excluded.\n‡ Includes 4 mm but excludes 1 mm.\nPatient distribution by (A) anatomic location (H-N, head and neck) and (B) histological type (Un, unclassifiable).\n[SUBTITLE] Treatment [SUBSECTION] [SUBTITLE] Surgery [SUBSECTION] All patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\nAll patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\n[SUBTITLE] Adjuvant Therapy [SUBSECTION] Among patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\nAmong patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\n[SUBTITLE] Treatment for Stage IV Disease [SUBSECTION] All patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\nAll patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\n[SUBTITLE] Surgery [SUBSECTION] All patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\nAll patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\n[SUBTITLE] Adjuvant Therapy [SUBSECTION] Among patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\nAmong patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\n[SUBTITLE] Treatment for Stage IV Disease [SUBSECTION] All patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\nAll patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\n[SUBTITLE] Clinical staging [SUBSECTION] Since Breslow thickness was available in 180 of the 455 patients underwent surgery, clinical staging was based on the available knowledge of T-category, regional nodal status, and distant metastasis.\nThirty-two patients presented with limited primary disease (≤ 1.0 mm or ≤ 2.0 mm without ulceration based on pathology study or clinical evaluation) were staged as stage I disease. A total of 292 patients with more advanced T-disease on pathology or physical examination but no evidence of regional or distant metastasis were classified as stage II melanoma. And 131 patients with regional lymph adenopathy without distant metastasis, and 67 patients with distant metastasis were staged as stage III and stage IV diseases, respectively, according to the AJCC clinical classification for melanoma.\nSince close to 40% of all patients had stage III or IV diseases, and Breslow thickness were recorded in less than 70% of patients with stage I and II diseases, pathological TNM staging was not established in our database.\nSince Breslow thickness was available in 180 of the 455 patients underwent surgery, clinical staging was based on the available knowledge of T-category, regional nodal status, and distant metastasis.\nThirty-two patients presented with limited primary disease (≤ 1.0 mm or ≤ 2.0 mm without ulceration based on pathology study or clinical evaluation) were staged as stage I disease. A total of 292 patients with more advanced T-disease on pathology or physical examination but no evidence of regional or distant metastasis were classified as stage II melanoma. And 131 patients with regional lymph adenopathy without distant metastasis, and 67 patients with distant metastasis were staged as stage III and stage IV diseases, respectively, according to the AJCC clinical classification for melanoma.\nSince close to 40% of all patients had stage III or IV diseases, and Breslow thickness were recorded in less than 70% of patients with stage I and II diseases, pathological TNM staging was not established in our database.\n[SUBTITLE] Overall Survival [SUBSECTION] The 5-year overall survival rate of all 522 patients was 41.6%. The median survival time (MST) was 3.92 years (95% CI, 3.282 to 4.558) (Figure 2A).\nKaplan-Meier analyses of overall survival (OS) for the entire group of patients according to different stratums by prognostic factors (overall comparison was administered by Mantel-Cox log-rank test). Overall survival (A), OS based on stage (P < .001) (B), histology (P < .001) (C), MCM and non-MCM patients (P = .036) (D), Breslow thickness (P = .29) (E), and Ulceration status (P = .08) (F).\nKaplan-Meier analyses of disease-free survival (DFS) of 450 patients with stage I-III melanoma received definitive therapy according to different stratums by prognostic factors (overall comparison was performed by Mantel-Cox log-rank test): (A) Male vs. Female (P = .035), (B) DFS estimates according to stage (P < .001), (C) DFS of patients receiving different excision (EE for extended excision, LE for local excision, LND for lymph node dissection) of primary tumor (P < .001), (D) DFS of patients receiving different adjuvant therapy (P < .001).\nThe overall survival rates at 5-years stratified by stages at diagnosis were 94.1%, 44.0%, 38.4%, and 4.6%, respectively for stages I-IV diseases (Figure 2B) (P < 0.001); The median survival for stage I patients was not reached after a median follow-up of 12 months (range: 3-48 months). The median survival time were 4.25 years (95% CI, 3.557-4.943), 2.83 years (95% CI, 0.595-5.065) and 1.42 years (95% CI, 1.166-1.674), respectively for patients with stage II, III, and IV diseases.\nThe median survival time was 3.58 versus 4.67 years (P = 0.036) for patients with mucosal melanoma (MCM) versus non-MCM (Figure 2D). And the 5-year survival rates for MCM and ALM were 26.8% versus 53.9%, respectively (P = 0.003)\nNo significant difference in OS was observed among patients with different Breslow thickness of the primary tumor, although the 5-year OS of patients with disease ≤ 1 mm (92.3%) was higher than the other 2 groups (48.9% and 50.1% for disease of 1-4 mm and > 4 mm in thickness) (Figure 2E).\nUnivariate analyses revealed that histological type, clinical stage, as well as origin of the disease (mucosa vs. non-mucosa) were significant prognosticators for overall survival (Table 2).\nUnivariate analyses of prognostic factors for OS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\nThe 5-year overall survival rate of all 522 patients was 41.6%. The median survival time (MST) was 3.92 years (95% CI, 3.282 to 4.558) (Figure 2A).\nKaplan-Meier analyses of overall survival (OS) for the entire group of patients according to different stratums by prognostic factors (overall comparison was administered by Mantel-Cox log-rank test). Overall survival (A), OS based on stage (P < .001) (B), histology (P < .001) (C), MCM and non-MCM patients (P = .036) (D), Breslow thickness (P = .29) (E), and Ulceration status (P = .08) (F).\nKaplan-Meier analyses of disease-free survival (DFS) of 450 patients with stage I-III melanoma received definitive therapy according to different stratums by prognostic factors (overall comparison was performed by Mantel-Cox log-rank test): (A) Male vs. Female (P = .035), (B) DFS estimates according to stage (P < .001), (C) DFS of patients receiving different excision (EE for extended excision, LE for local excision, LND for lymph node dissection) of primary tumor (P < .001), (D) DFS of patients receiving different adjuvant therapy (P < .001).\nThe overall survival rates at 5-years stratified by stages at diagnosis were 94.1%, 44.0%, 38.4%, and 4.6%, respectively for stages I-IV diseases (Figure 2B) (P < 0.001); The median survival for stage I patients was not reached after a median follow-up of 12 months (range: 3-48 months). The median survival time were 4.25 years (95% CI, 3.557-4.943), 2.83 years (95% CI, 0.595-5.065) and 1.42 years (95% CI, 1.166-1.674), respectively for patients with stage II, III, and IV diseases.\nThe median survival time was 3.58 versus 4.67 years (P = 0.036) for patients with mucosal melanoma (MCM) versus non-MCM (Figure 2D). And the 5-year survival rates for MCM and ALM were 26.8% versus 53.9%, respectively (P = 0.003)\nNo significant difference in OS was observed among patients with different Breslow thickness of the primary tumor, although the 5-year OS of patients with disease ≤ 1 mm (92.3%) was higher than the other 2 groups (48.9% and 50.1% for disease of 1-4 mm and > 4 mm in thickness) (Figure 2E).\nUnivariate analyses revealed that histological type, clinical stage, as well as origin of the disease (mucosa vs. non-mucosa) were significant prognosticators for overall survival (Table 2).\nUnivariate analyses of prognostic factors for OS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\n[SUBTITLE] Disease-free Survival and Prognostic-free Survival [SUBSECTION] The 5-year disease-free survival (DFS) rate and the median survival time of 455 patients with stage I-III diseases were 12.3% (95% CI, 7.4%-17.2%) and 20 months, respectively.\nThe histology subtypes, Breslow thickness, presence of ulceration, location of the primary disease, and age had no significant value in predicting DFS in univariate analyses (Table 3).\nUnivariate analyses of prognostic factors for DFS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\nUnivariate analysis of all factors for progression-free survival (PFS) among patients with stage IV diseases was not significant.\nThe 5-year disease-free survival (DFS) rate and the median survival time of 455 patients with stage I-III diseases were 12.3% (95% CI, 7.4%-17.2%) and 20 months, respectively.\nThe histology subtypes, Breslow thickness, presence of ulceration, location of the primary disease, and age had no significant value in predicting DFS in univariate analyses (Table 3).\nUnivariate analyses of prognostic factors for DFS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\nUnivariate analysis of all factors for progression-free survival (PFS) among patients with stage IV diseases was not significant.\n[SUBTITLE] Prognostic Factors [SUBSECTION] Potential factors for overall survival (OS) of all patients, disease-free survival (DFS) for patients with stage I-III diseases indicated by univariate analyses described above that demonstrated significance or a trend (P < 0.10) were further analyzed in multivariate analysis to identify the independent prognostic factors.\nThe results of multivariate analysis indicated that stage at diagnosis and the presence of ulceration were two significant predictive factors, whereas anatomic origin of the disease and histological subtype provided no significant prognostic value for OS. Surgical modality (extended surgery vs. local excision) and the use of adjuvant therapy were significant prognosticators for DFS of patients with stage I-III malignant melanoma. Stage of the disease demonstrated a trend in predicting DFS in non-metastatic melanoma patients (Table 4).\nMultivariate analyses of prognostic factors for OS and DFS\n* DFS in non-metastatic patients who received definitive surgery.\nPotential factors for overall survival (OS) of all patients, disease-free survival (DFS) for patients with stage I-III diseases indicated by univariate analyses described above that demonstrated significance or a trend (P < 0.10) were further analyzed in multivariate analysis to identify the independent prognostic factors.\nThe results of multivariate analysis indicated that stage at diagnosis and the presence of ulceration were two significant predictive factors, whereas anatomic origin of the disease and histological subtype provided no significant prognostic value for OS. Surgical modality (extended surgery vs. local excision) and the use of adjuvant therapy were significant prognosticators for DFS of patients with stage I-III malignant melanoma. Stage of the disease demonstrated a trend in predicting DFS in non-metastatic melanoma patients (Table 4).\nMultivariate analyses of prognostic factors for OS and DFS\n* DFS in non-metastatic patients who received definitive surgery.", "[SUBTITLE] Surgery [SUBSECTION] All patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\nAll patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.\n[SUBTITLE] Adjuvant Therapy [SUBSECTION] Among patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\nAmong patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.\n[SUBTITLE] Treatment for Stage IV Disease [SUBSECTION] All patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.\nAll patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.", "All patients with stage I or II disease had undergone complete resection of their diseases: 95 and 229 cases received local or extended excision, respectively. Among the 131 patients with stage III diseases, local excision, extended excision, and extended excision with regional nodal dissection were performed in 35, 15, and 76 patients, respectively. All patients achieved complete resection except for one patient who had positive margin. In addition, 4 patients did not receive definitive surgery due to patients' preference or poor performance status, including 1 who had palliative surgery.\nBreslow thickness were measured and recorded in 180 cases of the cutaneous or mucosal disease in patients with stage I-III diseases after surgery. Measurements of the thickness of the disease were not performed for the rest of the group. Sentinel lymph node dissection was not utilized in the treatment of this group of patients as the majority had non-cutaneous and more advanced diseases.", "Among patients had stage I-III diseases, 155 patients had high-dose IFNα-2b treatment, 109 had adjuvant chemotherapy with or without radiation, and 2 patients received adjuvant radiotherapy only. In addition, 184 patients received no adjuvant therapy.", "All patients with stage IV were treated with systemic treatment using chemotherapy and or targeted therapy (on trial basis). One patient with unknown primary had definitive surgical resection to the solitary metastatic focus, and 62 cases had palliative surgery. The remaining 5 cases had no surgery.", "Since Breslow thickness was available in 180 of the 455 patients underwent surgery, clinical staging was based on the available knowledge of T-category, regional nodal status, and distant metastasis.\nThirty-two patients presented with limited primary disease (≤ 1.0 mm or ≤ 2.0 mm without ulceration based on pathology study or clinical evaluation) were staged as stage I disease. A total of 292 patients with more advanced T-disease on pathology or physical examination but no evidence of regional or distant metastasis were classified as stage II melanoma. And 131 patients with regional lymph adenopathy without distant metastasis, and 67 patients with distant metastasis were staged as stage III and stage IV diseases, respectively, according to the AJCC clinical classification for melanoma.\nSince close to 40% of all patients had stage III or IV diseases, and Breslow thickness were recorded in less than 70% of patients with stage I and II diseases, pathological TNM staging was not established in our database.", "The 5-year overall survival rate of all 522 patients was 41.6%. The median survival time (MST) was 3.92 years (95% CI, 3.282 to 4.558) (Figure 2A).\nKaplan-Meier analyses of overall survival (OS) for the entire group of patients according to different stratums by prognostic factors (overall comparison was administered by Mantel-Cox log-rank test). Overall survival (A), OS based on stage (P < .001) (B), histology (P < .001) (C), MCM and non-MCM patients (P = .036) (D), Breslow thickness (P = .29) (E), and Ulceration status (P = .08) (F).\nKaplan-Meier analyses of disease-free survival (DFS) of 450 patients with stage I-III melanoma received definitive therapy according to different stratums by prognostic factors (overall comparison was performed by Mantel-Cox log-rank test): (A) Male vs. Female (P = .035), (B) DFS estimates according to stage (P < .001), (C) DFS of patients receiving different excision (EE for extended excision, LE for local excision, LND for lymph node dissection) of primary tumor (P < .001), (D) DFS of patients receiving different adjuvant therapy (P < .001).\nThe overall survival rates at 5-years stratified by stages at diagnosis were 94.1%, 44.0%, 38.4%, and 4.6%, respectively for stages I-IV diseases (Figure 2B) (P < 0.001); The median survival for stage I patients was not reached after a median follow-up of 12 months (range: 3-48 months). The median survival time were 4.25 years (95% CI, 3.557-4.943), 2.83 years (95% CI, 0.595-5.065) and 1.42 years (95% CI, 1.166-1.674), respectively for patients with stage II, III, and IV diseases.\nThe median survival time was 3.58 versus 4.67 years (P = 0.036) for patients with mucosal melanoma (MCM) versus non-MCM (Figure 2D). And the 5-year survival rates for MCM and ALM were 26.8% versus 53.9%, respectively (P = 0.003)\nNo significant difference in OS was observed among patients with different Breslow thickness of the primary tumor, although the 5-year OS of patients with disease ≤ 1 mm (92.3%) was higher than the other 2 groups (48.9% and 50.1% for disease of 1-4 mm and > 4 mm in thickness) (Figure 2E).\nUnivariate analyses revealed that histological type, clinical stage, as well as origin of the disease (mucosa vs. non-mucosa) were significant prognosticators for overall survival (Table 2).\nUnivariate analyses of prognostic factors for OS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness", "The 5-year disease-free survival (DFS) rate and the median survival time of 455 patients with stage I-III diseases were 12.3% (95% CI, 7.4%-17.2%) and 20 months, respectively.\nThe histology subtypes, Breslow thickness, presence of ulceration, location of the primary disease, and age had no significant value in predicting DFS in univariate analyses (Table 3).\nUnivariate analyses of prognostic factors for DFS in patients with malignant melanoma\n*Limited to 180 cases with recorded Breslow thickness\nUnivariate analysis of all factors for progression-free survival (PFS) among patients with stage IV diseases was not significant.", "Potential factors for overall survival (OS) of all patients, disease-free survival (DFS) for patients with stage I-III diseases indicated by univariate analyses described above that demonstrated significance or a trend (P < 0.10) were further analyzed in multivariate analysis to identify the independent prognostic factors.\nThe results of multivariate analysis indicated that stage at diagnosis and the presence of ulceration were two significant predictive factors, whereas anatomic origin of the disease and histological subtype provided no significant prognostic value for OS. Surgical modality (extended surgery vs. local excision) and the use of adjuvant therapy were significant prognosticators for DFS of patients with stage I-III malignant melanoma. Stage of the disease demonstrated a trend in predicting DFS in non-metastatic melanoma patients (Table 4).\nMultivariate analyses of prognostic factors for OS and DFS\n* DFS in non-metastatic patients who received definitive surgery.", "In the current analyses of 522 Asian patients diagnosed with malignant melanoma, we discovered that histology subtypes of melanoma diagnosed in Asian patients differ substantially from those reported in Western populations. Specifically, the two most commonly diagnosed subtypes were acral lentiginous melanoma (ALM) and mucosal melanoma (MCM), which accounted for close to 65% of all patients collectively. The overall survival of Asian patients with melanoma was clearly suboptimal: The 5-year overall and disease-free survival (DFS) rates and median survival time were 41.6% and 43 months, and 12.3% and 20 months, respectively. These results were substantially worse than those observed in the SEER data from the United States (5-year survival rate of 91.4%) [12]. However, they were more comparable to those patients treated for high-risk disease at 37%/1.7 years and 26%/1.98 years, respectively for overall survival and DFS [13].\nA number of prognostic factors were analyzed, and demonstrated that in addition to the two most commonly observed prognostic factors emphasized by the current version of AJCC staging system, i.e., stage of the disease and the presence of ulceration, other factors including the pathology subtypes and the origin of the primary disease were not significant for predicting overall survival (OS) in multivariate analyses. Nevertheless, the use of adjuvant therapy and extent of surgery were found to be significant factors, and stage at diagnosis showed a clear trend, in predicting the disease-free survival (DFS) after treatment for patients with non-metastatic melanoma.\nStage and the extent of the primary disease (i.e., Breslow thickness) have been repeatedly confirmed to be the most important prognostic indicators for melanoma [3,14,15]. However, most evidences presented in the literatures were originated in endemic regions particularly Western countries, and the most commonly diagnosed subtype of malignant melanoma is superficial spreading melanoma (SMM). Due to the high prevalence of the disease, knowledge on diagnosis and screening is readily available and diagnoses are usually made in relatively earlier stages. The applicability of the AJCC/UICC TNM staging system in melanoma patients from non-endemic regions particularly for histological subtypes rarely observed in endemic regions has not been adequately addressed. Although malignant melanoma is a relatively rare disease in Asia, the incidence of melanoma is increasing. In a nationwide survey of 4495 cases of melanoma from 1992 to 1998 in Japan, the incidence of the disease increased 5-folds [16]. A similar pattern was observed in the metropolitan areas in China, although systemically established tumor registry is lacking.\nThe lack of knowledge of the disease in Asian patients clearly hampers the understanding thus the development of proper treatment strategy in this particular group of patients, as well as further research for more effective diagnosis and treatment. Therefore, we consider our finding important as the results confirmed the applicability of the updated AJCC staging to Asian patients with different subtypes of malignant melanoma. Both univariate and multivariate analyses of our data showed that clinical staging as well as presence of ulceration were two significant prognosticator for overall survival, in consistent with those reported in endemic regions including United States and Australia, regardless of histological subtypes [10-18]. In addition, a trend was demonstrated in our series for clinical staging in predicting DFS for patients with non-metastatic disease.\nAmong the five histological types in our series, ALM is the most common type in China accounting for nearly half of all patients, while the sum of NM, SMM and LMM is less than half of ALM. These results are consistent with the data from other Asian countries [7-9], but differed from those reported in the endemic regions where SSM, NM, LMM, and ALM account for > 70%, 15%, 13%, and 2-3%, respectively [3,15,18]. It is suggested that racial status play a key role in worldwide proportion of different histological types. In 1976, RJ Reed first described ALM and noted that this type of melanoma was the most common expression of melanoma in blacks [20]. Lately, ALM was found to be more commonly diagnosed in Asians and African Americans. In addition, African-Americans were found to have significantly shorter survival time as compared to their Caucasian counterparts in the United States [5]. Results from the series from Taiwan also suggested that ALM possess worse prognosis as compared to other types of melanoma such as SSM [9,19,21]. In the current series, we found that histological subtype was a significant prognosticator for overall survival in univariate analysis. However, when stage and presence of ulceration was added, multivariate analyses revealed that histological subtypes per se was not a significant predictor for OS. One of the potential reasons for this discovery was that ALM might have higher incidence of non-cutaneous (e.g., mucosal lesions) thereby a later stage presentation at diagnosis as compared to cutaneous SSM in endemic regions. Whereas in China, stage III and IV diseases were the majority thus the confounding effect of late diagnosis of ALM does not exist. Since the distribution of various subtypes of melanoma in our series was uneven, directly comparison for clinical presentation and outcome between ALM versus SSM, as well as other pathologies were not feasible.\nAs far as we know, this is the first large-scale observational study for malignant melanoma in Asian patients. Few reports have been published to address the epidemiology in Asia. In an updated survey from Japan published by Ishihara et al, malignant melanoma demonstrated a clear increasing incidence [22]. Furthermore, ALM and nodular melanoma (NM) were the cost common types and collectively accounted for close to 2/3 of all melanoma diagnosed in Japan. These data collide well with our results and those observed in Taiwan [9]. However, evaluation of the effects of certain treatment modality in the Japanese survey such as prophylactic lymph node dissection and adjuvant systemic therapy were inconclusive, and detailed knowledge on treatment as well as outcome is lacking due to the nature of a survey. On the other hand, our data demonstrated a clear benefit of adjuvant therapy, and suggested an efficacy of aggressive treatment in melanoma of Asian patients.\nOne of the prognostic factors that have not been addressed previously in literature is the mucosal origin of the primary. Mendenhall et al. reported the head and neck mucosal melanoma to be a rare entity comprising less than 1% for all Western melanomas. The likelihood of local recurrence after resection is approximately 50% with 5-year survival rates ranging from 20% to 50% [23]. Since mucosal melanoma occurred in 22.6% of all cases in our series, it was analyzed against non-MCM as a prognostic factor for OS and was found to be significant (P = 0.036). The primary lesions of MCM varied from nasal cavity, paranasal sinuses, oral cavity, choroids, to cervix and even vagina (not shown in this article). The features of occult onset, early recurrence after resection, and present at an advanced stage even with metastasis contribute to the poor survival of MCM.\nAlthough our study represented the largest clinical series and analyses of malignant melanoma in Asia, a number of limitations need to be addressed. Firstly, the database for the current series was prospectively designed and prepared based on the characteristics and factors published in the literature; however, the observational nature of the analysis precluded exhaustive record keeping for all the factors in all cases. As a cancer center specialized in the management of malignant melanoma, referral after surgical resection of the disease to our center is not uncommon. Malignant melanoma is among one of the rarest cancers in China, thus knowledge on proper diagnosis and treatment may not be readily available among the primary care physicians. As such, a number of parameters of the disease including Breslow thickness and the presence of ulceration were not recorded exhaustively. Thickness is the primary determinant of T-category in AJCC staging system of melanoma and in the 6th edition of AJCC in 2002, thickness thresholds were revised to 1.0, 2.0 and 4.0 mm due to a large statistic result of 17,600 patients [10,18]. Although our data revealed the 5-year OS of patients with thickness ≤ 1 mm (92.3%) was substantially higher than those of 1-4 mm (48.9%) and > 4 mm (50.1%), the incomplete data especially the extent of the primary disease may be the key reason that Breslow thickness was insignificant in predicting the overall survival in multivariate analyses. Another potential reason might be the advanced stage at presentation, i.e., the majority of patients had lymph node metastasis and distant metastasis.\nSince malignant melanoma is a relatively rare disease in Asia, the majority of patients initiated their treatment in our center were accrued into clinical trials, especially those with more advanced, i.e., stage II-IV, diseases. The effect of the heterogeneous treatment modalities used in clinical trials may also affect the treatment outcome. The results of a recently published prospective randomized trial showed that adjuvant radiation therapy to the regional lymph nodes significantly improved DFS in patients with high-risk melanoma after Lymphadenectomy [24]. In addition, adjuvant systemic treatment with interferon in patients with advanced disease is controversial, based on results of systemic reviews and meta-analyses [25,26]. Furthermore, most of these series were published in endemic regions and ALM and MCM melanoma are usually not a substantial component of the samples studied. Our analyses showed that adjuvant treatments and types of surgery in patients with non-metastatic disease might be a significant prognosticator for DFS. These findings appeared similar to those demonstrated in the literature based on other subtypes of melanoma. However, the selection of type of surgery and adjuvant treatment largely depended on the extent of disease, i.e. clinical staging in our series. The observation of an improved DFS but not OS by more extended surgical subtypes questioned whether improved local control could translate to survival in the more commonly diagnosed melanoma subtypes in Asia. Similar questions also need to be answered for the use of aggressive local and/or systemic adjuvant therapy. Clearly, more effective treatment is clearly needed. Unfortunately, the heterogeneity of our adjuvant treatment strategy precluded detailed and meaningful analyses of the efficacy of individual types of therapy.\nPatients with stage IV malignant melanoma are usually treated with systemic biological and/or chemotherapy. However, treatment for metastatic melanoma has always been a challenge, and aggressive treatment including combining chemotherapy and immunotherapy failed to show additive efficacy [27]. Furthermore, targeted therapy such as sorafenib or bevacizumab has been the focus of study for metastatic disease; however, most of the clinical trials revealed moderate or limited efficacy [28-30]. Currently, only one targeted agent demonstrated significant efficacy when used alone [31]. In our institute, stage IV melanoma patients are encouraged to participate clinical trial(s), and most of the 61 patients with stage IV disease were accrued in prospective studies including a combined molecular targeted therapy regimen. The outcomes of these clinical trials are pending.\nAlthough the current series presented the largest clinical data in Asia, epidemiology and etiology of malignant melanoma diagnosed in China were not the focus of the study. One of the interesting issues that await further research is the influence of sunlight to the incidence of cutaneous melanoma (not including ALM) in China. Currently there is no epidemiologic research reported from Asia that has addressed the cause-effect association between sun exposure and cutaneous melanoma. Although the protective effect of melanin in Asian population is well known, the anecdotal observation of the increasing incidence in China may be associated with the lifestyle changes. In addition, as a tertiary cancer institute specialized in melanoma treatment, the number of more advanced disease seen in our center may be overestimated. As such, a regional statistical investigation is needed and is currently under active planning to address the epidemiology of malignant in melanoma more effectively.", "Most malignant melanoma patients in China were diagnosed with locally advanced disease (i.e., stage II or above), and their prognoses were suboptimal. ALM and MCM are the two most commonly diagnosed pathological subtypes of malignant melanoma in China. Clinical staging and presence of ulceration was significantly associated with clinical outcome in terms of OS, while treatment strategy including extent of surgery and use of adjuvant therapy were significant predictors of DFS.", "The authors declare that they have no competing interests.", "ZC, SL, and JG participated in the study design. ZC, SL, XS, LS, CC, and MH participated in data collection and analysis. All authors participated in the interpretation and manuscript writing. SL, ZC, and JG participated in editing and proof reading. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2407/11/85/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Abdominal Cavity", "Animals", "Cecum", "Dextrans", "Female", "Hyaluronic Acid", "Male", "Membranes, Artificial", "Penicillamine", "Random Allocation", "Rats", "Rats, Wistar", "Shear Strength", "Tensile Strength", "Tissue Adhesions", "Wound Healing" ]

[ "Background", "Method for manufacturing novel penicillamine-bound membrane", "Animal model of abdominal adhesions", "Tissue preparation", "Measurement of adhesions degree and breaking strength of incision", "Immunohistochemistry", "Data analysis and statistics", "Results", "Penicillamine prevents the abdominal adhesions formation significantly", "Penicillamine has more preventive effects on abdominal adhesions than dextran (Dex) and sodium hyaluronate (SH)", "Penicillamine-bound membrane shows greater benefits in therapeutic prevention of abdominal adhesions than penicillamine", "Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Adhesions develop in over 90% of patients after abdominal operations [3,7] and can lead to significant postsurgical complications, including small bowel obstruction, infertility, chronic pelvic pain and difficult re-operative surgeries [2,8]. Adhesions formation is a dynamic and complex process, which involves a cascade of reactions of cellular, biochemical, immunological and biomechanical factors [9]. Unfortunately, there is no available marker to predict the occurrence or severity of adhesions preoperatively [10] and therapeutic prevention still remains a challenge.\nAt present, the prevention of adhesions formation after surgery has focused on minimizing peritoneal trauma and reducing the implantation of foreign materials into the peritoneal cavity, as they may aggravate the inflammatory response [11-14]. Numerous approaches have been attempted, including profibrinolytic agents and physical barriers [3,15], such as Dextran (Dex) [2] and sodium hyaluronate (SH)[3]. While the barrier products have been proven the most clinically successful [4-6], there is no effective method of preventing adhesions formation currently [1]\nPreviously, penicillamine was reported to prevent collagen fibers from crossing into non-soluble collagen tissue and inhibit the maturation of dissoluble collagen. Recent studies indicated the possibility of oral D-penicillamine-induced prevention on peritoneal adhesions band formation [16-18]. We therefore hypothesized that it can prevent the fibrin from converting into permanent fiber adhesions tissue. Hereby, we developed a novel membrane, which is composed of two regents- penicillamine and hyaluronic acid, and then applied this penicillamine-bound membrane to treat abdominal adhesions in an animal model, in order to identify its preventive and therapeutic potential for adhesions formation.", "Chitosan [2](Shanghai Qisheng Biologic Agent Company), or polylactic acid or hyaluronic acid [10](Center for New Drug Evaluation, Shandong University) was individually dissolved into saline at the concentration mentioned in previous literatures. Penicillamine (Catalog number: 000108, Shanghai, PR China) was dissolved into three different solutions. The solutions were drained into the flat bottom plastic container and dried thoroughly until polymerized. The thickness and dissolve time for the three different kinds of polymerized membranes were measured in order to select the best substrate of penicillamine-bound membrane. The release of penicillamine was defined by dissolving the membranes into saline solution. Eventually, hyaluronic acid was chosen for the substrate of penicillamine-bound membrane. Hyaluronic acid and aluminum chloride (at the concentration of 5%) were dissolved into autoclaved PBS to make solution 1. Carboxymethyl Cellulose was dissolved in double-distilled water (ddH2O) to make solution 2. And then solution 1 and 2 were 1:1 mixed thoroughly. 2.5 ml of 10% penicillamine was pipetted into 50 ml mixed solution in order to lead to cross-linking between penicillamine and substrate. The solutions were drained into the flat bottom plastic container and dried thoroughly for 4-7 days until fully polymerized in air. The thickness of penicillamine-bound membrane was about 0.1 mm, and the degradation of the membrane cost 5 days in corpore. We tested the concentration of penicillamine of the solution after the membrane was dissolved into saline at a different time.", "Total 150 rats (Wistar rats of both genders from animals facility of Shandong University) at 9 weeks of age, weighing 200~230 g, were involved in the present study with 25 animals per group in order to calculate the occurrence of abdominal adhesions. The rats were treated under the animal use guidelines of Institutional Animal Care and Use Committee (IACUC) at Qilu Hospital & College of Medicine, Shandong University. Rats were allowed to adapt to the new environment for 1-2 days prior to experimental study. The study was approved by the ethics committee of Qilu Hospital, Shandong University.\nAll animals were randomly divided into six groups (25 per group), including the vehicle group (A) and five treated groups (B, C, D, E and F). All groups were anesthetized with 10% chloral hydrate at the dosage of 300~350 mg/kg by intraperitoneal injection, and then underwent abdominal surgery through midline incision 1.5 cm in length. The caecum serosa was scratched with dry gauze at 2 cm × 2 cm [19] (Figure 1). One milliliter saline was put into the rats' abdominal cavity in group A, while 1 ml 40% Dextran (Dex), 0.5 ml sodium hyaluronate (SH) (Shandong Zhengda Freda Tragacanth Company) and 1 ml 3% penicillamine for group B, C and D. The scratched area on caecum serosa in group E was covered by penicillamine-bound membrane (Figure 1), and group F was covered by non-penicillamine-bound membrane. The membrane was not fixed and adhered to the scratched area naturally.\nAnimal model of abdominal adhesion. Arrow in (A) shows the caecum of normal animals. (B) Shows that the caecum is scratched with dry gauze at 2 cm × 2 cm, which is treated as the sham group. While the arrow in (C) shows that, the scratched caecum is covered with penicillamine-bound membrane.", "Half of the animals in each group (about 10 animals for every group and time point) were sacrificed at post-surgical day 7, while another half was sacrificed at post-surgical day 14. The adhesions tissue inside the abdominal cavity in different groups was removed and stored in 4% paraformalhyde, then subjected to immunohistochemistry staining. The incision, associated with the lateral skin tissue in different groups, was sheared at 4×0.5 cm to test its breaking strength.", "The occurrence of adhesions was calculated as the ratio between animals with adhesions tissue and total animals within that group (Table 1). The adhesions grade and score in different groups were defined by Bigatti's method [20](Table 2). The breaking strength of incision was measured by a strength-tester (See Additional file 1). After removing the stitches from the incision tissue, it was connected with a water container by a pulley and the breaking strength was defined by the gravity of the water (The Unit was gram), which was drained into the container when the incision tissue was broken. All of these measurements were performed by a blinded observer.\nComparison of the occurrence of adhesion, adhesion score and breaking strength of incision between control (A) and treated groups (B, C, D).\n(* means p < 0.05, compared to control level)\nAdhesion Score (Bigatti's method):", "Slices (40 μm) were made from 4% paraformalhyde-fixed adhesions tissue with a microtome, then transferred into 0.1 M phosphate buffer (PB) (pH = 7.4). The slices were incubated in 1:500-diluted polyclonal rabbit anti-collagen type I (Beijing Biosynthesis Biotechnology Co., Ltd., China) at 4°C overnight, and then washed in 0.1 M PB three times. Slices were then transferred into avidin-biotin-peroxidase complex and incubated for 20 min at 37°C, then washed with 0.1 M PB, incubated with 3, 3-diaminobenzidine (DAB) tetrahydrochloride for 5-15 min and washed three times in 0.1 M PB. Slices were then mounted onto gelatin-treated slides, dried overnight, and dehydrated serially with 50%, 70%, 95% ethanol once, and 100% ethanol and xylene twice. Slides were then coverslipped using the mounting solution and viewed under the microscope. Negative control experiment was performed by applying 0.1 M PB solution as the primary antibody.", "The occurrence of adhesion was compared by Chi-square test. The adhesion score, the breaking strength of incision were compared between different groups by one-way ANOVA with post hoc Tukey's test. Data was shown as percentage or mean ± SD. P< 0.05 was considered statistically significant. Data analyses were performed using SPSS statistical program version 16.0.", "[SUBTITLE] Penicillamine prevents the abdominal adhesions formation significantly [SUBSECTION] An animal model of abdominal adhesions was achieved with a success rate of 88.33%. About 24 animals in total died during surgery, about 4 animals per group for reasons such as bleeding or overdose of anesthesia. Animal death occurred across all groups, which could indicate no potential toxicity of any of the compounds used. Dead animals were discarded from our study. Additionally, our experiments showed no significant difference in adhesion occurrence and scores relating to the sex of the animal.\nThe occurrence of adhesions in group A and D was summarized in Table 1. Compared to the control group, the occurrence in group D was significantly lower at postsurgical day 7 and 14 (P = 0.0326, P < 0.05). Further, the adhesions score in group D was significantly decreased 7 days or 14 days after the surgery, compared to control level (P = 0.0064, P < 0.01). Immunohistochemical staining showed that more collagen fibers (Figure 2) and blood vessel hyperplasia (Figure 3) were observed in group A than in group D (Figure 2 at 10×, and Figure 3 at 40×magnification).\nComparison of adhesion formation in different groups by immunochemistry staining. Arrows in (A) and (B) show the stained collagen I. Significantly less collagen I and fibers are found in penicillamine-treated group (B) than in the sham group (A).\nComparison of wound healing in different groups by immunochemistry staining. Blood vessels hyperplasia is observed in the sham group (A), while not in the penicillamine-treated group (B).\nAn animal model of abdominal adhesions was achieved with a success rate of 88.33%. About 24 animals in total died during surgery, about 4 animals per group for reasons such as bleeding or overdose of anesthesia. Animal death occurred across all groups, which could indicate no potential toxicity of any of the compounds used. Dead animals were discarded from our study. Additionally, our experiments showed no significant difference in adhesion occurrence and scores relating to the sex of the animal.\nThe occurrence of adhesions in group A and D was summarized in Table 1. Compared to the control group, the occurrence in group D was significantly lower at postsurgical day 7 and 14 (P = 0.0326, P < 0.05). Further, the adhesions score in group D was significantly decreased 7 days or 14 days after the surgery, compared to control level (P = 0.0064, P < 0.01). Immunohistochemical staining showed that more collagen fibers (Figure 2) and blood vessel hyperplasia (Figure 3) were observed in group A than in group D (Figure 2 at 10×, and Figure 3 at 40×magnification).\nComparison of adhesion formation in different groups by immunochemistry staining. Arrows in (A) and (B) show the stained collagen I. Significantly less collagen I and fibers are found in penicillamine-treated group (B) than in the sham group (A).\nComparison of wound healing in different groups by immunochemistry staining. Blood vessels hyperplasia is observed in the sham group (A), while not in the penicillamine-treated group (B).\n[SUBTITLE] Penicillamine has more preventive effects on abdominal adhesions than dextran (Dex) and sodium hyaluronate (SH) [SUBSECTION] Compared to group B and C, penicillamine decreased the adhesions score in group D most significantly (P = 0.0326, P < 0.05). The occurrence of adhesions in group D was significantly lower at postsurgical day 7 (40% for D, 48% for B, 56% for C) and day 14 (48% for D, 80% for B, 80% for C).\nCompared to group B and C, penicillamine decreased the adhesions score in group D most significantly (P = 0.0326, P < 0.05). The occurrence of adhesions in group D was significantly lower at postsurgical day 7 (40% for D, 48% for B, 56% for C) and day 14 (48% for D, 80% for B, 80% for C).\n[SUBTITLE] Penicillamine-bound membrane shows greater benefits in therapeutic prevention of abdominal adhesions than penicillamine [SUBSECTION] The carrier for the penicillamine must be stable, non-toxic, non-irritating and not react with the drug. The Chitosan dissolve in acid solution, which has irritation. Nevertheless, the polylactic acid dissolve in organic solvents (chloroform, acetone, e.g.) in which penicillamine can't dissolve. Penicillamine-bound membrane was developed by using hyaluronic acid as the substrate. The membrane was made by only natural hyaluronic acid, which dissolved into the saline in 15 minutes, and so penicillamine was released thoroughly from the membrane. The new membrane was made by hyaluronic acid, aluminum chloride and carboxymethyl cellulose, which dissolved into the saline in 5 days. Penicillamine has a more prolonged period of action.\nThe concentration of penicillamine in the membrane is 1.501 ± 0.023 mg/cm2. The occurrence of abdominal adhesions in the penicillamine-bound membrane-treated group (40%) was significantly lower than control (100%), penicillamine-treated groups (48%) and non-penicillamine-treated groups (78%) (Table 1) on postsurgical day 14. Comparison of adhesions score (Table 1) showed a significance between control and treated groups, indicating that both penicillamine and penicillamine-bound membrane successfully prevented abdominal adhesions formation, which was confirmed by morphological observation (Figure 4). Moreover, penicillamine-bound membrane showed better effects than penicillamine itself and non-penicillamine-bound membrane (P = 0.0046, P < 0.01) (Table 3). Penicillamine-bound membrane could be loaded directly and locally onto the traumatic area, contributing to its advantages in clinical administration than penicillamine.\nComparison of adhesion formation in the sham and penicillamine-bound membrane-treated groups. Arrows in (A) and (B) show that abdominal adhesion tissue is found in the sham group, while the adhesion is not found in the penicillamine-bound membrane- treated group(C) and (D). Arrows in (C) and (D) show that the scratched caecum is recovered.\nComparison of occurrence of adhesion, adhesion score and breaking strength of incision between membranes with/without penicillamine- treated groups.\n(* means p < 0.05, compared to non-penicillamine -bound membrane- treated group)\nThe carrier for the penicillamine must be stable, non-toxic, non-irritating and not react with the drug. The Chitosan dissolve in acid solution, which has irritation. Nevertheless, the polylactic acid dissolve in organic solvents (chloroform, acetone, e.g.) in which penicillamine can't dissolve. Penicillamine-bound membrane was developed by using hyaluronic acid as the substrate. The membrane was made by only natural hyaluronic acid, which dissolved into the saline in 15 minutes, and so penicillamine was released thoroughly from the membrane. The new membrane was made by hyaluronic acid, aluminum chloride and carboxymethyl cellulose, which dissolved into the saline in 5 days. Penicillamine has a more prolonged period of action.\nThe concentration of penicillamine in the membrane is 1.501 ± 0.023 mg/cm2. The occurrence of abdominal adhesions in the penicillamine-bound membrane-treated group (40%) was significantly lower than control (100%), penicillamine-treated groups (48%) and non-penicillamine-treated groups (78%) (Table 1) on postsurgical day 14. Comparison of adhesions score (Table 1) showed a significance between control and treated groups, indicating that both penicillamine and penicillamine-bound membrane successfully prevented abdominal adhesions formation, which was confirmed by morphological observation (Figure 4). Moreover, penicillamine-bound membrane showed better effects than penicillamine itself and non-penicillamine-bound membrane (P = 0.0046, P < 0.01) (Table 3). Penicillamine-bound membrane could be loaded directly and locally onto the traumatic area, contributing to its advantages in clinical administration than penicillamine.\nComparison of adhesion formation in the sham and penicillamine-bound membrane-treated groups. Arrows in (A) and (B) show that abdominal adhesion tissue is found in the sham group, while the adhesion is not found in the penicillamine-bound membrane- treated group(C) and (D). Arrows in (C) and (D) show that the scratched caecum is recovered.\nComparison of occurrence of adhesion, adhesion score and breaking strength of incision between membranes with/without penicillamine- treated groups.\n(* means p < 0.05, compared to non-penicillamine -bound membrane- treated group)\n[SUBTITLE] Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision [SUBSECTION] Incision healing occurred very well in all groups. Compared to postsurgical day 7, the breaking strength of incision of each group was higher at postsurgical day 14. Compared to group A, the breaking strength of incision in group D, E and F were lower at postsurgical day 7 or day 14. The possibility exists that tensile strength of the abdominal wound might have been more affected if the membrane had been placed immediately deep to the laparotomy incision. Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision (Table 1, 3) (P > 0.05).\nIncision healing occurred very well in all groups. Compared to postsurgical day 7, the breaking strength of incision of each group was higher at postsurgical day 14. Compared to group A, the breaking strength of incision in group D, E and F were lower at postsurgical day 7 or day 14. The possibility exists that tensile strength of the abdominal wound might have been more affected if the membrane had been placed immediately deep to the laparotomy incision. Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision (Table 1, 3) (P > 0.05).", "An animal model of abdominal adhesions was achieved with a success rate of 88.33%. About 24 animals in total died during surgery, about 4 animals per group for reasons such as bleeding or overdose of anesthesia. Animal death occurred across all groups, which could indicate no potential toxicity of any of the compounds used. Dead animals were discarded from our study. Additionally, our experiments showed no significant difference in adhesion occurrence and scores relating to the sex of the animal.\nThe occurrence of adhesions in group A and D was summarized in Table 1. Compared to the control group, the occurrence in group D was significantly lower at postsurgical day 7 and 14 (P = 0.0326, P < 0.05). Further, the adhesions score in group D was significantly decreased 7 days or 14 days after the surgery, compared to control level (P = 0.0064, P < 0.01). Immunohistochemical staining showed that more collagen fibers (Figure 2) and blood vessel hyperplasia (Figure 3) were observed in group A than in group D (Figure 2 at 10×, and Figure 3 at 40×magnification).\nComparison of adhesion formation in different groups by immunochemistry staining. Arrows in (A) and (B) show the stained collagen I. Significantly less collagen I and fibers are found in penicillamine-treated group (B) than in the sham group (A).\nComparison of wound healing in different groups by immunochemistry staining. Blood vessels hyperplasia is observed in the sham group (A), while not in the penicillamine-treated group (B).", "Compared to group B and C, penicillamine decreased the adhesions score in group D most significantly (P = 0.0326, P < 0.05). The occurrence of adhesions in group D was significantly lower at postsurgical day 7 (40% for D, 48% for B, 56% for C) and day 14 (48% for D, 80% for B, 80% for C).", "The carrier for the penicillamine must be stable, non-toxic, non-irritating and not react with the drug. The Chitosan dissolve in acid solution, which has irritation. Nevertheless, the polylactic acid dissolve in organic solvents (chloroform, acetone, e.g.) in which penicillamine can't dissolve. Penicillamine-bound membrane was developed by using hyaluronic acid as the substrate. The membrane was made by only natural hyaluronic acid, which dissolved into the saline in 15 minutes, and so penicillamine was released thoroughly from the membrane. The new membrane was made by hyaluronic acid, aluminum chloride and carboxymethyl cellulose, which dissolved into the saline in 5 days. Penicillamine has a more prolonged period of action.\nThe concentration of penicillamine in the membrane is 1.501 ± 0.023 mg/cm2. The occurrence of abdominal adhesions in the penicillamine-bound membrane-treated group (40%) was significantly lower than control (100%), penicillamine-treated groups (48%) and non-penicillamine-treated groups (78%) (Table 1) on postsurgical day 14. Comparison of adhesions score (Table 1) showed a significance between control and treated groups, indicating that both penicillamine and penicillamine-bound membrane successfully prevented abdominal adhesions formation, which was confirmed by morphological observation (Figure 4). Moreover, penicillamine-bound membrane showed better effects than penicillamine itself and non-penicillamine-bound membrane (P = 0.0046, P < 0.01) (Table 3). Penicillamine-bound membrane could be loaded directly and locally onto the traumatic area, contributing to its advantages in clinical administration than penicillamine.\nComparison of adhesion formation in the sham and penicillamine-bound membrane-treated groups. Arrows in (A) and (B) show that abdominal adhesion tissue is found in the sham group, while the adhesion is not found in the penicillamine-bound membrane- treated group(C) and (D). Arrows in (C) and (D) show that the scratched caecum is recovered.\nComparison of occurrence of adhesion, adhesion score and breaking strength of incision between membranes with/without penicillamine- treated groups.\n(* means p < 0.05, compared to non-penicillamine -bound membrane- treated group)", "Incision healing occurred very well in all groups. Compared to postsurgical day 7, the breaking strength of incision of each group was higher at postsurgical day 14. Compared to group A, the breaking strength of incision in group D, E and F were lower at postsurgical day 7 or day 14. The possibility exists that tensile strength of the abdominal wound might have been more affected if the membrane had been placed immediately deep to the laparotomy incision. Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision (Table 1, 3) (P > 0.05).", "Postsurgical abdominal adhesions have a great impact on the quality of life of millions of people worldwide. Small bowel obstruction and others complications of adhesions are serious, causing not only morbidity but also mortality [8,21]. Adhesions are non-anatomic connections of fibrous tissue within normal peritoneal surfaces [7]. It may have a potential benefit, including neovascularization of ischaemic structures such anastomoses, but it is also responsible for various clinical problems [22].\nThe abdominal formation of fibrin is a common pathophysiological pathway for adhesions. Fibrin is formed after peritoneal injury, which can cause fibrinous adhesion. If the fibrinolytic system, which results in lysis of abdominal fibrin, is not activated, the adhesions will become fibrous [23]. This can be explained when the equilibrium between coagulation and fibrinolysis is disturbed [24-26]. Our present study confirmed this by the evidence of more collagen I fibers observed in the abdominal adhesions animal model than in the vehicle group.\nApart from the formation of fibrin, a complex interaction of biochemical components, including inflammation, fibrinolysis and wound healing, is involved in the pathological process of abdominal adhesions [27]. For instance, initially the deposition of fibrin is regulated and maintained by growth factors and cytokines [28]. After the first week and up to a month, the matrix is remodeled and replaced by persistent proteins, such as collagen, and revascularization occurs. Fibrinolysis stimulators, such as tissue plasminogen factor (t-PA), and urokinase and fibrinolysis inhibitors, such as plasminogen activator inhibitor type I (PAI-1), transforming growth factor (TGF)-β, a key molecular mediator of pathological fibrosis, have also been shown to play a role in adhesions pathogenesis [29,30]. The interrelationship between all the factors remains largely unknown, therefore, identifying the effective treatment or prevention for abdominal adhesions remains a big challenge.\nNumerous approaches have been used to prevent adhesions [15]. The three main principle pathways are: (1) decreasing the trauma to the peritoneum; (2) medical intervention in the fibrin formation/degradation balance, and (3) barriers (including fluid barrier and membranes) preventing organs from bridging over to other structures in the abdomen and thereby forming adhesions. Unfortunately none of these measures have proven uniformly effective under all surgical conditions. Barrier products, including hyaluronic acid-carboxymethyl cellulose membrane have been the most clinically successful in reducing adhesions formation by preventing the close apposition of injured tissues. However, treatment with these products induced many side effects, such as postponing wound healing. Furthermore, many treated models have a high standard deviation, which makes the relevance of results with only moderate effects questionable.\nPenicillamine can decrease the permeability of vessels by inhibiting aggregation of platelets, stabilizing lysosome and inhibiting releasing of lysomal enzymes. It can also attenuate immune reaction and decrease blood disk effusion and fibrin deposition by inhibiting generation of IgG and IgM and decreasing antigen-antibody complex in blood-serum, which blocks the first stage of abdominal adhesions. It is reported that D-penicillamine administration markedly reduces severe adhesions band formation without severe side effects [18]. Therefore, we hypothesized, based on these results, that a combination of penicillamine and barrier products may be a better treatment for adhesions.\nIn the present study, we developed a novel penicillamine- bound membrane, which used hyaluronic acid as the ideal substrate, and then found that both penicillamine and penicillamine-bound membrane have better therapeutic effects on preventing abdominal adhesions than Dextran (Dex), sodium hyaluronate (SH) and non-penicillamine-bound membrane. Both of them did not affect wound healing. Although they decreased the breaking strength of incision insignificantly at postsurgical day 7 but, this decrease was ameliorated at postsurgical day 14. Penicillamine-bound membrane showed greater benefits than penicillamine itself in preventive effects and local administration. Recent studies investigated that penicillamine can inhibit blood vessel hyperplasia[31], which plays an important role in adhesions generation, by inhibiting the proliferation of endangium and smooth muscle cell, and this was confirmed by our results.", "The present research indicated that penicillamine-bound membrane can be applied as an effective therapeutic intervention for abdominal adhesion with inconsequential side effects, but further studies on the detailed mechanisms for treating abdominal adhesion are still warranted.", "The authors declare that they have no competing interests.", "Q-YZ carried out the preparation of penicillamine-bound membrane, created the animal model of abdominal adhesions and drafted the manuscript. SM carried out the tissue preparation and measurement of adhesions degree and breaking strength of incision. DX carried out the Immunohistochemistry. W-TZ participated in the design of the study and performed the statistical analysis. A-WL conceived the study, and participated in its design and coordination.\nAll authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2482/11/5/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Method for manufacturing novel penicillamine-bound membrane", "Animal model of abdominal adhesions", "Tissue preparation", "Measurement of adhesions degree and breaking strength of incision", "Immunohistochemistry", "Data analysis and statistics", "Results", "Penicillamine prevents the abdominal adhesions formation significantly", "Penicillamine has more preventive effects on abdominal adhesions than dextran (Dex) and sodium hyaluronate (SH)", "Penicillamine-bound membrane shows greater benefits in therapeutic prevention of abdominal adhesions than penicillamine", "Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history", "Supplementary Material" ]

[ "Adhesions develop in over 90% of patients after abdominal operations [3,7] and can lead to significant postsurgical complications, including small bowel obstruction, infertility, chronic pelvic pain and difficult re-operative surgeries [2,8]. Adhesions formation is a dynamic and complex process, which involves a cascade of reactions of cellular, biochemical, immunological and biomechanical factors [9]. Unfortunately, there is no available marker to predict the occurrence or severity of adhesions preoperatively [10] and therapeutic prevention still remains a challenge.\nAt present, the prevention of adhesions formation after surgery has focused on minimizing peritoneal trauma and reducing the implantation of foreign materials into the peritoneal cavity, as they may aggravate the inflammatory response [11-14]. Numerous approaches have been attempted, including profibrinolytic agents and physical barriers [3,15], such as Dextran (Dex) [2] and sodium hyaluronate (SH)[3]. While the barrier products have been proven the most clinically successful [4-6], there is no effective method of preventing adhesions formation currently [1]\nPreviously, penicillamine was reported to prevent collagen fibers from crossing into non-soluble collagen tissue and inhibit the maturation of dissoluble collagen. Recent studies indicated the possibility of oral D-penicillamine-induced prevention on peritoneal adhesions band formation [16-18]. We therefore hypothesized that it can prevent the fibrin from converting into permanent fiber adhesions tissue. Hereby, we developed a novel membrane, which is composed of two regents- penicillamine and hyaluronic acid, and then applied this penicillamine-bound membrane to treat abdominal adhesions in an animal model, in order to identify its preventive and therapeutic potential for adhesions formation.", "[SUBTITLE] Method for manufacturing novel penicillamine-bound membrane [SUBSECTION] Chitosan [2](Shanghai Qisheng Biologic Agent Company), or polylactic acid or hyaluronic acid [10](Center for New Drug Evaluation, Shandong University) was individually dissolved into saline at the concentration mentioned in previous literatures. Penicillamine (Catalog number: 000108, Shanghai, PR China) was dissolved into three different solutions. The solutions were drained into the flat bottom plastic container and dried thoroughly until polymerized. The thickness and dissolve time for the three different kinds of polymerized membranes were measured in order to select the best substrate of penicillamine-bound membrane. The release of penicillamine was defined by dissolving the membranes into saline solution. Eventually, hyaluronic acid was chosen for the substrate of penicillamine-bound membrane. Hyaluronic acid and aluminum chloride (at the concentration of 5%) were dissolved into autoclaved PBS to make solution 1. Carboxymethyl Cellulose was dissolved in double-distilled water (ddH2O) to make solution 2. And then solution 1 and 2 were 1:1 mixed thoroughly. 2.5 ml of 10% penicillamine was pipetted into 50 ml mixed solution in order to lead to cross-linking between penicillamine and substrate. The solutions were drained into the flat bottom plastic container and dried thoroughly for 4-7 days until fully polymerized in air. The thickness of penicillamine-bound membrane was about 0.1 mm, and the degradation of the membrane cost 5 days in corpore. We tested the concentration of penicillamine of the solution after the membrane was dissolved into saline at a different time.\nChitosan [2](Shanghai Qisheng Biologic Agent Company), or polylactic acid or hyaluronic acid [10](Center for New Drug Evaluation, Shandong University) was individually dissolved into saline at the concentration mentioned in previous literatures. Penicillamine (Catalog number: 000108, Shanghai, PR China) was dissolved into three different solutions. The solutions were drained into the flat bottom plastic container and dried thoroughly until polymerized. The thickness and dissolve time for the three different kinds of polymerized membranes were measured in order to select the best substrate of penicillamine-bound membrane. The release of penicillamine was defined by dissolving the membranes into saline solution. Eventually, hyaluronic acid was chosen for the substrate of penicillamine-bound membrane. Hyaluronic acid and aluminum chloride (at the concentration of 5%) were dissolved into autoclaved PBS to make solution 1. Carboxymethyl Cellulose was dissolved in double-distilled water (ddH2O) to make solution 2. And then solution 1 and 2 were 1:1 mixed thoroughly. 2.5 ml of 10% penicillamine was pipetted into 50 ml mixed solution in order to lead to cross-linking between penicillamine and substrate. The solutions were drained into the flat bottom plastic container and dried thoroughly for 4-7 days until fully polymerized in air. The thickness of penicillamine-bound membrane was about 0.1 mm, and the degradation of the membrane cost 5 days in corpore. We tested the concentration of penicillamine of the solution after the membrane was dissolved into saline at a different time.\n[SUBTITLE] Animal model of abdominal adhesions [SUBSECTION] Total 150 rats (Wistar rats of both genders from animals facility of Shandong University) at 9 weeks of age, weighing 200~230 g, were involved in the present study with 25 animals per group in order to calculate the occurrence of abdominal adhesions. The rats were treated under the animal use guidelines of Institutional Animal Care and Use Committee (IACUC) at Qilu Hospital & College of Medicine, Shandong University. Rats were allowed to adapt to the new environment for 1-2 days prior to experimental study. The study was approved by the ethics committee of Qilu Hospital, Shandong University.\nAll animals were randomly divided into six groups (25 per group), including the vehicle group (A) and five treated groups (B, C, D, E and F). All groups were anesthetized with 10% chloral hydrate at the dosage of 300~350 mg/kg by intraperitoneal injection, and then underwent abdominal surgery through midline incision 1.5 cm in length. The caecum serosa was scratched with dry gauze at 2 cm × 2 cm [19] (Figure 1). One milliliter saline was put into the rats' abdominal cavity in group A, while 1 ml 40% Dextran (Dex), 0.5 ml sodium hyaluronate (SH) (Shandong Zhengda Freda Tragacanth Company) and 1 ml 3% penicillamine for group B, C and D. The scratched area on caecum serosa in group E was covered by penicillamine-bound membrane (Figure 1), and group F was covered by non-penicillamine-bound membrane. The membrane was not fixed and adhered to the scratched area naturally.\nAnimal model of abdominal adhesion. Arrow in (A) shows the caecum of normal animals. (B) Shows that the caecum is scratched with dry gauze at 2 cm × 2 cm, which is treated as the sham group. While the arrow in (C) shows that, the scratched caecum is covered with penicillamine-bound membrane.\nTotal 150 rats (Wistar rats of both genders from animals facility of Shandong University) at 9 weeks of age, weighing 200~230 g, were involved in the present study with 25 animals per group in order to calculate the occurrence of abdominal adhesions. The rats were treated under the animal use guidelines of Institutional Animal Care and Use Committee (IACUC) at Qilu Hospital & College of Medicine, Shandong University. Rats were allowed to adapt to the new environment for 1-2 days prior to experimental study. The study was approved by the ethics committee of Qilu Hospital, Shandong University.\nAll animals were randomly divided into six groups (25 per group), including the vehicle group (A) and five treated groups (B, C, D, E and F). All groups were anesthetized with 10% chloral hydrate at the dosage of 300~350 mg/kg by intraperitoneal injection, and then underwent abdominal surgery through midline incision 1.5 cm in length. The caecum serosa was scratched with dry gauze at 2 cm × 2 cm [19] (Figure 1). One milliliter saline was put into the rats' abdominal cavity in group A, while 1 ml 40% Dextran (Dex), 0.5 ml sodium hyaluronate (SH) (Shandong Zhengda Freda Tragacanth Company) and 1 ml 3% penicillamine for group B, C and D. The scratched area on caecum serosa in group E was covered by penicillamine-bound membrane (Figure 1), and group F was covered by non-penicillamine-bound membrane. The membrane was not fixed and adhered to the scratched area naturally.\nAnimal model of abdominal adhesion. Arrow in (A) shows the caecum of normal animals. (B) Shows that the caecum is scratched with dry gauze at 2 cm × 2 cm, which is treated as the sham group. While the arrow in (C) shows that, the scratched caecum is covered with penicillamine-bound membrane.\n[SUBTITLE] Tissue preparation [SUBSECTION] Half of the animals in each group (about 10 animals for every group and time point) were sacrificed at post-surgical day 7, while another half was sacrificed at post-surgical day 14. The adhesions tissue inside the abdominal cavity in different groups was removed and stored in 4% paraformalhyde, then subjected to immunohistochemistry staining. The incision, associated with the lateral skin tissue in different groups, was sheared at 4×0.5 cm to test its breaking strength.\nHalf of the animals in each group (about 10 animals for every group and time point) were sacrificed at post-surgical day 7, while another half was sacrificed at post-surgical day 14. The adhesions tissue inside the abdominal cavity in different groups was removed and stored in 4% paraformalhyde, then subjected to immunohistochemistry staining. The incision, associated with the lateral skin tissue in different groups, was sheared at 4×0.5 cm to test its breaking strength.\n[SUBTITLE] Measurement of adhesions degree and breaking strength of incision [SUBSECTION] The occurrence of adhesions was calculated as the ratio between animals with adhesions tissue and total animals within that group (Table 1). The adhesions grade and score in different groups were defined by Bigatti's method [20](Table 2). The breaking strength of incision was measured by a strength-tester (See Additional file 1). After removing the stitches from the incision tissue, it was connected with a water container by a pulley and the breaking strength was defined by the gravity of the water (The Unit was gram), which was drained into the container when the incision tissue was broken. All of these measurements were performed by a blinded observer.\nComparison of the occurrence of adhesion, adhesion score and breaking strength of incision between control (A) and treated groups (B, C, D).\n(* means p < 0.05, compared to control level)\nAdhesion Score (Bigatti's method):\nThe occurrence of adhesions was calculated as the ratio between animals with adhesions tissue and total animals within that group (Table 1). The adhesions grade and score in different groups were defined by Bigatti's method [20](Table 2). The breaking strength of incision was measured by a strength-tester (See Additional file 1). After removing the stitches from the incision tissue, it was connected with a water container by a pulley and the breaking strength was defined by the gravity of the water (The Unit was gram), which was drained into the container when the incision tissue was broken. All of these measurements were performed by a blinded observer.\nComparison of the occurrence of adhesion, adhesion score and breaking strength of incision between control (A) and treated groups (B, C, D).\n(* means p < 0.05, compared to control level)\nAdhesion Score (Bigatti's method):\n[SUBTITLE] Immunohistochemistry [SUBSECTION] Slices (40 μm) were made from 4% paraformalhyde-fixed adhesions tissue with a microtome, then transferred into 0.1 M phosphate buffer (PB) (pH = 7.4). The slices were incubated in 1:500-diluted polyclonal rabbit anti-collagen type I (Beijing Biosynthesis Biotechnology Co., Ltd., China) at 4°C overnight, and then washed in 0.1 M PB three times. Slices were then transferred into avidin-biotin-peroxidase complex and incubated for 20 min at 37°C, then washed with 0.1 M PB, incubated with 3, 3-diaminobenzidine (DAB) tetrahydrochloride for 5-15 min and washed three times in 0.1 M PB. Slices were then mounted onto gelatin-treated slides, dried overnight, and dehydrated serially with 50%, 70%, 95% ethanol once, and 100% ethanol and xylene twice. Slides were then coverslipped using the mounting solution and viewed under the microscope. Negative control experiment was performed by applying 0.1 M PB solution as the primary antibody.\nSlices (40 μm) were made from 4% paraformalhyde-fixed adhesions tissue with a microtome, then transferred into 0.1 M phosphate buffer (PB) (pH = 7.4). The slices were incubated in 1:500-diluted polyclonal rabbit anti-collagen type I (Beijing Biosynthesis Biotechnology Co., Ltd., China) at 4°C overnight, and then washed in 0.1 M PB three times. Slices were then transferred into avidin-biotin-peroxidase complex and incubated for 20 min at 37°C, then washed with 0.1 M PB, incubated with 3, 3-diaminobenzidine (DAB) tetrahydrochloride for 5-15 min and washed three times in 0.1 M PB. Slices were then mounted onto gelatin-treated slides, dried overnight, and dehydrated serially with 50%, 70%, 95% ethanol once, and 100% ethanol and xylene twice. Slides were then coverslipped using the mounting solution and viewed under the microscope. Negative control experiment was performed by applying 0.1 M PB solution as the primary antibody.\n[SUBTITLE] Data analysis and statistics [SUBSECTION] The occurrence of adhesion was compared by Chi-square test. The adhesion score, the breaking strength of incision were compared between different groups by one-way ANOVA with post hoc Tukey's test. Data was shown as percentage or mean ± SD. P< 0.05 was considered statistically significant. Data analyses were performed using SPSS statistical program version 16.0.\nThe occurrence of adhesion was compared by Chi-square test. The adhesion score, the breaking strength of incision were compared between different groups by one-way ANOVA with post hoc Tukey's test. Data was shown as percentage or mean ± SD. P< 0.05 was considered statistically significant. Data analyses were performed using SPSS statistical program version 16.0.", "Chitosan [2](Shanghai Qisheng Biologic Agent Company), or polylactic acid or hyaluronic acid [10](Center for New Drug Evaluation, Shandong University) was individually dissolved into saline at the concentration mentioned in previous literatures. Penicillamine (Catalog number: 000108, Shanghai, PR China) was dissolved into three different solutions. The solutions were drained into the flat bottom plastic container and dried thoroughly until polymerized. The thickness and dissolve time for the three different kinds of polymerized membranes were measured in order to select the best substrate of penicillamine-bound membrane. The release of penicillamine was defined by dissolving the membranes into saline solution. Eventually, hyaluronic acid was chosen for the substrate of penicillamine-bound membrane. Hyaluronic acid and aluminum chloride (at the concentration of 5%) were dissolved into autoclaved PBS to make solution 1. Carboxymethyl Cellulose was dissolved in double-distilled water (ddH2O) to make solution 2. And then solution 1 and 2 were 1:1 mixed thoroughly. 2.5 ml of 10% penicillamine was pipetted into 50 ml mixed solution in order to lead to cross-linking between penicillamine and substrate. The solutions were drained into the flat bottom plastic container and dried thoroughly for 4-7 days until fully polymerized in air. The thickness of penicillamine-bound membrane was about 0.1 mm, and the degradation of the membrane cost 5 days in corpore. We tested the concentration of penicillamine of the solution after the membrane was dissolved into saline at a different time.", "Total 150 rats (Wistar rats of both genders from animals facility of Shandong University) at 9 weeks of age, weighing 200~230 g, were involved in the present study with 25 animals per group in order to calculate the occurrence of abdominal adhesions. The rats were treated under the animal use guidelines of Institutional Animal Care and Use Committee (IACUC) at Qilu Hospital & College of Medicine, Shandong University. Rats were allowed to adapt to the new environment for 1-2 days prior to experimental study. The study was approved by the ethics committee of Qilu Hospital, Shandong University.\nAll animals were randomly divided into six groups (25 per group), including the vehicle group (A) and five treated groups (B, C, D, E and F). All groups were anesthetized with 10% chloral hydrate at the dosage of 300~350 mg/kg by intraperitoneal injection, and then underwent abdominal surgery through midline incision 1.5 cm in length. The caecum serosa was scratched with dry gauze at 2 cm × 2 cm [19] (Figure 1). One milliliter saline was put into the rats' abdominal cavity in group A, while 1 ml 40% Dextran (Dex), 0.5 ml sodium hyaluronate (SH) (Shandong Zhengda Freda Tragacanth Company) and 1 ml 3% penicillamine for group B, C and D. The scratched area on caecum serosa in group E was covered by penicillamine-bound membrane (Figure 1), and group F was covered by non-penicillamine-bound membrane. The membrane was not fixed and adhered to the scratched area naturally.\nAnimal model of abdominal adhesion. Arrow in (A) shows the caecum of normal animals. (B) Shows that the caecum is scratched with dry gauze at 2 cm × 2 cm, which is treated as the sham group. While the arrow in (C) shows that, the scratched caecum is covered with penicillamine-bound membrane.", "Half of the animals in each group (about 10 animals for every group and time point) were sacrificed at post-surgical day 7, while another half was sacrificed at post-surgical day 14. The adhesions tissue inside the abdominal cavity in different groups was removed and stored in 4% paraformalhyde, then subjected to immunohistochemistry staining. The incision, associated with the lateral skin tissue in different groups, was sheared at 4×0.5 cm to test its breaking strength.", "The occurrence of adhesions was calculated as the ratio between animals with adhesions tissue and total animals within that group (Table 1). The adhesions grade and score in different groups were defined by Bigatti's method [20](Table 2). The breaking strength of incision was measured by a strength-tester (See Additional file 1). After removing the stitches from the incision tissue, it was connected with a water container by a pulley and the breaking strength was defined by the gravity of the water (The Unit was gram), which was drained into the container when the incision tissue was broken. All of these measurements were performed by a blinded observer.\nComparison of the occurrence of adhesion, adhesion score and breaking strength of incision between control (A) and treated groups (B, C, D).\n(* means p < 0.05, compared to control level)\nAdhesion Score (Bigatti's method):", "Slices (40 μm) were made from 4% paraformalhyde-fixed adhesions tissue with a microtome, then transferred into 0.1 M phosphate buffer (PB) (pH = 7.4). The slices were incubated in 1:500-diluted polyclonal rabbit anti-collagen type I (Beijing Biosynthesis Biotechnology Co., Ltd., China) at 4°C overnight, and then washed in 0.1 M PB three times. Slices were then transferred into avidin-biotin-peroxidase complex and incubated for 20 min at 37°C, then washed with 0.1 M PB, incubated with 3, 3-diaminobenzidine (DAB) tetrahydrochloride for 5-15 min and washed three times in 0.1 M PB. Slices were then mounted onto gelatin-treated slides, dried overnight, and dehydrated serially with 50%, 70%, 95% ethanol once, and 100% ethanol and xylene twice. Slides were then coverslipped using the mounting solution and viewed under the microscope. Negative control experiment was performed by applying 0.1 M PB solution as the primary antibody.", "The occurrence of adhesion was compared by Chi-square test. The adhesion score, the breaking strength of incision were compared between different groups by one-way ANOVA with post hoc Tukey's test. Data was shown as percentage or mean ± SD. P< 0.05 was considered statistically significant. Data analyses were performed using SPSS statistical program version 16.0.", "[SUBTITLE] Penicillamine prevents the abdominal adhesions formation significantly [SUBSECTION] An animal model of abdominal adhesions was achieved with a success rate of 88.33%. About 24 animals in total died during surgery, about 4 animals per group for reasons such as bleeding or overdose of anesthesia. Animal death occurred across all groups, which could indicate no potential toxicity of any of the compounds used. Dead animals were discarded from our study. Additionally, our experiments showed no significant difference in adhesion occurrence and scores relating to the sex of the animal.\nThe occurrence of adhesions in group A and D was summarized in Table 1. Compared to the control group, the occurrence in group D was significantly lower at postsurgical day 7 and 14 (P = 0.0326, P < 0.05). Further, the adhesions score in group D was significantly decreased 7 days or 14 days after the surgery, compared to control level (P = 0.0064, P < 0.01). Immunohistochemical staining showed that more collagen fibers (Figure 2) and blood vessel hyperplasia (Figure 3) were observed in group A than in group D (Figure 2 at 10×, and Figure 3 at 40×magnification).\nComparison of adhesion formation in different groups by immunochemistry staining. Arrows in (A) and (B) show the stained collagen I. Significantly less collagen I and fibers are found in penicillamine-treated group (B) than in the sham group (A).\nComparison of wound healing in different groups by immunochemistry staining. Blood vessels hyperplasia is observed in the sham group (A), while not in the penicillamine-treated group (B).\nAn animal model of abdominal adhesions was achieved with a success rate of 88.33%. About 24 animals in total died during surgery, about 4 animals per group for reasons such as bleeding or overdose of anesthesia. Animal death occurred across all groups, which could indicate no potential toxicity of any of the compounds used. Dead animals were discarded from our study. Additionally, our experiments showed no significant difference in adhesion occurrence and scores relating to the sex of the animal.\nThe occurrence of adhesions in group A and D was summarized in Table 1. Compared to the control group, the occurrence in group D was significantly lower at postsurgical day 7 and 14 (P = 0.0326, P < 0.05). Further, the adhesions score in group D was significantly decreased 7 days or 14 days after the surgery, compared to control level (P = 0.0064, P < 0.01). Immunohistochemical staining showed that more collagen fibers (Figure 2) and blood vessel hyperplasia (Figure 3) were observed in group A than in group D (Figure 2 at 10×, and Figure 3 at 40×magnification).\nComparison of adhesion formation in different groups by immunochemistry staining. Arrows in (A) and (B) show the stained collagen I. Significantly less collagen I and fibers are found in penicillamine-treated group (B) than in the sham group (A).\nComparison of wound healing in different groups by immunochemistry staining. Blood vessels hyperplasia is observed in the sham group (A), while not in the penicillamine-treated group (B).\n[SUBTITLE] Penicillamine has more preventive effects on abdominal adhesions than dextran (Dex) and sodium hyaluronate (SH) [SUBSECTION] Compared to group B and C, penicillamine decreased the adhesions score in group D most significantly (P = 0.0326, P < 0.05). The occurrence of adhesions in group D was significantly lower at postsurgical day 7 (40% for D, 48% for B, 56% for C) and day 14 (48% for D, 80% for B, 80% for C).\nCompared to group B and C, penicillamine decreased the adhesions score in group D most significantly (P = 0.0326, P < 0.05). The occurrence of adhesions in group D was significantly lower at postsurgical day 7 (40% for D, 48% for B, 56% for C) and day 14 (48% for D, 80% for B, 80% for C).\n[SUBTITLE] Penicillamine-bound membrane shows greater benefits in therapeutic prevention of abdominal adhesions than penicillamine [SUBSECTION] The carrier for the penicillamine must be stable, non-toxic, non-irritating and not react with the drug. The Chitosan dissolve in acid solution, which has irritation. Nevertheless, the polylactic acid dissolve in organic solvents (chloroform, acetone, e.g.) in which penicillamine can't dissolve. Penicillamine-bound membrane was developed by using hyaluronic acid as the substrate. The membrane was made by only natural hyaluronic acid, which dissolved into the saline in 15 minutes, and so penicillamine was released thoroughly from the membrane. The new membrane was made by hyaluronic acid, aluminum chloride and carboxymethyl cellulose, which dissolved into the saline in 5 days. Penicillamine has a more prolonged period of action.\nThe concentration of penicillamine in the membrane is 1.501 ± 0.023 mg/cm2. The occurrence of abdominal adhesions in the penicillamine-bound membrane-treated group (40%) was significantly lower than control (100%), penicillamine-treated groups (48%) and non-penicillamine-treated groups (78%) (Table 1) on postsurgical day 14. Comparison of adhesions score (Table 1) showed a significance between control and treated groups, indicating that both penicillamine and penicillamine-bound membrane successfully prevented abdominal adhesions formation, which was confirmed by morphological observation (Figure 4). Moreover, penicillamine-bound membrane showed better effects than penicillamine itself and non-penicillamine-bound membrane (P = 0.0046, P < 0.01) (Table 3). Penicillamine-bound membrane could be loaded directly and locally onto the traumatic area, contributing to its advantages in clinical administration than penicillamine.\nComparison of adhesion formation in the sham and penicillamine-bound membrane-treated groups. Arrows in (A) and (B) show that abdominal adhesion tissue is found in the sham group, while the adhesion is not found in the penicillamine-bound membrane- treated group(C) and (D). Arrows in (C) and (D) show that the scratched caecum is recovered.\nComparison of occurrence of adhesion, adhesion score and breaking strength of incision between membranes with/without penicillamine- treated groups.\n(* means p < 0.05, compared to non-penicillamine -bound membrane- treated group)\nThe carrier for the penicillamine must be stable, non-toxic, non-irritating and not react with the drug. The Chitosan dissolve in acid solution, which has irritation. Nevertheless, the polylactic acid dissolve in organic solvents (chloroform, acetone, e.g.) in which penicillamine can't dissolve. Penicillamine-bound membrane was developed by using hyaluronic acid as the substrate. The membrane was made by only natural hyaluronic acid, which dissolved into the saline in 15 minutes, and so penicillamine was released thoroughly from the membrane. The new membrane was made by hyaluronic acid, aluminum chloride and carboxymethyl cellulose, which dissolved into the saline in 5 days. Penicillamine has a more prolonged period of action.\nThe concentration of penicillamine in the membrane is 1.501 ± 0.023 mg/cm2. The occurrence of abdominal adhesions in the penicillamine-bound membrane-treated group (40%) was significantly lower than control (100%), penicillamine-treated groups (48%) and non-penicillamine-treated groups (78%) (Table 1) on postsurgical day 14. Comparison of adhesions score (Table 1) showed a significance between control and treated groups, indicating that both penicillamine and penicillamine-bound membrane successfully prevented abdominal adhesions formation, which was confirmed by morphological observation (Figure 4). Moreover, penicillamine-bound membrane showed better effects than penicillamine itself and non-penicillamine-bound membrane (P = 0.0046, P < 0.01) (Table 3). Penicillamine-bound membrane could be loaded directly and locally onto the traumatic area, contributing to its advantages in clinical administration than penicillamine.\nComparison of adhesion formation in the sham and penicillamine-bound membrane-treated groups. Arrows in (A) and (B) show that abdominal adhesion tissue is found in the sham group, while the adhesion is not found in the penicillamine-bound membrane- treated group(C) and (D). Arrows in (C) and (D) show that the scratched caecum is recovered.\nComparison of occurrence of adhesion, adhesion score and breaking strength of incision between membranes with/without penicillamine- treated groups.\n(* means p < 0.05, compared to non-penicillamine -bound membrane- treated group)\n[SUBTITLE] Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision [SUBSECTION] Incision healing occurred very well in all groups. Compared to postsurgical day 7, the breaking strength of incision of each group was higher at postsurgical day 14. Compared to group A, the breaking strength of incision in group D, E and F were lower at postsurgical day 7 or day 14. The possibility exists that tensile strength of the abdominal wound might have been more affected if the membrane had been placed immediately deep to the laparotomy incision. Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision (Table 1, 3) (P > 0.05).\nIncision healing occurred very well in all groups. Compared to postsurgical day 7, the breaking strength of incision of each group was higher at postsurgical day 14. Compared to group A, the breaking strength of incision in group D, E and F were lower at postsurgical day 7 or day 14. The possibility exists that tensile strength of the abdominal wound might have been more affected if the membrane had been placed immediately deep to the laparotomy incision. Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision (Table 1, 3) (P > 0.05).", "An animal model of abdominal adhesions was achieved with a success rate of 88.33%. About 24 animals in total died during surgery, about 4 animals per group for reasons such as bleeding or overdose of anesthesia. Animal death occurred across all groups, which could indicate no potential toxicity of any of the compounds used. Dead animals were discarded from our study. Additionally, our experiments showed no significant difference in adhesion occurrence and scores relating to the sex of the animal.\nThe occurrence of adhesions in group A and D was summarized in Table 1. Compared to the control group, the occurrence in group D was significantly lower at postsurgical day 7 and 14 (P = 0.0326, P < 0.05). Further, the adhesions score in group D was significantly decreased 7 days or 14 days after the surgery, compared to control level (P = 0.0064, P < 0.01). Immunohistochemical staining showed that more collagen fibers (Figure 2) and blood vessel hyperplasia (Figure 3) were observed in group A than in group D (Figure 2 at 10×, and Figure 3 at 40×magnification).\nComparison of adhesion formation in different groups by immunochemistry staining. Arrows in (A) and (B) show the stained collagen I. Significantly less collagen I and fibers are found in penicillamine-treated group (B) than in the sham group (A).\nComparison of wound healing in different groups by immunochemistry staining. Blood vessels hyperplasia is observed in the sham group (A), while not in the penicillamine-treated group (B).", "Compared to group B and C, penicillamine decreased the adhesions score in group D most significantly (P = 0.0326, P < 0.05). The occurrence of adhesions in group D was significantly lower at postsurgical day 7 (40% for D, 48% for B, 56% for C) and day 14 (48% for D, 80% for B, 80% for C).", "The carrier for the penicillamine must be stable, non-toxic, non-irritating and not react with the drug. The Chitosan dissolve in acid solution, which has irritation. Nevertheless, the polylactic acid dissolve in organic solvents (chloroform, acetone, e.g.) in which penicillamine can't dissolve. Penicillamine-bound membrane was developed by using hyaluronic acid as the substrate. The membrane was made by only natural hyaluronic acid, which dissolved into the saline in 15 minutes, and so penicillamine was released thoroughly from the membrane. The new membrane was made by hyaluronic acid, aluminum chloride and carboxymethyl cellulose, which dissolved into the saline in 5 days. Penicillamine has a more prolonged period of action.\nThe concentration of penicillamine in the membrane is 1.501 ± 0.023 mg/cm2. The occurrence of abdominal adhesions in the penicillamine-bound membrane-treated group (40%) was significantly lower than control (100%), penicillamine-treated groups (48%) and non-penicillamine-treated groups (78%) (Table 1) on postsurgical day 14. Comparison of adhesions score (Table 1) showed a significance between control and treated groups, indicating that both penicillamine and penicillamine-bound membrane successfully prevented abdominal adhesions formation, which was confirmed by morphological observation (Figure 4). Moreover, penicillamine-bound membrane showed better effects than penicillamine itself and non-penicillamine-bound membrane (P = 0.0046, P < 0.01) (Table 3). Penicillamine-bound membrane could be loaded directly and locally onto the traumatic area, contributing to its advantages in clinical administration than penicillamine.\nComparison of adhesion formation in the sham and penicillamine-bound membrane-treated groups. Arrows in (A) and (B) show that abdominal adhesion tissue is found in the sham group, while the adhesion is not found in the penicillamine-bound membrane- treated group(C) and (D). Arrows in (C) and (D) show that the scratched caecum is recovered.\nComparison of occurrence of adhesion, adhesion score and breaking strength of incision between membranes with/without penicillamine- treated groups.\n(* means p < 0.05, compared to non-penicillamine -bound membrane- treated group)", "Incision healing occurred very well in all groups. Compared to postsurgical day 7, the breaking strength of incision of each group was higher at postsurgical day 14. Compared to group A, the breaking strength of incision in group D, E and F were lower at postsurgical day 7 or day 14. The possibility exists that tensile strength of the abdominal wound might have been more affected if the membrane had been placed immediately deep to the laparotomy incision. Penicillamine and penicillamine-bound membrane did not influence the incision healing, although they insignificantly decreased the breaking strength of incision (Table 1, 3) (P > 0.05).", "Postsurgical abdominal adhesions have a great impact on the quality of life of millions of people worldwide. Small bowel obstruction and others complications of adhesions are serious, causing not only morbidity but also mortality [8,21]. Adhesions are non-anatomic connections of fibrous tissue within normal peritoneal surfaces [7]. It may have a potential benefit, including neovascularization of ischaemic structures such anastomoses, but it is also responsible for various clinical problems [22].\nThe abdominal formation of fibrin is a common pathophysiological pathway for adhesions. Fibrin is formed after peritoneal injury, which can cause fibrinous adhesion. If the fibrinolytic system, which results in lysis of abdominal fibrin, is not activated, the adhesions will become fibrous [23]. This can be explained when the equilibrium between coagulation and fibrinolysis is disturbed [24-26]. Our present study confirmed this by the evidence of more collagen I fibers observed in the abdominal adhesions animal model than in the vehicle group.\nApart from the formation of fibrin, a complex interaction of biochemical components, including inflammation, fibrinolysis and wound healing, is involved in the pathological process of abdominal adhesions [27]. For instance, initially the deposition of fibrin is regulated and maintained by growth factors and cytokines [28]. After the first week and up to a month, the matrix is remodeled and replaced by persistent proteins, such as collagen, and revascularization occurs. Fibrinolysis stimulators, such as tissue plasminogen factor (t-PA), and urokinase and fibrinolysis inhibitors, such as plasminogen activator inhibitor type I (PAI-1), transforming growth factor (TGF)-β, a key molecular mediator of pathological fibrosis, have also been shown to play a role in adhesions pathogenesis [29,30]. The interrelationship between all the factors remains largely unknown, therefore, identifying the effective treatment or prevention for abdominal adhesions remains a big challenge.\nNumerous approaches have been used to prevent adhesions [15]. The three main principle pathways are: (1) decreasing the trauma to the peritoneum; (2) medical intervention in the fibrin formation/degradation balance, and (3) barriers (including fluid barrier and membranes) preventing organs from bridging over to other structures in the abdomen and thereby forming adhesions. Unfortunately none of these measures have proven uniformly effective under all surgical conditions. Barrier products, including hyaluronic acid-carboxymethyl cellulose membrane have been the most clinically successful in reducing adhesions formation by preventing the close apposition of injured tissues. However, treatment with these products induced many side effects, such as postponing wound healing. Furthermore, many treated models have a high standard deviation, which makes the relevance of results with only moderate effects questionable.\nPenicillamine can decrease the permeability of vessels by inhibiting aggregation of platelets, stabilizing lysosome and inhibiting releasing of lysomal enzymes. It can also attenuate immune reaction and decrease blood disk effusion and fibrin deposition by inhibiting generation of IgG and IgM and decreasing antigen-antibody complex in blood-serum, which blocks the first stage of abdominal adhesions. It is reported that D-penicillamine administration markedly reduces severe adhesions band formation without severe side effects [18]. Therefore, we hypothesized, based on these results, that a combination of penicillamine and barrier products may be a better treatment for adhesions.\nIn the present study, we developed a novel penicillamine- bound membrane, which used hyaluronic acid as the ideal substrate, and then found that both penicillamine and penicillamine-bound membrane have better therapeutic effects on preventing abdominal adhesions than Dextran (Dex), sodium hyaluronate (SH) and non-penicillamine-bound membrane. Both of them did not affect wound healing. Although they decreased the breaking strength of incision insignificantly at postsurgical day 7 but, this decrease was ameliorated at postsurgical day 14. Penicillamine-bound membrane showed greater benefits than penicillamine itself in preventive effects and local administration. Recent studies investigated that penicillamine can inhibit blood vessel hyperplasia[31], which plays an important role in adhesions generation, by inhibiting the proliferation of endangium and smooth muscle cell, and this was confirmed by our results.", "The present research indicated that penicillamine-bound membrane can be applied as an effective therapeutic intervention for abdominal adhesion with inconsequential side effects, but further studies on the detailed mechanisms for treating abdominal adhesion are still warranted.", "The authors declare that they have no competing interests.", "Q-YZ carried out the preparation of penicillamine-bound membrane, created the animal model of abdominal adhesions and drafted the manuscript. SM carried out the tissue preparation and measurement of adhesions degree and breaking strength of incision. DX carried out the Immunohistochemistry. W-TZ participated in the design of the study and performed the statistical analysis. A-WL conceived the study, and participated in its design and coordination.\nAll authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2482/11/5/prepub\n", "Test for breaking strength of incision. Step 1. Connecting the incision with an empty water container. Step 2. Draining the\nwater gradually into the container until the incision is broken, then calculating the breaking strength by this formula (Breaking strength = the gravity of total water).\nClick here for file" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Adolescent", "Adult", "Aged", "Aged, 80 and over", "Cancer Care Facilities", "Female", "Health Care Surveys", "Hospitals, Private", "Humans", "Male", "Middle Aged", "Multivariate Analysis", "Patient Satisfaction", "Quality of Health Care", "United States", "Young Adult" ]

[ "Background", "Study Population", "Questionnaire and Survey Administration", "Statistical Analysis", "Results", "Response Rate", "Baseline Patient Characteristics", "Service Quality Items", "Univariate Analysis - Predictors of Patient Willingness to Recommend", "Multivariate Analysis - Predictors of Patient Willingness to Recommend", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "As consumerism continues to increase in healthcare, there has been a rise in awareness about how patients perceive the quality of the services they receive at a health-care institution [1,2]. In addition, the web offers patients the opportunity to shop for the best places for care due to the rise in transparency of provider information on service quality and patient experience. As a consequence, patient satisfaction with service quality is becoming an increasingly important tool for providers to demonstrate patient focus and differentiation in the marketplace, as well as enhance patient loyalty. Furthermore, providers are using the information to make important decisions regarding operational and treatment plans [3].\nEvaluations of service quality provide important data on the patient's perception of the quality of care and treatment delivered by physicians, paramedical staff and the hospital as a whole [4]. Health providers can use data on service quality to design and track quality improvement over time and compare themselves to other health providers when the same measures are used, as well as recognize and expeditiously resolve service problems in real-time [5,6]. Measuring service quality also helps health care providers identify specific, and often unmet needs of patients, which has been a large focus of our work, and demonstrated by other research [7].\nSimilar to other acute health-care settings, the assessment of service quality, as perceived by patients, is critical in the oncology setting as well. Advances in diagnostics, treatment, supportive care and rehabilitation all necessitate continued monitoring to determine whether patients are satisfied with the increasingly complex and multidisciplinary nature of health care services that they are receiving, and to identify areas in which improvement is needed. Similar to other health-care disciplines, evaluation of perceived service quality in an acute care oncology setting, involves a diverse array of methodologies including longitudinal surveys, in-depth interviews, focus-group discussions, patient panels, consultation of voluntary groups, and analyses of patient feedback and concerns, followed by quick improvements to operations to help patients while they are undergoing care throughout the full cycle of treatment and follow up, as well as to help future patients. Patient-reported service quality surveys still continue to be the most widely used method of objectively and systematically determining a cancer patient's perception of the healthcare received. Cancer patients should be surveyed regularly due to the often aggressive nature of the disease and treatment. The modes of therapies have their own side effects and often result in difficult patient compliance. As a result, considerable demands are placed on health care providers to satisfy the complex healthcare needs of cancer patients.\nThe literature shows that perceived service quality can act as a marker for patient willingness to comply with the treatment plan as well as to predict a patient's willingness to recommend a provider to friends and relatives [8,9]. This is especially important in many countries where service quality data are not readily published and recommendations from family or friends becomes an important source of information for selecting a provider [9,10]. There are several studies in the literature that have evaluated service quality in cancers like gastro esophageal [11], breast [5,12], colorectal [13], lung, prostate [14] and gynecological [15,16]. Collectively, these studies have found that satisfaction with the information provided by medical staff about a patient's illness and the course of treatment is important. This is followed closely by the time spent with the physician and the interpersonal skills of the physician. Other key factors are waiting time to get an appointment, empathy of staff with the patient, the continuity of care provided, and satisfaction with the nursing staff [17]. We are unaware of any information in the oncology literature demonstrating a link between perceived service quality and patient willingness to recommend a provider. In light of the importance of this information to the healthcare industry, as well as with the goal of taking the existing research in this area to the next level, we designed a study to investigate the relationship between perceived service quality and patient willingness to recommend at a network of national oncology hospitals.", "All returning treating patients were eligible for inclusion in this study. Patients with all stages of all cancer types were eligible for the study. Specifically, patients who participated in the study were randomly selected from a population that had not responded to a service quality questionnaire within 60 days of the start of the study. The selected patients were approached onsite for survey administration. The surveyed cohort included 2018 randomly-selected returning cancer patients who had been treated at one of three Cancer Treatment Centers of America® (CTCA) hospitals between July 2007 and September 2009. The study was approved by the CTCA Institutional Review Board.", "The service quality questionnaire used in this study was first developed and implemented by the Research team at CTCA in August 2006. The questionnaire was developed based on a patient-centered approach that used questions that patients view as important in their treatment experience. In addition to patient focus groups, survey dimensions were collated from several existing studies or questionnaires of oncology patients [18-21]. This service quality questionnaire covers the following dimensions of patient satisfaction: hospital operations and services, physicians and staff, and patient endorsements for others (friends and associates). After the patient consented to complete the survey, the Survey Research Associate completed the \"office use only section\" on the last page of the survey which includes unique patient identifiers. The survey was then given to the patient. The Survey Research Associate then opened and explained the survey, specifically describing the rating scale and the open-ended questions. Next, the Survey Research Associate informed the patient that he/she will return to collect the survey and/or explained the option of the comment/suggestion drop box. Throughout the day, the Survey Research Associate updated the survey tracking list to note the following: patients contacted, surveys returned, surveys declined, and missed patients.", "Patient willingness to recommend, \"will you recommend this facility to friends and associates?\" was used as the dependent variable and was measured on an 11-point scale ranging from \"not at all likely\" to \"extremely likely\". This question is used to calculate the Net Promoter Score [22,23], a measure that has been shown in a number of industries to effectively measure customer loyalty, with increasing use in healthcare, including our hospitals as a management tool. For the purpose of this analysis, as well as in accordance with previously reported research [9,10], data were dichotomized into 2 categories: top box response (10) versus all others (0-9). The service quality items that were used as independent variables in this study were the ease of the admission (registration) process, the speed of the admission (registration) process, the timeliness with which care was delivered, team helping you understand your medical condition, team explaining your treatment options, team involving you in decision making, the amount of time spent team with you, team calling you by your name, team genuinely caring for you as an individual, team providing you with a sense of well-being, our team's \"whole person\" approach to patient care and the CTCA medical oncologist (patient's primary physician). These items were measured on a 7-point Likert-type scale ranging from \"completely dissatisfied\" to \"completely satisfied.\" Each service quality item was also dichotomized into 2 categories: \"completely satisfied\" (7) and \"not completely satisfied\" (1-6). Other control variables that were investigated for their relationship with patient willingness to recommend were age at diagnosis, prior treatment history and gender. The prior treatment history variable categorized patients into those who have received definitive cancer treatment elsewhere before coming to CTCA and those who were newly diagnosed at CTCA. The multivariate analysis also adjusted for the effects of CTCA center and survey year with dummy variables representing these categories.\nDescriptive statistics and frequencies were computed for each service quality item in the questionnaire. The relationship between perceived service quality and \"willingness to recommend\" was initially assessed via Kendall's tau b correlation and univariate logistic regression. Kendall's tau b is an appropriate measure of association for categorical variables and is commonly used when both variables have the same number of categories. Logistic regression was then employed to develop a multivariate model to predict patient willingness to recommend. Potential multicollinearity was assessed in two steps. Large values (above 0.70) of tau b were used as an initial screen for pairs of service quality measures, with one member of the pair not entered into the multivariate model (the measure that was more meaningful or actionable was retained). As a second check, the variance inflation factor was used with the final model to verify that multicollinearity was not significantly influencing model coefficients.\nThe effect of perceived service quality on patient willingness to recommend was expressed as odds ratios (ORs) with 95% confidence intervals. A difference was considered to be statistically significant if the p value was less than or equal to 0.05. All data were analyzed using SPSS version 17.0 (SPSS, Chicago, IL, USA).", "[SUBTITLE] Response Rate [SUBSECTION] A total of 2754 returning patients were contacted at all three centers combined to participate in the survey between July 2007 and September 2009. However, only 2018 patients responded. As a result, the response rate for this study was 73.3%.\nA total of 2754 returning patients were contacted at all three centers combined to participate in the survey between July 2007 and September 2009. However, only 2018 patients responded. As a result, the response rate for this study was 73.3%.\n[SUBTITLE] Baseline Patient Characteristics [SUBSECTION] Table 1 displays baseline patient characteristics across the entire study population (N = 2018). The most frequent diagnoses were breast (N = 412), lung (N = 294), prostate (N = 260), colorectal (N = 179) and pancreatic (N = 169) cancer.\nBaseline Patient Characteristics (N = 2018)\nTable 1 displays baseline patient characteristics across the entire study population (N = 2018). The most frequent diagnoses were breast (N = 412), lung (N = 294), prostate (N = 260), colorectal (N = 179) and pancreatic (N = 169) cancer.\nBaseline Patient Characteristics (N = 2018)\n[SUBTITLE] Service Quality Items [SUBSECTION] Table 2 describes the level of patient satisfaction with service quality items concerning CTCA operations and services. Table 3 describes the level of patient satisfaction with service quality items concerning CTCA's multidisciplinary patient care team. Table 4 reports the patient willingness to recommend CTCA to friends and associates. 1553 (77.0%) patients said they were \"extremely likely\" to recommend CTCA to friends and associates.\nService Quality Items: Operations and Services\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nService Quality Items: Multidisciplinary Patient Care Team\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nPatient endorsement of CTCA for themselves and others (N = 1963)\nTable 2 describes the level of patient satisfaction with service quality items concerning CTCA operations and services. Table 3 describes the level of patient satisfaction with service quality items concerning CTCA's multidisciplinary patient care team. Table 4 reports the patient willingness to recommend CTCA to friends and associates. 1553 (77.0%) patients said they were \"extremely likely\" to recommend CTCA to friends and associates.\nService Quality Items: Operations and Services\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nService Quality Items: Multidisciplinary Patient Care Team\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nPatient endorsement of CTCA for themselves and others (N = 1963)\n[SUBTITLE] Univariate Analysis - Predictors of Patient Willingness to Recommend [SUBSECTION] Kendall's tau b correlations between the service quality measures and willingness to recommend were all significant at p < .05, with values ranging from 0.20 to 0.40 (see Table 5). Univariate logistic regression analyses were also all significant at p < .05, with odds ratios ranging from 3.2 to 9.5 (see Table 6)., In addition, prior treatment history was found to be predictive of patient willingness to recommend such that newly diagnosed patients were more likely to recommend as compared to those who had been previously treated. Age and gender were not significant.\nAssociation between Patient Endorsement of CTCA and Service Quality Measures\nUnivariate Logistic Regression Analysis\nKendall's tau b correlations between the service quality measures and willingness to recommend were all significant at p < .05, with values ranging from 0.20 to 0.40 (see Table 5). Univariate logistic regression analyses were also all significant at p < .05, with odds ratios ranging from 3.2 to 9.5 (see Table 6)., In addition, prior treatment history was found to be predictive of patient willingness to recommend such that newly diagnosed patients were more likely to recommend as compared to those who had been previously treated. Age and gender were not significant.\nAssociation between Patient Endorsement of CTCA and Service Quality Measures\nUnivariate Logistic Regression Analysis\n[SUBTITLE] Multivariate Analysis - Predictors of Patient Willingness to Recommend [SUBSECTION] Before proceeding with multivariate analysis, we checked the bivariate Kendall's tau b correlation among the service quality predictors to screen for observable multicollinearity. Speed of admission and ease of admission were highly correlated (tau b = 0.74). \"Explaining treatment options\" was highly correlated with several items (\"helping you understand your condition\", tau b = 0.77 and \"involving you in decision-making\", tau b = 0.74). \"Providing a sense of well being\" and \"caring for you as an individual\" were highly correlated (tau b = 0.70). \"Ease of admission\", \"explaining treatment options\", and \"providing a sense of well being\" were accordingly not used in the multivariate model. \"Ease of admission\" and \"providing a sense of well being\" were dropped because we believe they may not have been consistently interpreted by patients. \"Explaining treatment options\" was dropped because it was highly correlated with several items and so dropping it was the most parsimonious approach.\nTable 7 displays the results of the multivariate logistic regression. The overall model was significant (chi-square 426.0, df = 16, p < .001). The service quality items that were significant in the final model were \"team helping you understand your medical condition\", \"staff genuinely caring for you as an individual\" \"whole person approach to patient care\" and \"CTCA medical oncologist.\" Odds ratios ranged from about 2.0 to 2.2 for these service quality measures. Gender, treatment history, CTCA center and survey year were also significant. Males had lower willingness to recommend than females. Newly diagnosed patients had higher willingness to recommend as compared to those who were previously treated. Patients surveyed in 2009 were more like to recommend as compared to those surveyed in 2007. Patients surveyed at CTCA Southwestern in Tulsa, OK, were more likely to recommend as compared to those treated at CTCA Midwestern in Zion, IL, and CTCA Eastern at Philadelphia, PA. Finally, the type of cancer diagnosis was not found to influence patient \"willingness to recommend\" in the multivariate model. VIF values for the service quality measures ranged from 1.3 to 2.5, none of which indicate a significant problem with multicollinearity [24,25].\nMultivariate Logistic Regression Analysis\nBefore proceeding with multivariate analysis, we checked the bivariate Kendall's tau b correlation among the service quality predictors to screen for observable multicollinearity. Speed of admission and ease of admission were highly correlated (tau b = 0.74). \"Explaining treatment options\" was highly correlated with several items (\"helping you understand your condition\", tau b = 0.77 and \"involving you in decision-making\", tau b = 0.74). \"Providing a sense of well being\" and \"caring for you as an individual\" were highly correlated (tau b = 0.70). \"Ease of admission\", \"explaining treatment options\", and \"providing a sense of well being\" were accordingly not used in the multivariate model. \"Ease of admission\" and \"providing a sense of well being\" were dropped because we believe they may not have been consistently interpreted by patients. \"Explaining treatment options\" was dropped because it was highly correlated with several items and so dropping it was the most parsimonious approach.\nTable 7 displays the results of the multivariate logistic regression. The overall model was significant (chi-square 426.0, df = 16, p < .001). The service quality items that were significant in the final model were \"team helping you understand your medical condition\", \"staff genuinely caring for you as an individual\" \"whole person approach to patient care\" and \"CTCA medical oncologist.\" Odds ratios ranged from about 2.0 to 2.2 for these service quality measures. Gender, treatment history, CTCA center and survey year were also significant. Males had lower willingness to recommend than females. Newly diagnosed patients had higher willingness to recommend as compared to those who were previously treated. Patients surveyed in 2009 were more like to recommend as compared to those surveyed in 2007. Patients surveyed at CTCA Southwestern in Tulsa, OK, were more likely to recommend as compared to those treated at CTCA Midwestern in Zion, IL, and CTCA Eastern at Philadelphia, PA. Finally, the type of cancer diagnosis was not found to influence patient \"willingness to recommend\" in the multivariate model. VIF values for the service quality measures ranged from 1.3 to 2.5, none of which indicate a significant problem with multicollinearity [24,25].\nMultivariate Logistic Regression Analysis", "A total of 2754 returning patients were contacted at all three centers combined to participate in the survey between July 2007 and September 2009. However, only 2018 patients responded. As a result, the response rate for this study was 73.3%.", "Table 1 displays baseline patient characteristics across the entire study population (N = 2018). The most frequent diagnoses were breast (N = 412), lung (N = 294), prostate (N = 260), colorectal (N = 179) and pancreatic (N = 169) cancer.\nBaseline Patient Characteristics (N = 2018)", "Table 2 describes the level of patient satisfaction with service quality items concerning CTCA operations and services. Table 3 describes the level of patient satisfaction with service quality items concerning CTCA's multidisciplinary patient care team. Table 4 reports the patient willingness to recommend CTCA to friends and associates. 1553 (77.0%) patients said they were \"extremely likely\" to recommend CTCA to friends and associates.\nService Quality Items: Operations and Services\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nService Quality Items: Multidisciplinary Patient Care Team\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nPatient endorsement of CTCA for themselves and others (N = 1963)", "Kendall's tau b correlations between the service quality measures and willingness to recommend were all significant at p < .05, with values ranging from 0.20 to 0.40 (see Table 5). Univariate logistic regression analyses were also all significant at p < .05, with odds ratios ranging from 3.2 to 9.5 (see Table 6)., In addition, prior treatment history was found to be predictive of patient willingness to recommend such that newly diagnosed patients were more likely to recommend as compared to those who had been previously treated. Age and gender were not significant.\nAssociation between Patient Endorsement of CTCA and Service Quality Measures\nUnivariate Logistic Regression Analysis", "Before proceeding with multivariate analysis, we checked the bivariate Kendall's tau b correlation among the service quality predictors to screen for observable multicollinearity. Speed of admission and ease of admission were highly correlated (tau b = 0.74). \"Explaining treatment options\" was highly correlated with several items (\"helping you understand your condition\", tau b = 0.77 and \"involving you in decision-making\", tau b = 0.74). \"Providing a sense of well being\" and \"caring for you as an individual\" were highly correlated (tau b = 0.70). \"Ease of admission\", \"explaining treatment options\", and \"providing a sense of well being\" were accordingly not used in the multivariate model. \"Ease of admission\" and \"providing a sense of well being\" were dropped because we believe they may not have been consistently interpreted by patients. \"Explaining treatment options\" was dropped because it was highly correlated with several items and so dropping it was the most parsimonious approach.\nTable 7 displays the results of the multivariate logistic regression. The overall model was significant (chi-square 426.0, df = 16, p < .001). The service quality items that were significant in the final model were \"team helping you understand your medical condition\", \"staff genuinely caring for you as an individual\" \"whole person approach to patient care\" and \"CTCA medical oncologist.\" Odds ratios ranged from about 2.0 to 2.2 for these service quality measures. Gender, treatment history, CTCA center and survey year were also significant. Males had lower willingness to recommend than females. Newly diagnosed patients had higher willingness to recommend as compared to those who were previously treated. Patients surveyed in 2009 were more like to recommend as compared to those surveyed in 2007. Patients surveyed at CTCA Southwestern in Tulsa, OK, were more likely to recommend as compared to those treated at CTCA Midwestern in Zion, IL, and CTCA Eastern at Philadelphia, PA. Finally, the type of cancer diagnosis was not found to influence patient \"willingness to recommend\" in the multivariate model. VIF values for the service quality measures ranged from 1.3 to 2.5, none of which indicate a significant problem with multicollinearity [24,25].\nMultivariate Logistic Regression Analysis", "Patient-reported service quality assesses the extent to which an individual's health care experiences match his or her expectations which in turn can influence a patient's willingness to recommend a health care provider to friends and associates. The present study investigates this association in an acute care national oncology hospital network.\nOur findings show that helping a patient to understand her/his condition, caring for a patient as an individual, a whole-person approach to care, and satisfaction with the medical oncologist all contribute to willingness to recommend CTCA to friends and associates. On the other hand, speed of admission, timeliness with which care was delivered, involving a patient in decision-making, calling a patient by their name, and the amount of time spent with a patient may not be as critical in willingness to recommend, relative to the other measures studied. These findings suggest that service quality that is central to the patient experience is critical for patient loyalty. The only seeming exceptions to this are amount of time spent with a patient and involving a patient in decision-making, but in this population, patients find these of lesser import than the quality of care itself. Further studies would need to examine these factors in new patient populations.\nIn order to put our study in context, we review here a few available studies in the healthcare literature which have investigated service quality predictors of patient willingness to recommend a healthcare provider. A study conducted in 1910 patients in clinics throughout Taiwan investigated whether attributes of perceived clinic quality and patient education were associated with patient satisfaction and recommendation of a primary care provider. Patient recommendation was measured on a five-point Likert scale using the question 'When your family, relatives or friends need to see a doctor, would you recommend this clinic?' The study found doctor's technical skill to be the most critical attribute of primary care quality for both overall satisfaction and recommendation, followed by doctor's interpersonal skill [9]. Another study conducted in 4945 patients in 126 Taiwanese hospitals examined the correlation of patient satisfaction with and recommendation of a hospital to patient ratings of interpersonal and technical performance of the hospital. Patient recommendation was measured on a five-point Likert scale using the question 'If someone asks you about the hospital, would you recommend it?' The study found that technical competence was a more influential predictor for recommendation [10]. Another study conducted in 2160 consecutive adult patients treated within 36 family practice clinics in Slovenia investigated factors influencing patients' recommendation of doctor. Patients' responded to the statement \"I can strongly recommend my family doctor to my friends\" on a five-point scale, from strongly disagree to strongly agree. Higher satisfaction with doctor's working style and organization of the health care system predicted patient recommendation [26].\nThe results of our study do not compare directly with above mentioned studies because of the differences in study design, patient population studied, questionnaire used and factors adjusted for. Nevertheless, our study adds useful information to the growing body of literature on the importance of assessing patient perception about service quality as a predictor of patient willingness to recommend a hospital.\nAlthough this study reports on a relatively uncommon analysis of predicting patient willingness to recommend with perceived service quality, several limitations of the study require acknowledgment. The patient cohort was limited to only those patients who were English speakers, so this study sample is therefore not broadly representative of cancer patients in general. As a result, the generalizability of this study is limited. The data we used for this study were not primarily meant for research purposes. CTCA is a unique medical center. It specializes in treating only cancer patients, and it has an intense focus on patient-centered care. Compared to other centers, patients report very high levels of service quality at CTCA. Our study, which is hypothesis generating by nature, used a non-validated patient satisfaction questionnaire. However, it is reasonable to use a non-validated survey if the intent of the study is hypothesis generation rather than hypothesis testing. It might be argued that patients do not have the ability to judge a hospital's performance; however, patient perception is a key factor for hospital selection. This was the main goal of our study - to show the effects of patient perception about service quality on patient recommendation of a hospital. Finally, a response rate of 73.3% could potentially introduce a selection bias in our study. The baseline characteristics of patients who did not respond are not available for us to evaluate any systematic differences between responders and non-responders.\nMore and more health care consumers are using the web to research and shop for the best health care providers, especially for complex medical conditions. In addition, in our own experience, we hear about more and more patients traveling great distances to receive care, due in large part, to a strong recommendation of the provider from the patient's friend, associate, colleague, or family member. And as the asymmetry in information between providers and consumers decreases, we can and should expect to see more informed consumers shopping for the best available healthcare.\nThe strengths of our study include: a large sample size, the fact that we measured service quality as close to the time service was delivered as possible, and the fact that we used willingness to recommend (using the question and scale most commonly used in industry) as our dependent variable, which has been previously validated through research in many industries. To the best of our knowledge, this study is the first in the health care literature to report on the positive correlation between patient-reported service quality and patient willingness to recommend a provider in a large sample of cancer patients.\nDuring a time in which quality and value are becoming increasingly prominent themes in healthcare reform, we believe the patient's perspective of the key drivers of loyalty should be given greater national attention. In most of American medicine, we assess the relative importance of quality and patient satisfaction measures from expert panels and other traditional research methods developing institutionalized views of the attributes of health care that are most important. But what has been largely missing in the conversation is the patient's perspective (and perception) of the relative importance of key aspects of service quality in the health care delivery cycle. These are areas that require further research and at the time of this writing, our organization is seeking partners to conduct national population-based research on the key drivers of value in oncology, with service quality being an important dimension to be studied. As the health care legislation continues to be implemented, the entire health care system would benefit from a greater focus on the key drivers of value from the consumer's perspective - the patient, as well as their caregivers and families. These are important areas of research that will lend greater focus to where, when, and how we should apply our scarce resources to deliver the most valuable care to our patients - the ultimate consumer.\nNext steps in our research include linking data on service quality to patient outcomes. We are unaware of any literature linking service quality to data on patient quality of life, length of life, and overall satisfaction with health. Research is also underway at our center to explore the relationship between patient willingness to recommend and actual patient return (behavior) as well as how changes in patients' clinical condition affect their willingness to recommend a provider, controlling for all other known variables. With respect to population-based research, we do plan on conducting national research on the patient's perspective of value in oncologic care, data that has largely been missing from the health policy discussions.", "In this multi-center study, we demonstrate the predictive significance of perceived service quality as it relates to patient willingness to recommend an oncology service provider. We identified four key service quality drivers of patient loyalty: \"team helping you understand your medical condition\", \"staff genuinely caring for you as an individual\" \"whole person approach to patient care\" and \"CTCA medical oncologist\".", "The authors declare that they have no competing interests.", "CGL participated in concept, design, writing, statistical analysis and data interpretation. MR participated in concept, statistical analysis and data interpretation. DG participated in concept, statistical analysis, data interpretation and writing. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1472-6963/11/46/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study Population", "Questionnaire and Survey Administration", "Statistical Analysis", "Results", "Response Rate", "Baseline Patient Characteristics", "Service Quality Items", "Univariate Analysis - Predictors of Patient Willingness to Recommend", "Multivariate Analysis - Predictors of Patient Willingness to Recommend", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "As consumerism continues to increase in healthcare, there has been a rise in awareness about how patients perceive the quality of the services they receive at a health-care institution [1,2]. In addition, the web offers patients the opportunity to shop for the best places for care due to the rise in transparency of provider information on service quality and patient experience. As a consequence, patient satisfaction with service quality is becoming an increasingly important tool for providers to demonstrate patient focus and differentiation in the marketplace, as well as enhance patient loyalty. Furthermore, providers are using the information to make important decisions regarding operational and treatment plans [3].\nEvaluations of service quality provide important data on the patient's perception of the quality of care and treatment delivered by physicians, paramedical staff and the hospital as a whole [4]. Health providers can use data on service quality to design and track quality improvement over time and compare themselves to other health providers when the same measures are used, as well as recognize and expeditiously resolve service problems in real-time [5,6]. Measuring service quality also helps health care providers identify specific, and often unmet needs of patients, which has been a large focus of our work, and demonstrated by other research [7].\nSimilar to other acute health-care settings, the assessment of service quality, as perceived by patients, is critical in the oncology setting as well. Advances in diagnostics, treatment, supportive care and rehabilitation all necessitate continued monitoring to determine whether patients are satisfied with the increasingly complex and multidisciplinary nature of health care services that they are receiving, and to identify areas in which improvement is needed. Similar to other health-care disciplines, evaluation of perceived service quality in an acute care oncology setting, involves a diverse array of methodologies including longitudinal surveys, in-depth interviews, focus-group discussions, patient panels, consultation of voluntary groups, and analyses of patient feedback and concerns, followed by quick improvements to operations to help patients while they are undergoing care throughout the full cycle of treatment and follow up, as well as to help future patients. Patient-reported service quality surveys still continue to be the most widely used method of objectively and systematically determining a cancer patient's perception of the healthcare received. Cancer patients should be surveyed regularly due to the often aggressive nature of the disease and treatment. The modes of therapies have their own side effects and often result in difficult patient compliance. As a result, considerable demands are placed on health care providers to satisfy the complex healthcare needs of cancer patients.\nThe literature shows that perceived service quality can act as a marker for patient willingness to comply with the treatment plan as well as to predict a patient's willingness to recommend a provider to friends and relatives [8,9]. This is especially important in many countries where service quality data are not readily published and recommendations from family or friends becomes an important source of information for selecting a provider [9,10]. There are several studies in the literature that have evaluated service quality in cancers like gastro esophageal [11], breast [5,12], colorectal [13], lung, prostate [14] and gynecological [15,16]. Collectively, these studies have found that satisfaction with the information provided by medical staff about a patient's illness and the course of treatment is important. This is followed closely by the time spent with the physician and the interpersonal skills of the physician. Other key factors are waiting time to get an appointment, empathy of staff with the patient, the continuity of care provided, and satisfaction with the nursing staff [17]. We are unaware of any information in the oncology literature demonstrating a link between perceived service quality and patient willingness to recommend a provider. In light of the importance of this information to the healthcare industry, as well as with the goal of taking the existing research in this area to the next level, we designed a study to investigate the relationship between perceived service quality and patient willingness to recommend at a network of national oncology hospitals.", "[SUBTITLE] Study Population [SUBSECTION] All returning treating patients were eligible for inclusion in this study. Patients with all stages of all cancer types were eligible for the study. Specifically, patients who participated in the study were randomly selected from a population that had not responded to a service quality questionnaire within 60 days of the start of the study. The selected patients were approached onsite for survey administration. The surveyed cohort included 2018 randomly-selected returning cancer patients who had been treated at one of three Cancer Treatment Centers of America® (CTCA) hospitals between July 2007 and September 2009. The study was approved by the CTCA Institutional Review Board.\nAll returning treating patients were eligible for inclusion in this study. Patients with all stages of all cancer types were eligible for the study. Specifically, patients who participated in the study were randomly selected from a population that had not responded to a service quality questionnaire within 60 days of the start of the study. The selected patients were approached onsite for survey administration. The surveyed cohort included 2018 randomly-selected returning cancer patients who had been treated at one of three Cancer Treatment Centers of America® (CTCA) hospitals between July 2007 and September 2009. The study was approved by the CTCA Institutional Review Board.\n[SUBTITLE] Questionnaire and Survey Administration [SUBSECTION] The service quality questionnaire used in this study was first developed and implemented by the Research team at CTCA in August 2006. The questionnaire was developed based on a patient-centered approach that used questions that patients view as important in their treatment experience. In addition to patient focus groups, survey dimensions were collated from several existing studies or questionnaires of oncology patients [18-21]. This service quality questionnaire covers the following dimensions of patient satisfaction: hospital operations and services, physicians and staff, and patient endorsements for others (friends and associates). After the patient consented to complete the survey, the Survey Research Associate completed the \"office use only section\" on the last page of the survey which includes unique patient identifiers. The survey was then given to the patient. The Survey Research Associate then opened and explained the survey, specifically describing the rating scale and the open-ended questions. Next, the Survey Research Associate informed the patient that he/she will return to collect the survey and/or explained the option of the comment/suggestion drop box. Throughout the day, the Survey Research Associate updated the survey tracking list to note the following: patients contacted, surveys returned, surveys declined, and missed patients.\nThe service quality questionnaire used in this study was first developed and implemented by the Research team at CTCA in August 2006. The questionnaire was developed based on a patient-centered approach that used questions that patients view as important in their treatment experience. In addition to patient focus groups, survey dimensions were collated from several existing studies or questionnaires of oncology patients [18-21]. This service quality questionnaire covers the following dimensions of patient satisfaction: hospital operations and services, physicians and staff, and patient endorsements for others (friends and associates). After the patient consented to complete the survey, the Survey Research Associate completed the \"office use only section\" on the last page of the survey which includes unique patient identifiers. The survey was then given to the patient. The Survey Research Associate then opened and explained the survey, specifically describing the rating scale and the open-ended questions. Next, the Survey Research Associate informed the patient that he/she will return to collect the survey and/or explained the option of the comment/suggestion drop box. Throughout the day, the Survey Research Associate updated the survey tracking list to note the following: patients contacted, surveys returned, surveys declined, and missed patients.\n[SUBTITLE] Statistical Analysis [SUBSECTION] Patient willingness to recommend, \"will you recommend this facility to friends and associates?\" was used as the dependent variable and was measured on an 11-point scale ranging from \"not at all likely\" to \"extremely likely\". This question is used to calculate the Net Promoter Score [22,23], a measure that has been shown in a number of industries to effectively measure customer loyalty, with increasing use in healthcare, including our hospitals as a management tool. For the purpose of this analysis, as well as in accordance with previously reported research [9,10], data were dichotomized into 2 categories: top box response (10) versus all others (0-9). The service quality items that were used as independent variables in this study were the ease of the admission (registration) process, the speed of the admission (registration) process, the timeliness with which care was delivered, team helping you understand your medical condition, team explaining your treatment options, team involving you in decision making, the amount of time spent team with you, team calling you by your name, team genuinely caring for you as an individual, team providing you with a sense of well-being, our team's \"whole person\" approach to patient care and the CTCA medical oncologist (patient's primary physician). These items were measured on a 7-point Likert-type scale ranging from \"completely dissatisfied\" to \"completely satisfied.\" Each service quality item was also dichotomized into 2 categories: \"completely satisfied\" (7) and \"not completely satisfied\" (1-6). Other control variables that were investigated for their relationship with patient willingness to recommend were age at diagnosis, prior treatment history and gender. The prior treatment history variable categorized patients into those who have received definitive cancer treatment elsewhere before coming to CTCA and those who were newly diagnosed at CTCA. The multivariate analysis also adjusted for the effects of CTCA center and survey year with dummy variables representing these categories.\nDescriptive statistics and frequencies were computed for each service quality item in the questionnaire. The relationship between perceived service quality and \"willingness to recommend\" was initially assessed via Kendall's tau b correlation and univariate logistic regression. Kendall's tau b is an appropriate measure of association for categorical variables and is commonly used when both variables have the same number of categories. Logistic regression was then employed to develop a multivariate model to predict patient willingness to recommend. Potential multicollinearity was assessed in two steps. Large values (above 0.70) of tau b were used as an initial screen for pairs of service quality measures, with one member of the pair not entered into the multivariate model (the measure that was more meaningful or actionable was retained). As a second check, the variance inflation factor was used with the final model to verify that multicollinearity was not significantly influencing model coefficients.\nThe effect of perceived service quality on patient willingness to recommend was expressed as odds ratios (ORs) with 95% confidence intervals. A difference was considered to be statistically significant if the p value was less than or equal to 0.05. All data were analyzed using SPSS version 17.0 (SPSS, Chicago, IL, USA).\nPatient willingness to recommend, \"will you recommend this facility to friends and associates?\" was used as the dependent variable and was measured on an 11-point scale ranging from \"not at all likely\" to \"extremely likely\". This question is used to calculate the Net Promoter Score [22,23], a measure that has been shown in a number of industries to effectively measure customer loyalty, with increasing use in healthcare, including our hospitals as a management tool. For the purpose of this analysis, as well as in accordance with previously reported research [9,10], data were dichotomized into 2 categories: top box response (10) versus all others (0-9). The service quality items that were used as independent variables in this study were the ease of the admission (registration) process, the speed of the admission (registration) process, the timeliness with which care was delivered, team helping you understand your medical condition, team explaining your treatment options, team involving you in decision making, the amount of time spent team with you, team calling you by your name, team genuinely caring for you as an individual, team providing you with a sense of well-being, our team's \"whole person\" approach to patient care and the CTCA medical oncologist (patient's primary physician). These items were measured on a 7-point Likert-type scale ranging from \"completely dissatisfied\" to \"completely satisfied.\" Each service quality item was also dichotomized into 2 categories: \"completely satisfied\" (7) and \"not completely satisfied\" (1-6). Other control variables that were investigated for their relationship with patient willingness to recommend were age at diagnosis, prior treatment history and gender. The prior treatment history variable categorized patients into those who have received definitive cancer treatment elsewhere before coming to CTCA and those who were newly diagnosed at CTCA. The multivariate analysis also adjusted for the effects of CTCA center and survey year with dummy variables representing these categories.\nDescriptive statistics and frequencies were computed for each service quality item in the questionnaire. The relationship between perceived service quality and \"willingness to recommend\" was initially assessed via Kendall's tau b correlation and univariate logistic regression. Kendall's tau b is an appropriate measure of association for categorical variables and is commonly used when both variables have the same number of categories. Logistic regression was then employed to develop a multivariate model to predict patient willingness to recommend. Potential multicollinearity was assessed in two steps. Large values (above 0.70) of tau b were used as an initial screen for pairs of service quality measures, with one member of the pair not entered into the multivariate model (the measure that was more meaningful or actionable was retained). As a second check, the variance inflation factor was used with the final model to verify that multicollinearity was not significantly influencing model coefficients.\nThe effect of perceived service quality on patient willingness to recommend was expressed as odds ratios (ORs) with 95% confidence intervals. A difference was considered to be statistically significant if the p value was less than or equal to 0.05. All data were analyzed using SPSS version 17.0 (SPSS, Chicago, IL, USA).", "All returning treating patients were eligible for inclusion in this study. Patients with all stages of all cancer types were eligible for the study. Specifically, patients who participated in the study were randomly selected from a population that had not responded to a service quality questionnaire within 60 days of the start of the study. The selected patients were approached onsite for survey administration. The surveyed cohort included 2018 randomly-selected returning cancer patients who had been treated at one of three Cancer Treatment Centers of America® (CTCA) hospitals between July 2007 and September 2009. The study was approved by the CTCA Institutional Review Board.", "The service quality questionnaire used in this study was first developed and implemented by the Research team at CTCA in August 2006. The questionnaire was developed based on a patient-centered approach that used questions that patients view as important in their treatment experience. In addition to patient focus groups, survey dimensions were collated from several existing studies or questionnaires of oncology patients [18-21]. This service quality questionnaire covers the following dimensions of patient satisfaction: hospital operations and services, physicians and staff, and patient endorsements for others (friends and associates). After the patient consented to complete the survey, the Survey Research Associate completed the \"office use only section\" on the last page of the survey which includes unique patient identifiers. The survey was then given to the patient. The Survey Research Associate then opened and explained the survey, specifically describing the rating scale and the open-ended questions. Next, the Survey Research Associate informed the patient that he/she will return to collect the survey and/or explained the option of the comment/suggestion drop box. Throughout the day, the Survey Research Associate updated the survey tracking list to note the following: patients contacted, surveys returned, surveys declined, and missed patients.", "Patient willingness to recommend, \"will you recommend this facility to friends and associates?\" was used as the dependent variable and was measured on an 11-point scale ranging from \"not at all likely\" to \"extremely likely\". This question is used to calculate the Net Promoter Score [22,23], a measure that has been shown in a number of industries to effectively measure customer loyalty, with increasing use in healthcare, including our hospitals as a management tool. For the purpose of this analysis, as well as in accordance with previously reported research [9,10], data were dichotomized into 2 categories: top box response (10) versus all others (0-9). The service quality items that were used as independent variables in this study were the ease of the admission (registration) process, the speed of the admission (registration) process, the timeliness with which care was delivered, team helping you understand your medical condition, team explaining your treatment options, team involving you in decision making, the amount of time spent team with you, team calling you by your name, team genuinely caring for you as an individual, team providing you with a sense of well-being, our team's \"whole person\" approach to patient care and the CTCA medical oncologist (patient's primary physician). These items were measured on a 7-point Likert-type scale ranging from \"completely dissatisfied\" to \"completely satisfied.\" Each service quality item was also dichotomized into 2 categories: \"completely satisfied\" (7) and \"not completely satisfied\" (1-6). Other control variables that were investigated for their relationship with patient willingness to recommend were age at diagnosis, prior treatment history and gender. The prior treatment history variable categorized patients into those who have received definitive cancer treatment elsewhere before coming to CTCA and those who were newly diagnosed at CTCA. The multivariate analysis also adjusted for the effects of CTCA center and survey year with dummy variables representing these categories.\nDescriptive statistics and frequencies were computed for each service quality item in the questionnaire. The relationship between perceived service quality and \"willingness to recommend\" was initially assessed via Kendall's tau b correlation and univariate logistic regression. Kendall's tau b is an appropriate measure of association for categorical variables and is commonly used when both variables have the same number of categories. Logistic regression was then employed to develop a multivariate model to predict patient willingness to recommend. Potential multicollinearity was assessed in two steps. Large values (above 0.70) of tau b were used as an initial screen for pairs of service quality measures, with one member of the pair not entered into the multivariate model (the measure that was more meaningful or actionable was retained). As a second check, the variance inflation factor was used with the final model to verify that multicollinearity was not significantly influencing model coefficients.\nThe effect of perceived service quality on patient willingness to recommend was expressed as odds ratios (ORs) with 95% confidence intervals. A difference was considered to be statistically significant if the p value was less than or equal to 0.05. All data were analyzed using SPSS version 17.0 (SPSS, Chicago, IL, USA).", "[SUBTITLE] Response Rate [SUBSECTION] A total of 2754 returning patients were contacted at all three centers combined to participate in the survey between July 2007 and September 2009. However, only 2018 patients responded. As a result, the response rate for this study was 73.3%.\nA total of 2754 returning patients were contacted at all three centers combined to participate in the survey between July 2007 and September 2009. However, only 2018 patients responded. As a result, the response rate for this study was 73.3%.\n[SUBTITLE] Baseline Patient Characteristics [SUBSECTION] Table 1 displays baseline patient characteristics across the entire study population (N = 2018). The most frequent diagnoses were breast (N = 412), lung (N = 294), prostate (N = 260), colorectal (N = 179) and pancreatic (N = 169) cancer.\nBaseline Patient Characteristics (N = 2018)\nTable 1 displays baseline patient characteristics across the entire study population (N = 2018). The most frequent diagnoses were breast (N = 412), lung (N = 294), prostate (N = 260), colorectal (N = 179) and pancreatic (N = 169) cancer.\nBaseline Patient Characteristics (N = 2018)\n[SUBTITLE] Service Quality Items [SUBSECTION] Table 2 describes the level of patient satisfaction with service quality items concerning CTCA operations and services. Table 3 describes the level of patient satisfaction with service quality items concerning CTCA's multidisciplinary patient care team. Table 4 reports the patient willingness to recommend CTCA to friends and associates. 1553 (77.0%) patients said they were \"extremely likely\" to recommend CTCA to friends and associates.\nService Quality Items: Operations and Services\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nService Quality Items: Multidisciplinary Patient Care Team\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nPatient endorsement of CTCA for themselves and others (N = 1963)\nTable 2 describes the level of patient satisfaction with service quality items concerning CTCA operations and services. Table 3 describes the level of patient satisfaction with service quality items concerning CTCA's multidisciplinary patient care team. Table 4 reports the patient willingness to recommend CTCA to friends and associates. 1553 (77.0%) patients said they were \"extremely likely\" to recommend CTCA to friends and associates.\nService Quality Items: Operations and Services\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nService Quality Items: Multidisciplinary Patient Care Team\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nPatient endorsement of CTCA for themselves and others (N = 1963)\n[SUBTITLE] Univariate Analysis - Predictors of Patient Willingness to Recommend [SUBSECTION] Kendall's tau b correlations between the service quality measures and willingness to recommend were all significant at p < .05, with values ranging from 0.20 to 0.40 (see Table 5). Univariate logistic regression analyses were also all significant at p < .05, with odds ratios ranging from 3.2 to 9.5 (see Table 6)., In addition, prior treatment history was found to be predictive of patient willingness to recommend such that newly diagnosed patients were more likely to recommend as compared to those who had been previously treated. Age and gender were not significant.\nAssociation between Patient Endorsement of CTCA and Service Quality Measures\nUnivariate Logistic Regression Analysis\nKendall's tau b correlations between the service quality measures and willingness to recommend were all significant at p < .05, with values ranging from 0.20 to 0.40 (see Table 5). Univariate logistic regression analyses were also all significant at p < .05, with odds ratios ranging from 3.2 to 9.5 (see Table 6)., In addition, prior treatment history was found to be predictive of patient willingness to recommend such that newly diagnosed patients were more likely to recommend as compared to those who had been previously treated. Age and gender were not significant.\nAssociation between Patient Endorsement of CTCA and Service Quality Measures\nUnivariate Logistic Regression Analysis\n[SUBTITLE] Multivariate Analysis - Predictors of Patient Willingness to Recommend [SUBSECTION] Before proceeding with multivariate analysis, we checked the bivariate Kendall's tau b correlation among the service quality predictors to screen for observable multicollinearity. Speed of admission and ease of admission were highly correlated (tau b = 0.74). \"Explaining treatment options\" was highly correlated with several items (\"helping you understand your condition\", tau b = 0.77 and \"involving you in decision-making\", tau b = 0.74). \"Providing a sense of well being\" and \"caring for you as an individual\" were highly correlated (tau b = 0.70). \"Ease of admission\", \"explaining treatment options\", and \"providing a sense of well being\" were accordingly not used in the multivariate model. \"Ease of admission\" and \"providing a sense of well being\" were dropped because we believe they may not have been consistently interpreted by patients. \"Explaining treatment options\" was dropped because it was highly correlated with several items and so dropping it was the most parsimonious approach.\nTable 7 displays the results of the multivariate logistic regression. The overall model was significant (chi-square 426.0, df = 16, p < .001). The service quality items that were significant in the final model were \"team helping you understand your medical condition\", \"staff genuinely caring for you as an individual\" \"whole person approach to patient care\" and \"CTCA medical oncologist.\" Odds ratios ranged from about 2.0 to 2.2 for these service quality measures. Gender, treatment history, CTCA center and survey year were also significant. Males had lower willingness to recommend than females. Newly diagnosed patients had higher willingness to recommend as compared to those who were previously treated. Patients surveyed in 2009 were more like to recommend as compared to those surveyed in 2007. Patients surveyed at CTCA Southwestern in Tulsa, OK, were more likely to recommend as compared to those treated at CTCA Midwestern in Zion, IL, and CTCA Eastern at Philadelphia, PA. Finally, the type of cancer diagnosis was not found to influence patient \"willingness to recommend\" in the multivariate model. VIF values for the service quality measures ranged from 1.3 to 2.5, none of which indicate a significant problem with multicollinearity [24,25].\nMultivariate Logistic Regression Analysis\nBefore proceeding with multivariate analysis, we checked the bivariate Kendall's tau b correlation among the service quality predictors to screen for observable multicollinearity. Speed of admission and ease of admission were highly correlated (tau b = 0.74). \"Explaining treatment options\" was highly correlated with several items (\"helping you understand your condition\", tau b = 0.77 and \"involving you in decision-making\", tau b = 0.74). \"Providing a sense of well being\" and \"caring for you as an individual\" were highly correlated (tau b = 0.70). \"Ease of admission\", \"explaining treatment options\", and \"providing a sense of well being\" were accordingly not used in the multivariate model. \"Ease of admission\" and \"providing a sense of well being\" were dropped because we believe they may not have been consistently interpreted by patients. \"Explaining treatment options\" was dropped because it was highly correlated with several items and so dropping it was the most parsimonious approach.\nTable 7 displays the results of the multivariate logistic regression. The overall model was significant (chi-square 426.0, df = 16, p < .001). The service quality items that were significant in the final model were \"team helping you understand your medical condition\", \"staff genuinely caring for you as an individual\" \"whole person approach to patient care\" and \"CTCA medical oncologist.\" Odds ratios ranged from about 2.0 to 2.2 for these service quality measures. Gender, treatment history, CTCA center and survey year were also significant. Males had lower willingness to recommend than females. Newly diagnosed patients had higher willingness to recommend as compared to those who were previously treated. Patients surveyed in 2009 were more like to recommend as compared to those surveyed in 2007. Patients surveyed at CTCA Southwestern in Tulsa, OK, were more likely to recommend as compared to those treated at CTCA Midwestern in Zion, IL, and CTCA Eastern at Philadelphia, PA. Finally, the type of cancer diagnosis was not found to influence patient \"willingness to recommend\" in the multivariate model. VIF values for the service quality measures ranged from 1.3 to 2.5, none of which indicate a significant problem with multicollinearity [24,25].\nMultivariate Logistic Regression Analysis", "A total of 2754 returning patients were contacted at all three centers combined to participate in the survey between July 2007 and September 2009. However, only 2018 patients responded. As a result, the response rate for this study was 73.3%.", "Table 1 displays baseline patient characteristics across the entire study population (N = 2018). The most frequent diagnoses were breast (N = 412), lung (N = 294), prostate (N = 260), colorectal (N = 179) and pancreatic (N = 169) cancer.\nBaseline Patient Characteristics (N = 2018)", "Table 2 describes the level of patient satisfaction with service quality items concerning CTCA operations and services. Table 3 describes the level of patient satisfaction with service quality items concerning CTCA's multidisciplinary patient care team. Table 4 reports the patient willingness to recommend CTCA to friends and associates. 1553 (77.0%) patients said they were \"extremely likely\" to recommend CTCA to friends and associates.\nService Quality Items: Operations and Services\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nService Quality Items: Multidisciplinary Patient Care Team\nItems were dichotomized into 2 groups of \"completely satisfied (7)\" and \"not completely satisfied (1-6)\"\nPatient endorsement of CTCA for themselves and others (N = 1963)", "Kendall's tau b correlations between the service quality measures and willingness to recommend were all significant at p < .05, with values ranging from 0.20 to 0.40 (see Table 5). Univariate logistic regression analyses were also all significant at p < .05, with odds ratios ranging from 3.2 to 9.5 (see Table 6)., In addition, prior treatment history was found to be predictive of patient willingness to recommend such that newly diagnosed patients were more likely to recommend as compared to those who had been previously treated. Age and gender were not significant.\nAssociation between Patient Endorsement of CTCA and Service Quality Measures\nUnivariate Logistic Regression Analysis", "Before proceeding with multivariate analysis, we checked the bivariate Kendall's tau b correlation among the service quality predictors to screen for observable multicollinearity. Speed of admission and ease of admission were highly correlated (tau b = 0.74). \"Explaining treatment options\" was highly correlated with several items (\"helping you understand your condition\", tau b = 0.77 and \"involving you in decision-making\", tau b = 0.74). \"Providing a sense of well being\" and \"caring for you as an individual\" were highly correlated (tau b = 0.70). \"Ease of admission\", \"explaining treatment options\", and \"providing a sense of well being\" were accordingly not used in the multivariate model. \"Ease of admission\" and \"providing a sense of well being\" were dropped because we believe they may not have been consistently interpreted by patients. \"Explaining treatment options\" was dropped because it was highly correlated with several items and so dropping it was the most parsimonious approach.\nTable 7 displays the results of the multivariate logistic regression. The overall model was significant (chi-square 426.0, df = 16, p < .001). The service quality items that were significant in the final model were \"team helping you understand your medical condition\", \"staff genuinely caring for you as an individual\" \"whole person approach to patient care\" and \"CTCA medical oncologist.\" Odds ratios ranged from about 2.0 to 2.2 for these service quality measures. Gender, treatment history, CTCA center and survey year were also significant. Males had lower willingness to recommend than females. Newly diagnosed patients had higher willingness to recommend as compared to those who were previously treated. Patients surveyed in 2009 were more like to recommend as compared to those surveyed in 2007. Patients surveyed at CTCA Southwestern in Tulsa, OK, were more likely to recommend as compared to those treated at CTCA Midwestern in Zion, IL, and CTCA Eastern at Philadelphia, PA. Finally, the type of cancer diagnosis was not found to influence patient \"willingness to recommend\" in the multivariate model. VIF values for the service quality measures ranged from 1.3 to 2.5, none of which indicate a significant problem with multicollinearity [24,25].\nMultivariate Logistic Regression Analysis", "Patient-reported service quality assesses the extent to which an individual's health care experiences match his or her expectations which in turn can influence a patient's willingness to recommend a health care provider to friends and associates. The present study investigates this association in an acute care national oncology hospital network.\nOur findings show that helping a patient to understand her/his condition, caring for a patient as an individual, a whole-person approach to care, and satisfaction with the medical oncologist all contribute to willingness to recommend CTCA to friends and associates. On the other hand, speed of admission, timeliness with which care was delivered, involving a patient in decision-making, calling a patient by their name, and the amount of time spent with a patient may not be as critical in willingness to recommend, relative to the other measures studied. These findings suggest that service quality that is central to the patient experience is critical for patient loyalty. The only seeming exceptions to this are amount of time spent with a patient and involving a patient in decision-making, but in this population, patients find these of lesser import than the quality of care itself. Further studies would need to examine these factors in new patient populations.\nIn order to put our study in context, we review here a few available studies in the healthcare literature which have investigated service quality predictors of patient willingness to recommend a healthcare provider. A study conducted in 1910 patients in clinics throughout Taiwan investigated whether attributes of perceived clinic quality and patient education were associated with patient satisfaction and recommendation of a primary care provider. Patient recommendation was measured on a five-point Likert scale using the question 'When your family, relatives or friends need to see a doctor, would you recommend this clinic?' The study found doctor's technical skill to be the most critical attribute of primary care quality for both overall satisfaction and recommendation, followed by doctor's interpersonal skill [9]. Another study conducted in 4945 patients in 126 Taiwanese hospitals examined the correlation of patient satisfaction with and recommendation of a hospital to patient ratings of interpersonal and technical performance of the hospital. Patient recommendation was measured on a five-point Likert scale using the question 'If someone asks you about the hospital, would you recommend it?' The study found that technical competence was a more influential predictor for recommendation [10]. Another study conducted in 2160 consecutive adult patients treated within 36 family practice clinics in Slovenia investigated factors influencing patients' recommendation of doctor. Patients' responded to the statement \"I can strongly recommend my family doctor to my friends\" on a five-point scale, from strongly disagree to strongly agree. Higher satisfaction with doctor's working style and organization of the health care system predicted patient recommendation [26].\nThe results of our study do not compare directly with above mentioned studies because of the differences in study design, patient population studied, questionnaire used and factors adjusted for. Nevertheless, our study adds useful information to the growing body of literature on the importance of assessing patient perception about service quality as a predictor of patient willingness to recommend a hospital.\nAlthough this study reports on a relatively uncommon analysis of predicting patient willingness to recommend with perceived service quality, several limitations of the study require acknowledgment. The patient cohort was limited to only those patients who were English speakers, so this study sample is therefore not broadly representative of cancer patients in general. As a result, the generalizability of this study is limited. The data we used for this study were not primarily meant for research purposes. CTCA is a unique medical center. It specializes in treating only cancer patients, and it has an intense focus on patient-centered care. Compared to other centers, patients report very high levels of service quality at CTCA. Our study, which is hypothesis generating by nature, used a non-validated patient satisfaction questionnaire. However, it is reasonable to use a non-validated survey if the intent of the study is hypothesis generation rather than hypothesis testing. It might be argued that patients do not have the ability to judge a hospital's performance; however, patient perception is a key factor for hospital selection. This was the main goal of our study - to show the effects of patient perception about service quality on patient recommendation of a hospital. Finally, a response rate of 73.3% could potentially introduce a selection bias in our study. The baseline characteristics of patients who did not respond are not available for us to evaluate any systematic differences between responders and non-responders.\nMore and more health care consumers are using the web to research and shop for the best health care providers, especially for complex medical conditions. In addition, in our own experience, we hear about more and more patients traveling great distances to receive care, due in large part, to a strong recommendation of the provider from the patient's friend, associate, colleague, or family member. And as the asymmetry in information between providers and consumers decreases, we can and should expect to see more informed consumers shopping for the best available healthcare.\nThe strengths of our study include: a large sample size, the fact that we measured service quality as close to the time service was delivered as possible, and the fact that we used willingness to recommend (using the question and scale most commonly used in industry) as our dependent variable, which has been previously validated through research in many industries. To the best of our knowledge, this study is the first in the health care literature to report on the positive correlation between patient-reported service quality and patient willingness to recommend a provider in a large sample of cancer patients.\nDuring a time in which quality and value are becoming increasingly prominent themes in healthcare reform, we believe the patient's perspective of the key drivers of loyalty should be given greater national attention. In most of American medicine, we assess the relative importance of quality and patient satisfaction measures from expert panels and other traditional research methods developing institutionalized views of the attributes of health care that are most important. But what has been largely missing in the conversation is the patient's perspective (and perception) of the relative importance of key aspects of service quality in the health care delivery cycle. These are areas that require further research and at the time of this writing, our organization is seeking partners to conduct national population-based research on the key drivers of value in oncology, with service quality being an important dimension to be studied. As the health care legislation continues to be implemented, the entire health care system would benefit from a greater focus on the key drivers of value from the consumer's perspective - the patient, as well as their caregivers and families. These are important areas of research that will lend greater focus to where, when, and how we should apply our scarce resources to deliver the most valuable care to our patients - the ultimate consumer.\nNext steps in our research include linking data on service quality to patient outcomes. We are unaware of any literature linking service quality to data on patient quality of life, length of life, and overall satisfaction with health. Research is also underway at our center to explore the relationship between patient willingness to recommend and actual patient return (behavior) as well as how changes in patients' clinical condition affect their willingness to recommend a provider, controlling for all other known variables. With respect to population-based research, we do plan on conducting national research on the patient's perspective of value in oncologic care, data that has largely been missing from the health policy discussions.", "In this multi-center study, we demonstrate the predictive significance of perceived service quality as it relates to patient willingness to recommend an oncology service provider. We identified four key service quality drivers of patient loyalty: \"team helping you understand your medical condition\", \"staff genuinely caring for you as an individual\" \"whole person approach to patient care\" and \"CTCA medical oncologist\".", "The authors declare that they have no competing interests.", "CGL participated in concept, design, writing, statistical analysis and data interpretation. MR participated in concept, statistical analysis and data interpretation. DG participated in concept, statistical analysis, data interpretation and writing. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1472-6963/11/46/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Aged", "Biomarkers", "Blood Glucose", "Chi-Square Distribution", "China", "Coronary Angiography", "Coronary Stenosis", "Cross-Sectional Studies", "Diabetes Complications", "Diabetes Mellitus, Type 2", "Female", "Glycated Hemoglobin", "Humans", "Logistic Models", "Male", "Middle Aged", "Monitoring, Ambulatory", "Odds Ratio", "Predictive Value of Tests", "ROC Curve", "Risk Assessment", "Risk Factors", "Severity of Illness Index", "Time Factors" ]

[ "Background", "Study population", "Continuous glucose monitoring", "Coronary Angiography", "Biochemical Investigations", "Statistical Analysis", "Results", "Clinical Characteristics", "Glycemic variability, HbA1c and Fasting Plasma Glucose Levels", "Relationship between Gensini Score and Glycemic variability, Age and Biochemical Parameters", "Logistic Regression Analysis for Independent Determinants of CAD", "ROC Curve for MAGE and HbA1c in Predicting CAD in T2DM", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Cardiovascular diseases, including coronary artery disease (CAD), are the major causes of morbidity and cardiovascular death in patients with type 2 diabetes mellitus (T2DM) [1,2]. Diabetic patients usually present various factors contributing to the risk of cardiovascular diseases, which include hyperglycemia, fluctuation of blood glucose, central obesity, hyperlipidemia and hypertension and so on [2]. Glycemic disorders are important components of these risk factors.\nInterventional studies have established that cardiovascular complications are mainly or partly dependent on sustained chronic hyperglycemia [3,4]. This glycemic disorder can be estimated as a whole from the determination of hemoglobin A1c (HbA1c) level, which integrates both basal and postprandial hyperglycemia [5,6]. The incidence of cardiovascular complications has been identified as depending on HbA1c and on fasting and/or postprandial hyperglycemia, whether these parameters were investigated concomitantly or separately [7,8]. However, the glycemic disorders in T2DM are not solely limited to sustained chronic hyperglycemia but can be extended to the glycemic variability that includes both upward and downward acute glucose changes [9]. Some studies have showed that fluctuations of glucose seem to have more deleterious effects than sustained hyperglycemia in the development of diabetic complications as acute glucose swings activate the oxidative stress [10,11]. Recent studies have indicated that glycemic variability might play a role in the pathogenesis of atherosclerosis and may be an independent risk factor for cardiovascular complications in diabetic patients [10-13]. Therefore, in order to assess the risk of diabetes, it may be necessary to evaluate not only the mean level of glycemic control, but also the extent of glucose excursions. However, there have been no sufficient studies presented so far that specifically evaluated the relationship between glycemic variability and coronary artery disease in diabetic patients.\nIn this study, we examined the parameters of glucose profile using continuous glucose monitoring system (CGMS) in T2DM patients with CAD, and established a correlation between glycemic variability and the severity of coronary artery disease assessed by coronary angiogram, using the Gensini score.", "We studied consecutive T2DM patients with chest pain, who underwent coronary angiography at the Department of Cardiology, Beijing Anzhen Hospital of Capital Medical University. The inclusion criteria were: (i) referral to coronary angiography, due to chest pain; (ii) admission glucose < 16.7 mmol/L, and without diabetic ketosis or nonketotic hyperosmolar coma. Patients' anti-hyperglycemic therapy would be maintained as usual until CGMS monitoring completed. Otherwise, the patient would be excluded from the study. In addition, patients with acute coronary syndrome, infectious diseases, previous coronary artery bypass graft surgery and previous percutaneous coronary intervention were not included. Thus, 344 patients with complete data were included in the final analysis. Hypertension was defined as systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg or treatment with oral anti-hypertension drugs. Hyperlipidemia was diagnosed according to guideline of the National Cholesterol Education Program (ATP III). T2DM was diagnosed according to the American Diabetes Association criteria [14]. Renal insufficiency was defined as estimated glomerular filtration rate (eGFR) < 60 (ml/min/1.73 m2). eGFR value was calculated by MDRD equation [15]. The study was approved beforehand by the Ethics Committee of Beijing Anzhen Hospital of Capital Medical University and the procedures followed were in accordance with the institutional guidelines. The study complied with the declaration of Helsinki and informed consent was obtained from all patients.", "All patients were equipped with CGMS (Medtronic MiniMed, USA), and were monitored for 72 consecutive hours after admission. A CGMS sensor was inserted into the subcutaneous abdominal fat tissue, calibrated according to the standard Medtronic MiniMed operating guidelines. During CGMS monitoring, patients checked their blood glucose level with a self-monitoring of blood glucose (SMBG) device (Medisafe Mini, Terumo, Japan) at least 4 times per day. Then, they entered the SMBG data and time of each meal into the CGMS. After monitoring for 72 hours, the recorded data were downloaded into a personal computer for analysis of the glucose profile and glucose excursion parameters with MiniMed Solutions software. Analysis was limited to the data obtained from the intermediate 48 hours of recording to avoid bias due to insertion and removal of the CGMS or insufficient stability of the monitoring system. Since measurable range of glucose by CGMS was mechanically limited from 2.2 to 22.2 mmol/L, the case showing the data out of this range was excluded from the study. After downloading the recorded data, three indices of glycemic variability were analyzed from the intermediate 48 hours of data [16]: the mean amplitude of glycemic excursions (MAGE), the mean of daily differences (MODD) and postprandial glucose excursion (PPGE). The MAGE was calculated by measuring the arithmetic mean of the differences between consecutive peaks and nadirs, provided that the differences are greater than one standard deviation of the mean glucose value. The MODD was calculated as the mean of the absolute differences between glucose values at the same time of two consecutive days. The PPGE was obtained by calculating the post-breakfast increments in blood glucose.", "After CGMS monitoring finished, coronary artery angiography was performed by using standard Judkins techniques or a radial approach. During cardiac catheterization, nitroglycerine was administrated routinely in all cases suspected of having coronary spasm. Angiographic analysis was carried out by two experienced interventional cardiologists who were blinded to the study protocol. Angiography results were divided into CAD (≥50% obstruction in ≥1 coronary artery) group and non-CAD group. Gensini score assesses the severity of coronary artery disease: it grades narrowing of the lumen of the coronary artery and scores it as 1 for 1-25% narrowing, 2 for 26-50% narrowing, 4 for 51-75%, 8 for 76-90%, 16 for 91-99% and 32 for a completely occluded artery. This score is then multiplied by a factor according to the importance of the coronary artery. The multiplication factor for a left main stem (LMS) lesion is 5. It is 2.5 for proximal left anterior descending artery (LAD) and proximal circumflex artery (CX) lesions, 1.5 for a mid-LAD lesion, and 1 for distal LAD, mid/distal CX and right coronary artery lesions. The multiplication factor for any other branch is 0.5.", "Blood samples were collected after overnight fasting and stored at -70°C prior to analysis. Serum creatinine, fasting plasma glucose (FPG), high-sensitive C-reactive protein (hs-CRP), total cholesterol(TC), low-density lipoprotein-cholesterol (LDL-C), high-density lipoprotein-cholesterol (HDL-C) and triglyceride (TG) levels were measured by automatic biochemical analyzer (Hitachi 747, Tokyo, Japan). Serum concentration of hemoglobin A1c (HbA1c) was determined by high-performance liquid chromatographic method using automatic HbA1c analyzer (Tosoh HLC-723G7, Japan).", "All statistical analyses were performed by using SPSS for Windows 13.0 (SPSS Inc, Chicago, IL, USA). Data are presented as frequencies and percentages for categorical variables and mean ± SD for continuous variables, unless otherwise indicated. Differences between two groups were assessed by using the Chi-square and unpaired t-tests. Correlation between continuous variables was determined by Pearson correlation coefficients. Binary logistic regression analysis was performed to identify the relative risk of the glycemic variability for the presence of CAD, expressed as odds ratios (OR) with 95% confidence intervals (CI). The predictive values of MAGE and HbA1c for the presence of CAD were calculated by constructing receiver-operating characteristic (ROC) curves. To ascertain the independent contribution to severity of CAD, multivariate stepwise linear regression analysis was made with Gensini score as the dependent variable control for FPG, duration of diabetes, blood pressure, age, MAGE, MODD, PPGE, HbA1c, BMI, hs-CRP and TC. A value of p < 0.05 was considered statistically significant.", "[SUBTITLE] Clinical Characteristics [SUBSECTION] Among the 344 participants, 252 patients had angiographically-proven CAD and 92 had almost normal coronary arteries (Non-CAD group). Baseline characteristics of the two groups are shown in Table 1. The Gensini score ranged from 0 to 162 with a mean of 67.8 ± 34.1, as shown in Figure 1. Diabetic patients with CAD were older, and more were male and cigarette smokers compared with the patients without CAD. The CAD group had significantly higher levels of hs-CRP and lower levels of eGFR, but no significant differences in blood pressure, hyperlipidemia, BMI, TC, HDL-C, LDL-C and TG levels. In addition, most of the diabetic patients in the CAD group and Non-CAD group were receiving statin medication (69.4% vs. 61.9%, p > 0.05), and did not differ significantly with the use of insulin (40.9% vs. 35.1%, p > 0.05). There was no difference in treatment with other antidiabetic agents in diabetic patients with and without CAD.\nBaseline characteristics of diabetic patients with (CAD) and without (Non-CAD) coronary artery disease, as defined by coronary angiography\nAbbreviations: CAD, coronary artery disease; LDL-C, lower-density lipoprotein-cholesterol; HDL-C, high-density lipoprotein-cholesterol; eGFR, estimated glomerular filtration rate; MAGE, the mean amplitude of glycemic excursions; MODD, the mean of daily differences; PPGE, postprandial blood glucose excursions; BMI, body mass index; hs-CRP, highsensitive-C reactive protein.\nData are mean ± SD and number (%).\nDistribution of Gensini score among participants.\nAmong the 344 participants, 252 patients had angiographically-proven CAD and 92 had almost normal coronary arteries (Non-CAD group). Baseline characteristics of the two groups are shown in Table 1. The Gensini score ranged from 0 to 162 with a mean of 67.8 ± 34.1, as shown in Figure 1. Diabetic patients with CAD were older, and more were male and cigarette smokers compared with the patients without CAD. The CAD group had significantly higher levels of hs-CRP and lower levels of eGFR, but no significant differences in blood pressure, hyperlipidemia, BMI, TC, HDL-C, LDL-C and TG levels. In addition, most of the diabetic patients in the CAD group and Non-CAD group were receiving statin medication (69.4% vs. 61.9%, p > 0.05), and did not differ significantly with the use of insulin (40.9% vs. 35.1%, p > 0.05). There was no difference in treatment with other antidiabetic agents in diabetic patients with and without CAD.\nBaseline characteristics of diabetic patients with (CAD) and without (Non-CAD) coronary artery disease, as defined by coronary angiography\nAbbreviations: CAD, coronary artery disease; LDL-C, lower-density lipoprotein-cholesterol; HDL-C, high-density lipoprotein-cholesterol; eGFR, estimated glomerular filtration rate; MAGE, the mean amplitude of glycemic excursions; MODD, the mean of daily differences; PPGE, postprandial blood glucose excursions; BMI, body mass index; hs-CRP, highsensitive-C reactive protein.\nData are mean ± SD and number (%).\nDistribution of Gensini score among participants.\n[SUBTITLE] Glycemic variability, HbA1c and Fasting Plasma Glucose Levels [SUBSECTION] MAGE and PPGE were significantly higher in patients with CAD than in patients without CAD, but MODD did not significantly differ. CAD patients had longer duration of diabetes. There was also no significant difference in the HbA1c and FPG levels between two groups (Table 1).\nMAGE and PPGE were significantly higher in patients with CAD than in patients without CAD, but MODD did not significantly differ. CAD patients had longer duration of diabetes. There was also no significant difference in the HbA1c and FPG levels between two groups (Table 1).\n[SUBTITLE] Relationship between Gensini Score and Glycemic variability, Age and Biochemical Parameters [SUBSECTION] Pearson correlation analysis showed that Gensini score was closely related to MAGE(r = 0.277, p < 0.001), age (r = 0.288, p < 0.001), PPGE (r = 0.167, p = 0.002) and the level of HbA1c (r = 0.136, p = 0.011) (Figure 2), and correlated significantly with the levels of hs-CRP (r = 0.132, p = 0.014) and TC (r = 0.108, p = 0.045), but not with MODD, eGFR, duration of diabetes, blood pressure and other factors. In the final multivariate linear regression analysis model that explained 19.1% (adjusted multiple R2 = 0.191) of the variation in Gensini score, the independent determinants were age (p < 0.001), MAGE (p < 0.001), hs-CRP (p = 0.002) and HbA1c (p = 0.022) (Table 2).\nSimple linear correlation of Gensini score and age, MAGE, PPGE and hemoglobin A1c in patients with type 2 diabetes.\nMultivariate analysis of determinants of Gensini score\nAbbreviations: MAGE, the mean amplitude of glycemic excursions; hs-CRP, highsensitive-C reactive protein; HbA1c, Hemoglobin A1c.\nPearson correlation analysis showed that Gensini score was closely related to MAGE(r = 0.277, p < 0.001), age (r = 0.288, p < 0.001), PPGE (r = 0.167, p = 0.002) and the level of HbA1c (r = 0.136, p = 0.011) (Figure 2), and correlated significantly with the levels of hs-CRP (r = 0.132, p = 0.014) and TC (r = 0.108, p = 0.045), but not with MODD, eGFR, duration of diabetes, blood pressure and other factors. In the final multivariate linear regression analysis model that explained 19.1% (adjusted multiple R2 = 0.191) of the variation in Gensini score, the independent determinants were age (p < 0.001), MAGE (p < 0.001), hs-CRP (p = 0.002) and HbA1c (p = 0.022) (Table 2).\nSimple linear correlation of Gensini score and age, MAGE, PPGE and hemoglobin A1c in patients with type 2 diabetes.\nMultivariate analysis of determinants of Gensini score\nAbbreviations: MAGE, the mean amplitude of glycemic excursions; hs-CRP, highsensitive-C reactive protein; HbA1c, Hemoglobin A1c.\n[SUBTITLE] Logistic Regression Analysis for Independent Determinants of CAD [SUBSECTION] Cigarette smoking, male, older age (≥65 years), high MAGE level (≥3.4 mmol/L) and high hs-CRP level (> 5 mg/L) were found to be independent risk factors for the presence of CAD in T2DM patients, having OR 2.492 (p = 0.005), 1.936 (p = 0.036), 2.516 (p = 0.002), 2.612 (p = 0.002), and 2.851 (p = 0.009), respectively (Figure 3).\nMultivariate analysis for independent determinants of coronary artery disease (CAD). Smoking, male, older age, MAGE and hs-CRP were independent risk factors for CAD.\nCigarette smoking, male, older age (≥65 years), high MAGE level (≥3.4 mmol/L) and high hs-CRP level (> 5 mg/L) were found to be independent risk factors for the presence of CAD in T2DM patients, having OR 2.492 (p = 0.005), 1.936 (p = 0.036), 2.516 (p = 0.002), 2.612 (p = 0.002), and 2.851 (p = 0.009), respectively (Figure 3).\nMultivariate analysis for independent determinants of coronary artery disease (CAD). Smoking, male, older age, MAGE and hs-CRP were independent risk factors for CAD.\n[SUBTITLE] ROC Curve for MAGE and HbA1c in Predicting CAD in T2DM [SUBSECTION] The area under the ROC curve for MAGE (0.618, 95%CI 0.555 to 0.680, p = 0.001) was superior to that for HbA1c (0.554, 95%CI 0.487 to 0.620, p = 0.129) (Figure 4).\nReceiver-operating characteristic (ROC) curve for MAGE and hemoglobin A1c (HbA1c) in predicting coronary artery disease (CAD) in patients with type 2 diabetes (T2DM). Area under the receiver-operating characteristic curve: MAGE 0.618 (95% CI 0.555, 0.680), p = 0.001; HbA1c 0.554 (95% CI 0.487, 0.620), p = 0.129. MAGE, but not HbA1c, displayed significant value in predicting CAD in patients with T2DM.\nThe area under the ROC curve for MAGE (0.618, 95%CI 0.555 to 0.680, p = 0.001) was superior to that for HbA1c (0.554, 95%CI 0.487 to 0.620, p = 0.129) (Figure 4).\nReceiver-operating characteristic (ROC) curve for MAGE and hemoglobin A1c (HbA1c) in predicting coronary artery disease (CAD) in patients with type 2 diabetes (T2DM). Area under the receiver-operating characteristic curve: MAGE 0.618 (95% CI 0.555, 0.680), p = 0.001; HbA1c 0.554 (95% CI 0.487, 0.620), p = 0.129. MAGE, but not HbA1c, displayed significant value in predicting CAD in patients with T2DM.", "Among the 344 participants, 252 patients had angiographically-proven CAD and 92 had almost normal coronary arteries (Non-CAD group). Baseline characteristics of the two groups are shown in Table 1. The Gensini score ranged from 0 to 162 with a mean of 67.8 ± 34.1, as shown in Figure 1. Diabetic patients with CAD were older, and more were male and cigarette smokers compared with the patients without CAD. The CAD group had significantly higher levels of hs-CRP and lower levels of eGFR, but no significant differences in blood pressure, hyperlipidemia, BMI, TC, HDL-C, LDL-C and TG levels. In addition, most of the diabetic patients in the CAD group and Non-CAD group were receiving statin medication (69.4% vs. 61.9%, p > 0.05), and did not differ significantly with the use of insulin (40.9% vs. 35.1%, p > 0.05). There was no difference in treatment with other antidiabetic agents in diabetic patients with and without CAD.\nBaseline characteristics of diabetic patients with (CAD) and without (Non-CAD) coronary artery disease, as defined by coronary angiography\nAbbreviations: CAD, coronary artery disease; LDL-C, lower-density lipoprotein-cholesterol; HDL-C, high-density lipoprotein-cholesterol; eGFR, estimated glomerular filtration rate; MAGE, the mean amplitude of glycemic excursions; MODD, the mean of daily differences; PPGE, postprandial blood glucose excursions; BMI, body mass index; hs-CRP, highsensitive-C reactive protein.\nData are mean ± SD and number (%).\nDistribution of Gensini score among participants.", "MAGE and PPGE were significantly higher in patients with CAD than in patients without CAD, but MODD did not significantly differ. CAD patients had longer duration of diabetes. There was also no significant difference in the HbA1c and FPG levels between two groups (Table 1).", "Pearson correlation analysis showed that Gensini score was closely related to MAGE(r = 0.277, p < 0.001), age (r = 0.288, p < 0.001), PPGE (r = 0.167, p = 0.002) and the level of HbA1c (r = 0.136, p = 0.011) (Figure 2), and correlated significantly with the levels of hs-CRP (r = 0.132, p = 0.014) and TC (r = 0.108, p = 0.045), but not with MODD, eGFR, duration of diabetes, blood pressure and other factors. In the final multivariate linear regression analysis model that explained 19.1% (adjusted multiple R2 = 0.191) of the variation in Gensini score, the independent determinants were age (p < 0.001), MAGE (p < 0.001), hs-CRP (p = 0.002) and HbA1c (p = 0.022) (Table 2).\nSimple linear correlation of Gensini score and age, MAGE, PPGE and hemoglobin A1c in patients with type 2 diabetes.\nMultivariate analysis of determinants of Gensini score\nAbbreviations: MAGE, the mean amplitude of glycemic excursions; hs-CRP, highsensitive-C reactive protein; HbA1c, Hemoglobin A1c.", "Cigarette smoking, male, older age (≥65 years), high MAGE level (≥3.4 mmol/L) and high hs-CRP level (> 5 mg/L) were found to be independent risk factors for the presence of CAD in T2DM patients, having OR 2.492 (p = 0.005), 1.936 (p = 0.036), 2.516 (p = 0.002), 2.612 (p = 0.002), and 2.851 (p = 0.009), respectively (Figure 3).\nMultivariate analysis for independent determinants of coronary artery disease (CAD). Smoking, male, older age, MAGE and hs-CRP were independent risk factors for CAD.", "The area under the ROC curve for MAGE (0.618, 95%CI 0.555 to 0.680, p = 0.001) was superior to that for HbA1c (0.554, 95%CI 0.487 to 0.620, p = 0.129) (Figure 4).\nReceiver-operating characteristic (ROC) curve for MAGE and hemoglobin A1c (HbA1c) in predicting coronary artery disease (CAD) in patients with type 2 diabetes (T2DM). Area under the receiver-operating characteristic curve: MAGE 0.618 (95% CI 0.555, 0.680), p = 0.001; HbA1c 0.554 (95% CI 0.487, 0.620), p = 0.129. MAGE, but not HbA1c, displayed significant value in predicting CAD in patients with T2DM.", "Blood glucose level continuously fluctuates within a certain range in the human body. In diabetic patients, glycemic disorders include both sustained chronic hyperglycemia and glucose excursions. Patients with similar mean glucose or HbA1c levels can have markedly different glycemic excursions. Our study shows levels of FPG and HbA1c were not higher in CAD group patients than in the controls (p > 0.05). According to the traditional view, the two groups should have similar blood glucose control. However, we found that MAGE and PPGE levels in T2DM patients with CAD were higher than in T2DM patients without CAD. This result indicates that glucose excursions can not be neglected as an important aspect of glucose disorders and suggests it may be associated with coronary artery disease in T2DM patients.\nAlthough the influence of glucose control, as assessed primarily by HbA1c levels, on the development of diabetes complications has been proven in numerous large-scale epidemiological studies [3,7,17], there is still an extensive debate about glucose variability as a risk factor for complications independent of HbA1c in diabetes[18,19]. A retrospective analysis from the Diabetes Control Complications Trial (DCCT) showed that HbA1c, as well as blood glucose fluctuation, seem to be associated with the microvascular complications of type1 diabetes [20]. A single study in T2DM demonstrated that fasting plasma glucose variability is a predictor of the onset of retinopathy in patients [21]. However, the controversial results were concluded from Kilpatrick et al. They independently performed analysis of the data of the DCCT showing that blood glucose variability was not related to the development or progression of either retinopathy or nephropathy in type1 diabetes [22]. Thirteen years later, the DCCT statisticians themselves corrected their previous findings and refuted the relation of glucose fluctuation and microvascular complications [23]. However, the DCCT was not designed to determine the impact of glycemic variability on the risk of the vascular complications of diabetes, and in fact the investigators required two attempts to report an acceptable statistical model. Since this was a post hoc analysis, at best it is hypothesis-generating.\nIn fact, more and more evidences have been found that glycemic variability may be an important parameter used to resolve potential clinical problems in diabetic patients [24]. Some researchers identified several important associations between postchallenge glucose excursions and known risk factors for atherosclerosis, and suggested that postchallenge glucose excursion is independently related to carotid intima-media thickness and may contribute to the development of atherosclerosis in individuals with T2DM independent of other risk factors [12,13]. The Verona Diabetes study reported that long-term variability of fasting glucose is an independent predictor of mortality in T2DM patients [25]. Some studies concluded that glucose variability was a significant predictor of mortality in critically ill patients independently from mean glucose level and severity of illness [26-28]. In the present study, MAGE≥3.4 mmol/L was found to be an independent predictor for the presence of CAD (OR 2.612; p = 0.002). Furthermore, the ROC plot also demonstrated that the MAGE level was a significant predictor for the presence of CAD (p = 0.001), whereas HbA1c was not (p = 0.129). These results indicate that intraday glucose excursion might contribute to generation of atherosclerosis even more specifically than sustained chronic hyperglycemia.\nAt present, the identified role of glucose variability in pathogenesis of atherosclerosis is not clear. Hyperglycemia is thought to induce oxidative stress and interfere with normal endothelial function by overproduction of reactive oxygen species, which results in atherosclerosis through several molecular mechanisms. In addition, glucose variability might contribute to these processes as well. Recent in vitro studies indicate that glucose fluctuations can activate nuclear factor-κB and protein kinase C (PKC) pathway, leading to a greater expression of the adhesion molecules and excess formation of advanced glycation end-products than stable high glucose [29,30]. Some studies reported that intermittent hyperglycemia induced a higher degree of apoptosis in endothelial cells than chronic hyperglycemia [10,31]. Quagliaro et al. showed that the apoptosis of endothelial cells exposed to intermittent high glucose may be related to a reactive oxygen species (ROS) overproduction, through PKC-dependent activation of nicotinamide adenine dinucleotide phosphate (NADPH)-oxidase [30]. NADPH oxidase is a major cause of atherosclerosis. That suggests blood glucose excursion plays an important role in the occurrence and acceleration of atherosclerosis in diabetes. However, the relationship between glycemic variability and oxidative stress was not consistently reproduced in human studies [11,32]. These discrepant findings might be explained by differences in duration and frequency of the periods with alternating glycemia as well as differences in methods used for oxidative stress quantification. Furthermore, severe glycemic disorders may adversely affect circadian variation of cardiac autonomic modulation and circadian blood pressure variability which are associated with mortality and morbidity of cardiovascular disease [33,34]. Takei et al. find acute hyperglycemia may induce sympathetic dysfunction through multiple mechanisms, including hyperglycemia, hyperinsulinemia, and increased oxidative stress, whereas coronary microvascular dysfunction is closely related to sympathetic activation [35].\nIn our study, Pearson correlation analysis showed that Gensini score correlated positively with the level of MAGE(r = 0.277, p < 0.001), PPGE (r = 0.167, p = 0.002) and HbA1c(r = 0.136, p = 0.011), indicating a pronounced proatherogenic effect of worse glycemic disorders, which is consistent with previous findings. Furthermore, multivariate regression analysis revealed that MAGE and HbA1c were independent risk factors for the severity of CAD. These results indicate that intraday glucose excursion is an important contributing factor in the severity of coronary artery disease, which is independent of the average level of blood glucose. Our results further indicate that current practice of relying mainly on HbA1c within a target range is inadequate for timely therapeutic adjustments and reducing the risk of coronary artery disease. To accomplish this goal, the CGMS appears to be a very useful tool in primary care and its use should be expanded for effective management of T2DM.\nA few limitations of this study should be recognized. Firstly, the sample size was relatively small in this study, so that some subgroup comparisons may have lacked power to detect significant differences for selected variables. Secondly, although we had maintained the patients' anti-hyperglycemic therapy as usual and avoided glucose infusion during CGMS monitoring period, some factors, such as different diets, physical and emotional stress etc., which may affect levels of admission glucose fluctuations couldn't be all prevented. Thirdly, lack of microvascular complications data, we didn't include those risk factors in study. Fourthly, due to the fact that the present study was a cross-sectional design, our results only show the association between glycemic variability and prevalent CAD rather than incident CAD. All subjects in the present study were scheduled for coronary angiography for their suffering chest pain in cardiovascular department of Beijing Anzhen hospital. Therefore, it is likely that enrolled T2DM patients had greater risk for CAD than ordinary T2DM patients.", "The present study shows that the intraday glycemic variability is associated with the presence and severity of CAD in patients with T2DM. Recent accumulated evidence suggested that blood glucose excursions may play an important role in the occurrence and development of atherosclerosis in diabetes. There should be an emphasis on efforts to control all aspects of blood glucose disorders including HbA1c, FPG, postprandial glucose, as well as glucose fluctuations. Further well-designed studies are warranted to examine if reduction of glucose excursions has a substantial impact on CAD development in patients with T2DM.", "The authors declare that they have no competing interests.", "All authors listed on the manuscript participated in the design and coordination of the study and made substantial contribution to the intellectual content of the project to be included as authors. They also read and approved the final manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study population", "Continuous glucose monitoring", "Coronary Angiography", "Biochemical Investigations", "Statistical Analysis", "Results", "Clinical Characteristics", "Glycemic variability, HbA1c and Fasting Plasma Glucose Levels", "Relationship between Gensini Score and Glycemic variability, Age and Biochemical Parameters", "Logistic Regression Analysis for Independent Determinants of CAD", "ROC Curve for MAGE and HbA1c in Predicting CAD in T2DM", "Discussion", "Conclusions", "Competing interests", "Authors' contributions" ]

[ "Cardiovascular diseases, including coronary artery disease (CAD), are the major causes of morbidity and cardiovascular death in patients with type 2 diabetes mellitus (T2DM) [1,2]. Diabetic patients usually present various factors contributing to the risk of cardiovascular diseases, which include hyperglycemia, fluctuation of blood glucose, central obesity, hyperlipidemia and hypertension and so on [2]. Glycemic disorders are important components of these risk factors.\nInterventional studies have established that cardiovascular complications are mainly or partly dependent on sustained chronic hyperglycemia [3,4]. This glycemic disorder can be estimated as a whole from the determination of hemoglobin A1c (HbA1c) level, which integrates both basal and postprandial hyperglycemia [5,6]. The incidence of cardiovascular complications has been identified as depending on HbA1c and on fasting and/or postprandial hyperglycemia, whether these parameters were investigated concomitantly or separately [7,8]. However, the glycemic disorders in T2DM are not solely limited to sustained chronic hyperglycemia but can be extended to the glycemic variability that includes both upward and downward acute glucose changes [9]. Some studies have showed that fluctuations of glucose seem to have more deleterious effects than sustained hyperglycemia in the development of diabetic complications as acute glucose swings activate the oxidative stress [10,11]. Recent studies have indicated that glycemic variability might play a role in the pathogenesis of atherosclerosis and may be an independent risk factor for cardiovascular complications in diabetic patients [10-13]. Therefore, in order to assess the risk of diabetes, it may be necessary to evaluate not only the mean level of glycemic control, but also the extent of glucose excursions. However, there have been no sufficient studies presented so far that specifically evaluated the relationship between glycemic variability and coronary artery disease in diabetic patients.\nIn this study, we examined the parameters of glucose profile using continuous glucose monitoring system (CGMS) in T2DM patients with CAD, and established a correlation between glycemic variability and the severity of coronary artery disease assessed by coronary angiogram, using the Gensini score.", "[SUBTITLE] Study population [SUBSECTION] We studied consecutive T2DM patients with chest pain, who underwent coronary angiography at the Department of Cardiology, Beijing Anzhen Hospital of Capital Medical University. The inclusion criteria were: (i) referral to coronary angiography, due to chest pain; (ii) admission glucose < 16.7 mmol/L, and without diabetic ketosis or nonketotic hyperosmolar coma. Patients' anti-hyperglycemic therapy would be maintained as usual until CGMS monitoring completed. Otherwise, the patient would be excluded from the study. In addition, patients with acute coronary syndrome, infectious diseases, previous coronary artery bypass graft surgery and previous percutaneous coronary intervention were not included. Thus, 344 patients with complete data were included in the final analysis. Hypertension was defined as systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg or treatment with oral anti-hypertension drugs. Hyperlipidemia was diagnosed according to guideline of the National Cholesterol Education Program (ATP III). T2DM was diagnosed according to the American Diabetes Association criteria [14]. Renal insufficiency was defined as estimated glomerular filtration rate (eGFR) < 60 (ml/min/1.73 m2). eGFR value was calculated by MDRD equation [15]. The study was approved beforehand by the Ethics Committee of Beijing Anzhen Hospital of Capital Medical University and the procedures followed were in accordance with the institutional guidelines. The study complied with the declaration of Helsinki and informed consent was obtained from all patients.\nWe studied consecutive T2DM patients with chest pain, who underwent coronary angiography at the Department of Cardiology, Beijing Anzhen Hospital of Capital Medical University. The inclusion criteria were: (i) referral to coronary angiography, due to chest pain; (ii) admission glucose < 16.7 mmol/L, and without diabetic ketosis or nonketotic hyperosmolar coma. Patients' anti-hyperglycemic therapy would be maintained as usual until CGMS monitoring completed. Otherwise, the patient would be excluded from the study. In addition, patients with acute coronary syndrome, infectious diseases, previous coronary artery bypass graft surgery and previous percutaneous coronary intervention were not included. Thus, 344 patients with complete data were included in the final analysis. Hypertension was defined as systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg or treatment with oral anti-hypertension drugs. Hyperlipidemia was diagnosed according to guideline of the National Cholesterol Education Program (ATP III). T2DM was diagnosed according to the American Diabetes Association criteria [14]. Renal insufficiency was defined as estimated glomerular filtration rate (eGFR) < 60 (ml/min/1.73 m2). eGFR value was calculated by MDRD equation [15]. The study was approved beforehand by the Ethics Committee of Beijing Anzhen Hospital of Capital Medical University and the procedures followed were in accordance with the institutional guidelines. The study complied with the declaration of Helsinki and informed consent was obtained from all patients.\n[SUBTITLE] Continuous glucose monitoring [SUBSECTION] All patients were equipped with CGMS (Medtronic MiniMed, USA), and were monitored for 72 consecutive hours after admission. A CGMS sensor was inserted into the subcutaneous abdominal fat tissue, calibrated according to the standard Medtronic MiniMed operating guidelines. During CGMS monitoring, patients checked their blood glucose level with a self-monitoring of blood glucose (SMBG) device (Medisafe Mini, Terumo, Japan) at least 4 times per day. Then, they entered the SMBG data and time of each meal into the CGMS. After monitoring for 72 hours, the recorded data were downloaded into a personal computer for analysis of the glucose profile and glucose excursion parameters with MiniMed Solutions software. Analysis was limited to the data obtained from the intermediate 48 hours of recording to avoid bias due to insertion and removal of the CGMS or insufficient stability of the monitoring system. Since measurable range of glucose by CGMS was mechanically limited from 2.2 to 22.2 mmol/L, the case showing the data out of this range was excluded from the study. After downloading the recorded data, three indices of glycemic variability were analyzed from the intermediate 48 hours of data [16]: the mean amplitude of glycemic excursions (MAGE), the mean of daily differences (MODD) and postprandial glucose excursion (PPGE). The MAGE was calculated by measuring the arithmetic mean of the differences between consecutive peaks and nadirs, provided that the differences are greater than one standard deviation of the mean glucose value. The MODD was calculated as the mean of the absolute differences between glucose values at the same time of two consecutive days. The PPGE was obtained by calculating the post-breakfast increments in blood glucose.\nAll patients were equipped with CGMS (Medtronic MiniMed, USA), and were monitored for 72 consecutive hours after admission. A CGMS sensor was inserted into the subcutaneous abdominal fat tissue, calibrated according to the standard Medtronic MiniMed operating guidelines. During CGMS monitoring, patients checked their blood glucose level with a self-monitoring of blood glucose (SMBG) device (Medisafe Mini, Terumo, Japan) at least 4 times per day. Then, they entered the SMBG data and time of each meal into the CGMS. After monitoring for 72 hours, the recorded data were downloaded into a personal computer for analysis of the glucose profile and glucose excursion parameters with MiniMed Solutions software. Analysis was limited to the data obtained from the intermediate 48 hours of recording to avoid bias due to insertion and removal of the CGMS or insufficient stability of the monitoring system. Since measurable range of glucose by CGMS was mechanically limited from 2.2 to 22.2 mmol/L, the case showing the data out of this range was excluded from the study. After downloading the recorded data, three indices of glycemic variability were analyzed from the intermediate 48 hours of data [16]: the mean amplitude of glycemic excursions (MAGE), the mean of daily differences (MODD) and postprandial glucose excursion (PPGE). The MAGE was calculated by measuring the arithmetic mean of the differences between consecutive peaks and nadirs, provided that the differences are greater than one standard deviation of the mean glucose value. The MODD was calculated as the mean of the absolute differences between glucose values at the same time of two consecutive days. The PPGE was obtained by calculating the post-breakfast increments in blood glucose.\n[SUBTITLE] Coronary Angiography [SUBSECTION] After CGMS monitoring finished, coronary artery angiography was performed by using standard Judkins techniques or a radial approach. During cardiac catheterization, nitroglycerine was administrated routinely in all cases suspected of having coronary spasm. Angiographic analysis was carried out by two experienced interventional cardiologists who were blinded to the study protocol. Angiography results were divided into CAD (≥50% obstruction in ≥1 coronary artery) group and non-CAD group. Gensini score assesses the severity of coronary artery disease: it grades narrowing of the lumen of the coronary artery and scores it as 1 for 1-25% narrowing, 2 for 26-50% narrowing, 4 for 51-75%, 8 for 76-90%, 16 for 91-99% and 32 for a completely occluded artery. This score is then multiplied by a factor according to the importance of the coronary artery. The multiplication factor for a left main stem (LMS) lesion is 5. It is 2.5 for proximal left anterior descending artery (LAD) and proximal circumflex artery (CX) lesions, 1.5 for a mid-LAD lesion, and 1 for distal LAD, mid/distal CX and right coronary artery lesions. The multiplication factor for any other branch is 0.5.\nAfter CGMS monitoring finished, coronary artery angiography was performed by using standard Judkins techniques or a radial approach. During cardiac catheterization, nitroglycerine was administrated routinely in all cases suspected of having coronary spasm. Angiographic analysis was carried out by two experienced interventional cardiologists who were blinded to the study protocol. Angiography results were divided into CAD (≥50% obstruction in ≥1 coronary artery) group and non-CAD group. Gensini score assesses the severity of coronary artery disease: it grades narrowing of the lumen of the coronary artery and scores it as 1 for 1-25% narrowing, 2 for 26-50% narrowing, 4 for 51-75%, 8 for 76-90%, 16 for 91-99% and 32 for a completely occluded artery. This score is then multiplied by a factor according to the importance of the coronary artery. The multiplication factor for a left main stem (LMS) lesion is 5. It is 2.5 for proximal left anterior descending artery (LAD) and proximal circumflex artery (CX) lesions, 1.5 for a mid-LAD lesion, and 1 for distal LAD, mid/distal CX and right coronary artery lesions. The multiplication factor for any other branch is 0.5.\n[SUBTITLE] Biochemical Investigations [SUBSECTION] Blood samples were collected after overnight fasting and stored at -70°C prior to analysis. Serum creatinine, fasting plasma glucose (FPG), high-sensitive C-reactive protein (hs-CRP), total cholesterol(TC), low-density lipoprotein-cholesterol (LDL-C), high-density lipoprotein-cholesterol (HDL-C) and triglyceride (TG) levels were measured by automatic biochemical analyzer (Hitachi 747, Tokyo, Japan). Serum concentration of hemoglobin A1c (HbA1c) was determined by high-performance liquid chromatographic method using automatic HbA1c analyzer (Tosoh HLC-723G7, Japan).\nBlood samples were collected after overnight fasting and stored at -70°C prior to analysis. Serum creatinine, fasting plasma glucose (FPG), high-sensitive C-reactive protein (hs-CRP), total cholesterol(TC), low-density lipoprotein-cholesterol (LDL-C), high-density lipoprotein-cholesterol (HDL-C) and triglyceride (TG) levels were measured by automatic biochemical analyzer (Hitachi 747, Tokyo, Japan). Serum concentration of hemoglobin A1c (HbA1c) was determined by high-performance liquid chromatographic method using automatic HbA1c analyzer (Tosoh HLC-723G7, Japan).\n[SUBTITLE] Statistical Analysis [SUBSECTION] All statistical analyses were performed by using SPSS for Windows 13.0 (SPSS Inc, Chicago, IL, USA). Data are presented as frequencies and percentages for categorical variables and mean ± SD for continuous variables, unless otherwise indicated. Differences between two groups were assessed by using the Chi-square and unpaired t-tests. Correlation between continuous variables was determined by Pearson correlation coefficients. Binary logistic regression analysis was performed to identify the relative risk of the glycemic variability for the presence of CAD, expressed as odds ratios (OR) with 95% confidence intervals (CI). The predictive values of MAGE and HbA1c for the presence of CAD were calculated by constructing receiver-operating characteristic (ROC) curves. To ascertain the independent contribution to severity of CAD, multivariate stepwise linear regression analysis was made with Gensini score as the dependent variable control for FPG, duration of diabetes, blood pressure, age, MAGE, MODD, PPGE, HbA1c, BMI, hs-CRP and TC. A value of p < 0.05 was considered statistically significant.\nAll statistical analyses were performed by using SPSS for Windows 13.0 (SPSS Inc, Chicago, IL, USA). Data are presented as frequencies and percentages for categorical variables and mean ± SD for continuous variables, unless otherwise indicated. Differences between two groups were assessed by using the Chi-square and unpaired t-tests. Correlation between continuous variables was determined by Pearson correlation coefficients. Binary logistic regression analysis was performed to identify the relative risk of the glycemic variability for the presence of CAD, expressed as odds ratios (OR) with 95% confidence intervals (CI). The predictive values of MAGE and HbA1c for the presence of CAD were calculated by constructing receiver-operating characteristic (ROC) curves. To ascertain the independent contribution to severity of CAD, multivariate stepwise linear regression analysis was made with Gensini score as the dependent variable control for FPG, duration of diabetes, blood pressure, age, MAGE, MODD, PPGE, HbA1c, BMI, hs-CRP and TC. A value of p < 0.05 was considered statistically significant.", "We studied consecutive T2DM patients with chest pain, who underwent coronary angiography at the Department of Cardiology, Beijing Anzhen Hospital of Capital Medical University. The inclusion criteria were: (i) referral to coronary angiography, due to chest pain; (ii) admission glucose < 16.7 mmol/L, and without diabetic ketosis or nonketotic hyperosmolar coma. Patients' anti-hyperglycemic therapy would be maintained as usual until CGMS monitoring completed. Otherwise, the patient would be excluded from the study. In addition, patients with acute coronary syndrome, infectious diseases, previous coronary artery bypass graft surgery and previous percutaneous coronary intervention were not included. Thus, 344 patients with complete data were included in the final analysis. Hypertension was defined as systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg or treatment with oral anti-hypertension drugs. Hyperlipidemia was diagnosed according to guideline of the National Cholesterol Education Program (ATP III). T2DM was diagnosed according to the American Diabetes Association criteria [14]. Renal insufficiency was defined as estimated glomerular filtration rate (eGFR) < 60 (ml/min/1.73 m2). eGFR value was calculated by MDRD equation [15]. The study was approved beforehand by the Ethics Committee of Beijing Anzhen Hospital of Capital Medical University and the procedures followed were in accordance with the institutional guidelines. The study complied with the declaration of Helsinki and informed consent was obtained from all patients.", "All patients were equipped with CGMS (Medtronic MiniMed, USA), and were monitored for 72 consecutive hours after admission. A CGMS sensor was inserted into the subcutaneous abdominal fat tissue, calibrated according to the standard Medtronic MiniMed operating guidelines. During CGMS monitoring, patients checked their blood glucose level with a self-monitoring of blood glucose (SMBG) device (Medisafe Mini, Terumo, Japan) at least 4 times per day. Then, they entered the SMBG data and time of each meal into the CGMS. After monitoring for 72 hours, the recorded data were downloaded into a personal computer for analysis of the glucose profile and glucose excursion parameters with MiniMed Solutions software. Analysis was limited to the data obtained from the intermediate 48 hours of recording to avoid bias due to insertion and removal of the CGMS or insufficient stability of the monitoring system. Since measurable range of glucose by CGMS was mechanically limited from 2.2 to 22.2 mmol/L, the case showing the data out of this range was excluded from the study. After downloading the recorded data, three indices of glycemic variability were analyzed from the intermediate 48 hours of data [16]: the mean amplitude of glycemic excursions (MAGE), the mean of daily differences (MODD) and postprandial glucose excursion (PPGE). The MAGE was calculated by measuring the arithmetic mean of the differences between consecutive peaks and nadirs, provided that the differences are greater than one standard deviation of the mean glucose value. The MODD was calculated as the mean of the absolute differences between glucose values at the same time of two consecutive days. The PPGE was obtained by calculating the post-breakfast increments in blood glucose.", "After CGMS monitoring finished, coronary artery angiography was performed by using standard Judkins techniques or a radial approach. During cardiac catheterization, nitroglycerine was administrated routinely in all cases suspected of having coronary spasm. Angiographic analysis was carried out by two experienced interventional cardiologists who were blinded to the study protocol. Angiography results were divided into CAD (≥50% obstruction in ≥1 coronary artery) group and non-CAD group. Gensini score assesses the severity of coronary artery disease: it grades narrowing of the lumen of the coronary artery and scores it as 1 for 1-25% narrowing, 2 for 26-50% narrowing, 4 for 51-75%, 8 for 76-90%, 16 for 91-99% and 32 for a completely occluded artery. This score is then multiplied by a factor according to the importance of the coronary artery. The multiplication factor for a left main stem (LMS) lesion is 5. It is 2.5 for proximal left anterior descending artery (LAD) and proximal circumflex artery (CX) lesions, 1.5 for a mid-LAD lesion, and 1 for distal LAD, mid/distal CX and right coronary artery lesions. The multiplication factor for any other branch is 0.5.", "Blood samples were collected after overnight fasting and stored at -70°C prior to analysis. Serum creatinine, fasting plasma glucose (FPG), high-sensitive C-reactive protein (hs-CRP), total cholesterol(TC), low-density lipoprotein-cholesterol (LDL-C), high-density lipoprotein-cholesterol (HDL-C) and triglyceride (TG) levels were measured by automatic biochemical analyzer (Hitachi 747, Tokyo, Japan). Serum concentration of hemoglobin A1c (HbA1c) was determined by high-performance liquid chromatographic method using automatic HbA1c analyzer (Tosoh HLC-723G7, Japan).", "All statistical analyses were performed by using SPSS for Windows 13.0 (SPSS Inc, Chicago, IL, USA). Data are presented as frequencies and percentages for categorical variables and mean ± SD for continuous variables, unless otherwise indicated. Differences between two groups were assessed by using the Chi-square and unpaired t-tests. Correlation between continuous variables was determined by Pearson correlation coefficients. Binary logistic regression analysis was performed to identify the relative risk of the glycemic variability for the presence of CAD, expressed as odds ratios (OR) with 95% confidence intervals (CI). The predictive values of MAGE and HbA1c for the presence of CAD were calculated by constructing receiver-operating characteristic (ROC) curves. To ascertain the independent contribution to severity of CAD, multivariate stepwise linear regression analysis was made with Gensini score as the dependent variable control for FPG, duration of diabetes, blood pressure, age, MAGE, MODD, PPGE, HbA1c, BMI, hs-CRP and TC. A value of p < 0.05 was considered statistically significant.", "[SUBTITLE] Clinical Characteristics [SUBSECTION] Among the 344 participants, 252 patients had angiographically-proven CAD and 92 had almost normal coronary arteries (Non-CAD group). Baseline characteristics of the two groups are shown in Table 1. The Gensini score ranged from 0 to 162 with a mean of 67.8 ± 34.1, as shown in Figure 1. Diabetic patients with CAD were older, and more were male and cigarette smokers compared with the patients without CAD. The CAD group had significantly higher levels of hs-CRP and lower levels of eGFR, but no significant differences in blood pressure, hyperlipidemia, BMI, TC, HDL-C, LDL-C and TG levels. In addition, most of the diabetic patients in the CAD group and Non-CAD group were receiving statin medication (69.4% vs. 61.9%, p > 0.05), and did not differ significantly with the use of insulin (40.9% vs. 35.1%, p > 0.05). There was no difference in treatment with other antidiabetic agents in diabetic patients with and without CAD.\nBaseline characteristics of diabetic patients with (CAD) and without (Non-CAD) coronary artery disease, as defined by coronary angiography\nAbbreviations: CAD, coronary artery disease; LDL-C, lower-density lipoprotein-cholesterol; HDL-C, high-density lipoprotein-cholesterol; eGFR, estimated glomerular filtration rate; MAGE, the mean amplitude of glycemic excursions; MODD, the mean of daily differences; PPGE, postprandial blood glucose excursions; BMI, body mass index; hs-CRP, highsensitive-C reactive protein.\nData are mean ± SD and number (%).\nDistribution of Gensini score among participants.\nAmong the 344 participants, 252 patients had angiographically-proven CAD and 92 had almost normal coronary arteries (Non-CAD group). Baseline characteristics of the two groups are shown in Table 1. The Gensini score ranged from 0 to 162 with a mean of 67.8 ± 34.1, as shown in Figure 1. Diabetic patients with CAD were older, and more were male and cigarette smokers compared with the patients without CAD. The CAD group had significantly higher levels of hs-CRP and lower levels of eGFR, but no significant differences in blood pressure, hyperlipidemia, BMI, TC, HDL-C, LDL-C and TG levels. In addition, most of the diabetic patients in the CAD group and Non-CAD group were receiving statin medication (69.4% vs. 61.9%, p > 0.05), and did not differ significantly with the use of insulin (40.9% vs. 35.1%, p > 0.05). There was no difference in treatment with other antidiabetic agents in diabetic patients with and without CAD.\nBaseline characteristics of diabetic patients with (CAD) and without (Non-CAD) coronary artery disease, as defined by coronary angiography\nAbbreviations: CAD, coronary artery disease; LDL-C, lower-density lipoprotein-cholesterol; HDL-C, high-density lipoprotein-cholesterol; eGFR, estimated glomerular filtration rate; MAGE, the mean amplitude of glycemic excursions; MODD, the mean of daily differences; PPGE, postprandial blood glucose excursions; BMI, body mass index; hs-CRP, highsensitive-C reactive protein.\nData are mean ± SD and number (%).\nDistribution of Gensini score among participants.\n[SUBTITLE] Glycemic variability, HbA1c and Fasting Plasma Glucose Levels [SUBSECTION] MAGE and PPGE were significantly higher in patients with CAD than in patients without CAD, but MODD did not significantly differ. CAD patients had longer duration of diabetes. There was also no significant difference in the HbA1c and FPG levels between two groups (Table 1).\nMAGE and PPGE were significantly higher in patients with CAD than in patients without CAD, but MODD did not significantly differ. CAD patients had longer duration of diabetes. There was also no significant difference in the HbA1c and FPG levels between two groups (Table 1).\n[SUBTITLE] Relationship between Gensini Score and Glycemic variability, Age and Biochemical Parameters [SUBSECTION] Pearson correlation analysis showed that Gensini score was closely related to MAGE(r = 0.277, p < 0.001), age (r = 0.288, p < 0.001), PPGE (r = 0.167, p = 0.002) and the level of HbA1c (r = 0.136, p = 0.011) (Figure 2), and correlated significantly with the levels of hs-CRP (r = 0.132, p = 0.014) and TC (r = 0.108, p = 0.045), but not with MODD, eGFR, duration of diabetes, blood pressure and other factors. In the final multivariate linear regression analysis model that explained 19.1% (adjusted multiple R2 = 0.191) of the variation in Gensini score, the independent determinants were age (p < 0.001), MAGE (p < 0.001), hs-CRP (p = 0.002) and HbA1c (p = 0.022) (Table 2).\nSimple linear correlation of Gensini score and age, MAGE, PPGE and hemoglobin A1c in patients with type 2 diabetes.\nMultivariate analysis of determinants of Gensini score\nAbbreviations: MAGE, the mean amplitude of glycemic excursions; hs-CRP, highsensitive-C reactive protein; HbA1c, Hemoglobin A1c.\nPearson correlation analysis showed that Gensini score was closely related to MAGE(r = 0.277, p < 0.001), age (r = 0.288, p < 0.001), PPGE (r = 0.167, p = 0.002) and the level of HbA1c (r = 0.136, p = 0.011) (Figure 2), and correlated significantly with the levels of hs-CRP (r = 0.132, p = 0.014) and TC (r = 0.108, p = 0.045), but not with MODD, eGFR, duration of diabetes, blood pressure and other factors. In the final multivariate linear regression analysis model that explained 19.1% (adjusted multiple R2 = 0.191) of the variation in Gensini score, the independent determinants were age (p < 0.001), MAGE (p < 0.001), hs-CRP (p = 0.002) and HbA1c (p = 0.022) (Table 2).\nSimple linear correlation of Gensini score and age, MAGE, PPGE and hemoglobin A1c in patients with type 2 diabetes.\nMultivariate analysis of determinants of Gensini score\nAbbreviations: MAGE, the mean amplitude of glycemic excursions; hs-CRP, highsensitive-C reactive protein; HbA1c, Hemoglobin A1c.\n[SUBTITLE] Logistic Regression Analysis for Independent Determinants of CAD [SUBSECTION] Cigarette smoking, male, older age (≥65 years), high MAGE level (≥3.4 mmol/L) and high hs-CRP level (> 5 mg/L) were found to be independent risk factors for the presence of CAD in T2DM patients, having OR 2.492 (p = 0.005), 1.936 (p = 0.036), 2.516 (p = 0.002), 2.612 (p = 0.002), and 2.851 (p = 0.009), respectively (Figure 3).\nMultivariate analysis for independent determinants of coronary artery disease (CAD). Smoking, male, older age, MAGE and hs-CRP were independent risk factors for CAD.\nCigarette smoking, male, older age (≥65 years), high MAGE level (≥3.4 mmol/L) and high hs-CRP level (> 5 mg/L) were found to be independent risk factors for the presence of CAD in T2DM patients, having OR 2.492 (p = 0.005), 1.936 (p = 0.036), 2.516 (p = 0.002), 2.612 (p = 0.002), and 2.851 (p = 0.009), respectively (Figure 3).\nMultivariate analysis for independent determinants of coronary artery disease (CAD). Smoking, male, older age, MAGE and hs-CRP were independent risk factors for CAD.\n[SUBTITLE] ROC Curve for MAGE and HbA1c in Predicting CAD in T2DM [SUBSECTION] The area under the ROC curve for MAGE (0.618, 95%CI 0.555 to 0.680, p = 0.001) was superior to that for HbA1c (0.554, 95%CI 0.487 to 0.620, p = 0.129) (Figure 4).\nReceiver-operating characteristic (ROC) curve for MAGE and hemoglobin A1c (HbA1c) in predicting coronary artery disease (CAD) in patients with type 2 diabetes (T2DM). Area under the receiver-operating characteristic curve: MAGE 0.618 (95% CI 0.555, 0.680), p = 0.001; HbA1c 0.554 (95% CI 0.487, 0.620), p = 0.129. MAGE, but not HbA1c, displayed significant value in predicting CAD in patients with T2DM.\nThe area under the ROC curve for MAGE (0.618, 95%CI 0.555 to 0.680, p = 0.001) was superior to that for HbA1c (0.554, 95%CI 0.487 to 0.620, p = 0.129) (Figure 4).\nReceiver-operating characteristic (ROC) curve for MAGE and hemoglobin A1c (HbA1c) in predicting coronary artery disease (CAD) in patients with type 2 diabetes (T2DM). Area under the receiver-operating characteristic curve: MAGE 0.618 (95% CI 0.555, 0.680), p = 0.001; HbA1c 0.554 (95% CI 0.487, 0.620), p = 0.129. MAGE, but not HbA1c, displayed significant value in predicting CAD in patients with T2DM.", "Among the 344 participants, 252 patients had angiographically-proven CAD and 92 had almost normal coronary arteries (Non-CAD group). Baseline characteristics of the two groups are shown in Table 1. The Gensini score ranged from 0 to 162 with a mean of 67.8 ± 34.1, as shown in Figure 1. Diabetic patients with CAD were older, and more were male and cigarette smokers compared with the patients without CAD. The CAD group had significantly higher levels of hs-CRP and lower levels of eGFR, but no significant differences in blood pressure, hyperlipidemia, BMI, TC, HDL-C, LDL-C and TG levels. In addition, most of the diabetic patients in the CAD group and Non-CAD group were receiving statin medication (69.4% vs. 61.9%, p > 0.05), and did not differ significantly with the use of insulin (40.9% vs. 35.1%, p > 0.05). There was no difference in treatment with other antidiabetic agents in diabetic patients with and without CAD.\nBaseline characteristics of diabetic patients with (CAD) and without (Non-CAD) coronary artery disease, as defined by coronary angiography\nAbbreviations: CAD, coronary artery disease; LDL-C, lower-density lipoprotein-cholesterol; HDL-C, high-density lipoprotein-cholesterol; eGFR, estimated glomerular filtration rate; MAGE, the mean amplitude of glycemic excursions; MODD, the mean of daily differences; PPGE, postprandial blood glucose excursions; BMI, body mass index; hs-CRP, highsensitive-C reactive protein.\nData are mean ± SD and number (%).\nDistribution of Gensini score among participants.", "MAGE and PPGE were significantly higher in patients with CAD than in patients without CAD, but MODD did not significantly differ. CAD patients had longer duration of diabetes. There was also no significant difference in the HbA1c and FPG levels between two groups (Table 1).", "Pearson correlation analysis showed that Gensini score was closely related to MAGE(r = 0.277, p < 0.001), age (r = 0.288, p < 0.001), PPGE (r = 0.167, p = 0.002) and the level of HbA1c (r = 0.136, p = 0.011) (Figure 2), and correlated significantly with the levels of hs-CRP (r = 0.132, p = 0.014) and TC (r = 0.108, p = 0.045), but not with MODD, eGFR, duration of diabetes, blood pressure and other factors. In the final multivariate linear regression analysis model that explained 19.1% (adjusted multiple R2 = 0.191) of the variation in Gensini score, the independent determinants were age (p < 0.001), MAGE (p < 0.001), hs-CRP (p = 0.002) and HbA1c (p = 0.022) (Table 2).\nSimple linear correlation of Gensini score and age, MAGE, PPGE and hemoglobin A1c in patients with type 2 diabetes.\nMultivariate analysis of determinants of Gensini score\nAbbreviations: MAGE, the mean amplitude of glycemic excursions; hs-CRP, highsensitive-C reactive protein; HbA1c, Hemoglobin A1c.", "Cigarette smoking, male, older age (≥65 years), high MAGE level (≥3.4 mmol/L) and high hs-CRP level (> 5 mg/L) were found to be independent risk factors for the presence of CAD in T2DM patients, having OR 2.492 (p = 0.005), 1.936 (p = 0.036), 2.516 (p = 0.002), 2.612 (p = 0.002), and 2.851 (p = 0.009), respectively (Figure 3).\nMultivariate analysis for independent determinants of coronary artery disease (CAD). Smoking, male, older age, MAGE and hs-CRP were independent risk factors for CAD.", "The area under the ROC curve for MAGE (0.618, 95%CI 0.555 to 0.680, p = 0.001) was superior to that for HbA1c (0.554, 95%CI 0.487 to 0.620, p = 0.129) (Figure 4).\nReceiver-operating characteristic (ROC) curve for MAGE and hemoglobin A1c (HbA1c) in predicting coronary artery disease (CAD) in patients with type 2 diabetes (T2DM). Area under the receiver-operating characteristic curve: MAGE 0.618 (95% CI 0.555, 0.680), p = 0.001; HbA1c 0.554 (95% CI 0.487, 0.620), p = 0.129. MAGE, but not HbA1c, displayed significant value in predicting CAD in patients with T2DM.", "Blood glucose level continuously fluctuates within a certain range in the human body. In diabetic patients, glycemic disorders include both sustained chronic hyperglycemia and glucose excursions. Patients with similar mean glucose or HbA1c levels can have markedly different glycemic excursions. Our study shows levels of FPG and HbA1c were not higher in CAD group patients than in the controls (p > 0.05). According to the traditional view, the two groups should have similar blood glucose control. However, we found that MAGE and PPGE levels in T2DM patients with CAD were higher than in T2DM patients without CAD. This result indicates that glucose excursions can not be neglected as an important aspect of glucose disorders and suggests it may be associated with coronary artery disease in T2DM patients.\nAlthough the influence of glucose control, as assessed primarily by HbA1c levels, on the development of diabetes complications has been proven in numerous large-scale epidemiological studies [3,7,17], there is still an extensive debate about glucose variability as a risk factor for complications independent of HbA1c in diabetes[18,19]. A retrospective analysis from the Diabetes Control Complications Trial (DCCT) showed that HbA1c, as well as blood glucose fluctuation, seem to be associated with the microvascular complications of type1 diabetes [20]. A single study in T2DM demonstrated that fasting plasma glucose variability is a predictor of the onset of retinopathy in patients [21]. However, the controversial results were concluded from Kilpatrick et al. They independently performed analysis of the data of the DCCT showing that blood glucose variability was not related to the development or progression of either retinopathy or nephropathy in type1 diabetes [22]. Thirteen years later, the DCCT statisticians themselves corrected their previous findings and refuted the relation of glucose fluctuation and microvascular complications [23]. However, the DCCT was not designed to determine the impact of glycemic variability on the risk of the vascular complications of diabetes, and in fact the investigators required two attempts to report an acceptable statistical model. Since this was a post hoc analysis, at best it is hypothesis-generating.\nIn fact, more and more evidences have been found that glycemic variability may be an important parameter used to resolve potential clinical problems in diabetic patients [24]. Some researchers identified several important associations between postchallenge glucose excursions and known risk factors for atherosclerosis, and suggested that postchallenge glucose excursion is independently related to carotid intima-media thickness and may contribute to the development of atherosclerosis in individuals with T2DM independent of other risk factors [12,13]. The Verona Diabetes study reported that long-term variability of fasting glucose is an independent predictor of mortality in T2DM patients [25]. Some studies concluded that glucose variability was a significant predictor of mortality in critically ill patients independently from mean glucose level and severity of illness [26-28]. In the present study, MAGE≥3.4 mmol/L was found to be an independent predictor for the presence of CAD (OR 2.612; p = 0.002). Furthermore, the ROC plot also demonstrated that the MAGE level was a significant predictor for the presence of CAD (p = 0.001), whereas HbA1c was not (p = 0.129). These results indicate that intraday glucose excursion might contribute to generation of atherosclerosis even more specifically than sustained chronic hyperglycemia.\nAt present, the identified role of glucose variability in pathogenesis of atherosclerosis is not clear. Hyperglycemia is thought to induce oxidative stress and interfere with normal endothelial function by overproduction of reactive oxygen species, which results in atherosclerosis through several molecular mechanisms. In addition, glucose variability might contribute to these processes as well. Recent in vitro studies indicate that glucose fluctuations can activate nuclear factor-κB and protein kinase C (PKC) pathway, leading to a greater expression of the adhesion molecules and excess formation of advanced glycation end-products than stable high glucose [29,30]. Some studies reported that intermittent hyperglycemia induced a higher degree of apoptosis in endothelial cells than chronic hyperglycemia [10,31]. Quagliaro et al. showed that the apoptosis of endothelial cells exposed to intermittent high glucose may be related to a reactive oxygen species (ROS) overproduction, through PKC-dependent activation of nicotinamide adenine dinucleotide phosphate (NADPH)-oxidase [30]. NADPH oxidase is a major cause of atherosclerosis. That suggests blood glucose excursion plays an important role in the occurrence and acceleration of atherosclerosis in diabetes. However, the relationship between glycemic variability and oxidative stress was not consistently reproduced in human studies [11,32]. These discrepant findings might be explained by differences in duration and frequency of the periods with alternating glycemia as well as differences in methods used for oxidative stress quantification. Furthermore, severe glycemic disorders may adversely affect circadian variation of cardiac autonomic modulation and circadian blood pressure variability which are associated with mortality and morbidity of cardiovascular disease [33,34]. Takei et al. find acute hyperglycemia may induce sympathetic dysfunction through multiple mechanisms, including hyperglycemia, hyperinsulinemia, and increased oxidative stress, whereas coronary microvascular dysfunction is closely related to sympathetic activation [35].\nIn our study, Pearson correlation analysis showed that Gensini score correlated positively with the level of MAGE(r = 0.277, p < 0.001), PPGE (r = 0.167, p = 0.002) and HbA1c(r = 0.136, p = 0.011), indicating a pronounced proatherogenic effect of worse glycemic disorders, which is consistent with previous findings. Furthermore, multivariate regression analysis revealed that MAGE and HbA1c were independent risk factors for the severity of CAD. These results indicate that intraday glucose excursion is an important contributing factor in the severity of coronary artery disease, which is independent of the average level of blood glucose. Our results further indicate that current practice of relying mainly on HbA1c within a target range is inadequate for timely therapeutic adjustments and reducing the risk of coronary artery disease. To accomplish this goal, the CGMS appears to be a very useful tool in primary care and its use should be expanded for effective management of T2DM.\nA few limitations of this study should be recognized. Firstly, the sample size was relatively small in this study, so that some subgroup comparisons may have lacked power to detect significant differences for selected variables. Secondly, although we had maintained the patients' anti-hyperglycemic therapy as usual and avoided glucose infusion during CGMS monitoring period, some factors, such as different diets, physical and emotional stress etc., which may affect levels of admission glucose fluctuations couldn't be all prevented. Thirdly, lack of microvascular complications data, we didn't include those risk factors in study. Fourthly, due to the fact that the present study was a cross-sectional design, our results only show the association between glycemic variability and prevalent CAD rather than incident CAD. All subjects in the present study were scheduled for coronary angiography for their suffering chest pain in cardiovascular department of Beijing Anzhen hospital. Therefore, it is likely that enrolled T2DM patients had greater risk for CAD than ordinary T2DM patients.", "The present study shows that the intraday glycemic variability is associated with the presence and severity of CAD in patients with T2DM. Recent accumulated evidence suggested that blood glucose excursions may play an important role in the occurrence and development of atherosclerosis in diabetes. There should be an emphasis on efforts to control all aspects of blood glucose disorders including HbA1c, FPG, postprandial glucose, as well as glucose fluctuations. Further well-designed studies are warranted to examine if reduction of glucose excursions has a substantial impact on CAD development in patients with T2DM.", "The authors declare that they have no competing interests.", "All authors listed on the manuscript participated in the design and coordination of the study and made substantial contribution to the intellectual content of the project to be included as authors. They also read and approved the final manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adolescent", "Adult", "Aged", "Analysis of Variance", "Cardiomyopathy, Restrictive", "China", "Contrast Media", "Diagnosis, Differential", "Female", "Gadolinium DTPA", "Heart Atria", "Humans", "Magnetic Resonance Imaging, Cine", "Male", "Middle Aged", "Pericarditis, Constrictive", "Predictive Value of Tests", "ROC Curve", "Sensitivity and Specificity", "Young Adult" ]

[ "Background", "Study population", "CMR protocol", "CMR Analysis", "Statistical analysis", "Results", "Patient characteristics", "CMR characteristics", "Sensitivity and specificity", "Discussion", "Study limitations", "Conclusions", "List of abbreviations used", "Competing interests", "Authors' contributions" ]

[ "Clinical and hemodynamic features are often similar in constrictive pericarditis (CP) and restrictive cardiomyopathy (RCM), but differentiation of these 2 conditions is crucial because CP requires surgical treatment and is usually curable, while RCM, short of cardiac transplantation, is treatable only by medical means and often responds unsatisfactorily [1-3]. At present, clinical evaluation, measurement of pericardial thickness, analysis of septal motion, quantitative assessment of systolic and diastolic myocardial function, invasive pressure measurement, and endomyocardial biopsy have been useful in this differential diagnosis, but no one diagnostic method can be relied upon to make the distinction by itself [4-7].\nCardiovascular magnetic resonance (CMR) provides high-resolution imaging of the pericardium and associated structures in any imaging plane. It fuses excellent anatomic detail and tissue characterization with accurate evaluation of cardiac function and assessment of the haemodynamic consequences of pericardial constraint on cardiac filling [8-12]. Compared with echocardiography and computed tomography, CMR with late gadolinium enhancement (LGE) is the only method that can depict the presence of myocardial fibrosis, which may well facilitate diagnosis of RCM resulting from infiltrative myocardial disease and have important prognostic implications [13-17].\nThe aim of the present study was to describe the clinical utility of CMR for distinguishing CP form RCM. We sought to determine the diagnostic accuracy of the relative atrial volume ratio (RAR) for the detection of CP and its possible use as a screening tool to aid in the differentiation between CP and RCM.", "The study population consists of 45 consecutive patients who were referred for CMR, including 23 surgically documented CP patients and 22 RCM patients. All patients had been underwent previously systematic clinical evaluation, including history and examination, electrocardiography, chest radiography, and echocardiography. In each CP case, surgical confirmation was obtained by the presence of an obliterated pericardial space, an adhesive pericarditis with bulging of the heart out of the pericardial incision at pericardiectomy and pathological confirmation. The diagnosis of RCM was confirmed by pathological specimens or based on impaired cardiac filling (i.e., increased filling pressures and no echo-Doppler evidence of respiratory-dependent ventricular coupling) in combination with pericardial thickness < 2 mm. All patients were referred to rule out any other cardiovascular diseases such as coronary artery disease, hypertension, valvular and congenital heart disease, and other cardiomyopathy. As a control group, 25 normal subjects without a history of cardiovascular symptoms or risk factors were also included in this study.\nThe study was approved by the institutional ethics committee, and all subjects gave written informed consent.", "Cardiac magnetic resonance imaging was performed in all patients by using a 1.5-T unit (Magnetom Avanto; Siemens Medical Solutions, Erlangen, Germany) with a high-performance gradient system (maximum gradient amplitude 45 mT/m; maximum slew rate 200-μs rise time), a 12-element-body phased-array coil system and electrocardiographic triggering. The CMR examinations began with the acquisition of survey images in three orthogonal planes (transverse, coronal, and sagittal) to localize the heart within the chest. Next, we studied the heart by performing a dark blood half-Fourier acquisition single-shot turbo spin echo (HASTE: repetition time [TR]/echo time [TE] = 700/26 ms, slice thickness = 6 mm, flip angle = 160°, field of view [FOV] = 350 mm) and turbo spin-echo (TSE) T1- (TR/TE = 700/20 ms, slice thickness = 6 mm, flip angle = 180°, matrix = 256 × 156; FOV = 350 mm) and T2-weighted (TR/TE = 800/77 ms, slice thickness = 6 mm, matrix = 256 ×190, FOV = 350 mm, flip angle = 180°) CMR sequences. Left ventricular (LV) short-axis, horizontal long-axis, and vertical long-axis views were used to evaluate cardiac function on cine CMR sequences. Cine CMR were acquired using generalized autocalibrating partially parallel acquisitions (GRAPPA: TR/TE = 45.9/1.07 ms, slice thickness = 6 mm, matrix = 109 × 192, FOV = 350 mm, flip angle = 80°) or time-adaptive sensitivity encoding (TSENSE: TR/TE = 41.7/1.39 ms, slice thickness = 6 mm, matrix = 109 × 192, FOV = 350 mm, flip angle = 70°) with true fast imaging with steady-state precession (TrueFISP) cine sequences. 15 to 20 minutes after injection of 0.2 mmol/kg of gadolinium diethylenetriamine pentaacetic acid (Magnevist, Schering, Berlin, Germany), the images of LGE were obtained in standard short axis covering the entire ventricle, and in long axis views to detect areas of LGE using a phase-sensitive inversion recovery (PISR) spoiled gradient echo sequence (TR/TE = 8.7/3.4 ms, slice thickness = 6 mm, imaging matrix = 256 × 256, FOV = 350 mm, flip angle = 15°).", "All CMR images were transferred to workstation (Siemens medical systems) for analysis. Qualitative assessments were performed independently by three readers. If there was a discrepancy, majority opinion was used. Quantitative measurements were performed independently by two readers. All observers were blinded to the diagnosis. For morphological evaluation of the pericardium, TSE and HASTE images were employed (Figure 1) to assess the maximum pericardial thickness. Septal motion was evaluated on a short-axis cine function view 1 cm beneath the atrioventricular valves on a visual basis and described as normal and the early diastolic septal bounce. The biventricular volumes and ejective fraction were obtained using Argus analytical software (version VE36A). Endocardial margins of the LV and right ventricular (RV) were semi-automatically contoured on end-diastolic and end-systolic images. End-diastolic and end-systolic frames were defined on the basis of the respective image frames demonstrating the largest and smallest cavity size. For the left atrial (LA) volume, the biplane area-length method was used. For the right atrial (RA) volume, the monoplane area-length formula was used. Atrial diastole was determined by selecting the last frame in ventricular systole before mitral valve opening. The measurements were made according to published methods [18]. The long-axis length of the LA and RA was defined by measuring the distance from the center of the mitral annulus to the posterior atrial wall. The atrial endocardial area was manually traced to exclude the atrial appendages and pulmonary or caval veins. Body weight and body height were measured and the body surface area was calculated. Subsequently, division with body surface area indexed all CMR variables apart from the ejection fraction. The RAR was defined as the LA volume divided by RA volume. LGE was considered present only if myocardial enhancement was confirmed on both short-axis and matching long-axis locations using a signal intensity threshold of > 2 standard deviation (SD) above a remote reference region in the same image.\nDiffuse thickened pericardium. HASTE (A), T1- (B) and T2-weighted (C) TSE images showed diffuse pericardial thickening (white arrows) which is most pronounced over the RV and RA and moderate right-sided pleural effusion (*).", "All values were given as mean ± SD or counts (percentage). Categorical values were compared by chi-square test or Fisher exact test as appropriate. Comparisons of normally distributed continuous variables between the different groups were performed by using one-way analysis of variance (ANOVA) with Fisher's least significant difference (LSD) posttest. The Kruskal-Wallis and Mann-Whitney U test were used to compare non-normally distributed continuous variables. The inter-observer agreement for the RAR was tested using intra-class correlation coefficient and limits of agreement using Bland Altman plots. Receiver operating characteristic (ROC) curve analysis was used to test the ability of different variables in differentiating CP from RCM. The area under the ROC curve (AUC) for each variable was calculated and compared. The statistical comparison of the ROC curves was performed using MedCalc (11.4.4, MedCalc, Belgium). Other statistical analyses were performed using SPSS for Windows (version 16.0; SPSS Inc., Chicago, IL, USA). P values < 0.05 were considered significant.", "[SUBTITLE] Patient characteristics [SUBSECTION] There were 18 men and 5 women in CP group, with a mean age of 43.0 ± 20.2 years (range 15 to 77 years). The aetiology of CP was unknown in 10 patients, whereas 4 patients had previous cardiac surgery, 7 had tuberculous infection, and 2 had history of an inflammatory infection. The RCM group included 22 patients (12 men, 10 women) with a mean age of 47.5 ± 18.5 years (range 14 to 72 years). Five RCM patients underwent heart transplantation, and surgical pathology specimens showed the presence of cardiac amyloidosis in 3 patients and nonspecific findings in 2 patients. Endomyocardial biopsy was performed in other 10 RCM patients. Cardiac amyloidosis was confirmed in 2 patients, mixed connective tissue disease in 1 patient, and idiopathic forms in 7 patients. There were no significant differences between RCM patients and the two other groups in terms of gender, age or BSA distribution. The demographic and clinical characteristics in each group and their comparison are shown in Table 1.\nBaseline and clinical characteristics\nThere were 18 men and 5 women in CP group, with a mean age of 43.0 ± 20.2 years (range 15 to 77 years). The aetiology of CP was unknown in 10 patients, whereas 4 patients had previous cardiac surgery, 7 had tuberculous infection, and 2 had history of an inflammatory infection. The RCM group included 22 patients (12 men, 10 women) with a mean age of 47.5 ± 18.5 years (range 14 to 72 years). Five RCM patients underwent heart transplantation, and surgical pathology specimens showed the presence of cardiac amyloidosis in 3 patients and nonspecific findings in 2 patients. Endomyocardial biopsy was performed in other 10 RCM patients. Cardiac amyloidosis was confirmed in 2 patients, mixed connective tissue disease in 1 patient, and idiopathic forms in 7 patients. There were no significant differences between RCM patients and the two other groups in terms of gender, age or BSA distribution. The demographic and clinical characteristics in each group and their comparison are shown in Table 1.\nBaseline and clinical characteristics\n[SUBTITLE] CMR characteristics [SUBSECTION] The maximal pericardial thickness in CP patients (6.9 ± 2.6 mm, range 4-12 mm) was significantly larger than in normal subjects (1.5 ± 0.4 mm, range 0.9-2.7 mm, p < 0.001) and RCM patients (2.0 ± 0.7 mm, range 1.0-3.4 mm, p < 0.001). There were no differences among the three groups in biventricular end-systolic volume index. There were no differences between CP and RCM patients in biventricular end-diastolic volume index, stroke volume index, and EF, although these values were significantly smaller in RCM and CP patients compared with normal subjects. The RA volume index (RAI) in RCM patients (90.5 ± 35.3 mL/m2) was significantly larger than in CP patients (71.4 ± 15.7 mL/m2, p = 0.006) and normal subjects (38.1 ± 9.0 mL/m2, p < 0.001). Although the LA volume index (LAI) yielded significantly increased values in RCM (96.0 ± 37.0 mL/m2) and CP patients (105.6 ± 25.1 mL/m2) compared with normal subjects (39.5 ± 9.5 mL/m2, p < 0.001 for all), no statistical significance were reached between CP and RCM patients (p = 0.200) (Figure 2). The RAR in CP patients (1.50 ± 0.29) was significantly larger than in normal subjects (1.06 ± 0.20, p < 0.001) and RCM patients (1.12 ± 0.33, p < 0.001). There were no differences between RCM patients and normal subjects in the RAR (p = 0.452) (Figure 3). The intra-class correlation coefficient [equal to 0.917 with 95% confidence interval (CI), 0.84-0.92] showed that there was an excellent inter-observer agreement on the measurement of the RAR. The Bland-Altman plots showing the limits of agreement are shown in Figure 4.\nLeft and right atrial volume indices. Comparisons of LAI and RAI between CP, RCM patients and normal subjects.\nError bar of the RAR. Data are presented as means (squares) and 95% confidence interval (whiskers). **p < 0.001.\nBland Altman plot of the RAR values. Bland-Altman analysis showed excellent inter-observer agreement for the RAR.\nThe analysis of septal movement during early diastole revealed a septal bounce in 22 CP patients. One CP patient, all RCM patients and normal subjects had a normal septal configuration during diastole. Comparative results of CMR parameters are independently shown in Table 2. LGE was present in 7 of 22 RCM patients (31.8%) and absence in all CP patients and normal subjects. Several different patterns of LGE were present in RCM patients. In 4 of 5 patients with histopathologically proven cardiac amyloidosis, LGE was distributed over the entire subendocardial circumference, extending in various degrees into the neighboring myocardium and the papillary muscles. In remaining 1 patient with cardiac amyloidosis, diffuse transmural LGE was found in the LV wall (Figure 5). Two idiopathic RCM patients had focal areas of LGE in various locations of the LV myocardium.\nCardiac magnetic resonance imaging findings\nLAI = left atrial volume index; RAI = right atrial volume index; RAR = relative atrial volume ratio; EDV = end-diastolic volume; ESV = end-systolic volume; SV = stroke volume; EF = ejective fraction; Comparison with RCM *P < 0.05; Comparison with CP #P < 0.05\nLGE of cardiac amyloidosis. CMR demonstrated global LV and RV wall hypertrophy (a and c) (white arrows), diffuse transmural LGE (b and d) (black arrows), and mild left-sided pleural effusion (*) in a 39-year-old male patient with cardiac amyloidosis who underwent cardiac transplantation and was proven by surgical pathology specimen.\nThe maximal pericardial thickness in CP patients (6.9 ± 2.6 mm, range 4-12 mm) was significantly larger than in normal subjects (1.5 ± 0.4 mm, range 0.9-2.7 mm, p < 0.001) and RCM patients (2.0 ± 0.7 mm, range 1.0-3.4 mm, p < 0.001). There were no differences among the three groups in biventricular end-systolic volume index. There were no differences between CP and RCM patients in biventricular end-diastolic volume index, stroke volume index, and EF, although these values were significantly smaller in RCM and CP patients compared with normal subjects. The RA volume index (RAI) in RCM patients (90.5 ± 35.3 mL/m2) was significantly larger than in CP patients (71.4 ± 15.7 mL/m2, p = 0.006) and normal subjects (38.1 ± 9.0 mL/m2, p < 0.001). Although the LA volume index (LAI) yielded significantly increased values in RCM (96.0 ± 37.0 mL/m2) and CP patients (105.6 ± 25.1 mL/m2) compared with normal subjects (39.5 ± 9.5 mL/m2, p < 0.001 for all), no statistical significance were reached between CP and RCM patients (p = 0.200) (Figure 2). The RAR in CP patients (1.50 ± 0.29) was significantly larger than in normal subjects (1.06 ± 0.20, p < 0.001) and RCM patients (1.12 ± 0.33, p < 0.001). There were no differences between RCM patients and normal subjects in the RAR (p = 0.452) (Figure 3). The intra-class correlation coefficient [equal to 0.917 with 95% confidence interval (CI), 0.84-0.92] showed that there was an excellent inter-observer agreement on the measurement of the RAR. The Bland-Altman plots showing the limits of agreement are shown in Figure 4.\nLeft and right atrial volume indices. Comparisons of LAI and RAI between CP, RCM patients and normal subjects.\nError bar of the RAR. Data are presented as means (squares) and 95% confidence interval (whiskers). **p < 0.001.\nBland Altman plot of the RAR values. Bland-Altman analysis showed excellent inter-observer agreement for the RAR.\nThe analysis of septal movement during early diastole revealed a septal bounce in 22 CP patients. One CP patient, all RCM patients and normal subjects had a normal septal configuration during diastole. Comparative results of CMR parameters are independently shown in Table 2. LGE was present in 7 of 22 RCM patients (31.8%) and absence in all CP patients and normal subjects. Several different patterns of LGE were present in RCM patients. In 4 of 5 patients with histopathologically proven cardiac amyloidosis, LGE was distributed over the entire subendocardial circumference, extending in various degrees into the neighboring myocardium and the papillary muscles. In remaining 1 patient with cardiac amyloidosis, diffuse transmural LGE was found in the LV wall (Figure 5). Two idiopathic RCM patients had focal areas of LGE in various locations of the LV myocardium.\nCardiac magnetic resonance imaging findings\nLAI = left atrial volume index; RAI = right atrial volume index; RAR = relative atrial volume ratio; EDV = end-diastolic volume; ESV = end-systolic volume; SV = stroke volume; EF = ejective fraction; Comparison with RCM *P < 0.05; Comparison with CP #P < 0.05\nLGE of cardiac amyloidosis. CMR demonstrated global LV and RV wall hypertrophy (a and c) (white arrows), diffuse transmural LGE (b and d) (black arrows), and mild left-sided pleural effusion (*) in a 39-year-old male patient with cardiac amyloidosis who underwent cardiac transplantation and was proven by surgical pathology specimen.\n[SUBTITLE] Sensitivity and specificity [SUBSECTION] Receiver operating characteristic curve analysis was used for comparison of discriminative capacity between different indices (Table 3 and Figure 6). The RAR [AUC 0.83 (95% CI 0.69-0.93)] had higher accuracy than LAI [AUC 0.64 (95% CI 0.48-0.78), p = 0.0216] and RAI [AUC 0.63 (95% CI 0.47-0.77), p = 0.0378] for predicting CP. There were no differences between the LAI and RAI (p = 0.950) in identifying CP. At a cut-off value of 1.32 for the RAR, the sensitivity was 82.6%, and the specificity was 86.4% in the detection of CP. Cut-offs of LAI > 83.4 mL/m2 and RAI ≤ 81.1 mL/m2 had sensitivities of 82.6% and 54.6%, respectively, and specificities of 87.0% and 50.0%, respectively.\nDiagnostic accuracy of individual parameters to distinguish between CP and RCM\nAUC = area under receiver operating characteristics curve; 95% CI = 95% confidence interval; other abbreviations as in Table 2; *Statistically significant, p < 0.001\nReceiver operating characteristics curves. ROC curves described the performance of different variables in differentiating CP from RCM.\nReceiver operating characteristic curve analysis was used for comparison of discriminative capacity between different indices (Table 3 and Figure 6). The RAR [AUC 0.83 (95% CI 0.69-0.93)] had higher accuracy than LAI [AUC 0.64 (95% CI 0.48-0.78), p = 0.0216] and RAI [AUC 0.63 (95% CI 0.47-0.77), p = 0.0378] for predicting CP. There were no differences between the LAI and RAI (p = 0.950) in identifying CP. At a cut-off value of 1.32 for the RAR, the sensitivity was 82.6%, and the specificity was 86.4% in the detection of CP. Cut-offs of LAI > 83.4 mL/m2 and RAI ≤ 81.1 mL/m2 had sensitivities of 82.6% and 54.6%, respectively, and specificities of 87.0% and 50.0%, respectively.\nDiagnostic accuracy of individual parameters to distinguish between CP and RCM\nAUC = area under receiver operating characteristics curve; 95% CI = 95% confidence interval; other abbreviations as in Table 2; *Statistically significant, p < 0.001\nReceiver operating characteristics curves. ROC curves described the performance of different variables in differentiating CP from RCM.", "There were 18 men and 5 women in CP group, with a mean age of 43.0 ± 20.2 years (range 15 to 77 years). The aetiology of CP was unknown in 10 patients, whereas 4 patients had previous cardiac surgery, 7 had tuberculous infection, and 2 had history of an inflammatory infection. The RCM group included 22 patients (12 men, 10 women) with a mean age of 47.5 ± 18.5 years (range 14 to 72 years). Five RCM patients underwent heart transplantation, and surgical pathology specimens showed the presence of cardiac amyloidosis in 3 patients and nonspecific findings in 2 patients. Endomyocardial biopsy was performed in other 10 RCM patients. Cardiac amyloidosis was confirmed in 2 patients, mixed connective tissue disease in 1 patient, and idiopathic forms in 7 patients. There were no significant differences between RCM patients and the two other groups in terms of gender, age or BSA distribution. The demographic and clinical characteristics in each group and their comparison are shown in Table 1.\nBaseline and clinical characteristics", "The maximal pericardial thickness in CP patients (6.9 ± 2.6 mm, range 4-12 mm) was significantly larger than in normal subjects (1.5 ± 0.4 mm, range 0.9-2.7 mm, p < 0.001) and RCM patients (2.0 ± 0.7 mm, range 1.0-3.4 mm, p < 0.001). There were no differences among the three groups in biventricular end-systolic volume index. There were no differences between CP and RCM patients in biventricular end-diastolic volume index, stroke volume index, and EF, although these values were significantly smaller in RCM and CP patients compared with normal subjects. The RA volume index (RAI) in RCM patients (90.5 ± 35.3 mL/m2) was significantly larger than in CP patients (71.4 ± 15.7 mL/m2, p = 0.006) and normal subjects (38.1 ± 9.0 mL/m2, p < 0.001). Although the LA volume index (LAI) yielded significantly increased values in RCM (96.0 ± 37.0 mL/m2) and CP patients (105.6 ± 25.1 mL/m2) compared with normal subjects (39.5 ± 9.5 mL/m2, p < 0.001 for all), no statistical significance were reached between CP and RCM patients (p = 0.200) (Figure 2). The RAR in CP patients (1.50 ± 0.29) was significantly larger than in normal subjects (1.06 ± 0.20, p < 0.001) and RCM patients (1.12 ± 0.33, p < 0.001). There were no differences between RCM patients and normal subjects in the RAR (p = 0.452) (Figure 3). The intra-class correlation coefficient [equal to 0.917 with 95% confidence interval (CI), 0.84-0.92] showed that there was an excellent inter-observer agreement on the measurement of the RAR. The Bland-Altman plots showing the limits of agreement are shown in Figure 4.\nLeft and right atrial volume indices. Comparisons of LAI and RAI between CP, RCM patients and normal subjects.\nError bar of the RAR. Data are presented as means (squares) and 95% confidence interval (whiskers). **p < 0.001.\nBland Altman plot of the RAR values. Bland-Altman analysis showed excellent inter-observer agreement for the RAR.\nThe analysis of septal movement during early diastole revealed a septal bounce in 22 CP patients. One CP patient, all RCM patients and normal subjects had a normal septal configuration during diastole. Comparative results of CMR parameters are independently shown in Table 2. LGE was present in 7 of 22 RCM patients (31.8%) and absence in all CP patients and normal subjects. Several different patterns of LGE were present in RCM patients. In 4 of 5 patients with histopathologically proven cardiac amyloidosis, LGE was distributed over the entire subendocardial circumference, extending in various degrees into the neighboring myocardium and the papillary muscles. In remaining 1 patient with cardiac amyloidosis, diffuse transmural LGE was found in the LV wall (Figure 5). Two idiopathic RCM patients had focal areas of LGE in various locations of the LV myocardium.\nCardiac magnetic resonance imaging findings\nLAI = left atrial volume index; RAI = right atrial volume index; RAR = relative atrial volume ratio; EDV = end-diastolic volume; ESV = end-systolic volume; SV = stroke volume; EF = ejective fraction; Comparison with RCM *P < 0.05; Comparison with CP #P < 0.05\nLGE of cardiac amyloidosis. CMR demonstrated global LV and RV wall hypertrophy (a and c) (white arrows), diffuse transmural LGE (b and d) (black arrows), and mild left-sided pleural effusion (*) in a 39-year-old male patient with cardiac amyloidosis who underwent cardiac transplantation and was proven by surgical pathology specimen.", "Receiver operating characteristic curve analysis was used for comparison of discriminative capacity between different indices (Table 3 and Figure 6). The RAR [AUC 0.83 (95% CI 0.69-0.93)] had higher accuracy than LAI [AUC 0.64 (95% CI 0.48-0.78), p = 0.0216] and RAI [AUC 0.63 (95% CI 0.47-0.77), p = 0.0378] for predicting CP. There were no differences between the LAI and RAI (p = 0.950) in identifying CP. At a cut-off value of 1.32 for the RAR, the sensitivity was 82.6%, and the specificity was 86.4% in the detection of CP. Cut-offs of LAI > 83.4 mL/m2 and RAI ≤ 81.1 mL/m2 had sensitivities of 82.6% and 54.6%, respectively, and specificities of 87.0% and 50.0%, respectively.\nDiagnostic accuracy of individual parameters to distinguish between CP and RCM\nAUC = area under receiver operating characteristics curve; 95% CI = 95% confidence interval; other abbreviations as in Table 2; *Statistically significant, p < 0.001\nReceiver operating characteristics curves. ROC curves described the performance of different variables in differentiating CP from RCM.", "Both RCM and CP are often characterized by normal or decreased volume of both ventricles associated with biatrial enlargement, normal LV wall thickness and atrioventricular valves, impaired ventricular filling with restrictive physiology, and normal (or near normal) systolic function. Echocardiography, computed tomography, CMR and invasive cardiac catheterization have been useful in this differential diagnosis [4-7,10,12,13], but the diagnosis remains equivocal after extensive testing in a subset of patients.\nThe principal finding of this study demonstrated that the RAR was higher in CP patients than in RCM patients. The pathophysiological hallmarks of pericardial constriction, which are caused by confinement of the cardiac chambers by the rigid, fixed pericardial volume, are limitation of outward expansion of cardiac chambers. The pericardial oblique sinus lies behind the LA so that the posterior wall of the LA is actually separated from the pericardial space. Compared with the RA, the outward expansion of the LA may be less limited by the rigid and fixed pericardium in CP patients, which can lead to out-of-proportion expansion of the LA and RA. In RCM patients, the restrictive physiology caused by decreased myocardial compliance affects both ventricles, while the normally compliant pericardium allows for significantly prominent expansion of the LA and RA at the same time. In this study, the RAR in CP patients was significantly larger than in normal subjects and RCM patients. There were no differences between RCM patients and normal subjects in the RAR. The AUC of RAR was greater than those of the other parameters, while the AUC between the LAI and RAI did not show a difference. These results suggest that the RAR is a useful index for differentiating CP from RCM. These findings are of clinical significance, as substantial differentiation between CP and RCM could not be often made from extensive clinical and noninvasive testing.\nThe early diastolic septal bounce, a brief rapid motion of the ventricular septum toward the RV in early diastole, is considered a reliable echocardiopraphic and CMR sign of pericardial constriction [1,12]. As shown in other studies as well as herein, abnormal diastolic septal bounce had a sensitivity of 96%, a specificity of 100% for the prediction of surgically proven CP.\nThe versatility of CMR for CP not only enables accurate, noninvasive, quantitative and qualitative assessment of the pericardium and its associated structures, but also facilitates differentiation from a restrictive physiology that could be challenging clinically. These findings consist of a thickened, fibrotic, and/or calcified pericardium, a sigmoid-shaped septum, a restrictive filling pattern with an enhanced early filling, a respiratory-related variation in the position of the interventricular septum, and an extension of the fibrocalcific process into the underlying myocardium [8,10-13,19]. Moreover, CMR is helpful by its ability to characterize tissues, especially the demonstration of interstitial or nodular fibrosis based on the underlying etiology. The recent studies have showed that CMR has been used to characterize the type of infiltrative RCM by the location and distribution of LGE and may well facilitate diagnosis of RCM resulting from infiltrative myocardial disease, for example, cardiac amyloidosis [14-17]. In patients with systemic amyloidosis, LGE is highly sensitive and specific for the identification of cardiac involvement. Ruberg FL et al.[17] demonstrated that the sensitivity, specificity, positive predictive value and negative predictive value of LGE for the identification of clinical cardiac involvement was 86%, 86%, 95%, and 67% respectively. In the group with histologically proven cardiac amyloidosis, our CMR findings are in line with the recent report. Five RCM patients with LGE were histologically proven cardiac amyloidosis, 4 of whom had the entire subendocardial circumference enhancement. Vogelsberg H et al. [14] reported that patients with biopsy-proven cardiac amyloidosis had a distinct pattern of LGE, which was distributed over the entire subendocardial circumference, extending in various degrees into the neighboring myocardium. They concluded that using this pattern as a diagnostic criterion, the sensitivity of CMR for diagnosing cardiac amyloidosis was 80%, yielding a specificity of 94%. The positive predictive value was 92%, and the negative predictive value was 85%.\nIn addition, CMR techniques might be used to study the hemodynamic. Francone M et al. [13] reported that real-time cine CMR can easily depict increased ventricular coupling, which may be helpful to better differentiate between CP and RCM patients, especially in patients with normal or minimally thickened pericardium. Recently, Bauner K et al. [20] reported that velocity-encoded flow measurements with calculation of transtricuspid e- to a-wave ratios are a valuable tool for detection of diastolic dysfunction in CP patients.", "The study is limited by a small sample size, which may cause a statistical bias, and larger numbers of patients should be addressed in a future study. RCM patients shared a lot of characteristics with CP patients. In this study, we applied strict standardized diagnostic criteria of RCM. However, histological proof was not available for all RCM patients, and therefore, CP patients with normal pericardial thickness may have been included in RCM patients [21].", "In conclusion, CMR has the potential to enable precise assessment of morphology, function, and tissue characteristics of the heart, which can facilitate differential diagnosis between CP and RCM. If the differential diagnosis between CP and RCM could not be made from extensive clinical and noninvasive testing, a further analysis of both the RAR and LGE is helpful in distinguishing CP from RCM.", "(CP): Constrictive pericarditis; (RCM): Restrictive cardiomyopathy; (CMR): Cardiovascular magnetic resonance; (LGE): Late gadolinium enhancement; (RAR): Relative atrial volume ratio; (LV): Left ventricular; (RV): Right ventricular; (LA): Left atrial; (LAI): Left artial volume index; (RA): Right atrial; (RAI): Right atrial index; (SD): Standard deviation; (ROC): Receiver operating characteristics; (AUC): Area under receiver operating characteristics curve; (CI): Confidence interval.", "The authors declare that they have no competing interests.", "HC drafted both the text and figure file. SZ and SJ conceived and designed this study. JR and JL revised the manuscript. ML, CY and YZ were involved with data acquisition. ZH and NM helped draft the manuscript. GY and QL assisted with data analysis and statistics. All authors take responsibility for the entire content of this study and have read and approved the submission of this manuscript." ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study population", "CMR protocol", "CMR Analysis", "Statistical analysis", "Results", "Patient characteristics", "CMR characteristics", "Sensitivity and specificity", "Discussion", "Study limitations", "Conclusions", "List of abbreviations used", "Competing interests", "Authors' contributions" ]

[ "Clinical and hemodynamic features are often similar in constrictive pericarditis (CP) and restrictive cardiomyopathy (RCM), but differentiation of these 2 conditions is crucial because CP requires surgical treatment and is usually curable, while RCM, short of cardiac transplantation, is treatable only by medical means and often responds unsatisfactorily [1-3]. At present, clinical evaluation, measurement of pericardial thickness, analysis of septal motion, quantitative assessment of systolic and diastolic myocardial function, invasive pressure measurement, and endomyocardial biopsy have been useful in this differential diagnosis, but no one diagnostic method can be relied upon to make the distinction by itself [4-7].\nCardiovascular magnetic resonance (CMR) provides high-resolution imaging of the pericardium and associated structures in any imaging plane. It fuses excellent anatomic detail and tissue characterization with accurate evaluation of cardiac function and assessment of the haemodynamic consequences of pericardial constraint on cardiac filling [8-12]. Compared with echocardiography and computed tomography, CMR with late gadolinium enhancement (LGE) is the only method that can depict the presence of myocardial fibrosis, which may well facilitate diagnosis of RCM resulting from infiltrative myocardial disease and have important prognostic implications [13-17].\nThe aim of the present study was to describe the clinical utility of CMR for distinguishing CP form RCM. We sought to determine the diagnostic accuracy of the relative atrial volume ratio (RAR) for the detection of CP and its possible use as a screening tool to aid in the differentiation between CP and RCM.", "[SUBTITLE] Study population [SUBSECTION] The study population consists of 45 consecutive patients who were referred for CMR, including 23 surgically documented CP patients and 22 RCM patients. All patients had been underwent previously systematic clinical evaluation, including history and examination, electrocardiography, chest radiography, and echocardiography. In each CP case, surgical confirmation was obtained by the presence of an obliterated pericardial space, an adhesive pericarditis with bulging of the heart out of the pericardial incision at pericardiectomy and pathological confirmation. The diagnosis of RCM was confirmed by pathological specimens or based on impaired cardiac filling (i.e., increased filling pressures and no echo-Doppler evidence of respiratory-dependent ventricular coupling) in combination with pericardial thickness < 2 mm. All patients were referred to rule out any other cardiovascular diseases such as coronary artery disease, hypertension, valvular and congenital heart disease, and other cardiomyopathy. As a control group, 25 normal subjects without a history of cardiovascular symptoms or risk factors were also included in this study.\nThe study was approved by the institutional ethics committee, and all subjects gave written informed consent.\nThe study population consists of 45 consecutive patients who were referred for CMR, including 23 surgically documented CP patients and 22 RCM patients. All patients had been underwent previously systematic clinical evaluation, including history and examination, electrocardiography, chest radiography, and echocardiography. In each CP case, surgical confirmation was obtained by the presence of an obliterated pericardial space, an adhesive pericarditis with bulging of the heart out of the pericardial incision at pericardiectomy and pathological confirmation. The diagnosis of RCM was confirmed by pathological specimens or based on impaired cardiac filling (i.e., increased filling pressures and no echo-Doppler evidence of respiratory-dependent ventricular coupling) in combination with pericardial thickness < 2 mm. All patients were referred to rule out any other cardiovascular diseases such as coronary artery disease, hypertension, valvular and congenital heart disease, and other cardiomyopathy. As a control group, 25 normal subjects without a history of cardiovascular symptoms or risk factors were also included in this study.\nThe study was approved by the institutional ethics committee, and all subjects gave written informed consent.\n[SUBTITLE] CMR protocol [SUBSECTION] Cardiac magnetic resonance imaging was performed in all patients by using a 1.5-T unit (Magnetom Avanto; Siemens Medical Solutions, Erlangen, Germany) with a high-performance gradient system (maximum gradient amplitude 45 mT/m; maximum slew rate 200-μs rise time), a 12-element-body phased-array coil system and electrocardiographic triggering. The CMR examinations began with the acquisition of survey images in three orthogonal planes (transverse, coronal, and sagittal) to localize the heart within the chest. Next, we studied the heart by performing a dark blood half-Fourier acquisition single-shot turbo spin echo (HASTE: repetition time [TR]/echo time [TE] = 700/26 ms, slice thickness = 6 mm, flip angle = 160°, field of view [FOV] = 350 mm) and turbo spin-echo (TSE) T1- (TR/TE = 700/20 ms, slice thickness = 6 mm, flip angle = 180°, matrix = 256 × 156; FOV = 350 mm) and T2-weighted (TR/TE = 800/77 ms, slice thickness = 6 mm, matrix = 256 ×190, FOV = 350 mm, flip angle = 180°) CMR sequences. Left ventricular (LV) short-axis, horizontal long-axis, and vertical long-axis views were used to evaluate cardiac function on cine CMR sequences. Cine CMR were acquired using generalized autocalibrating partially parallel acquisitions (GRAPPA: TR/TE = 45.9/1.07 ms, slice thickness = 6 mm, matrix = 109 × 192, FOV = 350 mm, flip angle = 80°) or time-adaptive sensitivity encoding (TSENSE: TR/TE = 41.7/1.39 ms, slice thickness = 6 mm, matrix = 109 × 192, FOV = 350 mm, flip angle = 70°) with true fast imaging with steady-state precession (TrueFISP) cine sequences. 15 to 20 minutes after injection of 0.2 mmol/kg of gadolinium diethylenetriamine pentaacetic acid (Magnevist, Schering, Berlin, Germany), the images of LGE were obtained in standard short axis covering the entire ventricle, and in long axis views to detect areas of LGE using a phase-sensitive inversion recovery (PISR) spoiled gradient echo sequence (TR/TE = 8.7/3.4 ms, slice thickness = 6 mm, imaging matrix = 256 × 256, FOV = 350 mm, flip angle = 15°).\nCardiac magnetic resonance imaging was performed in all patients by using a 1.5-T unit (Magnetom Avanto; Siemens Medical Solutions, Erlangen, Germany) with a high-performance gradient system (maximum gradient amplitude 45 mT/m; maximum slew rate 200-μs rise time), a 12-element-body phased-array coil system and electrocardiographic triggering. The CMR examinations began with the acquisition of survey images in three orthogonal planes (transverse, coronal, and sagittal) to localize the heart within the chest. Next, we studied the heart by performing a dark blood half-Fourier acquisition single-shot turbo spin echo (HASTE: repetition time [TR]/echo time [TE] = 700/26 ms, slice thickness = 6 mm, flip angle = 160°, field of view [FOV] = 350 mm) and turbo spin-echo (TSE) T1- (TR/TE = 700/20 ms, slice thickness = 6 mm, flip angle = 180°, matrix = 256 × 156; FOV = 350 mm) and T2-weighted (TR/TE = 800/77 ms, slice thickness = 6 mm, matrix = 256 ×190, FOV = 350 mm, flip angle = 180°) CMR sequences. Left ventricular (LV) short-axis, horizontal long-axis, and vertical long-axis views were used to evaluate cardiac function on cine CMR sequences. Cine CMR were acquired using generalized autocalibrating partially parallel acquisitions (GRAPPA: TR/TE = 45.9/1.07 ms, slice thickness = 6 mm, matrix = 109 × 192, FOV = 350 mm, flip angle = 80°) or time-adaptive sensitivity encoding (TSENSE: TR/TE = 41.7/1.39 ms, slice thickness = 6 mm, matrix = 109 × 192, FOV = 350 mm, flip angle = 70°) with true fast imaging with steady-state precession (TrueFISP) cine sequences. 15 to 20 minutes after injection of 0.2 mmol/kg of gadolinium diethylenetriamine pentaacetic acid (Magnevist, Schering, Berlin, Germany), the images of LGE were obtained in standard short axis covering the entire ventricle, and in long axis views to detect areas of LGE using a phase-sensitive inversion recovery (PISR) spoiled gradient echo sequence (TR/TE = 8.7/3.4 ms, slice thickness = 6 mm, imaging matrix = 256 × 256, FOV = 350 mm, flip angle = 15°).\n[SUBTITLE] CMR Analysis [SUBSECTION] All CMR images were transferred to workstation (Siemens medical systems) for analysis. Qualitative assessments were performed independently by three readers. If there was a discrepancy, majority opinion was used. Quantitative measurements were performed independently by two readers. All observers were blinded to the diagnosis. For morphological evaluation of the pericardium, TSE and HASTE images were employed (Figure 1) to assess the maximum pericardial thickness. Septal motion was evaluated on a short-axis cine function view 1 cm beneath the atrioventricular valves on a visual basis and described as normal and the early diastolic septal bounce. The biventricular volumes and ejective fraction were obtained using Argus analytical software (version VE36A). Endocardial margins of the LV and right ventricular (RV) were semi-automatically contoured on end-diastolic and end-systolic images. End-diastolic and end-systolic frames were defined on the basis of the respective image frames demonstrating the largest and smallest cavity size. For the left atrial (LA) volume, the biplane area-length method was used. For the right atrial (RA) volume, the monoplane area-length formula was used. Atrial diastole was determined by selecting the last frame in ventricular systole before mitral valve opening. The measurements were made according to published methods [18]. The long-axis length of the LA and RA was defined by measuring the distance from the center of the mitral annulus to the posterior atrial wall. The atrial endocardial area was manually traced to exclude the atrial appendages and pulmonary or caval veins. Body weight and body height were measured and the body surface area was calculated. Subsequently, division with body surface area indexed all CMR variables apart from the ejection fraction. The RAR was defined as the LA volume divided by RA volume. LGE was considered present only if myocardial enhancement was confirmed on both short-axis and matching long-axis locations using a signal intensity threshold of > 2 standard deviation (SD) above a remote reference region in the same image.\nDiffuse thickened pericardium. HASTE (A), T1- (B) and T2-weighted (C) TSE images showed diffuse pericardial thickening (white arrows) which is most pronounced over the RV and RA and moderate right-sided pleural effusion (*).\nAll CMR images were transferred to workstation (Siemens medical systems) for analysis. Qualitative assessments were performed independently by three readers. If there was a discrepancy, majority opinion was used. Quantitative measurements were performed independently by two readers. All observers were blinded to the diagnosis. For morphological evaluation of the pericardium, TSE and HASTE images were employed (Figure 1) to assess the maximum pericardial thickness. Septal motion was evaluated on a short-axis cine function view 1 cm beneath the atrioventricular valves on a visual basis and described as normal and the early diastolic septal bounce. The biventricular volumes and ejective fraction were obtained using Argus analytical software (version VE36A). Endocardial margins of the LV and right ventricular (RV) were semi-automatically contoured on end-diastolic and end-systolic images. End-diastolic and end-systolic frames were defined on the basis of the respective image frames demonstrating the largest and smallest cavity size. For the left atrial (LA) volume, the biplane area-length method was used. For the right atrial (RA) volume, the monoplane area-length formula was used. Atrial diastole was determined by selecting the last frame in ventricular systole before mitral valve opening. The measurements were made according to published methods [18]. The long-axis length of the LA and RA was defined by measuring the distance from the center of the mitral annulus to the posterior atrial wall. The atrial endocardial area was manually traced to exclude the atrial appendages and pulmonary or caval veins. Body weight and body height were measured and the body surface area was calculated. Subsequently, division with body surface area indexed all CMR variables apart from the ejection fraction. The RAR was defined as the LA volume divided by RA volume. LGE was considered present only if myocardial enhancement was confirmed on both short-axis and matching long-axis locations using a signal intensity threshold of > 2 standard deviation (SD) above a remote reference region in the same image.\nDiffuse thickened pericardium. HASTE (A), T1- (B) and T2-weighted (C) TSE images showed diffuse pericardial thickening (white arrows) which is most pronounced over the RV and RA and moderate right-sided pleural effusion (*).\n[SUBTITLE] Statistical analysis [SUBSECTION] All values were given as mean ± SD or counts (percentage). Categorical values were compared by chi-square test or Fisher exact test as appropriate. Comparisons of normally distributed continuous variables between the different groups were performed by using one-way analysis of variance (ANOVA) with Fisher's least significant difference (LSD) posttest. The Kruskal-Wallis and Mann-Whitney U test were used to compare non-normally distributed continuous variables. The inter-observer agreement for the RAR was tested using intra-class correlation coefficient and limits of agreement using Bland Altman plots. Receiver operating characteristic (ROC) curve analysis was used to test the ability of different variables in differentiating CP from RCM. The area under the ROC curve (AUC) for each variable was calculated and compared. The statistical comparison of the ROC curves was performed using MedCalc (11.4.4, MedCalc, Belgium). Other statistical analyses were performed using SPSS for Windows (version 16.0; SPSS Inc., Chicago, IL, USA). P values < 0.05 were considered significant.\nAll values were given as mean ± SD or counts (percentage). Categorical values were compared by chi-square test or Fisher exact test as appropriate. Comparisons of normally distributed continuous variables between the different groups were performed by using one-way analysis of variance (ANOVA) with Fisher's least significant difference (LSD) posttest. The Kruskal-Wallis and Mann-Whitney U test were used to compare non-normally distributed continuous variables. The inter-observer agreement for the RAR was tested using intra-class correlation coefficient and limits of agreement using Bland Altman plots. Receiver operating characteristic (ROC) curve analysis was used to test the ability of different variables in differentiating CP from RCM. The area under the ROC curve (AUC) for each variable was calculated and compared. The statistical comparison of the ROC curves was performed using MedCalc (11.4.4, MedCalc, Belgium). Other statistical analyses were performed using SPSS for Windows (version 16.0; SPSS Inc., Chicago, IL, USA). P values < 0.05 were considered significant.", "The study population consists of 45 consecutive patients who were referred for CMR, including 23 surgically documented CP patients and 22 RCM patients. All patients had been underwent previously systematic clinical evaluation, including history and examination, electrocardiography, chest radiography, and echocardiography. In each CP case, surgical confirmation was obtained by the presence of an obliterated pericardial space, an adhesive pericarditis with bulging of the heart out of the pericardial incision at pericardiectomy and pathological confirmation. The diagnosis of RCM was confirmed by pathological specimens or based on impaired cardiac filling (i.e., increased filling pressures and no echo-Doppler evidence of respiratory-dependent ventricular coupling) in combination with pericardial thickness < 2 mm. All patients were referred to rule out any other cardiovascular diseases such as coronary artery disease, hypertension, valvular and congenital heart disease, and other cardiomyopathy. As a control group, 25 normal subjects without a history of cardiovascular symptoms or risk factors were also included in this study.\nThe study was approved by the institutional ethics committee, and all subjects gave written informed consent.", "Cardiac magnetic resonance imaging was performed in all patients by using a 1.5-T unit (Magnetom Avanto; Siemens Medical Solutions, Erlangen, Germany) with a high-performance gradient system (maximum gradient amplitude 45 mT/m; maximum slew rate 200-μs rise time), a 12-element-body phased-array coil system and electrocardiographic triggering. The CMR examinations began with the acquisition of survey images in three orthogonal planes (transverse, coronal, and sagittal) to localize the heart within the chest. Next, we studied the heart by performing a dark blood half-Fourier acquisition single-shot turbo spin echo (HASTE: repetition time [TR]/echo time [TE] = 700/26 ms, slice thickness = 6 mm, flip angle = 160°, field of view [FOV] = 350 mm) and turbo spin-echo (TSE) T1- (TR/TE = 700/20 ms, slice thickness = 6 mm, flip angle = 180°, matrix = 256 × 156; FOV = 350 mm) and T2-weighted (TR/TE = 800/77 ms, slice thickness = 6 mm, matrix = 256 ×190, FOV = 350 mm, flip angle = 180°) CMR sequences. Left ventricular (LV) short-axis, horizontal long-axis, and vertical long-axis views were used to evaluate cardiac function on cine CMR sequences. Cine CMR were acquired using generalized autocalibrating partially parallel acquisitions (GRAPPA: TR/TE = 45.9/1.07 ms, slice thickness = 6 mm, matrix = 109 × 192, FOV = 350 mm, flip angle = 80°) or time-adaptive sensitivity encoding (TSENSE: TR/TE = 41.7/1.39 ms, slice thickness = 6 mm, matrix = 109 × 192, FOV = 350 mm, flip angle = 70°) with true fast imaging with steady-state precession (TrueFISP) cine sequences. 15 to 20 minutes after injection of 0.2 mmol/kg of gadolinium diethylenetriamine pentaacetic acid (Magnevist, Schering, Berlin, Germany), the images of LGE were obtained in standard short axis covering the entire ventricle, and in long axis views to detect areas of LGE using a phase-sensitive inversion recovery (PISR) spoiled gradient echo sequence (TR/TE = 8.7/3.4 ms, slice thickness = 6 mm, imaging matrix = 256 × 256, FOV = 350 mm, flip angle = 15°).", "All CMR images were transferred to workstation (Siemens medical systems) for analysis. Qualitative assessments were performed independently by three readers. If there was a discrepancy, majority opinion was used. Quantitative measurements were performed independently by two readers. All observers were blinded to the diagnosis. For morphological evaluation of the pericardium, TSE and HASTE images were employed (Figure 1) to assess the maximum pericardial thickness. Septal motion was evaluated on a short-axis cine function view 1 cm beneath the atrioventricular valves on a visual basis and described as normal and the early diastolic septal bounce. The biventricular volumes and ejective fraction were obtained using Argus analytical software (version VE36A). Endocardial margins of the LV and right ventricular (RV) were semi-automatically contoured on end-diastolic and end-systolic images. End-diastolic and end-systolic frames were defined on the basis of the respective image frames demonstrating the largest and smallest cavity size. For the left atrial (LA) volume, the biplane area-length method was used. For the right atrial (RA) volume, the monoplane area-length formula was used. Atrial diastole was determined by selecting the last frame in ventricular systole before mitral valve opening. The measurements were made according to published methods [18]. The long-axis length of the LA and RA was defined by measuring the distance from the center of the mitral annulus to the posterior atrial wall. The atrial endocardial area was manually traced to exclude the atrial appendages and pulmonary or caval veins. Body weight and body height were measured and the body surface area was calculated. Subsequently, division with body surface area indexed all CMR variables apart from the ejection fraction. The RAR was defined as the LA volume divided by RA volume. LGE was considered present only if myocardial enhancement was confirmed on both short-axis and matching long-axis locations using a signal intensity threshold of > 2 standard deviation (SD) above a remote reference region in the same image.\nDiffuse thickened pericardium. HASTE (A), T1- (B) and T2-weighted (C) TSE images showed diffuse pericardial thickening (white arrows) which is most pronounced over the RV and RA and moderate right-sided pleural effusion (*).", "All values were given as mean ± SD or counts (percentage). Categorical values were compared by chi-square test or Fisher exact test as appropriate. Comparisons of normally distributed continuous variables between the different groups were performed by using one-way analysis of variance (ANOVA) with Fisher's least significant difference (LSD) posttest. The Kruskal-Wallis and Mann-Whitney U test were used to compare non-normally distributed continuous variables. The inter-observer agreement for the RAR was tested using intra-class correlation coefficient and limits of agreement using Bland Altman plots. Receiver operating characteristic (ROC) curve analysis was used to test the ability of different variables in differentiating CP from RCM. The area under the ROC curve (AUC) for each variable was calculated and compared. The statistical comparison of the ROC curves was performed using MedCalc (11.4.4, MedCalc, Belgium). Other statistical analyses were performed using SPSS for Windows (version 16.0; SPSS Inc., Chicago, IL, USA). P values < 0.05 were considered significant.", "[SUBTITLE] Patient characteristics [SUBSECTION] There were 18 men and 5 women in CP group, with a mean age of 43.0 ± 20.2 years (range 15 to 77 years). The aetiology of CP was unknown in 10 patients, whereas 4 patients had previous cardiac surgery, 7 had tuberculous infection, and 2 had history of an inflammatory infection. The RCM group included 22 patients (12 men, 10 women) with a mean age of 47.5 ± 18.5 years (range 14 to 72 years). Five RCM patients underwent heart transplantation, and surgical pathology specimens showed the presence of cardiac amyloidosis in 3 patients and nonspecific findings in 2 patients. Endomyocardial biopsy was performed in other 10 RCM patients. Cardiac amyloidosis was confirmed in 2 patients, mixed connective tissue disease in 1 patient, and idiopathic forms in 7 patients. There were no significant differences between RCM patients and the two other groups in terms of gender, age or BSA distribution. The demographic and clinical characteristics in each group and their comparison are shown in Table 1.\nBaseline and clinical characteristics\nThere were 18 men and 5 women in CP group, with a mean age of 43.0 ± 20.2 years (range 15 to 77 years). The aetiology of CP was unknown in 10 patients, whereas 4 patients had previous cardiac surgery, 7 had tuberculous infection, and 2 had history of an inflammatory infection. The RCM group included 22 patients (12 men, 10 women) with a mean age of 47.5 ± 18.5 years (range 14 to 72 years). Five RCM patients underwent heart transplantation, and surgical pathology specimens showed the presence of cardiac amyloidosis in 3 patients and nonspecific findings in 2 patients. Endomyocardial biopsy was performed in other 10 RCM patients. Cardiac amyloidosis was confirmed in 2 patients, mixed connective tissue disease in 1 patient, and idiopathic forms in 7 patients. There were no significant differences between RCM patients and the two other groups in terms of gender, age or BSA distribution. The demographic and clinical characteristics in each group and their comparison are shown in Table 1.\nBaseline and clinical characteristics\n[SUBTITLE] CMR characteristics [SUBSECTION] The maximal pericardial thickness in CP patients (6.9 ± 2.6 mm, range 4-12 mm) was significantly larger than in normal subjects (1.5 ± 0.4 mm, range 0.9-2.7 mm, p < 0.001) and RCM patients (2.0 ± 0.7 mm, range 1.0-3.4 mm, p < 0.001). There were no differences among the three groups in biventricular end-systolic volume index. There were no differences between CP and RCM patients in biventricular end-diastolic volume index, stroke volume index, and EF, although these values were significantly smaller in RCM and CP patients compared with normal subjects. The RA volume index (RAI) in RCM patients (90.5 ± 35.3 mL/m2) was significantly larger than in CP patients (71.4 ± 15.7 mL/m2, p = 0.006) and normal subjects (38.1 ± 9.0 mL/m2, p < 0.001). Although the LA volume index (LAI) yielded significantly increased values in RCM (96.0 ± 37.0 mL/m2) and CP patients (105.6 ± 25.1 mL/m2) compared with normal subjects (39.5 ± 9.5 mL/m2, p < 0.001 for all), no statistical significance were reached between CP and RCM patients (p = 0.200) (Figure 2). The RAR in CP patients (1.50 ± 0.29) was significantly larger than in normal subjects (1.06 ± 0.20, p < 0.001) and RCM patients (1.12 ± 0.33, p < 0.001). There were no differences between RCM patients and normal subjects in the RAR (p = 0.452) (Figure 3). The intra-class correlation coefficient [equal to 0.917 with 95% confidence interval (CI), 0.84-0.92] showed that there was an excellent inter-observer agreement on the measurement of the RAR. The Bland-Altman plots showing the limits of agreement are shown in Figure 4.\nLeft and right atrial volume indices. Comparisons of LAI and RAI between CP, RCM patients and normal subjects.\nError bar of the RAR. Data are presented as means (squares) and 95% confidence interval (whiskers). **p < 0.001.\nBland Altman plot of the RAR values. Bland-Altman analysis showed excellent inter-observer agreement for the RAR.\nThe analysis of septal movement during early diastole revealed a septal bounce in 22 CP patients. One CP patient, all RCM patients and normal subjects had a normal septal configuration during diastole. Comparative results of CMR parameters are independently shown in Table 2. LGE was present in 7 of 22 RCM patients (31.8%) and absence in all CP patients and normal subjects. Several different patterns of LGE were present in RCM patients. In 4 of 5 patients with histopathologically proven cardiac amyloidosis, LGE was distributed over the entire subendocardial circumference, extending in various degrees into the neighboring myocardium and the papillary muscles. In remaining 1 patient with cardiac amyloidosis, diffuse transmural LGE was found in the LV wall (Figure 5). Two idiopathic RCM patients had focal areas of LGE in various locations of the LV myocardium.\nCardiac magnetic resonance imaging findings\nLAI = left atrial volume index; RAI = right atrial volume index; RAR = relative atrial volume ratio; EDV = end-diastolic volume; ESV = end-systolic volume; SV = stroke volume; EF = ejective fraction; Comparison with RCM *P < 0.05; Comparison with CP #P < 0.05\nLGE of cardiac amyloidosis. CMR demonstrated global LV and RV wall hypertrophy (a and c) (white arrows), diffuse transmural LGE (b and d) (black arrows), and mild left-sided pleural effusion (*) in a 39-year-old male patient with cardiac amyloidosis who underwent cardiac transplantation and was proven by surgical pathology specimen.\nThe maximal pericardial thickness in CP patients (6.9 ± 2.6 mm, range 4-12 mm) was significantly larger than in normal subjects (1.5 ± 0.4 mm, range 0.9-2.7 mm, p < 0.001) and RCM patients (2.0 ± 0.7 mm, range 1.0-3.4 mm, p < 0.001). There were no differences among the three groups in biventricular end-systolic volume index. There were no differences between CP and RCM patients in biventricular end-diastolic volume index, stroke volume index, and EF, although these values were significantly smaller in RCM and CP patients compared with normal subjects. The RA volume index (RAI) in RCM patients (90.5 ± 35.3 mL/m2) was significantly larger than in CP patients (71.4 ± 15.7 mL/m2, p = 0.006) and normal subjects (38.1 ± 9.0 mL/m2, p < 0.001). Although the LA volume index (LAI) yielded significantly increased values in RCM (96.0 ± 37.0 mL/m2) and CP patients (105.6 ± 25.1 mL/m2) compared with normal subjects (39.5 ± 9.5 mL/m2, p < 0.001 for all), no statistical significance were reached between CP and RCM patients (p = 0.200) (Figure 2). The RAR in CP patients (1.50 ± 0.29) was significantly larger than in normal subjects (1.06 ± 0.20, p < 0.001) and RCM patients (1.12 ± 0.33, p < 0.001). There were no differences between RCM patients and normal subjects in the RAR (p = 0.452) (Figure 3). The intra-class correlation coefficient [equal to 0.917 with 95% confidence interval (CI), 0.84-0.92] showed that there was an excellent inter-observer agreement on the measurement of the RAR. The Bland-Altman plots showing the limits of agreement are shown in Figure 4.\nLeft and right atrial volume indices. Comparisons of LAI and RAI between CP, RCM patients and normal subjects.\nError bar of the RAR. Data are presented as means (squares) and 95% confidence interval (whiskers). **p < 0.001.\nBland Altman plot of the RAR values. Bland-Altman analysis showed excellent inter-observer agreement for the RAR.\nThe analysis of septal movement during early diastole revealed a septal bounce in 22 CP patients. One CP patient, all RCM patients and normal subjects had a normal septal configuration during diastole. Comparative results of CMR parameters are independently shown in Table 2. LGE was present in 7 of 22 RCM patients (31.8%) and absence in all CP patients and normal subjects. Several different patterns of LGE were present in RCM patients. In 4 of 5 patients with histopathologically proven cardiac amyloidosis, LGE was distributed over the entire subendocardial circumference, extending in various degrees into the neighboring myocardium and the papillary muscles. In remaining 1 patient with cardiac amyloidosis, diffuse transmural LGE was found in the LV wall (Figure 5). Two idiopathic RCM patients had focal areas of LGE in various locations of the LV myocardium.\nCardiac magnetic resonance imaging findings\nLAI = left atrial volume index; RAI = right atrial volume index; RAR = relative atrial volume ratio; EDV = end-diastolic volume; ESV = end-systolic volume; SV = stroke volume; EF = ejective fraction; Comparison with RCM *P < 0.05; Comparison with CP #P < 0.05\nLGE of cardiac amyloidosis. CMR demonstrated global LV and RV wall hypertrophy (a and c) (white arrows), diffuse transmural LGE (b and d) (black arrows), and mild left-sided pleural effusion (*) in a 39-year-old male patient with cardiac amyloidosis who underwent cardiac transplantation and was proven by surgical pathology specimen.\n[SUBTITLE] Sensitivity and specificity [SUBSECTION] Receiver operating characteristic curve analysis was used for comparison of discriminative capacity between different indices (Table 3 and Figure 6). The RAR [AUC 0.83 (95% CI 0.69-0.93)] had higher accuracy than LAI [AUC 0.64 (95% CI 0.48-0.78), p = 0.0216] and RAI [AUC 0.63 (95% CI 0.47-0.77), p = 0.0378] for predicting CP. There were no differences between the LAI and RAI (p = 0.950) in identifying CP. At a cut-off value of 1.32 for the RAR, the sensitivity was 82.6%, and the specificity was 86.4% in the detection of CP. Cut-offs of LAI > 83.4 mL/m2 and RAI ≤ 81.1 mL/m2 had sensitivities of 82.6% and 54.6%, respectively, and specificities of 87.0% and 50.0%, respectively.\nDiagnostic accuracy of individual parameters to distinguish between CP and RCM\nAUC = area under receiver operating characteristics curve; 95% CI = 95% confidence interval; other abbreviations as in Table 2; *Statistically significant, p < 0.001\nReceiver operating characteristics curves. ROC curves described the performance of different variables in differentiating CP from RCM.\nReceiver operating characteristic curve analysis was used for comparison of discriminative capacity between different indices (Table 3 and Figure 6). The RAR [AUC 0.83 (95% CI 0.69-0.93)] had higher accuracy than LAI [AUC 0.64 (95% CI 0.48-0.78), p = 0.0216] and RAI [AUC 0.63 (95% CI 0.47-0.77), p = 0.0378] for predicting CP. There were no differences between the LAI and RAI (p = 0.950) in identifying CP. At a cut-off value of 1.32 for the RAR, the sensitivity was 82.6%, and the specificity was 86.4% in the detection of CP. Cut-offs of LAI > 83.4 mL/m2 and RAI ≤ 81.1 mL/m2 had sensitivities of 82.6% and 54.6%, respectively, and specificities of 87.0% and 50.0%, respectively.\nDiagnostic accuracy of individual parameters to distinguish between CP and RCM\nAUC = area under receiver operating characteristics curve; 95% CI = 95% confidence interval; other abbreviations as in Table 2; *Statistically significant, p < 0.001\nReceiver operating characteristics curves. ROC curves described the performance of different variables in differentiating CP from RCM.", "There were 18 men and 5 women in CP group, with a mean age of 43.0 ± 20.2 years (range 15 to 77 years). The aetiology of CP was unknown in 10 patients, whereas 4 patients had previous cardiac surgery, 7 had tuberculous infection, and 2 had history of an inflammatory infection. The RCM group included 22 patients (12 men, 10 women) with a mean age of 47.5 ± 18.5 years (range 14 to 72 years). Five RCM patients underwent heart transplantation, and surgical pathology specimens showed the presence of cardiac amyloidosis in 3 patients and nonspecific findings in 2 patients. Endomyocardial biopsy was performed in other 10 RCM patients. Cardiac amyloidosis was confirmed in 2 patients, mixed connective tissue disease in 1 patient, and idiopathic forms in 7 patients. There were no significant differences between RCM patients and the two other groups in terms of gender, age or BSA distribution. The demographic and clinical characteristics in each group and their comparison are shown in Table 1.\nBaseline and clinical characteristics", "The maximal pericardial thickness in CP patients (6.9 ± 2.6 mm, range 4-12 mm) was significantly larger than in normal subjects (1.5 ± 0.4 mm, range 0.9-2.7 mm, p < 0.001) and RCM patients (2.0 ± 0.7 mm, range 1.0-3.4 mm, p < 0.001). There were no differences among the three groups in biventricular end-systolic volume index. There were no differences between CP and RCM patients in biventricular end-diastolic volume index, stroke volume index, and EF, although these values were significantly smaller in RCM and CP patients compared with normal subjects. The RA volume index (RAI) in RCM patients (90.5 ± 35.3 mL/m2) was significantly larger than in CP patients (71.4 ± 15.7 mL/m2, p = 0.006) and normal subjects (38.1 ± 9.0 mL/m2, p < 0.001). Although the LA volume index (LAI) yielded significantly increased values in RCM (96.0 ± 37.0 mL/m2) and CP patients (105.6 ± 25.1 mL/m2) compared with normal subjects (39.5 ± 9.5 mL/m2, p < 0.001 for all), no statistical significance were reached between CP and RCM patients (p = 0.200) (Figure 2). The RAR in CP patients (1.50 ± 0.29) was significantly larger than in normal subjects (1.06 ± 0.20, p < 0.001) and RCM patients (1.12 ± 0.33, p < 0.001). There were no differences between RCM patients and normal subjects in the RAR (p = 0.452) (Figure 3). The intra-class correlation coefficient [equal to 0.917 with 95% confidence interval (CI), 0.84-0.92] showed that there was an excellent inter-observer agreement on the measurement of the RAR. The Bland-Altman plots showing the limits of agreement are shown in Figure 4.\nLeft and right atrial volume indices. Comparisons of LAI and RAI between CP, RCM patients and normal subjects.\nError bar of the RAR. Data are presented as means (squares) and 95% confidence interval (whiskers). **p < 0.001.\nBland Altman plot of the RAR values. Bland-Altman analysis showed excellent inter-observer agreement for the RAR.\nThe analysis of septal movement during early diastole revealed a septal bounce in 22 CP patients. One CP patient, all RCM patients and normal subjects had a normal septal configuration during diastole. Comparative results of CMR parameters are independently shown in Table 2. LGE was present in 7 of 22 RCM patients (31.8%) and absence in all CP patients and normal subjects. Several different patterns of LGE were present in RCM patients. In 4 of 5 patients with histopathologically proven cardiac amyloidosis, LGE was distributed over the entire subendocardial circumference, extending in various degrees into the neighboring myocardium and the papillary muscles. In remaining 1 patient with cardiac amyloidosis, diffuse transmural LGE was found in the LV wall (Figure 5). Two idiopathic RCM patients had focal areas of LGE in various locations of the LV myocardium.\nCardiac magnetic resonance imaging findings\nLAI = left atrial volume index; RAI = right atrial volume index; RAR = relative atrial volume ratio; EDV = end-diastolic volume; ESV = end-systolic volume; SV = stroke volume; EF = ejective fraction; Comparison with RCM *P < 0.05; Comparison with CP #P < 0.05\nLGE of cardiac amyloidosis. CMR demonstrated global LV and RV wall hypertrophy (a and c) (white arrows), diffuse transmural LGE (b and d) (black arrows), and mild left-sided pleural effusion (*) in a 39-year-old male patient with cardiac amyloidosis who underwent cardiac transplantation and was proven by surgical pathology specimen.", "Receiver operating characteristic curve analysis was used for comparison of discriminative capacity between different indices (Table 3 and Figure 6). The RAR [AUC 0.83 (95% CI 0.69-0.93)] had higher accuracy than LAI [AUC 0.64 (95% CI 0.48-0.78), p = 0.0216] and RAI [AUC 0.63 (95% CI 0.47-0.77), p = 0.0378] for predicting CP. There were no differences between the LAI and RAI (p = 0.950) in identifying CP. At a cut-off value of 1.32 for the RAR, the sensitivity was 82.6%, and the specificity was 86.4% in the detection of CP. Cut-offs of LAI > 83.4 mL/m2 and RAI ≤ 81.1 mL/m2 had sensitivities of 82.6% and 54.6%, respectively, and specificities of 87.0% and 50.0%, respectively.\nDiagnostic accuracy of individual parameters to distinguish between CP and RCM\nAUC = area under receiver operating characteristics curve; 95% CI = 95% confidence interval; other abbreviations as in Table 2; *Statistically significant, p < 0.001\nReceiver operating characteristics curves. ROC curves described the performance of different variables in differentiating CP from RCM.", "Both RCM and CP are often characterized by normal or decreased volume of both ventricles associated with biatrial enlargement, normal LV wall thickness and atrioventricular valves, impaired ventricular filling with restrictive physiology, and normal (or near normal) systolic function. Echocardiography, computed tomography, CMR and invasive cardiac catheterization have been useful in this differential diagnosis [4-7,10,12,13], but the diagnosis remains equivocal after extensive testing in a subset of patients.\nThe principal finding of this study demonstrated that the RAR was higher in CP patients than in RCM patients. The pathophysiological hallmarks of pericardial constriction, which are caused by confinement of the cardiac chambers by the rigid, fixed pericardial volume, are limitation of outward expansion of cardiac chambers. The pericardial oblique sinus lies behind the LA so that the posterior wall of the LA is actually separated from the pericardial space. Compared with the RA, the outward expansion of the LA may be less limited by the rigid and fixed pericardium in CP patients, which can lead to out-of-proportion expansion of the LA and RA. In RCM patients, the restrictive physiology caused by decreased myocardial compliance affects both ventricles, while the normally compliant pericardium allows for significantly prominent expansion of the LA and RA at the same time. In this study, the RAR in CP patients was significantly larger than in normal subjects and RCM patients. There were no differences between RCM patients and normal subjects in the RAR. The AUC of RAR was greater than those of the other parameters, while the AUC between the LAI and RAI did not show a difference. These results suggest that the RAR is a useful index for differentiating CP from RCM. These findings are of clinical significance, as substantial differentiation between CP and RCM could not be often made from extensive clinical and noninvasive testing.\nThe early diastolic septal bounce, a brief rapid motion of the ventricular septum toward the RV in early diastole, is considered a reliable echocardiopraphic and CMR sign of pericardial constriction [1,12]. As shown in other studies as well as herein, abnormal diastolic septal bounce had a sensitivity of 96%, a specificity of 100% for the prediction of surgically proven CP.\nThe versatility of CMR for CP not only enables accurate, noninvasive, quantitative and qualitative assessment of the pericardium and its associated structures, but also facilitates differentiation from a restrictive physiology that could be challenging clinically. These findings consist of a thickened, fibrotic, and/or calcified pericardium, a sigmoid-shaped septum, a restrictive filling pattern with an enhanced early filling, a respiratory-related variation in the position of the interventricular septum, and an extension of the fibrocalcific process into the underlying myocardium [8,10-13,19]. Moreover, CMR is helpful by its ability to characterize tissues, especially the demonstration of interstitial or nodular fibrosis based on the underlying etiology. The recent studies have showed that CMR has been used to characterize the type of infiltrative RCM by the location and distribution of LGE and may well facilitate diagnosis of RCM resulting from infiltrative myocardial disease, for example, cardiac amyloidosis [14-17]. In patients with systemic amyloidosis, LGE is highly sensitive and specific for the identification of cardiac involvement. Ruberg FL et al.[17] demonstrated that the sensitivity, specificity, positive predictive value and negative predictive value of LGE for the identification of clinical cardiac involvement was 86%, 86%, 95%, and 67% respectively. In the group with histologically proven cardiac amyloidosis, our CMR findings are in line with the recent report. Five RCM patients with LGE were histologically proven cardiac amyloidosis, 4 of whom had the entire subendocardial circumference enhancement. Vogelsberg H et al. [14] reported that patients with biopsy-proven cardiac amyloidosis had a distinct pattern of LGE, which was distributed over the entire subendocardial circumference, extending in various degrees into the neighboring myocardium. They concluded that using this pattern as a diagnostic criterion, the sensitivity of CMR for diagnosing cardiac amyloidosis was 80%, yielding a specificity of 94%. The positive predictive value was 92%, and the negative predictive value was 85%.\nIn addition, CMR techniques might be used to study the hemodynamic. Francone M et al. [13] reported that real-time cine CMR can easily depict increased ventricular coupling, which may be helpful to better differentiate between CP and RCM patients, especially in patients with normal or minimally thickened pericardium. Recently, Bauner K et al. [20] reported that velocity-encoded flow measurements with calculation of transtricuspid e- to a-wave ratios are a valuable tool for detection of diastolic dysfunction in CP patients.", "The study is limited by a small sample size, which may cause a statistical bias, and larger numbers of patients should be addressed in a future study. RCM patients shared a lot of characteristics with CP patients. In this study, we applied strict standardized diagnostic criteria of RCM. However, histological proof was not available for all RCM patients, and therefore, CP patients with normal pericardial thickness may have been included in RCM patients [21].", "In conclusion, CMR has the potential to enable precise assessment of morphology, function, and tissue characteristics of the heart, which can facilitate differential diagnosis between CP and RCM. If the differential diagnosis between CP and RCM could not be made from extensive clinical and noninvasive testing, a further analysis of both the RAR and LGE is helpful in distinguishing CP from RCM.", "(CP): Constrictive pericarditis; (RCM): Restrictive cardiomyopathy; (CMR): Cardiovascular magnetic resonance; (LGE): Late gadolinium enhancement; (RAR): Relative atrial volume ratio; (LV): Left ventricular; (RV): Right ventricular; (LA): Left atrial; (LAI): Left artial volume index; (RA): Right atrial; (RAI): Right atrial index; (SD): Standard deviation; (ROC): Receiver operating characteristics; (AUC): Area under receiver operating characteristics curve; (CI): Confidence interval.", "The authors declare that they have no competing interests.", "HC drafted both the text and figure file. SZ and SJ conceived and designed this study. JR and JL revised the manuscript. ML, CY and YZ were involved with data acquisition. ZH and NM helped draft the manuscript. GY and QL assisted with data analysis and statistics. All authors take responsibility for the entire content of this study and have read and approved the submission of this manuscript." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Asian People", "Base Sequence", "Case-Control Studies", "China", "Confidence Intervals", "DNA Primers", "Female", "Gene Frequency", "Genetic Predisposition to Disease", "Genetic Variation", "Genotype", "Haplotypes", "Hirschsprung Disease", "Humans", "Infant, Newborn", "Linkage Disequilibrium", "Logistic Models", "Male", "Molecular Epidemiology", "Odds Ratio", "Polymorphism, Single Nucleotide", "Proto-Oncogene Proteins c-ret", "Risk Factors" ]

[ "Background", "Study subjects", "Polymorphism Analysis", "Statistical analysis", "Results", "Association of RET SNPs with the risk of HSCR", "RET haplotypes and risk of HSCR", "Discussion", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Hirschsprung's disease (HSCR) is a congenital disorder characterized by the absence of enteric ganglion cells in the myenteric (Auerbach's) and submucosal (Meissner's) plexuses of the variable lengths of gastrointestinal tract. The lack of enteric neurons is attributed to the defective proliferation, differentiation, survival, and migration of enteric nervous system (ENS) towards the end of the gut, whereby neural crest cells (NCC) fail to innervate the lower gastrointestinal tract during embryonic development, presenting in clinic with failure to pass meconium, bilious vomiting, enterocolitis-associated diarrhoea, chronic severe constipation and progressive abdominal distension in the neonatal period [1,2]. Based on the length of the aganglionic segment, it can be classified into three subgroups: short-segment aganglionosis (S-HSCR), long-segment aganglionosis (L-HSCR), and total colonic aganglionsis (TCA). Noteworthy, there is significant racial variation in the incidence of the disease worldwide, which is most prevalent in Asians, including China, about 2.8 per 10,000 live births. The male to female ration (M:F) is around 4:1 among S-HSCR patients and 1:1 among L-HSCR patients. Generally, HSCR most commonly presents sporadically although it can be familial and manifests with low, sex-dependent penetrance and phenotypic variability [2].\nHSCR is a heterogenic disorder with a number of genes reported to be associated, most of which involved in the signaling pathway implicated in the proper ENS development. RET, a receptor tyrosine kinase, is the major susceptibility gene for HSCR, which locates in 10q11.2 and encodes a transmembrane tyrosine kinase receptor, responsible for triggering a number of downstream signaling transductions [3-5]. There is growing evidence showing that some potentially functional single nucleotide polymorphisms (SNPs) of RET gene could act as low susceptibility factors and modify the phenotype of HSCR, especially in certain combinations of alleles, haplotypes. Specifically, three common SNPs in the coding region of RET, c135G > A in exon2 (rs1800858, A45A), c1296A > G (rs1800860, A432A) in exon7 and c2307T > G (rs1800861, L769L) in exon13, were firstly reported to be associated with HSCR [6-8]. Two promoter polymorphisms -5G > A and -1A > C (rs10900296 and rs10900297) were also reported to be strongly associated with HSCR and have a significantly lower activity in an in vitro dual-luciferase expression assay conducted by Fitze G and his colleagues [2,9]. Garcia-Barceló, et al [10] further demonstrated that these two SNPs may disrupt a transcriptional factor, TTF1, binding site and then reduce transcription capacity from the promoter. However, to our best knowledge, looking at these 5 particular SNPs all at once in haplotype profile in the Chinese Han population and their effects on the risk of HSCR have not been fully described.\nIn this study, we conducted a large case-control study to evaluate the contribution of the genotypes and haplotypes of these aforementioned five SNPs in the pathogenesis of sporadic HSCR in a Southeastern Chinese population.", "This case-control study consisted of 123 sporadic HSCR (males: 75.6%) and 194 controls (males: 73.7%). Cases were histological diagnosed with either biopsy or surgical resection material for absence of enteric plexuses in Children's Hospital Affiliated to Zhejiang University Medical School, among which twelve (10%) of patients were affected with long-segment aganglionosis (LSA) and the rest with short-segment aganglionosis (SSA). Controls were normal children taking physical examination in the same hospital during the same time period as the patients were enrolled and frequency matched to cases by sex and age. Both cases and controls were Chinese Han in Zhejiang and informed consent was obtained from their parents. The study was approved by the Institutional Review Board of the Children's Hospital Affiliated to Zhejiang University Medical School.", "Genomic DNA was extracted from blood sample of all study subjects by using QIAamp-Blood Kit (Qiagen, Hilden, Germany). Genotypes were determined by PCR and direct sequencing without knowledge of the subjects' case/control status. Fifteen percent of the masked sample was randomly selected and genotyped twice by different people independently. The reproducibility was 100%.\nRET SNPs (-5G > A, -1A > C, c135G > A, c1296A > G, and c2307T > G) were genotyped as described previously [6,11,12]. PCR primers for RET -5G > A and -1A > C were 5'-CCCGCACTGAGCTCCTAC-3' and 5'-GGACGTCGCCTTCGCCAT-3', and 5'-TCT CGAGGGATGCTTACTGG-3' and 5'-GGCTCCGGTTAAGGTAGAGG-3' for c135G > A, 5'-CCCTGGTGGGGCATTTGTGC-3' and 5'-GGGTGGTTGGAAACTGATGG-3' for c1296 A > G and 5'-GAACTTGGGCAAGGCGATGC-3' and 5'-ACCCTGCAGCTGGCCTTAC-3' for c2307T > G, respectively. Amplication of the fragments was accomplished under a 25-ul reaction mixture with ~100 ng template DNA, 0.5 uM each primer, 0.2 mM each dNTP, 2.0 mM MgCl2 and 1.0 units of Tag DNA polymerase with 1 × reaction buffer (Applied Biosystems, Foster City, CA). The PCR profile consisted of an initial melting step of 5 minutes at 95°C, followed by 35 cycles of 40 seconds at 95°C, 30 seconds at 63°C (57°C for c135G > A and c2307T > G), and 45 second at 72°C, and a final elongation step of 7 minutes at 72°C. PCR products were then sequenced using an ABI PRISM® Big Dye™ Terminator v3.1 Cycle sequencing kit (Applied Biosystems, Foster City, CA) and an ABI 3130 automated sequencer according to the manufacturer's instructions.", "Chi-square test was used to compare the allele and genotype frequencies between cases and controls, as well as LSA and SSA. The associations between the genotype and the risk of HSCR were estimated by odds ratios (ORs) and their 95% confidence interval (CI), which was calculated by unconditional logistic regression adjusted for gender. PHASE v2.1 [13] was used to estimate the haplotype frequencies composed of five SNP pairs, respectively, and their difference between patients and controls. Diplotype were defined by two haplotypes of each individual. Bonferroni adjustment was used to adjust multiple comparison and p values were considered significantly at a level of 0.003, to allow for the testing of 5 partially correlated markers using 3 different inheritance model (dominant, recessive, and additive model). All of the statistical analyses were performed using Statistical Analysis System software (version 9.12; SAS Institute, Cary, NC).", "[SUBTITLE] Association of RET SNPs with the risk of HSCR [SUBSECTION] All the five markers in the controls conformed to HWE (P > 0.05, data not shown). The frequency of A allele in -5G > A, C allele in -1A > C, A allele in 135G > A, G allele in 1296A > G and G allele in 2307T > G in cases were 82.5%, 88.6%, 78.9%, 90.7% and 82.1%, which was higher than that of 45.5%, 63.1%, 43.8%, 80.4% and 49.7% in controls, with the P for Cochran-Armitage trend test of 1.12 × 10-16, 3.43 × 10-11, 4.20 × 10-15, 5.70 × 10-4, and 2.32 × 10-14, respectively. All those markers were shown to be strongly associated with the HSCR risk (Table 1). None of these SNPs were significantly different between LSA and SSA patients (data not shown).\nGenotype distribution of the RET gene polymorphisms in case-control\na p = 1.12 × 10-16, b p = 3.43 × 10-11, c p = 4.20 × 10-15, d p = 5.70 × 10-4, e p = 2.32 × 10-14, for Cochran-Armitage trend test.\nUnconditional logistic regression model was used to estimate RET SNPs effect (additive, dominant, or recessive) according to the smallest AIC (Akaike information criteria) value. Subjects carrying the RET -5AA, -1CC, 135AA or 2307GG genotype had a 17.75-fold (95% CI = 7.34-42.92), 10.89-fold (95% CI = 3.13-37.85), 13.61-fold (95% CI = 6.14-30.14) or 9.79-fold (95% CI = 4.28-22.43) elevated risk for the development of HSCR, compared with the counterpart wild genotype, respectively. The much higher OR value in the present study and previously published association studies indicated that HSCR is a classic oligogenic disorder, which is different from the common complex diseases, such as sporadic cancers, cardiovascular diseases and type 2 diabetes with OR less than 2,. Because the frequencies of wild genotype for 1296A > G were low, we combined it with heterozygous-type genotype together for analysis. Similarly, the homozygous RET 1296GG genotype was also linked to risk of HSCR, with the OR of 2.40 (95% CI = 1.38 - 4.18) compared with the AA or AG genotype.\nAll the five markers in the controls conformed to HWE (P > 0.05, data not shown). The frequency of A allele in -5G > A, C allele in -1A > C, A allele in 135G > A, G allele in 1296A > G and G allele in 2307T > G in cases were 82.5%, 88.6%, 78.9%, 90.7% and 82.1%, which was higher than that of 45.5%, 63.1%, 43.8%, 80.4% and 49.7% in controls, with the P for Cochran-Armitage trend test of 1.12 × 10-16, 3.43 × 10-11, 4.20 × 10-15, 5.70 × 10-4, and 2.32 × 10-14, respectively. All those markers were shown to be strongly associated with the HSCR risk (Table 1). None of these SNPs were significantly different between LSA and SSA patients (data not shown).\nGenotype distribution of the RET gene polymorphisms in case-control\na p = 1.12 × 10-16, b p = 3.43 × 10-11, c p = 4.20 × 10-15, d p = 5.70 × 10-4, e p = 2.32 × 10-14, for Cochran-Armitage trend test.\nUnconditional logistic regression model was used to estimate RET SNPs effect (additive, dominant, or recessive) according to the smallest AIC (Akaike information criteria) value. Subjects carrying the RET -5AA, -1CC, 135AA or 2307GG genotype had a 17.75-fold (95% CI = 7.34-42.92), 10.89-fold (95% CI = 3.13-37.85), 13.61-fold (95% CI = 6.14-30.14) or 9.79-fold (95% CI = 4.28-22.43) elevated risk for the development of HSCR, compared with the counterpart wild genotype, respectively. The much higher OR value in the present study and previously published association studies indicated that HSCR is a classic oligogenic disorder, which is different from the common complex diseases, such as sporadic cancers, cardiovascular diseases and type 2 diabetes with OR less than 2,. Because the frequencies of wild genotype for 1296A > G were low, we combined it with heterozygous-type genotype together for analysis. Similarly, the homozygous RET 1296GG genotype was also linked to risk of HSCR, with the OR of 2.40 (95% CI = 1.38 - 4.18) compared with the AA or AG genotype.\n[SUBTITLE] RET haplotypes and risk of HSCR [SUBSECTION] We further analyzed the association of the haplotypes (MAF > 0.05) comprising the five HSCR-associated SNPs with HSCR risk. The haplotype and diplotype frequencies are presented in Table 2 and LD analysis showed that the 5 SNPs are part of the same haplotype block in Chinese (D' > 0.6). The haplotype A-C-A-G-G, was highly associated with an increased risk of HSCR (OR = 5.06, 95% CI = 1.97-12.99, P = 0.001). Intriguingly, the haplotype composited with more HSCR-risk allele associated with more elevated risk of HSCR, indicating there is a cumulative effect of these 5 SNPs on the risk of HSCR (P = 7.33 × 10-17, for Cochran-Armitage trend test). We also investigated the associations between RET diplotypes and risk of HSCR. Since the low frequencies of haplotype G-A-G-A-T and G-A-G-G-T, we combined them as reference in the following analysis. In consistent with the results from haplotype, the genetic risk of HSCR was increased in subjects with less copies of the G-A-G-A(G)-T haplotype and more copies of the A-C-A-G-G haplotype, with the P values for the Cochran-Armitage test of 3.09 × 10-15. Only diplotype with two copy of risk haplotype, A-C-A-G-G/A-C-A-G-G, was found to be associated with increased risk of HSCR (OR = 21.08, 95% CI = 5.28-84.09), when compared with wild type of diplotype, suggesting a haplotype-dosage effect in the genetic susceptibility of HSCR.\nFrequencies and counts of RET haplotypes and diplotypes comprising -5G > A, -1A > C, c135G > A, c1296A > G and c2307T > G\na haplotypes with frequencies less than 5% were not shown\nb p = 7.33 × 10-17, c p = 3.09 × 10-15, for Cochran-Armitage trend test\nd represented not G-A-G-A(G)-T or A-C-A-G-G haplotypes\nWe further analyzed the association of the haplotypes (MAF > 0.05) comprising the five HSCR-associated SNPs with HSCR risk. The haplotype and diplotype frequencies are presented in Table 2 and LD analysis showed that the 5 SNPs are part of the same haplotype block in Chinese (D' > 0.6). The haplotype A-C-A-G-G, was highly associated with an increased risk of HSCR (OR = 5.06, 95% CI = 1.97-12.99, P = 0.001). Intriguingly, the haplotype composited with more HSCR-risk allele associated with more elevated risk of HSCR, indicating there is a cumulative effect of these 5 SNPs on the risk of HSCR (P = 7.33 × 10-17, for Cochran-Armitage trend test). We also investigated the associations between RET diplotypes and risk of HSCR. Since the low frequencies of haplotype G-A-G-A-T and G-A-G-G-T, we combined them as reference in the following analysis. In consistent with the results from haplotype, the genetic risk of HSCR was increased in subjects with less copies of the G-A-G-A(G)-T haplotype and more copies of the A-C-A-G-G haplotype, with the P values for the Cochran-Armitage test of 3.09 × 10-15. Only diplotype with two copy of risk haplotype, A-C-A-G-G/A-C-A-G-G, was found to be associated with increased risk of HSCR (OR = 21.08, 95% CI = 5.28-84.09), when compared with wild type of diplotype, suggesting a haplotype-dosage effect in the genetic susceptibility of HSCR.\nFrequencies and counts of RET haplotypes and diplotypes comprising -5G > A, -1A > C, c135G > A, c1296A > G and c2307T > G\na haplotypes with frequencies less than 5% were not shown\nb p = 7.33 × 10-17, c p = 3.09 × 10-15, for Cochran-Armitage trend test\nd represented not G-A-G-A(G)-T or A-C-A-G-G haplotypes", "All the five markers in the controls conformed to HWE (P > 0.05, data not shown). The frequency of A allele in -5G > A, C allele in -1A > C, A allele in 135G > A, G allele in 1296A > G and G allele in 2307T > G in cases were 82.5%, 88.6%, 78.9%, 90.7% and 82.1%, which was higher than that of 45.5%, 63.1%, 43.8%, 80.4% and 49.7% in controls, with the P for Cochran-Armitage trend test of 1.12 × 10-16, 3.43 × 10-11, 4.20 × 10-15, 5.70 × 10-4, and 2.32 × 10-14, respectively. All those markers were shown to be strongly associated with the HSCR risk (Table 1). None of these SNPs were significantly different between LSA and SSA patients (data not shown).\nGenotype distribution of the RET gene polymorphisms in case-control\na p = 1.12 × 10-16, b p = 3.43 × 10-11, c p = 4.20 × 10-15, d p = 5.70 × 10-4, e p = 2.32 × 10-14, for Cochran-Armitage trend test.\nUnconditional logistic regression model was used to estimate RET SNPs effect (additive, dominant, or recessive) according to the smallest AIC (Akaike information criteria) value. Subjects carrying the RET -5AA, -1CC, 135AA or 2307GG genotype had a 17.75-fold (95% CI = 7.34-42.92), 10.89-fold (95% CI = 3.13-37.85), 13.61-fold (95% CI = 6.14-30.14) or 9.79-fold (95% CI = 4.28-22.43) elevated risk for the development of HSCR, compared with the counterpart wild genotype, respectively. The much higher OR value in the present study and previously published association studies indicated that HSCR is a classic oligogenic disorder, which is different from the common complex diseases, such as sporadic cancers, cardiovascular diseases and type 2 diabetes with OR less than 2,. Because the frequencies of wild genotype for 1296A > G were low, we combined it with heterozygous-type genotype together for analysis. Similarly, the homozygous RET 1296GG genotype was also linked to risk of HSCR, with the OR of 2.40 (95% CI = 1.38 - 4.18) compared with the AA or AG genotype.", "We further analyzed the association of the haplotypes (MAF > 0.05) comprising the five HSCR-associated SNPs with HSCR risk. The haplotype and diplotype frequencies are presented in Table 2 and LD analysis showed that the 5 SNPs are part of the same haplotype block in Chinese (D' > 0.6). The haplotype A-C-A-G-G, was highly associated with an increased risk of HSCR (OR = 5.06, 95% CI = 1.97-12.99, P = 0.001). Intriguingly, the haplotype composited with more HSCR-risk allele associated with more elevated risk of HSCR, indicating there is a cumulative effect of these 5 SNPs on the risk of HSCR (P = 7.33 × 10-17, for Cochran-Armitage trend test). We also investigated the associations between RET diplotypes and risk of HSCR. Since the low frequencies of haplotype G-A-G-A-T and G-A-G-G-T, we combined them as reference in the following analysis. In consistent with the results from haplotype, the genetic risk of HSCR was increased in subjects with less copies of the G-A-G-A(G)-T haplotype and more copies of the A-C-A-G-G haplotype, with the P values for the Cochran-Armitage test of 3.09 × 10-15. Only diplotype with two copy of risk haplotype, A-C-A-G-G/A-C-A-G-G, was found to be associated with increased risk of HSCR (OR = 21.08, 95% CI = 5.28-84.09), when compared with wild type of diplotype, suggesting a haplotype-dosage effect in the genetic susceptibility of HSCR.\nFrequencies and counts of RET haplotypes and diplotypes comprising -5G > A, -1A > C, c135G > A, c1296A > G and c2307T > G\na haplotypes with frequencies less than 5% were not shown\nb p = 7.33 × 10-17, c p = 3.09 × 10-15, for Cochran-Armitage trend test\nd represented not G-A-G-A(G)-T or A-C-A-G-G haplotypes", "This genetic epidemiological study investigated whether genetic polymorphisms in RET, alone or in combination, could have effects on the risk of HSCR in a Southeastern Chinese population. In this case-control study including 123 HSCR cases and 168 age- and sex- frequency-matched controls, we demonstrated that five polymorphisms in RET gene, including two SNPs in the regulatory region and three SNPs in the coding region, had substantial impact on risk of HSCR, separately or collectively. Subjects carrying the homozygous variant genotypes of -5G > A, -1A > C, c135G > A, c1296A > G or c2307T > G SNPs were at an increased risk for developing HSCR. Moreover, we also observed cumulative effects of these five polymorphisms on HSCR risk, demonstrating the haplotype composited with more HSCR-risk alleles rendered the hosts more susceptible to HSCR. To our best knowledge, this is the first study to investigate the contribution of the haplotypes of these five SNPs to the pathogenesis of sporadic HSCR in Chinese population. It lends supports to the previous findings that these SNPs were associated with the risk of HSCR [1,2,7,10,14-16].\nLiu CP et al. reported C135 G/A and C1296 A/G at exon 2 and the combination of AG of these two SNPs were associated with susceptibility of HSCR in Southeast Chinese population [17,18]. Also, they observed these two SNPs were in LD with two other coding SNPs at exon 14 and exon 15 in the same population. Our results expanded their findings and found that C135G/A and C1296 A/G were in LD with the two functional SNPs (-5G > A and -1A > C) in the promoter region and C2307T > G in exon 13. Furthermore, we observed 54.4% of the cases but only 8.4% of the controls carried A-C-A-G-G/A-C-A-G-G diplotype, with a more than 21-fold increased risk of HSCR, indicating a cumulative effect of the A-C-A-G-G combination on the susceptibility of HSCR and A-C-A-G-G may represent the core of the susceptibility allele.\nIntriguingly, we found that the frequencies of risk alleles of-5G > A, -1A > C, c135G > A, c1296A > G or c2307T > G in Chinese Han were much higher than those in European [19], partially explaining the relatively higher incidence of HSCR in Chinese. In conclusion, we observed a strong association between a haplotype, comprising the 3 most common SNPs in coding region and two in promoter region, and HSCR risk. These results support the hypothesis that common variations in RET pathway play an important role in pathogenesis of HSCR and might provide clues to develop screening and surveillance strategies.", "These results support the hypothesis that common variations in RET pathway might play an important role in development of HSCR.", "The authors declare that they have no competing interests.", "JT, LW, LL, YW and RZ designed the study; WL, QX, QG, SD, HY and HL collected blood samples and epidemiological data and performed the experiments; JT and LW performed the statistical analysis and wrote the paper.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2350/12/32/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study subjects", "Polymorphism Analysis", "Statistical analysis", "Results", "Association of RET SNPs with the risk of HSCR", "RET haplotypes and risk of HSCR", "Discussion", "Conclusion", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Hirschsprung's disease (HSCR) is a congenital disorder characterized by the absence of enteric ganglion cells in the myenteric (Auerbach's) and submucosal (Meissner's) plexuses of the variable lengths of gastrointestinal tract. The lack of enteric neurons is attributed to the defective proliferation, differentiation, survival, and migration of enteric nervous system (ENS) towards the end of the gut, whereby neural crest cells (NCC) fail to innervate the lower gastrointestinal tract during embryonic development, presenting in clinic with failure to pass meconium, bilious vomiting, enterocolitis-associated diarrhoea, chronic severe constipation and progressive abdominal distension in the neonatal period [1,2]. Based on the length of the aganglionic segment, it can be classified into three subgroups: short-segment aganglionosis (S-HSCR), long-segment aganglionosis (L-HSCR), and total colonic aganglionsis (TCA). Noteworthy, there is significant racial variation in the incidence of the disease worldwide, which is most prevalent in Asians, including China, about 2.8 per 10,000 live births. The male to female ration (M:F) is around 4:1 among S-HSCR patients and 1:1 among L-HSCR patients. Generally, HSCR most commonly presents sporadically although it can be familial and manifests with low, sex-dependent penetrance and phenotypic variability [2].\nHSCR is a heterogenic disorder with a number of genes reported to be associated, most of which involved in the signaling pathway implicated in the proper ENS development. RET, a receptor tyrosine kinase, is the major susceptibility gene for HSCR, which locates in 10q11.2 and encodes a transmembrane tyrosine kinase receptor, responsible for triggering a number of downstream signaling transductions [3-5]. There is growing evidence showing that some potentially functional single nucleotide polymorphisms (SNPs) of RET gene could act as low susceptibility factors and modify the phenotype of HSCR, especially in certain combinations of alleles, haplotypes. Specifically, three common SNPs in the coding region of RET, c135G > A in exon2 (rs1800858, A45A), c1296A > G (rs1800860, A432A) in exon7 and c2307T > G (rs1800861, L769L) in exon13, were firstly reported to be associated with HSCR [6-8]. Two promoter polymorphisms -5G > A and -1A > C (rs10900296 and rs10900297) were also reported to be strongly associated with HSCR and have a significantly lower activity in an in vitro dual-luciferase expression assay conducted by Fitze G and his colleagues [2,9]. Garcia-Barceló, et al [10] further demonstrated that these two SNPs may disrupt a transcriptional factor, TTF1, binding site and then reduce transcription capacity from the promoter. However, to our best knowledge, looking at these 5 particular SNPs all at once in haplotype profile in the Chinese Han population and their effects on the risk of HSCR have not been fully described.\nIn this study, we conducted a large case-control study to evaluate the contribution of the genotypes and haplotypes of these aforementioned five SNPs in the pathogenesis of sporadic HSCR in a Southeastern Chinese population.", "[SUBTITLE] Study subjects [SUBSECTION] This case-control study consisted of 123 sporadic HSCR (males: 75.6%) and 194 controls (males: 73.7%). Cases were histological diagnosed with either biopsy or surgical resection material for absence of enteric plexuses in Children's Hospital Affiliated to Zhejiang University Medical School, among which twelve (10%) of patients were affected with long-segment aganglionosis (LSA) and the rest with short-segment aganglionosis (SSA). Controls were normal children taking physical examination in the same hospital during the same time period as the patients were enrolled and frequency matched to cases by sex and age. Both cases and controls were Chinese Han in Zhejiang and informed consent was obtained from their parents. The study was approved by the Institutional Review Board of the Children's Hospital Affiliated to Zhejiang University Medical School.\nThis case-control study consisted of 123 sporadic HSCR (males: 75.6%) and 194 controls (males: 73.7%). Cases were histological diagnosed with either biopsy or surgical resection material for absence of enteric plexuses in Children's Hospital Affiliated to Zhejiang University Medical School, among which twelve (10%) of patients were affected with long-segment aganglionosis (LSA) and the rest with short-segment aganglionosis (SSA). Controls were normal children taking physical examination in the same hospital during the same time period as the patients were enrolled and frequency matched to cases by sex and age. Both cases and controls were Chinese Han in Zhejiang and informed consent was obtained from their parents. The study was approved by the Institutional Review Board of the Children's Hospital Affiliated to Zhejiang University Medical School.\n[SUBTITLE] Polymorphism Analysis [SUBSECTION] Genomic DNA was extracted from blood sample of all study subjects by using QIAamp-Blood Kit (Qiagen, Hilden, Germany). Genotypes were determined by PCR and direct sequencing without knowledge of the subjects' case/control status. Fifteen percent of the masked sample was randomly selected and genotyped twice by different people independently. The reproducibility was 100%.\nRET SNPs (-5G > A, -1A > C, c135G > A, c1296A > G, and c2307T > G) were genotyped as described previously [6,11,12]. PCR primers for RET -5G > A and -1A > C were 5'-CCCGCACTGAGCTCCTAC-3' and 5'-GGACGTCGCCTTCGCCAT-3', and 5'-TCT CGAGGGATGCTTACTGG-3' and 5'-GGCTCCGGTTAAGGTAGAGG-3' for c135G > A, 5'-CCCTGGTGGGGCATTTGTGC-3' and 5'-GGGTGGTTGGAAACTGATGG-3' for c1296 A > G and 5'-GAACTTGGGCAAGGCGATGC-3' and 5'-ACCCTGCAGCTGGCCTTAC-3' for c2307T > G, respectively. Amplication of the fragments was accomplished under a 25-ul reaction mixture with ~100 ng template DNA, 0.5 uM each primer, 0.2 mM each dNTP, 2.0 mM MgCl2 and 1.0 units of Tag DNA polymerase with 1 × reaction buffer (Applied Biosystems, Foster City, CA). The PCR profile consisted of an initial melting step of 5 minutes at 95°C, followed by 35 cycles of 40 seconds at 95°C, 30 seconds at 63°C (57°C for c135G > A and c2307T > G), and 45 second at 72°C, and a final elongation step of 7 minutes at 72°C. PCR products were then sequenced using an ABI PRISM® Big Dye™ Terminator v3.1 Cycle sequencing kit (Applied Biosystems, Foster City, CA) and an ABI 3130 automated sequencer according to the manufacturer's instructions.\nGenomic DNA was extracted from blood sample of all study subjects by using QIAamp-Blood Kit (Qiagen, Hilden, Germany). Genotypes were determined by PCR and direct sequencing without knowledge of the subjects' case/control status. Fifteen percent of the masked sample was randomly selected and genotyped twice by different people independently. The reproducibility was 100%.\nRET SNPs (-5G > A, -1A > C, c135G > A, c1296A > G, and c2307T > G) were genotyped as described previously [6,11,12]. PCR primers for RET -5G > A and -1A > C were 5'-CCCGCACTGAGCTCCTAC-3' and 5'-GGACGTCGCCTTCGCCAT-3', and 5'-TCT CGAGGGATGCTTACTGG-3' and 5'-GGCTCCGGTTAAGGTAGAGG-3' for c135G > A, 5'-CCCTGGTGGGGCATTTGTGC-3' and 5'-GGGTGGTTGGAAACTGATGG-3' for c1296 A > G and 5'-GAACTTGGGCAAGGCGATGC-3' and 5'-ACCCTGCAGCTGGCCTTAC-3' for c2307T > G, respectively. Amplication of the fragments was accomplished under a 25-ul reaction mixture with ~100 ng template DNA, 0.5 uM each primer, 0.2 mM each dNTP, 2.0 mM MgCl2 and 1.0 units of Tag DNA polymerase with 1 × reaction buffer (Applied Biosystems, Foster City, CA). The PCR profile consisted of an initial melting step of 5 minutes at 95°C, followed by 35 cycles of 40 seconds at 95°C, 30 seconds at 63°C (57°C for c135G > A and c2307T > G), and 45 second at 72°C, and a final elongation step of 7 minutes at 72°C. PCR products were then sequenced using an ABI PRISM® Big Dye™ Terminator v3.1 Cycle sequencing kit (Applied Biosystems, Foster City, CA) and an ABI 3130 automated sequencer according to the manufacturer's instructions.\n[SUBTITLE] Statistical analysis [SUBSECTION] Chi-square test was used to compare the allele and genotype frequencies between cases and controls, as well as LSA and SSA. The associations between the genotype and the risk of HSCR were estimated by odds ratios (ORs) and their 95% confidence interval (CI), which was calculated by unconditional logistic regression adjusted for gender. PHASE v2.1 [13] was used to estimate the haplotype frequencies composed of five SNP pairs, respectively, and their difference between patients and controls. Diplotype were defined by two haplotypes of each individual. Bonferroni adjustment was used to adjust multiple comparison and p values were considered significantly at a level of 0.003, to allow for the testing of 5 partially correlated markers using 3 different inheritance model (dominant, recessive, and additive model). All of the statistical analyses were performed using Statistical Analysis System software (version 9.12; SAS Institute, Cary, NC).\nChi-square test was used to compare the allele and genotype frequencies between cases and controls, as well as LSA and SSA. The associations between the genotype and the risk of HSCR were estimated by odds ratios (ORs) and their 95% confidence interval (CI), which was calculated by unconditional logistic regression adjusted for gender. PHASE v2.1 [13] was used to estimate the haplotype frequencies composed of five SNP pairs, respectively, and their difference between patients and controls. Diplotype were defined by two haplotypes of each individual. Bonferroni adjustment was used to adjust multiple comparison and p values were considered significantly at a level of 0.003, to allow for the testing of 5 partially correlated markers using 3 different inheritance model (dominant, recessive, and additive model). All of the statistical analyses were performed using Statistical Analysis System software (version 9.12; SAS Institute, Cary, NC).", "This case-control study consisted of 123 sporadic HSCR (males: 75.6%) and 194 controls (males: 73.7%). Cases were histological diagnosed with either biopsy or surgical resection material for absence of enteric plexuses in Children's Hospital Affiliated to Zhejiang University Medical School, among which twelve (10%) of patients were affected with long-segment aganglionosis (LSA) and the rest with short-segment aganglionosis (SSA). Controls were normal children taking physical examination in the same hospital during the same time period as the patients were enrolled and frequency matched to cases by sex and age. Both cases and controls were Chinese Han in Zhejiang and informed consent was obtained from their parents. The study was approved by the Institutional Review Board of the Children's Hospital Affiliated to Zhejiang University Medical School.", "Genomic DNA was extracted from blood sample of all study subjects by using QIAamp-Blood Kit (Qiagen, Hilden, Germany). Genotypes were determined by PCR and direct sequencing without knowledge of the subjects' case/control status. Fifteen percent of the masked sample was randomly selected and genotyped twice by different people independently. The reproducibility was 100%.\nRET SNPs (-5G > A, -1A > C, c135G > A, c1296A > G, and c2307T > G) were genotyped as described previously [6,11,12]. PCR primers for RET -5G > A and -1A > C were 5'-CCCGCACTGAGCTCCTAC-3' and 5'-GGACGTCGCCTTCGCCAT-3', and 5'-TCT CGAGGGATGCTTACTGG-3' and 5'-GGCTCCGGTTAAGGTAGAGG-3' for c135G > A, 5'-CCCTGGTGGGGCATTTGTGC-3' and 5'-GGGTGGTTGGAAACTGATGG-3' for c1296 A > G and 5'-GAACTTGGGCAAGGCGATGC-3' and 5'-ACCCTGCAGCTGGCCTTAC-3' for c2307T > G, respectively. Amplication of the fragments was accomplished under a 25-ul reaction mixture with ~100 ng template DNA, 0.5 uM each primer, 0.2 mM each dNTP, 2.0 mM MgCl2 and 1.0 units of Tag DNA polymerase with 1 × reaction buffer (Applied Biosystems, Foster City, CA). The PCR profile consisted of an initial melting step of 5 minutes at 95°C, followed by 35 cycles of 40 seconds at 95°C, 30 seconds at 63°C (57°C for c135G > A and c2307T > G), and 45 second at 72°C, and a final elongation step of 7 minutes at 72°C. PCR products were then sequenced using an ABI PRISM® Big Dye™ Terminator v3.1 Cycle sequencing kit (Applied Biosystems, Foster City, CA) and an ABI 3130 automated sequencer according to the manufacturer's instructions.", "Chi-square test was used to compare the allele and genotype frequencies between cases and controls, as well as LSA and SSA. The associations between the genotype and the risk of HSCR were estimated by odds ratios (ORs) and their 95% confidence interval (CI), which was calculated by unconditional logistic regression adjusted for gender. PHASE v2.1 [13] was used to estimate the haplotype frequencies composed of five SNP pairs, respectively, and their difference between patients and controls. Diplotype were defined by two haplotypes of each individual. Bonferroni adjustment was used to adjust multiple comparison and p values were considered significantly at a level of 0.003, to allow for the testing of 5 partially correlated markers using 3 different inheritance model (dominant, recessive, and additive model). All of the statistical analyses were performed using Statistical Analysis System software (version 9.12; SAS Institute, Cary, NC).", "[SUBTITLE] Association of RET SNPs with the risk of HSCR [SUBSECTION] All the five markers in the controls conformed to HWE (P > 0.05, data not shown). The frequency of A allele in -5G > A, C allele in -1A > C, A allele in 135G > A, G allele in 1296A > G and G allele in 2307T > G in cases were 82.5%, 88.6%, 78.9%, 90.7% and 82.1%, which was higher than that of 45.5%, 63.1%, 43.8%, 80.4% and 49.7% in controls, with the P for Cochran-Armitage trend test of 1.12 × 10-16, 3.43 × 10-11, 4.20 × 10-15, 5.70 × 10-4, and 2.32 × 10-14, respectively. All those markers were shown to be strongly associated with the HSCR risk (Table 1). None of these SNPs were significantly different between LSA and SSA patients (data not shown).\nGenotype distribution of the RET gene polymorphisms in case-control\na p = 1.12 × 10-16, b p = 3.43 × 10-11, c p = 4.20 × 10-15, d p = 5.70 × 10-4, e p = 2.32 × 10-14, for Cochran-Armitage trend test.\nUnconditional logistic regression model was used to estimate RET SNPs effect (additive, dominant, or recessive) according to the smallest AIC (Akaike information criteria) value. Subjects carrying the RET -5AA, -1CC, 135AA or 2307GG genotype had a 17.75-fold (95% CI = 7.34-42.92), 10.89-fold (95% CI = 3.13-37.85), 13.61-fold (95% CI = 6.14-30.14) or 9.79-fold (95% CI = 4.28-22.43) elevated risk for the development of HSCR, compared with the counterpart wild genotype, respectively. The much higher OR value in the present study and previously published association studies indicated that HSCR is a classic oligogenic disorder, which is different from the common complex diseases, such as sporadic cancers, cardiovascular diseases and type 2 diabetes with OR less than 2,. Because the frequencies of wild genotype for 1296A > G were low, we combined it with heterozygous-type genotype together for analysis. Similarly, the homozygous RET 1296GG genotype was also linked to risk of HSCR, with the OR of 2.40 (95% CI = 1.38 - 4.18) compared with the AA or AG genotype.\nAll the five markers in the controls conformed to HWE (P > 0.05, data not shown). The frequency of A allele in -5G > A, C allele in -1A > C, A allele in 135G > A, G allele in 1296A > G and G allele in 2307T > G in cases were 82.5%, 88.6%, 78.9%, 90.7% and 82.1%, which was higher than that of 45.5%, 63.1%, 43.8%, 80.4% and 49.7% in controls, with the P for Cochran-Armitage trend test of 1.12 × 10-16, 3.43 × 10-11, 4.20 × 10-15, 5.70 × 10-4, and 2.32 × 10-14, respectively. All those markers were shown to be strongly associated with the HSCR risk (Table 1). None of these SNPs were significantly different between LSA and SSA patients (data not shown).\nGenotype distribution of the RET gene polymorphisms in case-control\na p = 1.12 × 10-16, b p = 3.43 × 10-11, c p = 4.20 × 10-15, d p = 5.70 × 10-4, e p = 2.32 × 10-14, for Cochran-Armitage trend test.\nUnconditional logistic regression model was used to estimate RET SNPs effect (additive, dominant, or recessive) according to the smallest AIC (Akaike information criteria) value. Subjects carrying the RET -5AA, -1CC, 135AA or 2307GG genotype had a 17.75-fold (95% CI = 7.34-42.92), 10.89-fold (95% CI = 3.13-37.85), 13.61-fold (95% CI = 6.14-30.14) or 9.79-fold (95% CI = 4.28-22.43) elevated risk for the development of HSCR, compared with the counterpart wild genotype, respectively. The much higher OR value in the present study and previously published association studies indicated that HSCR is a classic oligogenic disorder, which is different from the common complex diseases, such as sporadic cancers, cardiovascular diseases and type 2 diabetes with OR less than 2,. Because the frequencies of wild genotype for 1296A > G were low, we combined it with heterozygous-type genotype together for analysis. Similarly, the homozygous RET 1296GG genotype was also linked to risk of HSCR, with the OR of 2.40 (95% CI = 1.38 - 4.18) compared with the AA or AG genotype.\n[SUBTITLE] RET haplotypes and risk of HSCR [SUBSECTION] We further analyzed the association of the haplotypes (MAF > 0.05) comprising the five HSCR-associated SNPs with HSCR risk. The haplotype and diplotype frequencies are presented in Table 2 and LD analysis showed that the 5 SNPs are part of the same haplotype block in Chinese (D' > 0.6). The haplotype A-C-A-G-G, was highly associated with an increased risk of HSCR (OR = 5.06, 95% CI = 1.97-12.99, P = 0.001). Intriguingly, the haplotype composited with more HSCR-risk allele associated with more elevated risk of HSCR, indicating there is a cumulative effect of these 5 SNPs on the risk of HSCR (P = 7.33 × 10-17, for Cochran-Armitage trend test). We also investigated the associations between RET diplotypes and risk of HSCR. Since the low frequencies of haplotype G-A-G-A-T and G-A-G-G-T, we combined them as reference in the following analysis. In consistent with the results from haplotype, the genetic risk of HSCR was increased in subjects with less copies of the G-A-G-A(G)-T haplotype and more copies of the A-C-A-G-G haplotype, with the P values for the Cochran-Armitage test of 3.09 × 10-15. Only diplotype with two copy of risk haplotype, A-C-A-G-G/A-C-A-G-G, was found to be associated with increased risk of HSCR (OR = 21.08, 95% CI = 5.28-84.09), when compared with wild type of diplotype, suggesting a haplotype-dosage effect in the genetic susceptibility of HSCR.\nFrequencies and counts of RET haplotypes and diplotypes comprising -5G > A, -1A > C, c135G > A, c1296A > G and c2307T > G\na haplotypes with frequencies less than 5% were not shown\nb p = 7.33 × 10-17, c p = 3.09 × 10-15, for Cochran-Armitage trend test\nd represented not G-A-G-A(G)-T or A-C-A-G-G haplotypes\nWe further analyzed the association of the haplotypes (MAF > 0.05) comprising the five HSCR-associated SNPs with HSCR risk. The haplotype and diplotype frequencies are presented in Table 2 and LD analysis showed that the 5 SNPs are part of the same haplotype block in Chinese (D' > 0.6). The haplotype A-C-A-G-G, was highly associated with an increased risk of HSCR (OR = 5.06, 95% CI = 1.97-12.99, P = 0.001). Intriguingly, the haplotype composited with more HSCR-risk allele associated with more elevated risk of HSCR, indicating there is a cumulative effect of these 5 SNPs on the risk of HSCR (P = 7.33 × 10-17, for Cochran-Armitage trend test). We also investigated the associations between RET diplotypes and risk of HSCR. Since the low frequencies of haplotype G-A-G-A-T and G-A-G-G-T, we combined them as reference in the following analysis. In consistent with the results from haplotype, the genetic risk of HSCR was increased in subjects with less copies of the G-A-G-A(G)-T haplotype and more copies of the A-C-A-G-G haplotype, with the P values for the Cochran-Armitage test of 3.09 × 10-15. Only diplotype with two copy of risk haplotype, A-C-A-G-G/A-C-A-G-G, was found to be associated with increased risk of HSCR (OR = 21.08, 95% CI = 5.28-84.09), when compared with wild type of diplotype, suggesting a haplotype-dosage effect in the genetic susceptibility of HSCR.\nFrequencies and counts of RET haplotypes and diplotypes comprising -5G > A, -1A > C, c135G > A, c1296A > G and c2307T > G\na haplotypes with frequencies less than 5% were not shown\nb p = 7.33 × 10-17, c p = 3.09 × 10-15, for Cochran-Armitage trend test\nd represented not G-A-G-A(G)-T or A-C-A-G-G haplotypes", "All the five markers in the controls conformed to HWE (P > 0.05, data not shown). The frequency of A allele in -5G > A, C allele in -1A > C, A allele in 135G > A, G allele in 1296A > G and G allele in 2307T > G in cases were 82.5%, 88.6%, 78.9%, 90.7% and 82.1%, which was higher than that of 45.5%, 63.1%, 43.8%, 80.4% and 49.7% in controls, with the P for Cochran-Armitage trend test of 1.12 × 10-16, 3.43 × 10-11, 4.20 × 10-15, 5.70 × 10-4, and 2.32 × 10-14, respectively. All those markers were shown to be strongly associated with the HSCR risk (Table 1). None of these SNPs were significantly different between LSA and SSA patients (data not shown).\nGenotype distribution of the RET gene polymorphisms in case-control\na p = 1.12 × 10-16, b p = 3.43 × 10-11, c p = 4.20 × 10-15, d p = 5.70 × 10-4, e p = 2.32 × 10-14, for Cochran-Armitage trend test.\nUnconditional logistic regression model was used to estimate RET SNPs effect (additive, dominant, or recessive) according to the smallest AIC (Akaike information criteria) value. Subjects carrying the RET -5AA, -1CC, 135AA or 2307GG genotype had a 17.75-fold (95% CI = 7.34-42.92), 10.89-fold (95% CI = 3.13-37.85), 13.61-fold (95% CI = 6.14-30.14) or 9.79-fold (95% CI = 4.28-22.43) elevated risk for the development of HSCR, compared with the counterpart wild genotype, respectively. The much higher OR value in the present study and previously published association studies indicated that HSCR is a classic oligogenic disorder, which is different from the common complex diseases, such as sporadic cancers, cardiovascular diseases and type 2 diabetes with OR less than 2,. Because the frequencies of wild genotype for 1296A > G were low, we combined it with heterozygous-type genotype together for analysis. Similarly, the homozygous RET 1296GG genotype was also linked to risk of HSCR, with the OR of 2.40 (95% CI = 1.38 - 4.18) compared with the AA or AG genotype.", "We further analyzed the association of the haplotypes (MAF > 0.05) comprising the five HSCR-associated SNPs with HSCR risk. The haplotype and diplotype frequencies are presented in Table 2 and LD analysis showed that the 5 SNPs are part of the same haplotype block in Chinese (D' > 0.6). The haplotype A-C-A-G-G, was highly associated with an increased risk of HSCR (OR = 5.06, 95% CI = 1.97-12.99, P = 0.001). Intriguingly, the haplotype composited with more HSCR-risk allele associated with more elevated risk of HSCR, indicating there is a cumulative effect of these 5 SNPs on the risk of HSCR (P = 7.33 × 10-17, for Cochran-Armitage trend test). We also investigated the associations between RET diplotypes and risk of HSCR. Since the low frequencies of haplotype G-A-G-A-T and G-A-G-G-T, we combined them as reference in the following analysis. In consistent with the results from haplotype, the genetic risk of HSCR was increased in subjects with less copies of the G-A-G-A(G)-T haplotype and more copies of the A-C-A-G-G haplotype, with the P values for the Cochran-Armitage test of 3.09 × 10-15. Only diplotype with two copy of risk haplotype, A-C-A-G-G/A-C-A-G-G, was found to be associated with increased risk of HSCR (OR = 21.08, 95% CI = 5.28-84.09), when compared with wild type of diplotype, suggesting a haplotype-dosage effect in the genetic susceptibility of HSCR.\nFrequencies and counts of RET haplotypes and diplotypes comprising -5G > A, -1A > C, c135G > A, c1296A > G and c2307T > G\na haplotypes with frequencies less than 5% were not shown\nb p = 7.33 × 10-17, c p = 3.09 × 10-15, for Cochran-Armitage trend test\nd represented not G-A-G-A(G)-T or A-C-A-G-G haplotypes", "This genetic epidemiological study investigated whether genetic polymorphisms in RET, alone or in combination, could have effects on the risk of HSCR in a Southeastern Chinese population. In this case-control study including 123 HSCR cases and 168 age- and sex- frequency-matched controls, we demonstrated that five polymorphisms in RET gene, including two SNPs in the regulatory region and three SNPs in the coding region, had substantial impact on risk of HSCR, separately or collectively. Subjects carrying the homozygous variant genotypes of -5G > A, -1A > C, c135G > A, c1296A > G or c2307T > G SNPs were at an increased risk for developing HSCR. Moreover, we also observed cumulative effects of these five polymorphisms on HSCR risk, demonstrating the haplotype composited with more HSCR-risk alleles rendered the hosts more susceptible to HSCR. To our best knowledge, this is the first study to investigate the contribution of the haplotypes of these five SNPs to the pathogenesis of sporadic HSCR in Chinese population. It lends supports to the previous findings that these SNPs were associated with the risk of HSCR [1,2,7,10,14-16].\nLiu CP et al. reported C135 G/A and C1296 A/G at exon 2 and the combination of AG of these two SNPs were associated with susceptibility of HSCR in Southeast Chinese population [17,18]. Also, they observed these two SNPs were in LD with two other coding SNPs at exon 14 and exon 15 in the same population. Our results expanded their findings and found that C135G/A and C1296 A/G were in LD with the two functional SNPs (-5G > A and -1A > C) in the promoter region and C2307T > G in exon 13. Furthermore, we observed 54.4% of the cases but only 8.4% of the controls carried A-C-A-G-G/A-C-A-G-G diplotype, with a more than 21-fold increased risk of HSCR, indicating a cumulative effect of the A-C-A-G-G combination on the susceptibility of HSCR and A-C-A-G-G may represent the core of the susceptibility allele.\nIntriguingly, we found that the frequencies of risk alleles of-5G > A, -1A > C, c135G > A, c1296A > G or c2307T > G in Chinese Han were much higher than those in European [19], partially explaining the relatively higher incidence of HSCR in Chinese. In conclusion, we observed a strong association between a haplotype, comprising the 3 most common SNPs in coding region and two in promoter region, and HSCR risk. These results support the hypothesis that common variations in RET pathway play an important role in pathogenesis of HSCR and might provide clues to develop screening and surveillance strategies.", "These results support the hypothesis that common variations in RET pathway might play an important role in development of HSCR.", "The authors declare that they have no competing interests.", "JT, LW, LL, YW and RZ designed the study; WL, QX, QG, SD, HY and HL collected blood samples and epidemiological data and performed the experiments; JT and LW performed the statistical analysis and wrote the paper.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2350/12/32/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adult", "China", "Data Collection", "Female", "Humans", "Interviews as Topic", "Male", "Middle Aged", "Persuasive Communication", "Product Labeling", "Public Opinion", "Smoking", "Smoking Prevention", "Young Adult" ]

[ "Background", "Study sites", "Participants", "Smoking status and demographic variables", "Warning labels of cigarette packages", "The harm warning provided by the label", "The perceived impact of giving cigarettes as a gift", "The perceived impact on the decision to quit smoking", "Knowledge of the FCTC and its provision for cigarette packaging", "Statistic analysis", "Results", "General information", "The harm warning provided by the label", "The perceived impact of giving cigarettes as a gift", "The perceived impact on the decision to quit smoking", "Knowledge of the WHO FCTC and its provision for cigarette packaging", "Discussion and Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "China is the largest producer and consumer of tobacco in the world. The prevalence of cigarette smoking is above 60% for men aged 15 and above, and 50% of women in this age group and adolescent are passive smokers [1]. Tobacco smoking was estimated as being responsible for approximately 67,300 premature deaths in Chinese adults aged 40 years and over in 2005 [2]. The premature deaths, productivity losses, and a substantial number of preventable diseases and health care costs related to smoking have resulted in a significant economic burden in China [3-7]. However, few people are aware of the harm caused by smoking and passive smoking [8]. Instead, cigarettes are considered a good vehicle for communication, and a popular gift to relatives or friends in China, especially for holidays.\nThe World Health Organization Framework Convention on Tobacco Control (WHO FCTC), the world's first public health treaty, calls for warning labels to be displayed as large and clear health warnings covering 30% to 50% of the package in the form of pictures, pictograms or text. Every person should be informed of the health consequences, addictive nature, and mortal threat posed by tobacco use and exposure to tobacco smoke. This can be achieved by the warning labels on cigarette packages [9]. In 1991, the Chinese Congress enacted legislation requiring cigarette warnings to state 'smoking is harmful to your health' in Chinese. This warning appeared on one of the side panels of every cigarette package [10]. On 9th of January 2006, WHO FCTC was ratified in China. In 2008, China implemented new regulations according to the FCTC and the legislation. Cigarette warnings were moved from the side panels and covered at least 30% of the front and back of the pack, with Chinese on the front, and English on the back. The warnings include two rotating sets of text in Chinese and English. One set states 'smoking is harmful to your health', and 'quitting smoking reduces health risk'. Another states 'smoking is harmful to your health', and 'quitting smoking early is good for your health' [10].\nThe purpose of the current study was primarily to compare the difference in reactions to different types of warning labels on cigarette packages with a specific focus on whether the new warning label is better than the previous one and labels used abroad.", "The study was conducted in 2008 in Nantong and Zhangjiagang cities, Jiangsu Province, one of the economically booming areas in Eastern part of China. Nantong is an urban city, and is a moderate developed city in the Province. Zhangjiagang is a rural area, and belongs to Suzhou City, a highly developed city in the Province.", "Eligible study participants included in this survey were those aged 18 years and over and those working in hospitals, schools, bus/train stations, government offices, restaurants and bars. Altogether 1000 adults were approached, and 876 participants agreed to participate and finished the survey. All participants were asked to complete a face-to-face interview using a standard questionnaire; informed consent was sought prior to the interview being undertaken. The study was approved by the ethical board of Jiangsu Provincial Centre for Disease Control and Prevention. Verbal consent was obtained from each participant.", "Information was obtained from all participants about their smoking status. Smoking status was measured by asking if participants had ever smoked. Participants were grouped into two categories smokers and former smokers. Smokers were defined as those having smoked at least 100 cigarettes in their lifetime, and those having smoked at least one cigarette per day at the time of the survey. Former smokers were defined as individuals who had quit smoking at least one month prior to the survey and those who smoked at least one cigarette per day prior to quitting. Participants also reported their gender, age, and education level.", "Six warning labels were included in the interview questionnaire (Additional file 1). They were coded A-F. Label A was the old Chinese warning label, with 'smoking is harmful to your health' written in Chinese on one of the side panels of the pack (Figure 1). Label B was the new Chinese warning label, with 'smoking is harmful to your health', and 'quitting smoking reduces health risk' written on the front and back faces of the package, in Chinese and English respectively. Label C was a famous Chinese brand which is exported abroad and includes text, pictorials, and quitline information on the whole front face and 1/3 back face, respectively. The text on the pack reads 'smoking damages your blood vessels, which can prevent blood circulation, particularly to your legs or feet. This can result in blood clots, infection, gangrene, even amputation.' The other labels were foreign brands. Label D warned that cigarette smoking can result in mouth and oropharynx cancers by using text and pictures on 50% of the cigarette package. Label E, also used text and pictures to show that smoking when pregnant harms your baby. Label F indicated that smoking can lead to laryngeal cancer using both text and pictorials on 50% of the cigarette package. All English health warnings were translated into Chinese during the interview.\nSix cigarette warning labels. Label A was the old Chinese warning label, with 'smoking is harmful to your health' in Chinese on one of the side panels of the pack. Label B was the new Chinese warning label, with 'smoking is harmful to your health', and 'quit smoking reduces health risk' on the front and back faces of the package, in Chinese and English respectively. Label C was a famous Chinese brand exported abroad, with text, pictorials, and quitline on the whole front face and 1/3 back face, respectively. The text said 'smoking damages your blood vessels, which can prevent blood circulation, particularly to your legs or feet. This can result in blood clots, infection, gangrene, even amputation.' The other labels were foreign brands. Label D warned that cigarette smoking can result in mouth and oropharynx cancers in text and pictures on 50% of cigarette package. Label E showed smoking when pregnant harms your baby in large area with text and pictures. Label F indicated that smoking can lead to laryngeal cancer in text and pictorial on 50% of cigarette package.", "Making reference to Labels A to F, participants were asked if the each label gave them clear information on the harm which cigarette smoking can have on health and the specific diseases that occur related to cigarette smoking. Participants were also asked if Labels C, D, E, and F gave them clear information on specific diseases smoking can cause (as described above).", "Three questions on the perceived impact of giving cigarettes as a gift were presented. These included: 1) If you want to use cigarettes as a gift, do the following cigarette labels (A-F) make you change your mind and not do so? 2) If you want to give cigarettes as a gift, which warning label is least likely to stop you using cigarettes as a gift? 3) If you want to give cigarettes as a gift, which warning label is most likely to stop you using cigarettes as a gift?", "Participants were asked three questions about the perceived impact of quitting smoking. These include: 1) If you were a cigarette smoker, would the following labels (A-F) make you want to quit smoking? 2) If you were a cigarette smoker, which warning label is most likely to cause you to quit? 3) If you were a cigarette smoker, which warning label is least likely to cause you to quit?", "Participants were asked if they knew that China had ratified the WHO FCTC. If they answered yes, participants were then asked if they were aware of the FCTC provision that health warnings on cigarette packaging should be large, clear, visible and legible.", "Univariate and bivariate analyses were conducted to examine how much impact each of the different cigarette warning labels had and the knowledge of the FCTC by age groups, gender, education levels and smoking status. To compare the new Chinese label with international labels, Label C, D, E, and F were aggregated into one group. Chi-square tests were used to assess differences among groups where appropriate. All analyses were conducted using SPSS 13.0 (SPSS Inc., Chicago, IL, USA).", "[SUBTITLE] General information [SUBSECTION] Table 1 demonstrates the demographic characteristics of the sample. A total of 876 participants (374 male and 502 female) were involved in the study. The average age was 34.0 ± 11.0 years, a higher proportion of males reported that they were current smokers compared to females and 82.7% of participants had graduated from technical secondary school or higher.\nCharacteristics of the study sample\nTable 1 demonstrates the demographic characteristics of the sample. A total of 876 participants (374 male and 502 female) were involved in the study. The average age was 34.0 ± 11.0 years, a higher proportion of males reported that they were current smokers compared to females and 82.7% of participants had graduated from technical secondary school or higher.\nCharacteristics of the study sample\n[SUBTITLE] The harm warning provided by the label [SUBSECTION] Of the participants, 18.3% said Label A provided adequate information on the harm of cigarette smoking. Among them, 16.5% (14/85) of participants said Label A gave adequate information on the relationship between cigarette smoking and respiratory diseases, including lung cancer, and 16.5% and 3.5% respectively mentioned cancer and cardiovascular diseases. Overall, 31.2% said Label B gave adequate information on the harm of cigarette smoking. Among them, 36.6% said Label B provided adequate information on the relationship between cigarette smoking and respiratory diseases, 5.3% and 3.1% could identify the relationship of smoking with cancer and cardiovascular diseases respectively based on the information on label B. Similar percentage of participants said Label C-F gave adequate information on the harm of cigarette smoking, 90.5% for Label C, 92.7% for Label D, 92.4% for Label E and 92.7% for Label F. Compared to Label A, a higher proportion of participants said that Label B gave them clearer information on the harm of smoking across all subcategories, except those with low education level and current smokers. Labels C-F performed better than Label B in providing harm information for all sub-groups (Table 2).\nThe proportion of positive responses to the harm information provided by different cigarette labels by gender, age groups, educational levels and smoking status\nA positive answer means participants can understand the harm information provided by the label.\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\nOf the participants, 18.3% said Label A provided adequate information on the harm of cigarette smoking. Among them, 16.5% (14/85) of participants said Label A gave adequate information on the relationship between cigarette smoking and respiratory diseases, including lung cancer, and 16.5% and 3.5% respectively mentioned cancer and cardiovascular diseases. Overall, 31.2% said Label B gave adequate information on the harm of cigarette smoking. Among them, 36.6% said Label B provided adequate information on the relationship between cigarette smoking and respiratory diseases, 5.3% and 3.1% could identify the relationship of smoking with cancer and cardiovascular diseases respectively based on the information on label B. Similar percentage of participants said Label C-F gave adequate information on the harm of cigarette smoking, 90.5% for Label C, 92.7% for Label D, 92.4% for Label E and 92.7% for Label F. Compared to Label A, a higher proportion of participants said that Label B gave them clearer information on the harm of smoking across all subcategories, except those with low education level and current smokers. Labels C-F performed better than Label B in providing harm information for all sub-groups (Table 2).\nThe proportion of positive responses to the harm information provided by different cigarette labels by gender, age groups, educational levels and smoking status\nA positive answer means participants can understand the harm information provided by the label.\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\n[SUBTITLE] The perceived impact of giving cigarettes as a gift [SUBSECTION] Of the participants, 20.8% and 25.2% reported that they would not give cigarettes as a gift to somebody with Labels A and B (respectively) on the package. Over 80% of participants refused to give cigarettes as a gift if the package displayed warning Labels C-F. The proportion of those who would not give cigarettes as gift was higher among female, those who had never smoked and those having a higher educational level. Generally, there was no difference between the sub-groups, in terms of those who would not give cigarettes as a gift, for Label A and Label B, except the proportions were marginally higher among non-smokers and those aged between 30-40 for Label B. When comparing Label B to the combined labels, the proportion of respondents who would not give cigarettes as a gift was higher if any of Labels C-F were on the package (Table 3). The majority of participants (70.4%) considered that Label A was least likely to stop them using cigarettes as a gift, and the proportion was 20.2% for Label B. Almost half of participants (46.8%) considered that Label C was most likely to stop them using cigarettes as a gift, and the proportion was 5.7% and 3.4% for Label A and Label B respectively.\nThe perceived impact of not giving cigarette as a gift by gender, age groups, education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\nOf the participants, 20.8% and 25.2% reported that they would not give cigarettes as a gift to somebody with Labels A and B (respectively) on the package. Over 80% of participants refused to give cigarettes as a gift if the package displayed warning Labels C-F. The proportion of those who would not give cigarettes as gift was higher among female, those who had never smoked and those having a higher educational level. Generally, there was no difference between the sub-groups, in terms of those who would not give cigarettes as a gift, for Label A and Label B, except the proportions were marginally higher among non-smokers and those aged between 30-40 for Label B. When comparing Label B to the combined labels, the proportion of respondents who would not give cigarettes as a gift was higher if any of Labels C-F were on the package (Table 3). The majority of participants (70.4%) considered that Label A was least likely to stop them using cigarettes as a gift, and the proportion was 20.2% for Label B. Almost half of participants (46.8%) considered that Label C was most likely to stop them using cigarettes as a gift, and the proportion was 5.7% and 3.4% for Label A and Label B respectively.\nThe perceived impact of not giving cigarette as a gift by gender, age groups, education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\n[SUBTITLE] The perceived impact on the decision to quit smoking [SUBSECTION] There were 26.8% and 31.5% of the participants who reported thinking about quitting due to warning Label A and Label B, respectively. In addition, the proportions were all above 80% for Labels C-F. We asked non-smokers if they were smokers, if the labels would impact on a decision to quit smoking. Non-smokers were more likely to quit smoking due to Label C-F, compared to those who were smokers. It was shown that due to the warning on Label B, those more likely to quit were females, those with higher educational level and non-smokers when compared to Label A. Label B was less likely to make the participants quit smoking compared to Labels C-F combined (Table 4). Almost half of participants (43.3%) considered that Label C was most likely to cause them to quit. The proportion was only 4.5% and 3.7% for Label A and B, respectively. The majority of participants (69.9%) considered that Label A was least likely to cause them to quit, and the proportion was 20.2% for Label B. There was no significant difference between smoking status groups in terms of the impact Label A and Label B had on a decision on quit smoking.\nThe perceived impact on the decision to quit smoking by gender, age groups education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\nThere were 26.8% and 31.5% of the participants who reported thinking about quitting due to warning Label A and Label B, respectively. In addition, the proportions were all above 80% for Labels C-F. We asked non-smokers if they were smokers, if the labels would impact on a decision to quit smoking. Non-smokers were more likely to quit smoking due to Label C-F, compared to those who were smokers. It was shown that due to the warning on Label B, those more likely to quit were females, those with higher educational level and non-smokers when compared to Label A. Label B was less likely to make the participants quit smoking compared to Labels C-F combined (Table 4). Almost half of participants (43.3%) considered that Label C was most likely to cause them to quit. The proportion was only 4.5% and 3.7% for Label A and B, respectively. The majority of participants (69.9%) considered that Label A was least likely to cause them to quit, and the proportion was 20.2% for Label B. There was no significant difference between smoking status groups in terms of the impact Label A and Label B had on a decision on quit smoking.\nThe perceived impact on the decision to quit smoking by gender, age groups education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\n[SUBTITLE] Knowledge of the WHO FCTC and its provision for cigarette packaging [SUBSECTION] Overall 32.4% of participants knew of the FCTC. Among them, 77.1% and 72.4% were non-smokers and those with the highest educational level, respectively. Furthermore, 75.4% (214/284) knew that China have ratified the FCTC, and 77.5% knew the provision of the FCTC that health warnings on cigarette packaging should be large, clear, visible and legible.\nOverall 32.4% of participants knew of the FCTC. Among them, 77.1% and 72.4% were non-smokers and those with the highest educational level, respectively. Furthermore, 75.4% (214/284) knew that China have ratified the FCTC, and 77.5% knew the provision of the FCTC that health warnings on cigarette packaging should be large, clear, visible and legible.", "Table 1 demonstrates the demographic characteristics of the sample. A total of 876 participants (374 male and 502 female) were involved in the study. The average age was 34.0 ± 11.0 years, a higher proportion of males reported that they were current smokers compared to females and 82.7% of participants had graduated from technical secondary school or higher.\nCharacteristics of the study sample", "Of the participants, 18.3% said Label A provided adequate information on the harm of cigarette smoking. Among them, 16.5% (14/85) of participants said Label A gave adequate information on the relationship between cigarette smoking and respiratory diseases, including lung cancer, and 16.5% and 3.5% respectively mentioned cancer and cardiovascular diseases. Overall, 31.2% said Label B gave adequate information on the harm of cigarette smoking. Among them, 36.6% said Label B provided adequate information on the relationship between cigarette smoking and respiratory diseases, 5.3% and 3.1% could identify the relationship of smoking with cancer and cardiovascular diseases respectively based on the information on label B. Similar percentage of participants said Label C-F gave adequate information on the harm of cigarette smoking, 90.5% for Label C, 92.7% for Label D, 92.4% for Label E and 92.7% for Label F. Compared to Label A, a higher proportion of participants said that Label B gave them clearer information on the harm of smoking across all subcategories, except those with low education level and current smokers. Labels C-F performed better than Label B in providing harm information for all sub-groups (Table 2).\nThe proportion of positive responses to the harm information provided by different cigarette labels by gender, age groups, educational levels and smoking status\nA positive answer means participants can understand the harm information provided by the label.\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).", "Of the participants, 20.8% and 25.2% reported that they would not give cigarettes as a gift to somebody with Labels A and B (respectively) on the package. Over 80% of participants refused to give cigarettes as a gift if the package displayed warning Labels C-F. The proportion of those who would not give cigarettes as gift was higher among female, those who had never smoked and those having a higher educational level. Generally, there was no difference between the sub-groups, in terms of those who would not give cigarettes as a gift, for Label A and Label B, except the proportions were marginally higher among non-smokers and those aged between 30-40 for Label B. When comparing Label B to the combined labels, the proportion of respondents who would not give cigarettes as a gift was higher if any of Labels C-F were on the package (Table 3). The majority of participants (70.4%) considered that Label A was least likely to stop them using cigarettes as a gift, and the proportion was 20.2% for Label B. Almost half of participants (46.8%) considered that Label C was most likely to stop them using cigarettes as a gift, and the proportion was 5.7% and 3.4% for Label A and Label B respectively.\nThe perceived impact of not giving cigarette as a gift by gender, age groups, education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).", "There were 26.8% and 31.5% of the participants who reported thinking about quitting due to warning Label A and Label B, respectively. In addition, the proportions were all above 80% for Labels C-F. We asked non-smokers if they were smokers, if the labels would impact on a decision to quit smoking. Non-smokers were more likely to quit smoking due to Label C-F, compared to those who were smokers. It was shown that due to the warning on Label B, those more likely to quit were females, those with higher educational level and non-smokers when compared to Label A. Label B was less likely to make the participants quit smoking compared to Labels C-F combined (Table 4). Almost half of participants (43.3%) considered that Label C was most likely to cause them to quit. The proportion was only 4.5% and 3.7% for Label A and B, respectively. The majority of participants (69.9%) considered that Label A was least likely to cause them to quit, and the proportion was 20.2% for Label B. There was no significant difference between smoking status groups in terms of the impact Label A and Label B had on a decision on quit smoking.\nThe perceived impact on the decision to quit smoking by gender, age groups education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).", "Overall 32.4% of participants knew of the FCTC. Among them, 77.1% and 72.4% were non-smokers and those with the highest educational level, respectively. Furthermore, 75.4% (214/284) knew that China have ratified the FCTC, and 77.5% knew the provision of the FCTC that health warnings on cigarette packaging should be large, clear, visible and legible.", "Our study has shown that both the old and new Chinese warning labels have a low effect on the participants' knowledge of the harmful effects of smoking, on giving cigarettes as a gift, and quitting smoking. Labels used abroad were far more effective than the labels used in the Chinese market.\nOver 90% of the participants knew 'smoking was harmful to their health', while the knowledge of smoking-related disease, such as cardiovascular diseases, stroke etc. was relatively low. The result is consistent with another report from six cities in China [11]. From our survey, neither the old Chinese label nor the new one is able to provide details of smoking-related disease to smokers or nonsmokers, although there was a difference in the level of information related to the harm of smoking provided by Labels A and B, which may be due to their distinct location on the new pack. The result of no difference in low educational groups between Label A and B suggested that only text warnings cannot provide useful information to poor literacy population. In addition, the text-only labels cannot provide health warnings to current smokers, and smokers were failed to take notice of the difference between the old and new labels, even they take out cigarettes from packages every day.\nOur survey showed that text-plus-graphic warning labels were more effective than text-only labels, which is also consistent with other reports [12]. Graphic warnings can clearly express the consequences of smoking, and they are especially useful for populations with poor literacy and difficulty understanding text-based warnings. Moreover, graphic warning labels appear to be an important source of information regarding health risks for non-smokers, which may lead to increased pressure to quit from members of a smokers' social network [13]. More and more countries have mandated the inclusion of graphic imagery on cigarette warning labels (e.g., Australia, Brazil, Canada, Chile, Singapore, Thailand, Uruguay, and Venezuela), with other countries soon to follow (e.g., Belgium and New Zealand) [14].\nWarning labels can not only increase awareness of the health hazards, but also provide information on assistance for quitting and can promote interest in quitting. In our study, labels with detailed risk information and graphics had a more effective on the decision to quit. While both the old and the new Chinese labels had less effective with no information on specific smoking-related diseases, and no useful information on cessation. Canadian warning labels on cigarette packs are considered one of the most effective in the world, and are very useful for tobacco cessation. The requirements of the warning label with big, clear and direct health messages provides a strong incentive for smoking cessation [15-17]. Approximately one third of the smokers reported a likelihood of quitting and 20% of smokers reported smoking less, as a result of warning labels with graphic and detailed health risk and cessation information. Smokers were more likely to quit, make an attempt to quit, or reduce their smoking because of increased level of fear and disgust for the labels with text and large graphics [18]. Thus, graphic messages on warning labels appear more effective than text-only messages in promoting quitting [12,14,16,19]. Recent surveys have also shown increased cessation activities due to newly introduced text-and-graphic warnings in countries such as Australia, the United Kingdom, and the United States [19-21].\nAs a traditional culture, cigarettes are usually considered a valued gift to give, especially on special days, such as Chinese Spring Festival and other holidays. Chinese cigarette packages are always designed with beautiful brand names and graphics, and with one sentence of text warning about the harm but without information related to specific smoking-related diseases. Beautiful designed packs and high prevalence of cigarette smoking in male make cigarettes popular for giving in social communication. Giving cigarette is giving harm. While, the current Chinese warning labels have limited effect on not giving cigarettes as a gift. Compared with foreign warning label requirements, both the old Chinese warning labels and the new ones are relatively weak. The impact will increase if a country changes from smaller to larger and more contrasting warnings [19].\nA key limitation of this study concerns the use of a convenience sample which may not be representative of the Jiangsu population. However, purposive selection of groups in six types of work or public places in two cities enabled data collection from a wide range of population segments across a relatively small number of groups [14]. Another limitation was that the educational level of the participants was higher than the general population, thus this is not representative of the average education level of local residents. But, even in the population with higher educational levels, the proportion knowing the harm of smoking and WHO FCTC was not high. We estimate that the proportion is likely to be much lower in the general population. Dissemination of smoking-related knowledge needs be spread widely, especially in smokers and those with lower educational levels.\nAs the first report in Jiangsu Province, our findings suggest that the new Chinese warning labels are still not effective for this target population. People do not receive sufficient information on the harm of smoking and smoking-related diseases from these labels. In addition, the new warning labels do not effectively increase the desire to quit, or prevent individuals from giving cigarettes to others. The findings from this study indicate that cigarette packaging may benefit from more noticeable, readable, believable and memorable warnings in line with the WHO FCTC and this may be an important policy element in reducing the attractiveness of smoking especially among young adults and teenagers. Warning labels should be part of a larger public health education effort.", "('WHO FCTC'): World Health Organization Framework Convention on Tobacco Control;", "The authors declare that they have no competing interests.", "YQ contributed to the field work, data collection, quality control, analysis, and manuscript writing. MW, QX, and MZ contributed to the implementation in the field and gave advice on the manuscript writing. XP, JH, and ZG contributed to the field work, data collection and quality control. ZS contributed to the statistical advice and critical English review. All authors have read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2458/11/133/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study sites", "Participants", "Smoking status and demographic variables", "Warning labels of cigarette packages", "The harm warning provided by the label", "The perceived impact of giving cigarettes as a gift", "The perceived impact on the decision to quit smoking", "Knowledge of the FCTC and its provision for cigarette packaging", "Statistic analysis", "Results", "General information", "The harm warning provided by the label", "The perceived impact of giving cigarettes as a gift", "The perceived impact on the decision to quit smoking", "Knowledge of the WHO FCTC and its provision for cigarette packaging", "Discussion and Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history", "Supplementary Material" ]

[ "China is the largest producer and consumer of tobacco in the world. The prevalence of cigarette smoking is above 60% for men aged 15 and above, and 50% of women in this age group and adolescent are passive smokers [1]. Tobacco smoking was estimated as being responsible for approximately 67,300 premature deaths in Chinese adults aged 40 years and over in 2005 [2]. The premature deaths, productivity losses, and a substantial number of preventable diseases and health care costs related to smoking have resulted in a significant economic burden in China [3-7]. However, few people are aware of the harm caused by smoking and passive smoking [8]. Instead, cigarettes are considered a good vehicle for communication, and a popular gift to relatives or friends in China, especially for holidays.\nThe World Health Organization Framework Convention on Tobacco Control (WHO FCTC), the world's first public health treaty, calls for warning labels to be displayed as large and clear health warnings covering 30% to 50% of the package in the form of pictures, pictograms or text. Every person should be informed of the health consequences, addictive nature, and mortal threat posed by tobacco use and exposure to tobacco smoke. This can be achieved by the warning labels on cigarette packages [9]. In 1991, the Chinese Congress enacted legislation requiring cigarette warnings to state 'smoking is harmful to your health' in Chinese. This warning appeared on one of the side panels of every cigarette package [10]. On 9th of January 2006, WHO FCTC was ratified in China. In 2008, China implemented new regulations according to the FCTC and the legislation. Cigarette warnings were moved from the side panels and covered at least 30% of the front and back of the pack, with Chinese on the front, and English on the back. The warnings include two rotating sets of text in Chinese and English. One set states 'smoking is harmful to your health', and 'quitting smoking reduces health risk'. Another states 'smoking is harmful to your health', and 'quitting smoking early is good for your health' [10].\nThe purpose of the current study was primarily to compare the difference in reactions to different types of warning labels on cigarette packages with a specific focus on whether the new warning label is better than the previous one and labels used abroad.", "[SUBTITLE] Study sites [SUBSECTION] The study was conducted in 2008 in Nantong and Zhangjiagang cities, Jiangsu Province, one of the economically booming areas in Eastern part of China. Nantong is an urban city, and is a moderate developed city in the Province. Zhangjiagang is a rural area, and belongs to Suzhou City, a highly developed city in the Province.\nThe study was conducted in 2008 in Nantong and Zhangjiagang cities, Jiangsu Province, one of the economically booming areas in Eastern part of China. Nantong is an urban city, and is a moderate developed city in the Province. Zhangjiagang is a rural area, and belongs to Suzhou City, a highly developed city in the Province.\n[SUBTITLE] Participants [SUBSECTION] Eligible study participants included in this survey were those aged 18 years and over and those working in hospitals, schools, bus/train stations, government offices, restaurants and bars. Altogether 1000 adults were approached, and 876 participants agreed to participate and finished the survey. All participants were asked to complete a face-to-face interview using a standard questionnaire; informed consent was sought prior to the interview being undertaken. The study was approved by the ethical board of Jiangsu Provincial Centre for Disease Control and Prevention. Verbal consent was obtained from each participant.\nEligible study participants included in this survey were those aged 18 years and over and those working in hospitals, schools, bus/train stations, government offices, restaurants and bars. Altogether 1000 adults were approached, and 876 participants agreed to participate and finished the survey. All participants were asked to complete a face-to-face interview using a standard questionnaire; informed consent was sought prior to the interview being undertaken. The study was approved by the ethical board of Jiangsu Provincial Centre for Disease Control and Prevention. Verbal consent was obtained from each participant.\n[SUBTITLE] Smoking status and demographic variables [SUBSECTION] Information was obtained from all participants about their smoking status. Smoking status was measured by asking if participants had ever smoked. Participants were grouped into two categories smokers and former smokers. Smokers were defined as those having smoked at least 100 cigarettes in their lifetime, and those having smoked at least one cigarette per day at the time of the survey. Former smokers were defined as individuals who had quit smoking at least one month prior to the survey and those who smoked at least one cigarette per day prior to quitting. Participants also reported their gender, age, and education level.\nInformation was obtained from all participants about their smoking status. Smoking status was measured by asking if participants had ever smoked. Participants were grouped into two categories smokers and former smokers. Smokers were defined as those having smoked at least 100 cigarettes in their lifetime, and those having smoked at least one cigarette per day at the time of the survey. Former smokers were defined as individuals who had quit smoking at least one month prior to the survey and those who smoked at least one cigarette per day prior to quitting. Participants also reported their gender, age, and education level.\n[SUBTITLE] Warning labels of cigarette packages [SUBSECTION] Six warning labels were included in the interview questionnaire (Additional file 1). They were coded A-F. Label A was the old Chinese warning label, with 'smoking is harmful to your health' written in Chinese on one of the side panels of the pack (Figure 1). Label B was the new Chinese warning label, with 'smoking is harmful to your health', and 'quitting smoking reduces health risk' written on the front and back faces of the package, in Chinese and English respectively. Label C was a famous Chinese brand which is exported abroad and includes text, pictorials, and quitline information on the whole front face and 1/3 back face, respectively. The text on the pack reads 'smoking damages your blood vessels, which can prevent blood circulation, particularly to your legs or feet. This can result in blood clots, infection, gangrene, even amputation.' The other labels were foreign brands. Label D warned that cigarette smoking can result in mouth and oropharynx cancers by using text and pictures on 50% of the cigarette package. Label E, also used text and pictures to show that smoking when pregnant harms your baby. Label F indicated that smoking can lead to laryngeal cancer using both text and pictorials on 50% of the cigarette package. All English health warnings were translated into Chinese during the interview.\nSix cigarette warning labels. Label A was the old Chinese warning label, with 'smoking is harmful to your health' in Chinese on one of the side panels of the pack. Label B was the new Chinese warning label, with 'smoking is harmful to your health', and 'quit smoking reduces health risk' on the front and back faces of the package, in Chinese and English respectively. Label C was a famous Chinese brand exported abroad, with text, pictorials, and quitline on the whole front face and 1/3 back face, respectively. The text said 'smoking damages your blood vessels, which can prevent blood circulation, particularly to your legs or feet. This can result in blood clots, infection, gangrene, even amputation.' The other labels were foreign brands. Label D warned that cigarette smoking can result in mouth and oropharynx cancers in text and pictures on 50% of cigarette package. Label E showed smoking when pregnant harms your baby in large area with text and pictures. Label F indicated that smoking can lead to laryngeal cancer in text and pictorial on 50% of cigarette package.\nSix warning labels were included in the interview questionnaire (Additional file 1). They were coded A-F. Label A was the old Chinese warning label, with 'smoking is harmful to your health' written in Chinese on one of the side panels of the pack (Figure 1). Label B was the new Chinese warning label, with 'smoking is harmful to your health', and 'quitting smoking reduces health risk' written on the front and back faces of the package, in Chinese and English respectively. Label C was a famous Chinese brand which is exported abroad and includes text, pictorials, and quitline information on the whole front face and 1/3 back face, respectively. The text on the pack reads 'smoking damages your blood vessels, which can prevent blood circulation, particularly to your legs or feet. This can result in blood clots, infection, gangrene, even amputation.' The other labels were foreign brands. Label D warned that cigarette smoking can result in mouth and oropharynx cancers by using text and pictures on 50% of the cigarette package. Label E, also used text and pictures to show that smoking when pregnant harms your baby. Label F indicated that smoking can lead to laryngeal cancer using both text and pictorials on 50% of the cigarette package. All English health warnings were translated into Chinese during the interview.\nSix cigarette warning labels. Label A was the old Chinese warning label, with 'smoking is harmful to your health' in Chinese on one of the side panels of the pack. Label B was the new Chinese warning label, with 'smoking is harmful to your health', and 'quit smoking reduces health risk' on the front and back faces of the package, in Chinese and English respectively. Label C was a famous Chinese brand exported abroad, with text, pictorials, and quitline on the whole front face and 1/3 back face, respectively. The text said 'smoking damages your blood vessels, which can prevent blood circulation, particularly to your legs or feet. This can result in blood clots, infection, gangrene, even amputation.' The other labels were foreign brands. Label D warned that cigarette smoking can result in mouth and oropharynx cancers in text and pictures on 50% of cigarette package. Label E showed smoking when pregnant harms your baby in large area with text and pictures. Label F indicated that smoking can lead to laryngeal cancer in text and pictorial on 50% of cigarette package.\n[SUBTITLE] The harm warning provided by the label [SUBSECTION] Making reference to Labels A to F, participants were asked if the each label gave them clear information on the harm which cigarette smoking can have on health and the specific diseases that occur related to cigarette smoking. Participants were also asked if Labels C, D, E, and F gave them clear information on specific diseases smoking can cause (as described above).\nMaking reference to Labels A to F, participants were asked if the each label gave them clear information on the harm which cigarette smoking can have on health and the specific diseases that occur related to cigarette smoking. Participants were also asked if Labels C, D, E, and F gave them clear information on specific diseases smoking can cause (as described above).\n[SUBTITLE] The perceived impact of giving cigarettes as a gift [SUBSECTION] Three questions on the perceived impact of giving cigarettes as a gift were presented. These included: 1) If you want to use cigarettes as a gift, do the following cigarette labels (A-F) make you change your mind and not do so? 2) If you want to give cigarettes as a gift, which warning label is least likely to stop you using cigarettes as a gift? 3) If you want to give cigarettes as a gift, which warning label is most likely to stop you using cigarettes as a gift?\nThree questions on the perceived impact of giving cigarettes as a gift were presented. These included: 1) If you want to use cigarettes as a gift, do the following cigarette labels (A-F) make you change your mind and not do so? 2) If you want to give cigarettes as a gift, which warning label is least likely to stop you using cigarettes as a gift? 3) If you want to give cigarettes as a gift, which warning label is most likely to stop you using cigarettes as a gift?\n[SUBTITLE] The perceived impact on the decision to quit smoking [SUBSECTION] Participants were asked three questions about the perceived impact of quitting smoking. These include: 1) If you were a cigarette smoker, would the following labels (A-F) make you want to quit smoking? 2) If you were a cigarette smoker, which warning label is most likely to cause you to quit? 3) If you were a cigarette smoker, which warning label is least likely to cause you to quit?\nParticipants were asked three questions about the perceived impact of quitting smoking. These include: 1) If you were a cigarette smoker, would the following labels (A-F) make you want to quit smoking? 2) If you were a cigarette smoker, which warning label is most likely to cause you to quit? 3) If you were a cigarette smoker, which warning label is least likely to cause you to quit?\n[SUBTITLE] Knowledge of the FCTC and its provision for cigarette packaging [SUBSECTION] Participants were asked if they knew that China had ratified the WHO FCTC. If they answered yes, participants were then asked if they were aware of the FCTC provision that health warnings on cigarette packaging should be large, clear, visible and legible.\nParticipants were asked if they knew that China had ratified the WHO FCTC. If they answered yes, participants were then asked if they were aware of the FCTC provision that health warnings on cigarette packaging should be large, clear, visible and legible.\n[SUBTITLE] Statistic analysis [SUBSECTION] Univariate and bivariate analyses were conducted to examine how much impact each of the different cigarette warning labels had and the knowledge of the FCTC by age groups, gender, education levels and smoking status. To compare the new Chinese label with international labels, Label C, D, E, and F were aggregated into one group. Chi-square tests were used to assess differences among groups where appropriate. All analyses were conducted using SPSS 13.0 (SPSS Inc., Chicago, IL, USA).\nUnivariate and bivariate analyses were conducted to examine how much impact each of the different cigarette warning labels had and the knowledge of the FCTC by age groups, gender, education levels and smoking status. To compare the new Chinese label with international labels, Label C, D, E, and F were aggregated into one group. Chi-square tests were used to assess differences among groups where appropriate. All analyses were conducted using SPSS 13.0 (SPSS Inc., Chicago, IL, USA).", "The study was conducted in 2008 in Nantong and Zhangjiagang cities, Jiangsu Province, one of the economically booming areas in Eastern part of China. Nantong is an urban city, and is a moderate developed city in the Province. Zhangjiagang is a rural area, and belongs to Suzhou City, a highly developed city in the Province.", "Eligible study participants included in this survey were those aged 18 years and over and those working in hospitals, schools, bus/train stations, government offices, restaurants and bars. Altogether 1000 adults were approached, and 876 participants agreed to participate and finished the survey. All participants were asked to complete a face-to-face interview using a standard questionnaire; informed consent was sought prior to the interview being undertaken. The study was approved by the ethical board of Jiangsu Provincial Centre for Disease Control and Prevention. Verbal consent was obtained from each participant.", "Information was obtained from all participants about their smoking status. Smoking status was measured by asking if participants had ever smoked. Participants were grouped into two categories smokers and former smokers. Smokers were defined as those having smoked at least 100 cigarettes in their lifetime, and those having smoked at least one cigarette per day at the time of the survey. Former smokers were defined as individuals who had quit smoking at least one month prior to the survey and those who smoked at least one cigarette per day prior to quitting. Participants also reported their gender, age, and education level.", "Six warning labels were included in the interview questionnaire (Additional file 1). They were coded A-F. Label A was the old Chinese warning label, with 'smoking is harmful to your health' written in Chinese on one of the side panels of the pack (Figure 1). Label B was the new Chinese warning label, with 'smoking is harmful to your health', and 'quitting smoking reduces health risk' written on the front and back faces of the package, in Chinese and English respectively. Label C was a famous Chinese brand which is exported abroad and includes text, pictorials, and quitline information on the whole front face and 1/3 back face, respectively. The text on the pack reads 'smoking damages your blood vessels, which can prevent blood circulation, particularly to your legs or feet. This can result in blood clots, infection, gangrene, even amputation.' The other labels were foreign brands. Label D warned that cigarette smoking can result in mouth and oropharynx cancers by using text and pictures on 50% of the cigarette package. Label E, also used text and pictures to show that smoking when pregnant harms your baby. Label F indicated that smoking can lead to laryngeal cancer using both text and pictorials on 50% of the cigarette package. All English health warnings were translated into Chinese during the interview.\nSix cigarette warning labels. Label A was the old Chinese warning label, with 'smoking is harmful to your health' in Chinese on one of the side panels of the pack. Label B was the new Chinese warning label, with 'smoking is harmful to your health', and 'quit smoking reduces health risk' on the front and back faces of the package, in Chinese and English respectively. Label C was a famous Chinese brand exported abroad, with text, pictorials, and quitline on the whole front face and 1/3 back face, respectively. The text said 'smoking damages your blood vessels, which can prevent blood circulation, particularly to your legs or feet. This can result in blood clots, infection, gangrene, even amputation.' The other labels were foreign brands. Label D warned that cigarette smoking can result in mouth and oropharynx cancers in text and pictures on 50% of cigarette package. Label E showed smoking when pregnant harms your baby in large area with text and pictures. Label F indicated that smoking can lead to laryngeal cancer in text and pictorial on 50% of cigarette package.", "Making reference to Labels A to F, participants were asked if the each label gave them clear information on the harm which cigarette smoking can have on health and the specific diseases that occur related to cigarette smoking. Participants were also asked if Labels C, D, E, and F gave them clear information on specific diseases smoking can cause (as described above).", "Three questions on the perceived impact of giving cigarettes as a gift were presented. These included: 1) If you want to use cigarettes as a gift, do the following cigarette labels (A-F) make you change your mind and not do so? 2) If you want to give cigarettes as a gift, which warning label is least likely to stop you using cigarettes as a gift? 3) If you want to give cigarettes as a gift, which warning label is most likely to stop you using cigarettes as a gift?", "Participants were asked three questions about the perceived impact of quitting smoking. These include: 1) If you were a cigarette smoker, would the following labels (A-F) make you want to quit smoking? 2) If you were a cigarette smoker, which warning label is most likely to cause you to quit? 3) If you were a cigarette smoker, which warning label is least likely to cause you to quit?", "Participants were asked if they knew that China had ratified the WHO FCTC. If they answered yes, participants were then asked if they were aware of the FCTC provision that health warnings on cigarette packaging should be large, clear, visible and legible.", "Univariate and bivariate analyses were conducted to examine how much impact each of the different cigarette warning labels had and the knowledge of the FCTC by age groups, gender, education levels and smoking status. To compare the new Chinese label with international labels, Label C, D, E, and F were aggregated into one group. Chi-square tests were used to assess differences among groups where appropriate. All analyses were conducted using SPSS 13.0 (SPSS Inc., Chicago, IL, USA).", "[SUBTITLE] General information [SUBSECTION] Table 1 demonstrates the demographic characteristics of the sample. A total of 876 participants (374 male and 502 female) were involved in the study. The average age was 34.0 ± 11.0 years, a higher proportion of males reported that they were current smokers compared to females and 82.7% of participants had graduated from technical secondary school or higher.\nCharacteristics of the study sample\nTable 1 demonstrates the demographic characteristics of the sample. A total of 876 participants (374 male and 502 female) were involved in the study. The average age was 34.0 ± 11.0 years, a higher proportion of males reported that they were current smokers compared to females and 82.7% of participants had graduated from technical secondary school or higher.\nCharacteristics of the study sample\n[SUBTITLE] The harm warning provided by the label [SUBSECTION] Of the participants, 18.3% said Label A provided adequate information on the harm of cigarette smoking. Among them, 16.5% (14/85) of participants said Label A gave adequate information on the relationship between cigarette smoking and respiratory diseases, including lung cancer, and 16.5% and 3.5% respectively mentioned cancer and cardiovascular diseases. Overall, 31.2% said Label B gave adequate information on the harm of cigarette smoking. Among them, 36.6% said Label B provided adequate information on the relationship between cigarette smoking and respiratory diseases, 5.3% and 3.1% could identify the relationship of smoking with cancer and cardiovascular diseases respectively based on the information on label B. Similar percentage of participants said Label C-F gave adequate information on the harm of cigarette smoking, 90.5% for Label C, 92.7% for Label D, 92.4% for Label E and 92.7% for Label F. Compared to Label A, a higher proportion of participants said that Label B gave them clearer information on the harm of smoking across all subcategories, except those with low education level and current smokers. Labels C-F performed better than Label B in providing harm information for all sub-groups (Table 2).\nThe proportion of positive responses to the harm information provided by different cigarette labels by gender, age groups, educational levels and smoking status\nA positive answer means participants can understand the harm information provided by the label.\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\nOf the participants, 18.3% said Label A provided adequate information on the harm of cigarette smoking. Among them, 16.5% (14/85) of participants said Label A gave adequate information on the relationship between cigarette smoking and respiratory diseases, including lung cancer, and 16.5% and 3.5% respectively mentioned cancer and cardiovascular diseases. Overall, 31.2% said Label B gave adequate information on the harm of cigarette smoking. Among them, 36.6% said Label B provided adequate information on the relationship between cigarette smoking and respiratory diseases, 5.3% and 3.1% could identify the relationship of smoking with cancer and cardiovascular diseases respectively based on the information on label B. Similar percentage of participants said Label C-F gave adequate information on the harm of cigarette smoking, 90.5% for Label C, 92.7% for Label D, 92.4% for Label E and 92.7% for Label F. Compared to Label A, a higher proportion of participants said that Label B gave them clearer information on the harm of smoking across all subcategories, except those with low education level and current smokers. Labels C-F performed better than Label B in providing harm information for all sub-groups (Table 2).\nThe proportion of positive responses to the harm information provided by different cigarette labels by gender, age groups, educational levels and smoking status\nA positive answer means participants can understand the harm information provided by the label.\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\n[SUBTITLE] The perceived impact of giving cigarettes as a gift [SUBSECTION] Of the participants, 20.8% and 25.2% reported that they would not give cigarettes as a gift to somebody with Labels A and B (respectively) on the package. Over 80% of participants refused to give cigarettes as a gift if the package displayed warning Labels C-F. The proportion of those who would not give cigarettes as gift was higher among female, those who had never smoked and those having a higher educational level. Generally, there was no difference between the sub-groups, in terms of those who would not give cigarettes as a gift, for Label A and Label B, except the proportions were marginally higher among non-smokers and those aged between 30-40 for Label B. When comparing Label B to the combined labels, the proportion of respondents who would not give cigarettes as a gift was higher if any of Labels C-F were on the package (Table 3). The majority of participants (70.4%) considered that Label A was least likely to stop them using cigarettes as a gift, and the proportion was 20.2% for Label B. Almost half of participants (46.8%) considered that Label C was most likely to stop them using cigarettes as a gift, and the proportion was 5.7% and 3.4% for Label A and Label B respectively.\nThe perceived impact of not giving cigarette as a gift by gender, age groups, education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\nOf the participants, 20.8% and 25.2% reported that they would not give cigarettes as a gift to somebody with Labels A and B (respectively) on the package. Over 80% of participants refused to give cigarettes as a gift if the package displayed warning Labels C-F. The proportion of those who would not give cigarettes as gift was higher among female, those who had never smoked and those having a higher educational level. Generally, there was no difference between the sub-groups, in terms of those who would not give cigarettes as a gift, for Label A and Label B, except the proportions were marginally higher among non-smokers and those aged between 30-40 for Label B. When comparing Label B to the combined labels, the proportion of respondents who would not give cigarettes as a gift was higher if any of Labels C-F were on the package (Table 3). The majority of participants (70.4%) considered that Label A was least likely to stop them using cigarettes as a gift, and the proportion was 20.2% for Label B. Almost half of participants (46.8%) considered that Label C was most likely to stop them using cigarettes as a gift, and the proportion was 5.7% and 3.4% for Label A and Label B respectively.\nThe perceived impact of not giving cigarette as a gift by gender, age groups, education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\n[SUBTITLE] The perceived impact on the decision to quit smoking [SUBSECTION] There were 26.8% and 31.5% of the participants who reported thinking about quitting due to warning Label A and Label B, respectively. In addition, the proportions were all above 80% for Labels C-F. We asked non-smokers if they were smokers, if the labels would impact on a decision to quit smoking. Non-smokers were more likely to quit smoking due to Label C-F, compared to those who were smokers. It was shown that due to the warning on Label B, those more likely to quit were females, those with higher educational level and non-smokers when compared to Label A. Label B was less likely to make the participants quit smoking compared to Labels C-F combined (Table 4). Almost half of participants (43.3%) considered that Label C was most likely to cause them to quit. The proportion was only 4.5% and 3.7% for Label A and B, respectively. The majority of participants (69.9%) considered that Label A was least likely to cause them to quit, and the proportion was 20.2% for Label B. There was no significant difference between smoking status groups in terms of the impact Label A and Label B had on a decision on quit smoking.\nThe perceived impact on the decision to quit smoking by gender, age groups education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\nThere were 26.8% and 31.5% of the participants who reported thinking about quitting due to warning Label A and Label B, respectively. In addition, the proportions were all above 80% for Labels C-F. We asked non-smokers if they were smokers, if the labels would impact on a decision to quit smoking. Non-smokers were more likely to quit smoking due to Label C-F, compared to those who were smokers. It was shown that due to the warning on Label B, those more likely to quit were females, those with higher educational level and non-smokers when compared to Label A. Label B was less likely to make the participants quit smoking compared to Labels C-F combined (Table 4). Almost half of participants (43.3%) considered that Label C was most likely to cause them to quit. The proportion was only 4.5% and 3.7% for Label A and B, respectively. The majority of participants (69.9%) considered that Label A was least likely to cause them to quit, and the proportion was 20.2% for Label B. There was no significant difference between smoking status groups in terms of the impact Label A and Label B had on a decision on quit smoking.\nThe perceived impact on the decision to quit smoking by gender, age groups education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).\n[SUBTITLE] Knowledge of the WHO FCTC and its provision for cigarette packaging [SUBSECTION] Overall 32.4% of participants knew of the FCTC. Among them, 77.1% and 72.4% were non-smokers and those with the highest educational level, respectively. Furthermore, 75.4% (214/284) knew that China have ratified the FCTC, and 77.5% knew the provision of the FCTC that health warnings on cigarette packaging should be large, clear, visible and legible.\nOverall 32.4% of participants knew of the FCTC. Among them, 77.1% and 72.4% were non-smokers and those with the highest educational level, respectively. Furthermore, 75.4% (214/284) knew that China have ratified the FCTC, and 77.5% knew the provision of the FCTC that health warnings on cigarette packaging should be large, clear, visible and legible.", "Table 1 demonstrates the demographic characteristics of the sample. A total of 876 participants (374 male and 502 female) were involved in the study. The average age was 34.0 ± 11.0 years, a higher proportion of males reported that they were current smokers compared to females and 82.7% of participants had graduated from technical secondary school or higher.\nCharacteristics of the study sample", "Of the participants, 18.3% said Label A provided adequate information on the harm of cigarette smoking. Among them, 16.5% (14/85) of participants said Label A gave adequate information on the relationship between cigarette smoking and respiratory diseases, including lung cancer, and 16.5% and 3.5% respectively mentioned cancer and cardiovascular diseases. Overall, 31.2% said Label B gave adequate information on the harm of cigarette smoking. Among them, 36.6% said Label B provided adequate information on the relationship between cigarette smoking and respiratory diseases, 5.3% and 3.1% could identify the relationship of smoking with cancer and cardiovascular diseases respectively based on the information on label B. Similar percentage of participants said Label C-F gave adequate information on the harm of cigarette smoking, 90.5% for Label C, 92.7% for Label D, 92.4% for Label E and 92.7% for Label F. Compared to Label A, a higher proportion of participants said that Label B gave them clearer information on the harm of smoking across all subcategories, except those with low education level and current smokers. Labels C-F performed better than Label B in providing harm information for all sub-groups (Table 2).\nThe proportion of positive responses to the harm information provided by different cigarette labels by gender, age groups, educational levels and smoking status\nA positive answer means participants can understand the harm information provided by the label.\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).", "Of the participants, 20.8% and 25.2% reported that they would not give cigarettes as a gift to somebody with Labels A and B (respectively) on the package. Over 80% of participants refused to give cigarettes as a gift if the package displayed warning Labels C-F. The proportion of those who would not give cigarettes as gift was higher among female, those who had never smoked and those having a higher educational level. Generally, there was no difference between the sub-groups, in terms of those who would not give cigarettes as a gift, for Label A and Label B, except the proportions were marginally higher among non-smokers and those aged between 30-40 for Label B. When comparing Label B to the combined labels, the proportion of respondents who would not give cigarettes as a gift was higher if any of Labels C-F were on the package (Table 3). The majority of participants (70.4%) considered that Label A was least likely to stop them using cigarettes as a gift, and the proportion was 20.2% for Label B. Almost half of participants (46.8%) considered that Label C was most likely to stop them using cigarettes as a gift, and the proportion was 5.7% and 3.4% for Label A and Label B respectively.\nThe perceived impact of not giving cigarette as a gift by gender, age groups, education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).", "There were 26.8% and 31.5% of the participants who reported thinking about quitting due to warning Label A and Label B, respectively. In addition, the proportions were all above 80% for Labels C-F. We asked non-smokers if they were smokers, if the labels would impact on a decision to quit smoking. Non-smokers were more likely to quit smoking due to Label C-F, compared to those who were smokers. It was shown that due to the warning on Label B, those more likely to quit were females, those with higher educational level and non-smokers when compared to Label A. Label B was less likely to make the participants quit smoking compared to Labels C-F combined (Table 4). Almost half of participants (43.3%) considered that Label C was most likely to cause them to quit. The proportion was only 4.5% and 3.7% for Label A and B, respectively. The majority of participants (69.9%) considered that Label A was least likely to cause them to quit, and the proportion was 20.2% for Label B. There was no significant difference between smoking status groups in terms of the impact Label A and Label B had on a decision on quit smoking.\nThe perceived impact on the decision to quit smoking by gender, age groups education levels and smoking status\n*low, medium and high education refers to high school and below, technical secondary school, college and above, respectively.\n†By chi-square test.\n‡Combination of labels C-F (positive answer to all labels were regarded as positive).", "Overall 32.4% of participants knew of the FCTC. Among them, 77.1% and 72.4% were non-smokers and those with the highest educational level, respectively. Furthermore, 75.4% (214/284) knew that China have ratified the FCTC, and 77.5% knew the provision of the FCTC that health warnings on cigarette packaging should be large, clear, visible and legible.", "Our study has shown that both the old and new Chinese warning labels have a low effect on the participants' knowledge of the harmful effects of smoking, on giving cigarettes as a gift, and quitting smoking. Labels used abroad were far more effective than the labels used in the Chinese market.\nOver 90% of the participants knew 'smoking was harmful to their health', while the knowledge of smoking-related disease, such as cardiovascular diseases, stroke etc. was relatively low. The result is consistent with another report from six cities in China [11]. From our survey, neither the old Chinese label nor the new one is able to provide details of smoking-related disease to smokers or nonsmokers, although there was a difference in the level of information related to the harm of smoking provided by Labels A and B, which may be due to their distinct location on the new pack. The result of no difference in low educational groups between Label A and B suggested that only text warnings cannot provide useful information to poor literacy population. In addition, the text-only labels cannot provide health warnings to current smokers, and smokers were failed to take notice of the difference between the old and new labels, even they take out cigarettes from packages every day.\nOur survey showed that text-plus-graphic warning labels were more effective than text-only labels, which is also consistent with other reports [12]. Graphic warnings can clearly express the consequences of smoking, and they are especially useful for populations with poor literacy and difficulty understanding text-based warnings. Moreover, graphic warning labels appear to be an important source of information regarding health risks for non-smokers, which may lead to increased pressure to quit from members of a smokers' social network [13]. More and more countries have mandated the inclusion of graphic imagery on cigarette warning labels (e.g., Australia, Brazil, Canada, Chile, Singapore, Thailand, Uruguay, and Venezuela), with other countries soon to follow (e.g., Belgium and New Zealand) [14].\nWarning labels can not only increase awareness of the health hazards, but also provide information on assistance for quitting and can promote interest in quitting. In our study, labels with detailed risk information and graphics had a more effective on the decision to quit. While both the old and the new Chinese labels had less effective with no information on specific smoking-related diseases, and no useful information on cessation. Canadian warning labels on cigarette packs are considered one of the most effective in the world, and are very useful for tobacco cessation. The requirements of the warning label with big, clear and direct health messages provides a strong incentive for smoking cessation [15-17]. Approximately one third of the smokers reported a likelihood of quitting and 20% of smokers reported smoking less, as a result of warning labels with graphic and detailed health risk and cessation information. Smokers were more likely to quit, make an attempt to quit, or reduce their smoking because of increased level of fear and disgust for the labels with text and large graphics [18]. Thus, graphic messages on warning labels appear more effective than text-only messages in promoting quitting [12,14,16,19]. Recent surveys have also shown increased cessation activities due to newly introduced text-and-graphic warnings in countries such as Australia, the United Kingdom, and the United States [19-21].\nAs a traditional culture, cigarettes are usually considered a valued gift to give, especially on special days, such as Chinese Spring Festival and other holidays. Chinese cigarette packages are always designed with beautiful brand names and graphics, and with one sentence of text warning about the harm but without information related to specific smoking-related diseases. Beautiful designed packs and high prevalence of cigarette smoking in male make cigarettes popular for giving in social communication. Giving cigarette is giving harm. While, the current Chinese warning labels have limited effect on not giving cigarettes as a gift. Compared with foreign warning label requirements, both the old Chinese warning labels and the new ones are relatively weak. The impact will increase if a country changes from smaller to larger and more contrasting warnings [19].\nA key limitation of this study concerns the use of a convenience sample which may not be representative of the Jiangsu population. However, purposive selection of groups in six types of work or public places in two cities enabled data collection from a wide range of population segments across a relatively small number of groups [14]. Another limitation was that the educational level of the participants was higher than the general population, thus this is not representative of the average education level of local residents. But, even in the population with higher educational levels, the proportion knowing the harm of smoking and WHO FCTC was not high. We estimate that the proportion is likely to be much lower in the general population. Dissemination of smoking-related knowledge needs be spread widely, especially in smokers and those with lower educational levels.\nAs the first report in Jiangsu Province, our findings suggest that the new Chinese warning labels are still not effective for this target population. People do not receive sufficient information on the harm of smoking and smoking-related diseases from these labels. In addition, the new warning labels do not effectively increase the desire to quit, or prevent individuals from giving cigarettes to others. The findings from this study indicate that cigarette packaging may benefit from more noticeable, readable, believable and memorable warnings in line with the WHO FCTC and this may be an important policy element in reducing the attractiveness of smoking especially among young adults and teenagers. Warning labels should be part of a larger public health education effort.", "('WHO FCTC'): World Health Organization Framework Convention on Tobacco Control;", "The authors declare that they have no competing interests.", "YQ contributed to the field work, data collection, quality control, analysis, and manuscript writing. MW, QX, and MZ contributed to the implementation in the field and gave advice on the manuscript writing. XP, JH, and ZG contributed to the field work, data collection and quality control. ZS contributed to the statistical advice and critical English review. All authors have read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2458/11/133/prepub\n", "Questionnaire for warning labels of cigarette. As requested by the editor, the questionnaire used in the study was translated into English and presented for the readers. The questionnaire was designed by the China CDC-PUMC-JHSPH Project Group. Any use of it should be noticed to the Group and must be properly cited in any related research products.\nClick here for file" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, "supplementary-material" ]

[]

[ "Aged", "Aged, 80 and over", "Air Pollutants", "Boston", "Endothelium", "Humans", "Inflammation", "Intercellular Adhesion Molecule-1", "Male", "Middle Aged", "Regression Analysis", "Soot", "Vascular Cell Adhesion Molecule-1", "Vehicle Emissions" ]

[ "Study population", "Measurements of sICAM-1 and sVCAM-1", "BC exposure prediction", "Estimation of health effects", "Descriptive data", "Regression analysis", "TRAP and inflammatory response", "Mechanisms and interactions", "Exposure estimation", "Sources and components", "Generalizability" ]

[ "We studied participants in the Normative Aging Study (NAS), a longitudinal study established by the VA in 1963 (Bell et al. 1972). In brief, the NAS enrolled 2,280 men from the Greater Boston area who were initially free of known chronic medical conditions. All participants provided written informed consent, and the study was approved by the institutional review boards of all participating institutions. Participants visited the Boston VA Hospital study center every 3 years to undergo physical examinations. At each of these visits, blood samples and extensive physical examination, laboratory, anthropometric, and questionnaire data were collected. Information about cigarette smoking, medical history, and medication use were obtained by self-administered questionnaire. Each subject was interviewed to confirm the identity and purpose of medications used, and all new disease diagnoses were noted.\nDiabetes was defined as a physician diagnosis of diabetes, and obesity was defined as a body mass index (BMI) of at least 30 kg/m2. Self-reported data on diabetes status and statin use were updated at each study visit. In addition, BMI and obesity were updated based on height and weight measurements at each visit. Thus, NAS data reflect changes in disease status and medication use over time.\nMeasurements of sICAM-1 and sVCAM-1 began in 1999. For the present study, we included the 642 NAS participants with at least one measurement of sICAM-1 and sVCAM-1 and whose home address was in the Greater Boston area (1,423 total person-visits). Subjects who moved out of the area were excluded.", "Blood samples routinely collected during medical exam visits from 1999 through 2008 were analyzed for sICAM-1 and sVCAM-1 in N. Rafai’s laboratory at Children’s Hospital Boston (Boston, MA). Plasma sICAM-1 and sVCAM-1 concentrations were measured in duplicate using the enzyme-linked immunosorbent assay (ELISA) method (R&D Systems, Minneapolis, MN), with a sensitivity of 0.35 ng/mL for sICAM-1 and 2.0 ng/mL for sVCAM-1 (Lim et al. 1999).", "The BC exposure model and the stationary air monitors used to develop the model have been described in detail previously (Gryparis et al. 2007). Briefly, 82 sites were used; most sites measured BC continuously using aethalometers, and other sites collected particles on a filter over 24 hr and measured elemental carbon (EC) using reflectance analysis. The monitoring data used to develop our model included 6,031 observations from 2,079 unique exposure days.\nUsing a spatiotemporal model that we developed and validated previously (Gryparis et al. 2007), we estimated the 24-hr average BC concentration at each geocoded participant address. Predicted daily concentrations showed a > 3-fold range of variation in exposure across measurement sites (adjusted R2 = 0.83). A validation sample at 30 additional monitoring sites showed an average correlation of 0.59 between predicted and observed daily BC levels. We averaged the 24-hr predictions to form estimates for the 4, 8, and 12 weeks before each participant visit. We also averaged the 24-hr predictions to form estimates for the 4-week average during the 5–8 weeks before the study visit and the 4-week average during the 9–12 weeks before the study visit, which are components of the 8- and 12-week averages, to use as a sensitivity analysis.\nCovariates in the BC prediction model included measures of land use for each address (cumulative traffic density within 100 m, population density, distance to nearest major roadway, and percent urbanization), geographic information system (GIS) location (latitude, longitude), daily meteorological factors (apparent temperature, wind speed, and height of the planetary boundary layer), and other characteristics (day of week, day of season). The Boston central-site monitor was also included as a predictor to reflect average pollutant concentrations over the entire region on each day.\nSeparate models were fit for warm and cold seasons. Interaction terms between the temporal meteorological predictors and source-based geographic variables allowed for space–time interactions. Regression splines allowed main effect terms to nonlinearly predict exposure levels, and thin-plate splines modeled the residual spatial variability not explained by the spatial predictors. A latent variable framework was used to integrate BC and EC exposure data, where BC and EC measurements were treated as surrogates of some true, unobservable traffic exposure variable; see Gryparis et al. (2007) for further details.", "We log-transformed sICAM-1 and sVCAM-1 levels to increase normality and stabilize variance in the residuals. Model covariates were selected a priori, and all models included age, BMI, diabetes, smoking status, pack-years, and season. We used mixed models to account for correlation among measurements on the same subject across different medical visits. Mixed models have the form\nwhere Yij is log(sICAM-1) or log(sVCAM-1) in subject i on day j, and ui represents a subject-specific intercept that reflects unexplained heterogeneity in the outcome. BC averages and model covariates are modeled as fixed linear effects, and ui is modeled as a random effect. We assume that the ui are generated from a normal distribution with common variance, yielding the simple compound symmetry variance structure. This model requires estimation of two variance components, which represent between- and within-subject variation. Models with unbalanced data (i.e., varying numbers of repeated measurements on each subject) typically yield accurate estimates of within-subject variation, provided a sufficient number of repeated measurements contribute to the estimate.\nModels used to examine effect modification by obesity, diabetes, and statin use included interaction terms that allowed associations between BC and the outcomes to vary among subgroups. Diabetes, obesity, and statin use were treated as time-varying covariates, where the status was updated at each visit to reflect changes since the last visit. The percentages of subjects whose status changed for these factors over the study period were 3% for diabetes, 9% for obesity, and 21% for statin use.\nWe also performed a sensitivity analysis to investigate whether the interaction with statin use was heavily influenced by the people who began taking statins after the study period began. Specifically, we limited the interaction analysis to the participants who never used statins during the study period (n = 284) and those who used statins throughout the entire study period (n = 223).\nBecause the outcomes were log-transformed, effect estimates are reported as percent changes in sICAM-1 and sVCAM-1 concentrations associated with a 0.30-μg/m3 increase in BC, which corresponds to the interquartile range (IQR) for average BC exposures over all three time intervals (4, 8, and 12 weeks). A level of α = 0.05 was used to determine statistical significance.", "Subjects were elderly, with a mean age of 72 years (range, 56–100 years) at the first study visit (Table 1). Most subjects were overweight (median BMI, 27.5 kg/m2; range, 18–52 kg/m2). Average 4-week BC exposure estimates ranged from 0 to 3.26 μg/m3 (mean ± SD, 0.42 ± 0.30 μg/m3; median, 0.37 μg/m3; IQR, 0.30 μg/m3). For both 8- and 12-week exposures, exposure estimates were 0.42 ± 0.29 μg/m3 (median, 0.38 μg/m3; IQR, 0.30 μg/m3). Figure 1 illustrates the spatial distribution of predicted BC in the study area during an arbitrary 4-week period (February 2001). The distributions of BC exposures looked similar among subjects classified according to obesity, diabetes, statin use, or smoking status, and we found no significant differences in the mean BC exposure assessed at baseline.", "We found significant positive associations between sICAM-1 and BC averaged over the previous 4, 8, and 12 weeks (Table 2). Associations between average BC for 4, 8, and 12 weeks and sVCAM-1 were in the same direction, with slightly smaller effect size, but were not statistically significant. As a sensitivity analysis, we also estimated the effects of BC exposure in weeks 5–8 and weeks 9–12 separately. For sICAM-1, the association with the longer BC lags were positive with effect size slightly smaller than the effect size for weeks 1–4 and not quite statistically significant. A 0.30-μg/m3 increase in BC exposure during weeks 5–8 was associated with a 1.22% increase in sICAM-1 [95% confidence interval (CI), −0.10% to 3.00%], and during weeks 9–12, with a 1.09% increase in sICAM-1 (95% CI, −0.18% to 2.38%). For sVCAM-1, the estimated effects for the longer lags had the same effect size and direction as the effects for weeks 1–4 but were not statistically significant (data not shown). Overall, we saw the same pattern of effects in the lagged 4-week intervals, with only slight weakening in effect when the lags were considered individually rather than cumulatively.", "Several studies have examined the effects of short-term (i.e., < 24 hr to 3 weeks) TRAP and various markers of inflammation. For example, in a large epidemiological study, daily increases in ambient PM levels were positively associated with plasma fibrinogen levels, with the strongest association observed in participants with chronic obstructive lung disease (Schwartz 2001). In a study of a small (n = 29) elderly population, Delfino et al. (2008) examined inflammatory markers and PM components measured at each subject’s home for 1–9 days on average. The authors reported that several traffic-related components were significantly associated with CRP and interleukin-6, and observed positive but nonsignificant associations with sICAM-1 and sVCAM-1.\nIn other studies of NAS participants, traffic-related PM (BC and organic carbon) exposure has also been positively associated with plasma total homocysteine concentrations (Park et al. 2008), and this association is modified by polymorphisms in genes related to oxidative stress (Ren et al. 2010). PM2.5 and BC averages of 1–3 days measured at the central site have been associated with increased vascular cell adhesion in NAS participants (Madrigano et al. 2010). Traffic-related PM (particle number and BC) has also been positively associated with inflammatory markers (CRP, white blood cell count, sediment rate, and fibrinogen) in NAS participants, with stronger associations with particle numbers than with BC and stronger associations with BC averaged over 4 weeks than averaged over 48 hr or 1 week (Zeka et al. 2006).\nFew controlled studies of PM on human inflammatory response have been conducted, but one reported that plasma fibrinogen increased after exposure to urban PM (Ghio et al. 2000), and another reported that peripheral neutrophils, sVCAM-1, and sICAM-1 increased after a 1-hr exposure to diesel PM (Salvi et al. 1999).", "Diabetics have impaired endothelial function compared with nondiabetics (Calles-Escandon and Cipolla 2001), and there is increasing evidence that diabetics are more susceptible to the effects of air pollution. Inverse associations between 60-day average BC exposure and brachial artery flow-mediated dilation (FMD) were stronger in diabetics than in nondiabetics (O’Neill et al. 2005). Associations between 24-hr exposure to PM2.5 and FMD were stronger among diabetics with markers of severe insulin resistance compared with other diabetics (Schneider et al. 2008).\nIn contrast with our findings, obese NAS participants have been reported to have stronger associations between short-term BC exposure and plasma CRP, erythrocyte sediment rate (Zeka et al. 2006), and sVCAM-1 (Madrigano et al. 2010) than nonobese participants.\nOur finding that statin users are less susceptible than those not taking statins is generally consistent with other studies. In diabetics, stronger associations of PM2.5 and BC with sVCAM-1 were reported among those not using statins than in statin users (O’Neill et al. 2007). Associations between CRP and traffic PM exposures of 5 and 9 days were reported to be stronger among those using statins than those not using statins (Delfino et al. 2008). Statins promote endothelial nitric oxide release, which reduces cell adhesion, thus suggesting a mechanism for this association. However, we cannot rule out the possibility that statin use is an indicator of health status or some other unmeasured factor that may explain why BC did not appear to influence sICAM-1 and sVCAM-1 in statin users.\nThe question of which CAM is most closely associated with PM remains unresolved. Differences in the associations reported in epidemiological studies of the two CAMs could reflect differences in cell types that express the molecules or differences in the process of cleavage and shedding from endothelial cells. In the present study, the effects on both sICAM-1 and sVCAM-1 were consistent: The effect sizes were similar and the direction of the effects and the interactions were the same, even though we found differences in statistical significance. Thus, our study does not support the idea of a different underlying mechanism for the effects of TRAP on these two CAMs.\nThe expression of both ICAM-1 and VCAM-1 on the surface of vascular endothelial cells is associated with the formation of early atherosclerotic lesions. Although the relationship between the degree of cellular ICAM-1 and VCAM-1 expression and plasma concentrations of soluble forms is not entirely clear, multiple studies have shown that sICAM-1 and sVCAM-1 predict risk of cardiovascular disease (Albert and Ridker 1999; Pradhan et al. 2002; Rana et al. 2011; Ridker et al. 2003). In a recent study investigating the variability of sICAM-1 and sVCAM-1 measures over a 4-week period, Eschen et al. (2008) reported an estimated intrasubject variability of 7.6% for sICAM-1 and 9.5% for sVCAM-1, which suggests that these markers are relatively stable over a 4-week period.", "A major advantage of the present study is the use of a validated land-use regression model to characterize the individual-level differences in exposures instead of classifying exposure based on measurements at the nearest monitor, land-use regression models without BC measurements, or a weighted form of distance to roadway. Although the study is still limited by the lack of individual-level monitoring data at the home, our validation study suggests that our estimates are highly correlated with actual exposure measurements at locations in Boston other than those used to fit the model and are much more closely correlated with these measurements than are exposure estimates based on a central-site monitor. Although some exposure misclassification will still occur based on our model estimates, we expect that most of the residual error is Berkson type, based on a previous validation study for this model analyzing measurement error (Gryparis et al. 2009). Thus, we expect that the exposure misclassification will not bias effect estimates.\nThe lack of individual activity data also presents a limitation, but the NAS population consists of elderly men who spend a considerable amount of time at or near their home address compared with other population groups.\nOur sensitivity analysis of the block lags of weeks 5–8 and weeks 9–12 shows slight attenuation compared with the cumulative lags, suggesting that the cumulative effect may be dominated by the more recent weeks. The effect sizes for cumulative exposures do not attenuate when averaged up to 12 weeks. The 4-week block lags do not isolate the effects of those time windows because each week is correlated in both space and time; thus, future studies to model the specific contributions of different lag weeks would be beneficial in understanding these effects.", "Many studies have examined PM at different diameters, but fewer studies have looked at which components of PM are associated with adverse health effects. Our study links a health biomarker with BC, a specific component of PM2.5. In addition, we consider BC to be a surrogate for primary traffic PM (a specific source of air pollution), where the spatial variability of BC on a given day reflects the variability of traffic in the region, which is the basis of the BC model we developed and used for this study.\nWe did not examine other pollutants such as total PM2.5 or PM2.5 components other than BC in this study. We think it is unlikely that total PM2.5 or any other copollutant is driving the effect we observed for BC because the BC model is based largely on the daily spatial variation exhibited by BC. Other components of PM2.5, such as sulfates and organic PM, are more homogeneous over the study region. Although we cannot completely rule out confounding by copollutants or other factors, the correlation between the model-predicted 4-week BC and the corresponding 4-week averages of PM2.5 measured at the central site was low (0.108), which supports our belief that the effects observed are not driven by any temporal correlation with PM2.5 or one of its components.", "A limitation of this study is the restricted demographics of the study population. Study subjects were all elderly men, most of them white. Thus, we cannot generalize our results to other populations. However, the elderly represent a particularly susceptible population, and the growth in the number and proportion of older adults in the United States is unprecedented: by 2030, the proportion of the U.S. population ≥ 65 years of age will double to about 71 million older adults, or one in every five Americans (U.S. Census 2005)." ]

[ "methods", null, null, null, "methods", "methods", null, null, null, null, null ]

[ "Materials and Methods", "Study population", "Measurements of sICAM-1 and sVCAM-1", "BC exposure prediction", "Estimation of health effects", "Results", "Descriptive data", "Regression analysis", "Regression analysis and interactions", "Discussion", "TRAP and inflammatory response", "Mechanisms and interactions", "Exposure estimation", "Sources and components", "Generalizability", "Conclusions" ]

[ "[SUBTITLE] Study population [SUBSECTION] We studied participants in the Normative Aging Study (NAS), a longitudinal study established by the VA in 1963 (Bell et al. 1972). In brief, the NAS enrolled 2,280 men from the Greater Boston area who were initially free of known chronic medical conditions. All participants provided written informed consent, and the study was approved by the institutional review boards of all participating institutions. Participants visited the Boston VA Hospital study center every 3 years to undergo physical examinations. At each of these visits, blood samples and extensive physical examination, laboratory, anthropometric, and questionnaire data were collected. Information about cigarette smoking, medical history, and medication use were obtained by self-administered questionnaire. Each subject was interviewed to confirm the identity and purpose of medications used, and all new disease diagnoses were noted.\nDiabetes was defined as a physician diagnosis of diabetes, and obesity was defined as a body mass index (BMI) of at least 30 kg/m2. Self-reported data on diabetes status and statin use were updated at each study visit. In addition, BMI and obesity were updated based on height and weight measurements at each visit. Thus, NAS data reflect changes in disease status and medication use over time.\nMeasurements of sICAM-1 and sVCAM-1 began in 1999. For the present study, we included the 642 NAS participants with at least one measurement of sICAM-1 and sVCAM-1 and whose home address was in the Greater Boston area (1,423 total person-visits). Subjects who moved out of the area were excluded.\nWe studied participants in the Normative Aging Study (NAS), a longitudinal study established by the VA in 1963 (Bell et al. 1972). In brief, the NAS enrolled 2,280 men from the Greater Boston area who were initially free of known chronic medical conditions. All participants provided written informed consent, and the study was approved by the institutional review boards of all participating institutions. Participants visited the Boston VA Hospital study center every 3 years to undergo physical examinations. At each of these visits, blood samples and extensive physical examination, laboratory, anthropometric, and questionnaire data were collected. Information about cigarette smoking, medical history, and medication use were obtained by self-administered questionnaire. Each subject was interviewed to confirm the identity and purpose of medications used, and all new disease diagnoses were noted.\nDiabetes was defined as a physician diagnosis of diabetes, and obesity was defined as a body mass index (BMI) of at least 30 kg/m2. Self-reported data on diabetes status and statin use were updated at each study visit. In addition, BMI and obesity were updated based on height and weight measurements at each visit. Thus, NAS data reflect changes in disease status and medication use over time.\nMeasurements of sICAM-1 and sVCAM-1 began in 1999. For the present study, we included the 642 NAS participants with at least one measurement of sICAM-1 and sVCAM-1 and whose home address was in the Greater Boston area (1,423 total person-visits). Subjects who moved out of the area were excluded.\n[SUBTITLE] Measurements of sICAM-1 and sVCAM-1 [SUBSECTION] Blood samples routinely collected during medical exam visits from 1999 through 2008 were analyzed for sICAM-1 and sVCAM-1 in N. Rafai’s laboratory at Children’s Hospital Boston (Boston, MA). Plasma sICAM-1 and sVCAM-1 concentrations were measured in duplicate using the enzyme-linked immunosorbent assay (ELISA) method (R&D Systems, Minneapolis, MN), with a sensitivity of 0.35 ng/mL for sICAM-1 and 2.0 ng/mL for sVCAM-1 (Lim et al. 1999).\nBlood samples routinely collected during medical exam visits from 1999 through 2008 were analyzed for sICAM-1 and sVCAM-1 in N. Rafai’s laboratory at Children’s Hospital Boston (Boston, MA). Plasma sICAM-1 and sVCAM-1 concentrations were measured in duplicate using the enzyme-linked immunosorbent assay (ELISA) method (R&D Systems, Minneapolis, MN), with a sensitivity of 0.35 ng/mL for sICAM-1 and 2.0 ng/mL for sVCAM-1 (Lim et al. 1999).\n[SUBTITLE] BC exposure prediction [SUBSECTION] The BC exposure model and the stationary air monitors used to develop the model have been described in detail previously (Gryparis et al. 2007). Briefly, 82 sites were used; most sites measured BC continuously using aethalometers, and other sites collected particles on a filter over 24 hr and measured elemental carbon (EC) using reflectance analysis. The monitoring data used to develop our model included 6,031 observations from 2,079 unique exposure days.\nUsing a spatiotemporal model that we developed and validated previously (Gryparis et al. 2007), we estimated the 24-hr average BC concentration at each geocoded participant address. Predicted daily concentrations showed a > 3-fold range of variation in exposure across measurement sites (adjusted R2 = 0.83). A validation sample at 30 additional monitoring sites showed an average correlation of 0.59 between predicted and observed daily BC levels. We averaged the 24-hr predictions to form estimates for the 4, 8, and 12 weeks before each participant visit. We also averaged the 24-hr predictions to form estimates for the 4-week average during the 5–8 weeks before the study visit and the 4-week average during the 9–12 weeks before the study visit, which are components of the 8- and 12-week averages, to use as a sensitivity analysis.\nCovariates in the BC prediction model included measures of land use for each address (cumulative traffic density within 100 m, population density, distance to nearest major roadway, and percent urbanization), geographic information system (GIS) location (latitude, longitude), daily meteorological factors (apparent temperature, wind speed, and height of the planetary boundary layer), and other characteristics (day of week, day of season). The Boston central-site monitor was also included as a predictor to reflect average pollutant concentrations over the entire region on each day.\nSeparate models were fit for warm and cold seasons. Interaction terms between the temporal meteorological predictors and source-based geographic variables allowed for space–time interactions. Regression splines allowed main effect terms to nonlinearly predict exposure levels, and thin-plate splines modeled the residual spatial variability not explained by the spatial predictors. A latent variable framework was used to integrate BC and EC exposure data, where BC and EC measurements were treated as surrogates of some true, unobservable traffic exposure variable; see Gryparis et al. (2007) for further details.\nThe BC exposure model and the stationary air monitors used to develop the model have been described in detail previously (Gryparis et al. 2007). Briefly, 82 sites were used; most sites measured BC continuously using aethalometers, and other sites collected particles on a filter over 24 hr and measured elemental carbon (EC) using reflectance analysis. The monitoring data used to develop our model included 6,031 observations from 2,079 unique exposure days.\nUsing a spatiotemporal model that we developed and validated previously (Gryparis et al. 2007), we estimated the 24-hr average BC concentration at each geocoded participant address. Predicted daily concentrations showed a > 3-fold range of variation in exposure across measurement sites (adjusted R2 = 0.83). A validation sample at 30 additional monitoring sites showed an average correlation of 0.59 between predicted and observed daily BC levels. We averaged the 24-hr predictions to form estimates for the 4, 8, and 12 weeks before each participant visit. We also averaged the 24-hr predictions to form estimates for the 4-week average during the 5–8 weeks before the study visit and the 4-week average during the 9–12 weeks before the study visit, which are components of the 8- and 12-week averages, to use as a sensitivity analysis.\nCovariates in the BC prediction model included measures of land use for each address (cumulative traffic density within 100 m, population density, distance to nearest major roadway, and percent urbanization), geographic information system (GIS) location (latitude, longitude), daily meteorological factors (apparent temperature, wind speed, and height of the planetary boundary layer), and other characteristics (day of week, day of season). The Boston central-site monitor was also included as a predictor to reflect average pollutant concentrations over the entire region on each day.\nSeparate models were fit for warm and cold seasons. Interaction terms between the temporal meteorological predictors and source-based geographic variables allowed for space–time interactions. Regression splines allowed main effect terms to nonlinearly predict exposure levels, and thin-plate splines modeled the residual spatial variability not explained by the spatial predictors. A latent variable framework was used to integrate BC and EC exposure data, where BC and EC measurements were treated as surrogates of some true, unobservable traffic exposure variable; see Gryparis et al. (2007) for further details.\n[SUBTITLE] Estimation of health effects [SUBSECTION] We log-transformed sICAM-1 and sVCAM-1 levels to increase normality and stabilize variance in the residuals. Model covariates were selected a priori, and all models included age, BMI, diabetes, smoking status, pack-years, and season. We used mixed models to account for correlation among measurements on the same subject across different medical visits. Mixed models have the form\nwhere Yij is log(sICAM-1) or log(sVCAM-1) in subject i on day j, and ui represents a subject-specific intercept that reflects unexplained heterogeneity in the outcome. BC averages and model covariates are modeled as fixed linear effects, and ui is modeled as a random effect. We assume that the ui are generated from a normal distribution with common variance, yielding the simple compound symmetry variance structure. This model requires estimation of two variance components, which represent between- and within-subject variation. Models with unbalanced data (i.e., varying numbers of repeated measurements on each subject) typically yield accurate estimates of within-subject variation, provided a sufficient number of repeated measurements contribute to the estimate.\nModels used to examine effect modification by obesity, diabetes, and statin use included interaction terms that allowed associations between BC and the outcomes to vary among subgroups. Diabetes, obesity, and statin use were treated as time-varying covariates, where the status was updated at each visit to reflect changes since the last visit. The percentages of subjects whose status changed for these factors over the study period were 3% for diabetes, 9% for obesity, and 21% for statin use.\nWe also performed a sensitivity analysis to investigate whether the interaction with statin use was heavily influenced by the people who began taking statins after the study period began. Specifically, we limited the interaction analysis to the participants who never used statins during the study period (n = 284) and those who used statins throughout the entire study period (n = 223).\nBecause the outcomes were log-transformed, effect estimates are reported as percent changes in sICAM-1 and sVCAM-1 concentrations associated with a 0.30-μg/m3 increase in BC, which corresponds to the interquartile range (IQR) for average BC exposures over all three time intervals (4, 8, and 12 weeks). A level of α = 0.05 was used to determine statistical significance.\nWe log-transformed sICAM-1 and sVCAM-1 levels to increase normality and stabilize variance in the residuals. Model covariates were selected a priori, and all models included age, BMI, diabetes, smoking status, pack-years, and season. We used mixed models to account for correlation among measurements on the same subject across different medical visits. Mixed models have the form\nwhere Yij is log(sICAM-1) or log(sVCAM-1) in subject i on day j, and ui represents a subject-specific intercept that reflects unexplained heterogeneity in the outcome. BC averages and model covariates are modeled as fixed linear effects, and ui is modeled as a random effect. We assume that the ui are generated from a normal distribution with common variance, yielding the simple compound symmetry variance structure. This model requires estimation of two variance components, which represent between- and within-subject variation. Models with unbalanced data (i.e., varying numbers of repeated measurements on each subject) typically yield accurate estimates of within-subject variation, provided a sufficient number of repeated measurements contribute to the estimate.\nModels used to examine effect modification by obesity, diabetes, and statin use included interaction terms that allowed associations between BC and the outcomes to vary among subgroups. Diabetes, obesity, and statin use were treated as time-varying covariates, where the status was updated at each visit to reflect changes since the last visit. The percentages of subjects whose status changed for these factors over the study period were 3% for diabetes, 9% for obesity, and 21% for statin use.\nWe also performed a sensitivity analysis to investigate whether the interaction with statin use was heavily influenced by the people who began taking statins after the study period began. Specifically, we limited the interaction analysis to the participants who never used statins during the study period (n = 284) and those who used statins throughout the entire study period (n = 223).\nBecause the outcomes were log-transformed, effect estimates are reported as percent changes in sICAM-1 and sVCAM-1 concentrations associated with a 0.30-μg/m3 increase in BC, which corresponds to the interquartile range (IQR) for average BC exposures over all three time intervals (4, 8, and 12 weeks). A level of α = 0.05 was used to determine statistical significance.", "We studied participants in the Normative Aging Study (NAS), a longitudinal study established by the VA in 1963 (Bell et al. 1972). In brief, the NAS enrolled 2,280 men from the Greater Boston area who were initially free of known chronic medical conditions. All participants provided written informed consent, and the study was approved by the institutional review boards of all participating institutions. Participants visited the Boston VA Hospital study center every 3 years to undergo physical examinations. At each of these visits, blood samples and extensive physical examination, laboratory, anthropometric, and questionnaire data were collected. Information about cigarette smoking, medical history, and medication use were obtained by self-administered questionnaire. Each subject was interviewed to confirm the identity and purpose of medications used, and all new disease diagnoses were noted.\nDiabetes was defined as a physician diagnosis of diabetes, and obesity was defined as a body mass index (BMI) of at least 30 kg/m2. Self-reported data on diabetes status and statin use were updated at each study visit. In addition, BMI and obesity were updated based on height and weight measurements at each visit. Thus, NAS data reflect changes in disease status and medication use over time.\nMeasurements of sICAM-1 and sVCAM-1 began in 1999. For the present study, we included the 642 NAS participants with at least one measurement of sICAM-1 and sVCAM-1 and whose home address was in the Greater Boston area (1,423 total person-visits). Subjects who moved out of the area were excluded.", "Blood samples routinely collected during medical exam visits from 1999 through 2008 were analyzed for sICAM-1 and sVCAM-1 in N. Rafai’s laboratory at Children’s Hospital Boston (Boston, MA). Plasma sICAM-1 and sVCAM-1 concentrations were measured in duplicate using the enzyme-linked immunosorbent assay (ELISA) method (R&D Systems, Minneapolis, MN), with a sensitivity of 0.35 ng/mL for sICAM-1 and 2.0 ng/mL for sVCAM-1 (Lim et al. 1999).", "The BC exposure model and the stationary air monitors used to develop the model have been described in detail previously (Gryparis et al. 2007). Briefly, 82 sites were used; most sites measured BC continuously using aethalometers, and other sites collected particles on a filter over 24 hr and measured elemental carbon (EC) using reflectance analysis. The monitoring data used to develop our model included 6,031 observations from 2,079 unique exposure days.\nUsing a spatiotemporal model that we developed and validated previously (Gryparis et al. 2007), we estimated the 24-hr average BC concentration at each geocoded participant address. Predicted daily concentrations showed a > 3-fold range of variation in exposure across measurement sites (adjusted R2 = 0.83). A validation sample at 30 additional monitoring sites showed an average correlation of 0.59 between predicted and observed daily BC levels. We averaged the 24-hr predictions to form estimates for the 4, 8, and 12 weeks before each participant visit. We also averaged the 24-hr predictions to form estimates for the 4-week average during the 5–8 weeks before the study visit and the 4-week average during the 9–12 weeks before the study visit, which are components of the 8- and 12-week averages, to use as a sensitivity analysis.\nCovariates in the BC prediction model included measures of land use for each address (cumulative traffic density within 100 m, population density, distance to nearest major roadway, and percent urbanization), geographic information system (GIS) location (latitude, longitude), daily meteorological factors (apparent temperature, wind speed, and height of the planetary boundary layer), and other characteristics (day of week, day of season). The Boston central-site monitor was also included as a predictor to reflect average pollutant concentrations over the entire region on each day.\nSeparate models were fit for warm and cold seasons. Interaction terms between the temporal meteorological predictors and source-based geographic variables allowed for space–time interactions. Regression splines allowed main effect terms to nonlinearly predict exposure levels, and thin-plate splines modeled the residual spatial variability not explained by the spatial predictors. A latent variable framework was used to integrate BC and EC exposure data, where BC and EC measurements were treated as surrogates of some true, unobservable traffic exposure variable; see Gryparis et al. (2007) for further details.", "We log-transformed sICAM-1 and sVCAM-1 levels to increase normality and stabilize variance in the residuals. Model covariates were selected a priori, and all models included age, BMI, diabetes, smoking status, pack-years, and season. We used mixed models to account for correlation among measurements on the same subject across different medical visits. Mixed models have the form\nwhere Yij is log(sICAM-1) or log(sVCAM-1) in subject i on day j, and ui represents a subject-specific intercept that reflects unexplained heterogeneity in the outcome. BC averages and model covariates are modeled as fixed linear effects, and ui is modeled as a random effect. We assume that the ui are generated from a normal distribution with common variance, yielding the simple compound symmetry variance structure. This model requires estimation of two variance components, which represent between- and within-subject variation. Models with unbalanced data (i.e., varying numbers of repeated measurements on each subject) typically yield accurate estimates of within-subject variation, provided a sufficient number of repeated measurements contribute to the estimate.\nModels used to examine effect modification by obesity, diabetes, and statin use included interaction terms that allowed associations between BC and the outcomes to vary among subgroups. Diabetes, obesity, and statin use were treated as time-varying covariates, where the status was updated at each visit to reflect changes since the last visit. The percentages of subjects whose status changed for these factors over the study period were 3% for diabetes, 9% for obesity, and 21% for statin use.\nWe also performed a sensitivity analysis to investigate whether the interaction with statin use was heavily influenced by the people who began taking statins after the study period began. Specifically, we limited the interaction analysis to the participants who never used statins during the study period (n = 284) and those who used statins throughout the entire study period (n = 223).\nBecause the outcomes were log-transformed, effect estimates are reported as percent changes in sICAM-1 and sVCAM-1 concentrations associated with a 0.30-μg/m3 increase in BC, which corresponds to the interquartile range (IQR) for average BC exposures over all three time intervals (4, 8, and 12 weeks). A level of α = 0.05 was used to determine statistical significance.", "[SUBTITLE] Descriptive data [SUBSECTION] Subjects were elderly, with a mean age of 72 years (range, 56–100 years) at the first study visit (Table 1). Most subjects were overweight (median BMI, 27.5 kg/m2; range, 18–52 kg/m2). Average 4-week BC exposure estimates ranged from 0 to 3.26 μg/m3 (mean ± SD, 0.42 ± 0.30 μg/m3; median, 0.37 μg/m3; IQR, 0.30 μg/m3). For both 8- and 12-week exposures, exposure estimates were 0.42 ± 0.29 μg/m3 (median, 0.38 μg/m3; IQR, 0.30 μg/m3). Figure 1 illustrates the spatial distribution of predicted BC in the study area during an arbitrary 4-week period (February 2001). The distributions of BC exposures looked similar among subjects classified according to obesity, diabetes, statin use, or smoking status, and we found no significant differences in the mean BC exposure assessed at baseline.\nSubjects were elderly, with a mean age of 72 years (range, 56–100 years) at the first study visit (Table 1). Most subjects were overweight (median BMI, 27.5 kg/m2; range, 18–52 kg/m2). Average 4-week BC exposure estimates ranged from 0 to 3.26 μg/m3 (mean ± SD, 0.42 ± 0.30 μg/m3; median, 0.37 μg/m3; IQR, 0.30 μg/m3). For both 8- and 12-week exposures, exposure estimates were 0.42 ± 0.29 μg/m3 (median, 0.38 μg/m3; IQR, 0.30 μg/m3). Figure 1 illustrates the spatial distribution of predicted BC in the study area during an arbitrary 4-week period (February 2001). The distributions of BC exposures looked similar among subjects classified according to obesity, diabetes, statin use, or smoking status, and we found no significant differences in the mean BC exposure assessed at baseline.\n[SUBTITLE] Regression analysis [SUBSECTION] We found significant positive associations between sICAM-1 and BC averaged over the previous 4, 8, and 12 weeks (Table 2). Associations between average BC for 4, 8, and 12 weeks and sVCAM-1 were in the same direction, with slightly smaller effect size, but were not statistically significant. As a sensitivity analysis, we also estimated the effects of BC exposure in weeks 5–8 and weeks 9–12 separately. For sICAM-1, the association with the longer BC lags were positive with effect size slightly smaller than the effect size for weeks 1–4 and not quite statistically significant. A 0.30-μg/m3 increase in BC exposure during weeks 5–8 was associated with a 1.22% increase in sICAM-1 [95% confidence interval (CI), −0.10% to 3.00%], and during weeks 9–12, with a 1.09% increase in sICAM-1 (95% CI, −0.18% to 2.38%). For sVCAM-1, the estimated effects for the longer lags had the same effect size and direction as the effects for weeks 1–4 but were not statistically significant (data not shown). Overall, we saw the same pattern of effects in the lagged 4-week intervals, with only slight weakening in effect when the lags were considered individually rather than cumulatively.\nWe found significant positive associations between sICAM-1 and BC averaged over the previous 4, 8, and 12 weeks (Table 2). Associations between average BC for 4, 8, and 12 weeks and sVCAM-1 were in the same direction, with slightly smaller effect size, but were not statistically significant. As a sensitivity analysis, we also estimated the effects of BC exposure in weeks 5–8 and weeks 9–12 separately. For sICAM-1, the association with the longer BC lags were positive with effect size slightly smaller than the effect size for weeks 1–4 and not quite statistically significant. A 0.30-μg/m3 increase in BC exposure during weeks 5–8 was associated with a 1.22% increase in sICAM-1 [95% confidence interval (CI), −0.10% to 3.00%], and during weeks 9–12, with a 1.09% increase in sICAM-1 (95% CI, −0.18% to 2.38%). For sVCAM-1, the estimated effects for the longer lags had the same effect size and direction as the effects for weeks 1–4 but were not statistically significant (data not shown). Overall, we saw the same pattern of effects in the lagged 4-week intervals, with only slight weakening in effect when the lags were considered individually rather than cumulatively.\n[SUBTITLE] Regression analysis and interactions [SUBSECTION] When we examined the effect of BC on sICAM-1 and sVCAM-1 by diabetes status, we saw effects only among diabetics, whereas nondiabetics showed no effect (Figure 2; interaction p-values < 0.05 for all models except 12-week BC and sVCAM-1). When we examined the effects of BC by statin use, interaction models suggested that effects were present only in participants who did not use statins, with little evidence of associations among statin users (Figure 3; interaction p-values < 0.05 for sVCAM-1 only). In contrast, we found no difference in the estimated effect of BC by obesity status on either outcome (data not shown).\nWe restricted our sensitivity analysis of statin use interaction to participants who never used statins or always used statins during the study period. The results were comparable, with the effect sizes estimated to be slightly larger in all categories in the restricted model, and statistically significant (data not shown). These differences could reflect some overall difference in health between statin users and nonusers because statin use was not randomized.\nWhen we examined the effect of BC on sICAM-1 and sVCAM-1 by diabetes status, we saw effects only among diabetics, whereas nondiabetics showed no effect (Figure 2; interaction p-values < 0.05 for all models except 12-week BC and sVCAM-1). When we examined the effects of BC by statin use, interaction models suggested that effects were present only in participants who did not use statins, with little evidence of associations among statin users (Figure 3; interaction p-values < 0.05 for sVCAM-1 only). In contrast, we found no difference in the estimated effect of BC by obesity status on either outcome (data not shown).\nWe restricted our sensitivity analysis of statin use interaction to participants who never used statins or always used statins during the study period. The results were comparable, with the effect sizes estimated to be slightly larger in all categories in the restricted model, and statistically significant (data not shown). These differences could reflect some overall difference in health between statin users and nonusers because statin use was not randomized.", "Subjects were elderly, with a mean age of 72 years (range, 56–100 years) at the first study visit (Table 1). Most subjects were overweight (median BMI, 27.5 kg/m2; range, 18–52 kg/m2). Average 4-week BC exposure estimates ranged from 0 to 3.26 μg/m3 (mean ± SD, 0.42 ± 0.30 μg/m3; median, 0.37 μg/m3; IQR, 0.30 μg/m3). For both 8- and 12-week exposures, exposure estimates were 0.42 ± 0.29 μg/m3 (median, 0.38 μg/m3; IQR, 0.30 μg/m3). Figure 1 illustrates the spatial distribution of predicted BC in the study area during an arbitrary 4-week period (February 2001). The distributions of BC exposures looked similar among subjects classified according to obesity, diabetes, statin use, or smoking status, and we found no significant differences in the mean BC exposure assessed at baseline.", "We found significant positive associations between sICAM-1 and BC averaged over the previous 4, 8, and 12 weeks (Table 2). Associations between average BC for 4, 8, and 12 weeks and sVCAM-1 were in the same direction, with slightly smaller effect size, but were not statistically significant. As a sensitivity analysis, we also estimated the effects of BC exposure in weeks 5–8 and weeks 9–12 separately. For sICAM-1, the association with the longer BC lags were positive with effect size slightly smaller than the effect size for weeks 1–4 and not quite statistically significant. A 0.30-μg/m3 increase in BC exposure during weeks 5–8 was associated with a 1.22% increase in sICAM-1 [95% confidence interval (CI), −0.10% to 3.00%], and during weeks 9–12, with a 1.09% increase in sICAM-1 (95% CI, −0.18% to 2.38%). For sVCAM-1, the estimated effects for the longer lags had the same effect size and direction as the effects for weeks 1–4 but were not statistically significant (data not shown). Overall, we saw the same pattern of effects in the lagged 4-week intervals, with only slight weakening in effect when the lags were considered individually rather than cumulatively.", "When we examined the effect of BC on sICAM-1 and sVCAM-1 by diabetes status, we saw effects only among diabetics, whereas nondiabetics showed no effect (Figure 2; interaction p-values < 0.05 for all models except 12-week BC and sVCAM-1). When we examined the effects of BC by statin use, interaction models suggested that effects were present only in participants who did not use statins, with little evidence of associations among statin users (Figure 3; interaction p-values < 0.05 for sVCAM-1 only). In contrast, we found no difference in the estimated effect of BC by obesity status on either outcome (data not shown).\nWe restricted our sensitivity analysis of statin use interaction to participants who never used statins or always used statins during the study period. The results were comparable, with the effect sizes estimated to be slightly larger in all categories in the restricted model, and statistically significant (data not shown). These differences could reflect some overall difference in health between statin users and nonusers because statin use was not randomized.", "We found that address-specific BC exposures averaged over 4, 8, and 12 weeks were positively associated with markers of inflammation and endothelial dysfunction in this elderly cohort. Exposure estimates based on our validated land-use regression model are more accurate than estimates based on ambient monitoring, which are often used for cohorts of this size. Our analyses suggest that diabetics are more susceptible to adverse effects of TRAP than are nondiabetics, but we found no evidence of effect modification by obesity. In addition, we observed a null effect among participants who used statins and a positive association for those not taking statins, but further investigation is needed to clarify this potential effect.\n[SUBTITLE] TRAP and inflammatory response [SUBSECTION] Several studies have examined the effects of short-term (i.e., < 24 hr to 3 weeks) TRAP and various markers of inflammation. For example, in a large epidemiological study, daily increases in ambient PM levels were positively associated with plasma fibrinogen levels, with the strongest association observed in participants with chronic obstructive lung disease (Schwartz 2001). In a study of a small (n = 29) elderly population, Delfino et al. (2008) examined inflammatory markers and PM components measured at each subject’s home for 1–9 days on average. The authors reported that several traffic-related components were significantly associated with CRP and interleukin-6, and observed positive but nonsignificant associations with sICAM-1 and sVCAM-1.\nIn other studies of NAS participants, traffic-related PM (BC and organic carbon) exposure has also been positively associated with plasma total homocysteine concentrations (Park et al. 2008), and this association is modified by polymorphisms in genes related to oxidative stress (Ren et al. 2010). PM2.5 and BC averages of 1–3 days measured at the central site have been associated with increased vascular cell adhesion in NAS participants (Madrigano et al. 2010). Traffic-related PM (particle number and BC) has also been positively associated with inflammatory markers (CRP, white blood cell count, sediment rate, and fibrinogen) in NAS participants, with stronger associations with particle numbers than with BC and stronger associations with BC averaged over 4 weeks than averaged over 48 hr or 1 week (Zeka et al. 2006).\nFew controlled studies of PM on human inflammatory response have been conducted, but one reported that plasma fibrinogen increased after exposure to urban PM (Ghio et al. 2000), and another reported that peripheral neutrophils, sVCAM-1, and sICAM-1 increased after a 1-hr exposure to diesel PM (Salvi et al. 1999).\nSeveral studies have examined the effects of short-term (i.e., < 24 hr to 3 weeks) TRAP and various markers of inflammation. For example, in a large epidemiological study, daily increases in ambient PM levels were positively associated with plasma fibrinogen levels, with the strongest association observed in participants with chronic obstructive lung disease (Schwartz 2001). In a study of a small (n = 29) elderly population, Delfino et al. (2008) examined inflammatory markers and PM components measured at each subject’s home for 1–9 days on average. The authors reported that several traffic-related components were significantly associated with CRP and interleukin-6, and observed positive but nonsignificant associations with sICAM-1 and sVCAM-1.\nIn other studies of NAS participants, traffic-related PM (BC and organic carbon) exposure has also been positively associated with plasma total homocysteine concentrations (Park et al. 2008), and this association is modified by polymorphisms in genes related to oxidative stress (Ren et al. 2010). PM2.5 and BC averages of 1–3 days measured at the central site have been associated with increased vascular cell adhesion in NAS participants (Madrigano et al. 2010). Traffic-related PM (particle number and BC) has also been positively associated with inflammatory markers (CRP, white blood cell count, sediment rate, and fibrinogen) in NAS participants, with stronger associations with particle numbers than with BC and stronger associations with BC averaged over 4 weeks than averaged over 48 hr or 1 week (Zeka et al. 2006).\nFew controlled studies of PM on human inflammatory response have been conducted, but one reported that plasma fibrinogen increased after exposure to urban PM (Ghio et al. 2000), and another reported that peripheral neutrophils, sVCAM-1, and sICAM-1 increased after a 1-hr exposure to diesel PM (Salvi et al. 1999).\n[SUBTITLE] Mechanisms and interactions [SUBSECTION] Diabetics have impaired endothelial function compared with nondiabetics (Calles-Escandon and Cipolla 2001), and there is increasing evidence that diabetics are more susceptible to the effects of air pollution. Inverse associations between 60-day average BC exposure and brachial artery flow-mediated dilation (FMD) were stronger in diabetics than in nondiabetics (O’Neill et al. 2005). Associations between 24-hr exposure to PM2.5 and FMD were stronger among diabetics with markers of severe insulin resistance compared with other diabetics (Schneider et al. 2008).\nIn contrast with our findings, obese NAS participants have been reported to have stronger associations between short-term BC exposure and plasma CRP, erythrocyte sediment rate (Zeka et al. 2006), and sVCAM-1 (Madrigano et al. 2010) than nonobese participants.\nOur finding that statin users are less susceptible than those not taking statins is generally consistent with other studies. In diabetics, stronger associations of PM2.5 and BC with sVCAM-1 were reported among those not using statins than in statin users (O’Neill et al. 2007). Associations between CRP and traffic PM exposures of 5 and 9 days were reported to be stronger among those using statins than those not using statins (Delfino et al. 2008). Statins promote endothelial nitric oxide release, which reduces cell adhesion, thus suggesting a mechanism for this association. However, we cannot rule out the possibility that statin use is an indicator of health status or some other unmeasured factor that may explain why BC did not appear to influence sICAM-1 and sVCAM-1 in statin users.\nThe question of which CAM is most closely associated with PM remains unresolved. Differences in the associations reported in epidemiological studies of the two CAMs could reflect differences in cell types that express the molecules or differences in the process of cleavage and shedding from endothelial cells. In the present study, the effects on both sICAM-1 and sVCAM-1 were consistent: The effect sizes were similar and the direction of the effects and the interactions were the same, even though we found differences in statistical significance. Thus, our study does not support the idea of a different underlying mechanism for the effects of TRAP on these two CAMs.\nThe expression of both ICAM-1 and VCAM-1 on the surface of vascular endothelial cells is associated with the formation of early atherosclerotic lesions. Although the relationship between the degree of cellular ICAM-1 and VCAM-1 expression and plasma concentrations of soluble forms is not entirely clear, multiple studies have shown that sICAM-1 and sVCAM-1 predict risk of cardiovascular disease (Albert and Ridker 1999; Pradhan et al. 2002; Rana et al. 2011; Ridker et al. 2003). In a recent study investigating the variability of sICAM-1 and sVCAM-1 measures over a 4-week period, Eschen et al. (2008) reported an estimated intrasubject variability of 7.6% for sICAM-1 and 9.5% for sVCAM-1, which suggests that these markers are relatively stable over a 4-week period.\nDiabetics have impaired endothelial function compared with nondiabetics (Calles-Escandon and Cipolla 2001), and there is increasing evidence that diabetics are more susceptible to the effects of air pollution. Inverse associations between 60-day average BC exposure and brachial artery flow-mediated dilation (FMD) were stronger in diabetics than in nondiabetics (O’Neill et al. 2005). Associations between 24-hr exposure to PM2.5 and FMD were stronger among diabetics with markers of severe insulin resistance compared with other diabetics (Schneider et al. 2008).\nIn contrast with our findings, obese NAS participants have been reported to have stronger associations between short-term BC exposure and plasma CRP, erythrocyte sediment rate (Zeka et al. 2006), and sVCAM-1 (Madrigano et al. 2010) than nonobese participants.\nOur finding that statin users are less susceptible than those not taking statins is generally consistent with other studies. In diabetics, stronger associations of PM2.5 and BC with sVCAM-1 were reported among those not using statins than in statin users (O’Neill et al. 2007). Associations between CRP and traffic PM exposures of 5 and 9 days were reported to be stronger among those using statins than those not using statins (Delfino et al. 2008). Statins promote endothelial nitric oxide release, which reduces cell adhesion, thus suggesting a mechanism for this association. However, we cannot rule out the possibility that statin use is an indicator of health status or some other unmeasured factor that may explain why BC did not appear to influence sICAM-1 and sVCAM-1 in statin users.\nThe question of which CAM is most closely associated with PM remains unresolved. Differences in the associations reported in epidemiological studies of the two CAMs could reflect differences in cell types that express the molecules or differences in the process of cleavage and shedding from endothelial cells. In the present study, the effects on both sICAM-1 and sVCAM-1 were consistent: The effect sizes were similar and the direction of the effects and the interactions were the same, even though we found differences in statistical significance. Thus, our study does not support the idea of a different underlying mechanism for the effects of TRAP on these two CAMs.\nThe expression of both ICAM-1 and VCAM-1 on the surface of vascular endothelial cells is associated with the formation of early atherosclerotic lesions. Although the relationship between the degree of cellular ICAM-1 and VCAM-1 expression and plasma concentrations of soluble forms is not entirely clear, multiple studies have shown that sICAM-1 and sVCAM-1 predict risk of cardiovascular disease (Albert and Ridker 1999; Pradhan et al. 2002; Rana et al. 2011; Ridker et al. 2003). In a recent study investigating the variability of sICAM-1 and sVCAM-1 measures over a 4-week period, Eschen et al. (2008) reported an estimated intrasubject variability of 7.6% for sICAM-1 and 9.5% for sVCAM-1, which suggests that these markers are relatively stable over a 4-week period.\n[SUBTITLE] Exposure estimation [SUBSECTION] A major advantage of the present study is the use of a validated land-use regression model to characterize the individual-level differences in exposures instead of classifying exposure based on measurements at the nearest monitor, land-use regression models without BC measurements, or a weighted form of distance to roadway. Although the study is still limited by the lack of individual-level monitoring data at the home, our validation study suggests that our estimates are highly correlated with actual exposure measurements at locations in Boston other than those used to fit the model and are much more closely correlated with these measurements than are exposure estimates based on a central-site monitor. Although some exposure misclassification will still occur based on our model estimates, we expect that most of the residual error is Berkson type, based on a previous validation study for this model analyzing measurement error (Gryparis et al. 2009). Thus, we expect that the exposure misclassification will not bias effect estimates.\nThe lack of individual activity data also presents a limitation, but the NAS population consists of elderly men who spend a considerable amount of time at or near their home address compared with other population groups.\nOur sensitivity analysis of the block lags of weeks 5–8 and weeks 9–12 shows slight attenuation compared with the cumulative lags, suggesting that the cumulative effect may be dominated by the more recent weeks. The effect sizes for cumulative exposures do not attenuate when averaged up to 12 weeks. The 4-week block lags do not isolate the effects of those time windows because each week is correlated in both space and time; thus, future studies to model the specific contributions of different lag weeks would be beneficial in understanding these effects.\nA major advantage of the present study is the use of a validated land-use regression model to characterize the individual-level differences in exposures instead of classifying exposure based on measurements at the nearest monitor, land-use regression models without BC measurements, or a weighted form of distance to roadway. Although the study is still limited by the lack of individual-level monitoring data at the home, our validation study suggests that our estimates are highly correlated with actual exposure measurements at locations in Boston other than those used to fit the model and are much more closely correlated with these measurements than are exposure estimates based on a central-site monitor. Although some exposure misclassification will still occur based on our model estimates, we expect that most of the residual error is Berkson type, based on a previous validation study for this model analyzing measurement error (Gryparis et al. 2009). Thus, we expect that the exposure misclassification will not bias effect estimates.\nThe lack of individual activity data also presents a limitation, but the NAS population consists of elderly men who spend a considerable amount of time at or near their home address compared with other population groups.\nOur sensitivity analysis of the block lags of weeks 5–8 and weeks 9–12 shows slight attenuation compared with the cumulative lags, suggesting that the cumulative effect may be dominated by the more recent weeks. The effect sizes for cumulative exposures do not attenuate when averaged up to 12 weeks. The 4-week block lags do not isolate the effects of those time windows because each week is correlated in both space and time; thus, future studies to model the specific contributions of different lag weeks would be beneficial in understanding these effects.\n[SUBTITLE] Sources and components [SUBSECTION] Many studies have examined PM at different diameters, but fewer studies have looked at which components of PM are associated with adverse health effects. Our study links a health biomarker with BC, a specific component of PM2.5. In addition, we consider BC to be a surrogate for primary traffic PM (a specific source of air pollution), where the spatial variability of BC on a given day reflects the variability of traffic in the region, which is the basis of the BC model we developed and used for this study.\nWe did not examine other pollutants such as total PM2.5 or PM2.5 components other than BC in this study. We think it is unlikely that total PM2.5 or any other copollutant is driving the effect we observed for BC because the BC model is based largely on the daily spatial variation exhibited by BC. Other components of PM2.5, such as sulfates and organic PM, are more homogeneous over the study region. Although we cannot completely rule out confounding by copollutants or other factors, the correlation between the model-predicted 4-week BC and the corresponding 4-week averages of PM2.5 measured at the central site was low (0.108), which supports our belief that the effects observed are not driven by any temporal correlation with PM2.5 or one of its components.\nMany studies have examined PM at different diameters, but fewer studies have looked at which components of PM are associated with adverse health effects. Our study links a health biomarker with BC, a specific component of PM2.5. In addition, we consider BC to be a surrogate for primary traffic PM (a specific source of air pollution), where the spatial variability of BC on a given day reflects the variability of traffic in the region, which is the basis of the BC model we developed and used for this study.\nWe did not examine other pollutants such as total PM2.5 or PM2.5 components other than BC in this study. We think it is unlikely that total PM2.5 or any other copollutant is driving the effect we observed for BC because the BC model is based largely on the daily spatial variation exhibited by BC. Other components of PM2.5, such as sulfates and organic PM, are more homogeneous over the study region. Although we cannot completely rule out confounding by copollutants or other factors, the correlation between the model-predicted 4-week BC and the corresponding 4-week averages of PM2.5 measured at the central site was low (0.108), which supports our belief that the effects observed are not driven by any temporal correlation with PM2.5 or one of its components.\n[SUBTITLE] Generalizability [SUBSECTION] A limitation of this study is the restricted demographics of the study population. Study subjects were all elderly men, most of them white. Thus, we cannot generalize our results to other populations. However, the elderly represent a particularly susceptible population, and the growth in the number and proportion of older adults in the United States is unprecedented: by 2030, the proportion of the U.S. population ≥ 65 years of age will double to about 71 million older adults, or one in every five Americans (U.S. Census 2005).\nA limitation of this study is the restricted demographics of the study population. Study subjects were all elderly men, most of them white. Thus, we cannot generalize our results to other populations. However, the elderly represent a particularly susceptible population, and the growth in the number and proportion of older adults in the United States is unprecedented: by 2030, the proportion of the U.S. population ≥ 65 years of age will double to about 71 million older adults, or one in every five Americans (U.S. Census 2005).", "Several studies have examined the effects of short-term (i.e., < 24 hr to 3 weeks) TRAP and various markers of inflammation. For example, in a large epidemiological study, daily increases in ambient PM levels were positively associated with plasma fibrinogen levels, with the strongest association observed in participants with chronic obstructive lung disease (Schwartz 2001). In a study of a small (n = 29) elderly population, Delfino et al. (2008) examined inflammatory markers and PM components measured at each subject’s home for 1–9 days on average. The authors reported that several traffic-related components were significantly associated with CRP and interleukin-6, and observed positive but nonsignificant associations with sICAM-1 and sVCAM-1.\nIn other studies of NAS participants, traffic-related PM (BC and organic carbon) exposure has also been positively associated with plasma total homocysteine concentrations (Park et al. 2008), and this association is modified by polymorphisms in genes related to oxidative stress (Ren et al. 2010). PM2.5 and BC averages of 1–3 days measured at the central site have been associated with increased vascular cell adhesion in NAS participants (Madrigano et al. 2010). Traffic-related PM (particle number and BC) has also been positively associated with inflammatory markers (CRP, white blood cell count, sediment rate, and fibrinogen) in NAS participants, with stronger associations with particle numbers than with BC and stronger associations with BC averaged over 4 weeks than averaged over 48 hr or 1 week (Zeka et al. 2006).\nFew controlled studies of PM on human inflammatory response have been conducted, but one reported that plasma fibrinogen increased after exposure to urban PM (Ghio et al. 2000), and another reported that peripheral neutrophils, sVCAM-1, and sICAM-1 increased after a 1-hr exposure to diesel PM (Salvi et al. 1999).", "Diabetics have impaired endothelial function compared with nondiabetics (Calles-Escandon and Cipolla 2001), and there is increasing evidence that diabetics are more susceptible to the effects of air pollution. Inverse associations between 60-day average BC exposure and brachial artery flow-mediated dilation (FMD) were stronger in diabetics than in nondiabetics (O’Neill et al. 2005). Associations between 24-hr exposure to PM2.5 and FMD were stronger among diabetics with markers of severe insulin resistance compared with other diabetics (Schneider et al. 2008).\nIn contrast with our findings, obese NAS participants have been reported to have stronger associations between short-term BC exposure and plasma CRP, erythrocyte sediment rate (Zeka et al. 2006), and sVCAM-1 (Madrigano et al. 2010) than nonobese participants.\nOur finding that statin users are less susceptible than those not taking statins is generally consistent with other studies. In diabetics, stronger associations of PM2.5 and BC with sVCAM-1 were reported among those not using statins than in statin users (O’Neill et al. 2007). Associations between CRP and traffic PM exposures of 5 and 9 days were reported to be stronger among those using statins than those not using statins (Delfino et al. 2008). Statins promote endothelial nitric oxide release, which reduces cell adhesion, thus suggesting a mechanism for this association. However, we cannot rule out the possibility that statin use is an indicator of health status or some other unmeasured factor that may explain why BC did not appear to influence sICAM-1 and sVCAM-1 in statin users.\nThe question of which CAM is most closely associated with PM remains unresolved. Differences in the associations reported in epidemiological studies of the two CAMs could reflect differences in cell types that express the molecules or differences in the process of cleavage and shedding from endothelial cells. In the present study, the effects on both sICAM-1 and sVCAM-1 were consistent: The effect sizes were similar and the direction of the effects and the interactions were the same, even though we found differences in statistical significance. Thus, our study does not support the idea of a different underlying mechanism for the effects of TRAP on these two CAMs.\nThe expression of both ICAM-1 and VCAM-1 on the surface of vascular endothelial cells is associated with the formation of early atherosclerotic lesions. Although the relationship between the degree of cellular ICAM-1 and VCAM-1 expression and plasma concentrations of soluble forms is not entirely clear, multiple studies have shown that sICAM-1 and sVCAM-1 predict risk of cardiovascular disease (Albert and Ridker 1999; Pradhan et al. 2002; Rana et al. 2011; Ridker et al. 2003). In a recent study investigating the variability of sICAM-1 and sVCAM-1 measures over a 4-week period, Eschen et al. (2008) reported an estimated intrasubject variability of 7.6% for sICAM-1 and 9.5% for sVCAM-1, which suggests that these markers are relatively stable over a 4-week period.", "A major advantage of the present study is the use of a validated land-use regression model to characterize the individual-level differences in exposures instead of classifying exposure based on measurements at the nearest monitor, land-use regression models without BC measurements, or a weighted form of distance to roadway. Although the study is still limited by the lack of individual-level monitoring data at the home, our validation study suggests that our estimates are highly correlated with actual exposure measurements at locations in Boston other than those used to fit the model and are much more closely correlated with these measurements than are exposure estimates based on a central-site monitor. Although some exposure misclassification will still occur based on our model estimates, we expect that most of the residual error is Berkson type, based on a previous validation study for this model analyzing measurement error (Gryparis et al. 2009). Thus, we expect that the exposure misclassification will not bias effect estimates.\nThe lack of individual activity data also presents a limitation, but the NAS population consists of elderly men who spend a considerable amount of time at or near their home address compared with other population groups.\nOur sensitivity analysis of the block lags of weeks 5–8 and weeks 9–12 shows slight attenuation compared with the cumulative lags, suggesting that the cumulative effect may be dominated by the more recent weeks. The effect sizes for cumulative exposures do not attenuate when averaged up to 12 weeks. The 4-week block lags do not isolate the effects of those time windows because each week is correlated in both space and time; thus, future studies to model the specific contributions of different lag weeks would be beneficial in understanding these effects.", "Many studies have examined PM at different diameters, but fewer studies have looked at which components of PM are associated with adverse health effects. Our study links a health biomarker with BC, a specific component of PM2.5. In addition, we consider BC to be a surrogate for primary traffic PM (a specific source of air pollution), where the spatial variability of BC on a given day reflects the variability of traffic in the region, which is the basis of the BC model we developed and used for this study.\nWe did not examine other pollutants such as total PM2.5 or PM2.5 components other than BC in this study. We think it is unlikely that total PM2.5 or any other copollutant is driving the effect we observed for BC because the BC model is based largely on the daily spatial variation exhibited by BC. Other components of PM2.5, such as sulfates and organic PM, are more homogeneous over the study region. Although we cannot completely rule out confounding by copollutants or other factors, the correlation between the model-predicted 4-week BC and the corresponding 4-week averages of PM2.5 measured at the central site was low (0.108), which supports our belief that the effects observed are not driven by any temporal correlation with PM2.5 or one of its components.", "A limitation of this study is the restricted demographics of the study population. Study subjects were all elderly men, most of them white. Thus, we cannot generalize our results to other populations. However, the elderly represent a particularly susceptible population, and the growth in the number and proportion of older adults in the United States is unprecedented: by 2030, the proportion of the U.S. population ≥ 65 years of age will double to about 71 million older adults, or one in every five Americans (U.S. Census 2005).", "We observed positive associations between BC exposures and blood levels of sICAM-1 and sVCAM-1, with statistically significant effect estimates for sICAM-1 in the population as a whole and for sVCAM-1 among diabetics and participants who were not using statins. Effects of BC on both markers appeared to be limited to diabetics and possibly those not using statins. Overall, our results suggest that exposure to traffic PM over 4–12 weeks may induce an increased inflammatory and endothelial response, particularly among diabetics, and that statin use may be protective against this effect." ]

[ "materials|methods", "methods", null, null, null, "results", "methods", "methods", "methods", "discussion", null, null, null, null, null, "conclusions" ]

[ "adhesion molecules", "air", "cardiovascular", "environmental", "outdoor air", "roadway proximity" ]

[ "Adolescent", "Blood Pressure", "Body Height", "Body Weight", "Child", "Child, Preschool", "Cholesterol", "Cholesterol, HDL", "Cohort Studies", "DNA", "Female", "Humans", "Leukocytes", "Male", "Obesity", "Puberty", "Reverse Transcriptase Polymerase Chain Reaction", "Sex Characteristics", "Telomere" ]

[ "Cohorts", "Quantitative PCR", "Statistical analysis" ]

[ "Both cases and control subjects were participants in a genome-wide association study to identify loci associated with early-onset obesity (21), in which the threshold for childhood obesity was set as the 97th age- and gender-specific percentile of body mass index (BMI) from a French reference population (30). Control children were selected as having a BMI less than the 90th percentile for gender and age, the threshold for being overweight (31).\nInformed consent from all participants and ethical approval were obtained as detailed previously (32, 33). Genomic leukocyte DNA was extracted from peripheral blood samples using the salting-out method (34) for the controls, and using PURE-GENE D50K DNA isolation kits (Gentra Systems, Minneapolis, MN) for the cases. In the controls, serum levels of total cholesterol and fasting glucose were enzymatically determined using the AU 640 analyzer (Olympus, High Wycombe, UK), and serum concentrations of high-density lipoprotein (HDL)-cholesterol were measured using a COBAS-Mira analyzer (Roche Diagnostics, Rotkreuz, Switzerland). Fasting glucose levels in the cases were measured using the glucose oxidase procedure. For both cases and controls, low-density lipoprotein-cholesterol concentrations were calculated using the Friedewald formula (35). In addition to the anthropometric and biochemical parameters listed in Table 1, additional data for body fat content, age of onset of obesity, and eating behaviors were available for 217 of the cases.\nCharacteristics of the study cohorts\nAnthropometric and biochemical measurements are given as mean ± sd. Measurements for gender, age, height, weight, and BMI Z-score were available for all participants. Data on systolic and diastolic blood pressure were available for all nonobese and 370 of 471 obese subjects; data on fasting glucose were available for all nonobese and 418 of 471 obese subjects; data on HDL-cholesterol levels were available for all nonobese and 410 of 471 obese subjects and data on total cholesterol for all nonobese and 412 of 471 obese subjects.", "The mean LTL in genomic DNA samples prepared from peripheral blood lymphocytes was measured using multiplex quantitative real-time PCR (36). Duplicate quantitative real-time PCR reactions were carried out in a total volume of 12.5 μl, using approximately 15 ng of template DNA, with final concentrations of 1× iQ Sybr Green supermix (Bio-Rad Laboratories, Hemel Hempstead, UK), 900 nm of telg and telc primers, and 500 nm of single-copy gene primers (hbgd and hbgu) (for primer sequences, see Ref. 36). All the PCRs were carried out in white 384-well plates on a CFX384 real-time PCR detection system (Bio-Rad Laboratories). Five serial dilutions of a reference sample (leukocyte DNA from an adult female) spanning 5–50 ng were run in triplicate on each plate.\nAfter amplification and data collection, the CFX manager software (Bio-Rad Laboratories) was used to generate standard curves from the reference DNA dilutions, one for the telomere signal and one for the single-copy gene signal. The telomere amplicon signal (T) to single-copy gene control amplicon (S) ratios for each sample were calculated as T, the amount of reference DNA that matched the experimental sample for copy number of the telomere template, divided by S, that which matched the copy number of the single-copy gene template. The mean coefficient of variation for the T/S measurements of duplicate samples was 4.5%.", "All telomere measurements (obtained as T/S ratios) were log transformed before statistical analysis to ensure normal distribution of the data, as assessed both by visual inspection of a histogram of the plotted values, and the Shapiro-Wilk test for normality (Fig. 1). Unpaired two-tailed t tests were used to evaluate differences in telomere length between groups, using homoscedastic tests in which the variances were not significantly different (as determined by performing an F test). Pearson's correlation was used to test for association with individual variables, which were also log transformed when necessary. Multiple linear regression with stepwise removal of nonsignificant explanatory variables was carried out to investigate the relative contributions of selected variables to LTL. All analyses and plots were carried out using version 2.10.1 of the R statistical package (cran.r-project.org).\nDistribution of LTL for all samples (n = 793), expressed as log T/S values. Log T/S values are normally distributed (Shapiro-Wilk test for normality, P = 0.65)." ]

[ null, null, null ]

[ "Subjects and Methods", "Cohorts", "Quantitative PCR", "Statistical analysis", "Results", "Discussion" ]

[ "[SUBTITLE] Cohorts [SUBSECTION] Both cases and control subjects were participants in a genome-wide association study to identify loci associated with early-onset obesity (21), in which the threshold for childhood obesity was set as the 97th age- and gender-specific percentile of body mass index (BMI) from a French reference population (30). Control children were selected as having a BMI less than the 90th percentile for gender and age, the threshold for being overweight (31).\nInformed consent from all participants and ethical approval were obtained as detailed previously (32, 33). Genomic leukocyte DNA was extracted from peripheral blood samples using the salting-out method (34) for the controls, and using PURE-GENE D50K DNA isolation kits (Gentra Systems, Minneapolis, MN) for the cases. In the controls, serum levels of total cholesterol and fasting glucose were enzymatically determined using the AU 640 analyzer (Olympus, High Wycombe, UK), and serum concentrations of high-density lipoprotein (HDL)-cholesterol were measured using a COBAS-Mira analyzer (Roche Diagnostics, Rotkreuz, Switzerland). Fasting glucose levels in the cases were measured using the glucose oxidase procedure. For both cases and controls, low-density lipoprotein-cholesterol concentrations were calculated using the Friedewald formula (35). In addition to the anthropometric and biochemical parameters listed in Table 1, additional data for body fat content, age of onset of obesity, and eating behaviors were available for 217 of the cases.\nCharacteristics of the study cohorts\nAnthropometric and biochemical measurements are given as mean ± sd. Measurements for gender, age, height, weight, and BMI Z-score were available for all participants. Data on systolic and diastolic blood pressure were available for all nonobese and 370 of 471 obese subjects; data on fasting glucose were available for all nonobese and 418 of 471 obese subjects; data on HDL-cholesterol levels were available for all nonobese and 410 of 471 obese subjects and data on total cholesterol for all nonobese and 412 of 471 obese subjects.\nBoth cases and control subjects were participants in a genome-wide association study to identify loci associated with early-onset obesity (21), in which the threshold for childhood obesity was set as the 97th age- and gender-specific percentile of body mass index (BMI) from a French reference population (30). Control children were selected as having a BMI less than the 90th percentile for gender and age, the threshold for being overweight (31).\nInformed consent from all participants and ethical approval were obtained as detailed previously (32, 33). Genomic leukocyte DNA was extracted from peripheral blood samples using the salting-out method (34) for the controls, and using PURE-GENE D50K DNA isolation kits (Gentra Systems, Minneapolis, MN) for the cases. In the controls, serum levels of total cholesterol and fasting glucose were enzymatically determined using the AU 640 analyzer (Olympus, High Wycombe, UK), and serum concentrations of high-density lipoprotein (HDL)-cholesterol were measured using a COBAS-Mira analyzer (Roche Diagnostics, Rotkreuz, Switzerland). Fasting glucose levels in the cases were measured using the glucose oxidase procedure. For both cases and controls, low-density lipoprotein-cholesterol concentrations were calculated using the Friedewald formula (35). In addition to the anthropometric and biochemical parameters listed in Table 1, additional data for body fat content, age of onset of obesity, and eating behaviors were available for 217 of the cases.\nCharacteristics of the study cohorts\nAnthropometric and biochemical measurements are given as mean ± sd. Measurements for gender, age, height, weight, and BMI Z-score were available for all participants. Data on systolic and diastolic blood pressure were available for all nonobese and 370 of 471 obese subjects; data on fasting glucose were available for all nonobese and 418 of 471 obese subjects; data on HDL-cholesterol levels were available for all nonobese and 410 of 471 obese subjects and data on total cholesterol for all nonobese and 412 of 471 obese subjects.\n[SUBTITLE] Quantitative PCR [SUBSECTION] The mean LTL in genomic DNA samples prepared from peripheral blood lymphocytes was measured using multiplex quantitative real-time PCR (36). Duplicate quantitative real-time PCR reactions were carried out in a total volume of 12.5 μl, using approximately 15 ng of template DNA, with final concentrations of 1× iQ Sybr Green supermix (Bio-Rad Laboratories, Hemel Hempstead, UK), 900 nm of telg and telc primers, and 500 nm of single-copy gene primers (hbgd and hbgu) (for primer sequences, see Ref. 36). All the PCRs were carried out in white 384-well plates on a CFX384 real-time PCR detection system (Bio-Rad Laboratories). Five serial dilutions of a reference sample (leukocyte DNA from an adult female) spanning 5–50 ng were run in triplicate on each plate.\nAfter amplification and data collection, the CFX manager software (Bio-Rad Laboratories) was used to generate standard curves from the reference DNA dilutions, one for the telomere signal and one for the single-copy gene signal. The telomere amplicon signal (T) to single-copy gene control amplicon (S) ratios for each sample were calculated as T, the amount of reference DNA that matched the experimental sample for copy number of the telomere template, divided by S, that which matched the copy number of the single-copy gene template. The mean coefficient of variation for the T/S measurements of duplicate samples was 4.5%.\nThe mean LTL in genomic DNA samples prepared from peripheral blood lymphocytes was measured using multiplex quantitative real-time PCR (36). Duplicate quantitative real-time PCR reactions were carried out in a total volume of 12.5 μl, using approximately 15 ng of template DNA, with final concentrations of 1× iQ Sybr Green supermix (Bio-Rad Laboratories, Hemel Hempstead, UK), 900 nm of telg and telc primers, and 500 nm of single-copy gene primers (hbgd and hbgu) (for primer sequences, see Ref. 36). All the PCRs were carried out in white 384-well plates on a CFX384 real-time PCR detection system (Bio-Rad Laboratories). Five serial dilutions of a reference sample (leukocyte DNA from an adult female) spanning 5–50 ng were run in triplicate on each plate.\nAfter amplification and data collection, the CFX manager software (Bio-Rad Laboratories) was used to generate standard curves from the reference DNA dilutions, one for the telomere signal and one for the single-copy gene signal. The telomere amplicon signal (T) to single-copy gene control amplicon (S) ratios for each sample were calculated as T, the amount of reference DNA that matched the experimental sample for copy number of the telomere template, divided by S, that which matched the copy number of the single-copy gene template. The mean coefficient of variation for the T/S measurements of duplicate samples was 4.5%.\n[SUBTITLE] Statistical analysis [SUBSECTION] All telomere measurements (obtained as T/S ratios) were log transformed before statistical analysis to ensure normal distribution of the data, as assessed both by visual inspection of a histogram of the plotted values, and the Shapiro-Wilk test for normality (Fig. 1). Unpaired two-tailed t tests were used to evaluate differences in telomere length between groups, using homoscedastic tests in which the variances were not significantly different (as determined by performing an F test). Pearson's correlation was used to test for association with individual variables, which were also log transformed when necessary. Multiple linear regression with stepwise removal of nonsignificant explanatory variables was carried out to investigate the relative contributions of selected variables to LTL. All analyses and plots were carried out using version 2.10.1 of the R statistical package (cran.r-project.org).\nDistribution of LTL for all samples (n = 793), expressed as log T/S values. Log T/S values are normally distributed (Shapiro-Wilk test for normality, P = 0.65).\nAll telomere measurements (obtained as T/S ratios) were log transformed before statistical analysis to ensure normal distribution of the data, as assessed both by visual inspection of a histogram of the plotted values, and the Shapiro-Wilk test for normality (Fig. 1). Unpaired two-tailed t tests were used to evaluate differences in telomere length between groups, using homoscedastic tests in which the variances were not significantly different (as determined by performing an F test). Pearson's correlation was used to test for association with individual variables, which were also log transformed when necessary. Multiple linear regression with stepwise removal of nonsignificant explanatory variables was carried out to investigate the relative contributions of selected variables to LTL. All analyses and plots were carried out using version 2.10.1 of the R statistical package (cran.r-project.org).\nDistribution of LTL for all samples (n = 793), expressed as log T/S values. Log T/S values are normally distributed (Shapiro-Wilk test for normality, P = 0.65).", "Both cases and control subjects were participants in a genome-wide association study to identify loci associated with early-onset obesity (21), in which the threshold for childhood obesity was set as the 97th age- and gender-specific percentile of body mass index (BMI) from a French reference population (30). Control children were selected as having a BMI less than the 90th percentile for gender and age, the threshold for being overweight (31).\nInformed consent from all participants and ethical approval were obtained as detailed previously (32, 33). Genomic leukocyte DNA was extracted from peripheral blood samples using the salting-out method (34) for the controls, and using PURE-GENE D50K DNA isolation kits (Gentra Systems, Minneapolis, MN) for the cases. In the controls, serum levels of total cholesterol and fasting glucose were enzymatically determined using the AU 640 analyzer (Olympus, High Wycombe, UK), and serum concentrations of high-density lipoprotein (HDL)-cholesterol were measured using a COBAS-Mira analyzer (Roche Diagnostics, Rotkreuz, Switzerland). Fasting glucose levels in the cases were measured using the glucose oxidase procedure. For both cases and controls, low-density lipoprotein-cholesterol concentrations were calculated using the Friedewald formula (35). In addition to the anthropometric and biochemical parameters listed in Table 1, additional data for body fat content, age of onset of obesity, and eating behaviors were available for 217 of the cases.\nCharacteristics of the study cohorts\nAnthropometric and biochemical measurements are given as mean ± sd. Measurements for gender, age, height, weight, and BMI Z-score were available for all participants. Data on systolic and diastolic blood pressure were available for all nonobese and 370 of 471 obese subjects; data on fasting glucose were available for all nonobese and 418 of 471 obese subjects; data on HDL-cholesterol levels were available for all nonobese and 410 of 471 obese subjects and data on total cholesterol for all nonobese and 412 of 471 obese subjects.", "The mean LTL in genomic DNA samples prepared from peripheral blood lymphocytes was measured using multiplex quantitative real-time PCR (36). Duplicate quantitative real-time PCR reactions were carried out in a total volume of 12.5 μl, using approximately 15 ng of template DNA, with final concentrations of 1× iQ Sybr Green supermix (Bio-Rad Laboratories, Hemel Hempstead, UK), 900 nm of telg and telc primers, and 500 nm of single-copy gene primers (hbgd and hbgu) (for primer sequences, see Ref. 36). All the PCRs were carried out in white 384-well plates on a CFX384 real-time PCR detection system (Bio-Rad Laboratories). Five serial dilutions of a reference sample (leukocyte DNA from an adult female) spanning 5–50 ng were run in triplicate on each plate.\nAfter amplification and data collection, the CFX manager software (Bio-Rad Laboratories) was used to generate standard curves from the reference DNA dilutions, one for the telomere signal and one for the single-copy gene signal. The telomere amplicon signal (T) to single-copy gene control amplicon (S) ratios for each sample were calculated as T, the amount of reference DNA that matched the experimental sample for copy number of the telomere template, divided by S, that which matched the copy number of the single-copy gene template. The mean coefficient of variation for the T/S measurements of duplicate samples was 4.5%.", "All telomere measurements (obtained as T/S ratios) were log transformed before statistical analysis to ensure normal distribution of the data, as assessed both by visual inspection of a histogram of the plotted values, and the Shapiro-Wilk test for normality (Fig. 1). Unpaired two-tailed t tests were used to evaluate differences in telomere length between groups, using homoscedastic tests in which the variances were not significantly different (as determined by performing an F test). Pearson's correlation was used to test for association with individual variables, which were also log transformed when necessary. Multiple linear regression with stepwise removal of nonsignificant explanatory variables was carried out to investigate the relative contributions of selected variables to LTL. All analyses and plots were carried out using version 2.10.1 of the R statistical package (cran.r-project.org).\nDistribution of LTL for all samples (n = 793), expressed as log T/S values. Log T/S values are normally distributed (Shapiro-Wilk test for normality, P = 0.65).", "As expected, comparison of selected anthropometric and biochemical characteristics revealed highly significant differences between the cases and controls for several traits associated with adiposity, including circulating lipid levels and blood pressure (Table 1). There was no significant difference between the cohorts in fasting glucose levels which was therefore not included in later analyses.\nAverage LTL was ascertained in all 793 subjects using multiplex quantitative PCR (36). The measurements obtained using this method are expressed as T/S ratios, reflecting the telomere signal (T) relative to a single-copy gene (S), normalized to a single reference individual.\nStatistical analyses were carried out after log transformation of all T/S ratios to achieve a normal distribution (Fig. 1). The mean log T/S ratio in the obese children (0.072, se = 0.006) was significantly less than that in nonobese children (0.191, se 0.008, P < 0.0001) (Fig. 2A). This difference equates to a 23.9% decrease in mean T/S ratio in obese compared with nonobese children.\nComparison of LTL distribution in nonobese and obese subjects (A) and in male and female subjects (B). The box plots indicate the maximum and minimum, the lower and upper quartiles, and the median log T/S ratio value in each group. The mean log T/S ratio values of each group are given below. A, LTL (expressed as log T/S ratio) is significantly shorter in obese compared with nonobese study subjects. The mean log T/S ratio in all nonobese samples is 0.191 (se 0.008), the mean log T/S ratio in all obese samples is 0.072 (se 0.006, P < 0.0001). B, LTL (expressed as the log T/S ratio) is not significantly different in male study subjects compared with female study subjects (the mean log T/S ratio of males = 0.113, se 0.007; the mean log T/S ratio of females = 0.126, se 0.007, P = 0.19).\nTo investigate a possible gender effect, girls and boys were next analyzed separately. The mean log T/S ratio in obese girls (0.081, se 0.008) was significantly shorter than that in nonobese girls (0.199, se 0.012, P < 0.0001), a difference that equates to a 23.8% decrease in the T/S ratio. Similarly, the mean log T/S ratio in obese boys (0.061, se 0.008) was significantly shorter than in nonobese boys (0.183, se 0.010, P < 0.0001), a difference that equates to a 24.5% decrease in the T/S ratio. The overall mean log T/S ratio in all girls (0.126, se 0.007) was slightly greater than that in all boys (0.113, se 0.007), but this difference was not statistically significant (P = 0.19) (Fig. 2B).\nTo investigate the potential influence of pubertal stage on leukocyte telomere length, the analysis was carried out separately in those aged younger and older than 9 yr. In the children older than 9 yr, the mean log T/S ratio in the obese children (0.064, se 0.007, n = 364) was significantly shorter than that in nonobese children (0.187, se 0.008, n = 291, P < 0.0001), a difference that equates to a 24.7% decrease in the T/S ratio. The results for the children younger than 9 yr were very similar to those obtained in the older children: the mean log T/S ratio in the obese children (0.100, se 0.011, n = 109) was again significantly shorter than that in nonobese children (0.227, se 0.028, n = 31, P = 0.0002), a difference that equates to a 25.4% decrease in the T/S ratio. Separate analysis of girls and boys aged under/over 9 yr revealed no statistically significant gender differences in these subgroups.\nUnivariate analyses were performed, within each of the obese and nonobese groups separately, to investigate the potential contribution of selected variables to variation in LTL (Table 2). Explanatory variables were log transformed where necessary to ensure normality. As expected, significant inverse associations between the log T/S ratio and age were identified in both the cases and controls. The log T/S ratio was also inversely associated with height and weight in both groups.\nUnivariate analysis of selected biochemical and anthropometric variables\nSignificant P values shown in bold, whereas those approaching significance are in italics. In both nonobese and obese children, age, height, and log weight are inversely associated with mean telomere length (log T/S ratio). Borderline significant associations are also detected between log T/S ratio and log cholesterol levels in the controls and log HDL-cholesterol levels in the cases.\nThe high level of multicolinearity between age, height, and log weight meant that it was not possible to determine their relative contributions to telomere length. Multivariate analysis using all these variables, with LTL as the dependent variable, resulted in a model with age as the sole explanatory variable: adjusting log T/S values for any of age, height, or log weight abolished associations between LTL and the other two variables. The effect sizes of each of these variables on LTL were not significantly different in cases compared with controls. No improvement in the model was achieved by including, for obese subjects, time since onset of obesity as an additional parameter, because this variable showed the same multicolinearity with age, height, and log weight.\nAlthough not quite reaching nominal statistical significance, there were discernible trends between longer telomere length and increasing total cholesterol in the controls and with increasing HDL-cholesterol in the cases. No associations between LTL and BMI Z-score or either systolic or diastolic blood pressure were detected in either cohort (Table 2).\nAnalysis of additional obesity-related traits in a subset of 217 cases for which data on body fat content, age of onset of obesity, and eating behaviors revealed no significant associations with LTL that were separate from the effects of weight or height.", "We have demonstrated a highly significant inverse association between early-onset obesity and mean LTL in both boys and girls in a large case-control study of French children. Mean LTL in the obese cases is almost 25% less than in the controls, and LTL declines with increasing age, weight, and height in both groups, with comparable effect sizes for all three variables in both cases and controls.\nIn adulthood, shorter leukocyte telomeres are associated with BMI in women (24) and waist to hip ratio in both men and women (4). Furthermore, adipocytes from obese adults have telomeres approximately 17% shorter than those in adipocytes from nonobese adults (37). The mechanisms underlying the association between obesity and short telomeres in adults are also unknown, although an increase in oxidative stress and inflammation have both been suggested as possible explanations (23).\nPrevious studies investigating telomere length and childhood obesity proved inconclusive. Zannolli et al. (28) found no difference in LTL between obese and nonobese individuals in 53 Italian children, whereas Al-Attas et al. (29) investigated 148 Arab children and found that mean telomere length was shorter in obese boys compared with lean boys but found no difference between obese and lean girls. It is possible that the disparity between our findings and those of previous workers is due to undetermined genetic or nongenetic differences that exist between the populations examined. However, it is more likely that the increased statistical power afforded by the larger size of our study (n = 793), and the severe phenotype of the cases (above the 97th age and sex specific percentile of BMI), allowed us to detect a highly significant association between telomere length and obesity in both boys and girls.\nInterestingly, we found no significant difference between mean LTL in girls and boys in our study, which agrees with a previous report on LTL in newborns (2). This implies that the difference observed in adulthood, men having shorter mean leukocyte telomeres than women (38), may be due to factors that exert their influence after, rather than during, early childhood.\nDespite evidence that childhood obesity contributes to earlier onset of puberty (39), we found no evidence that this factor influenced our results. In the absence of data on pubertal stage, we subdivided our cohort into those younger and older than 9 yr of age. We found that shorter LTL was associated with obesity in both groups: in the children younger than 9 yr, mean telomere length in the obese children was 25.4% shorter than in the controls, whereas in the children older than 9 yr, it was 24.7% shorter.\nIn agreement with numerous previous studies in both adults and children, we found an inverse association between age and telomere length, within each of the case and control cohorts. However, we also detected inverse associations between LTL and height and weight. All of these variables are indicators of body size in childhood, and although it was not possible to determine the relative contribution of each, this supports the hypothesis that telomere length in childhood chronicles the expansion of the hematopoietic stem and progenitor cell populations, which in turn reflect body size (40).\nThe highly significant difference between telomere length in obese and nonobese children may be partly or wholly due to a general overgrowth compared with their normal-weight peers of a similar chronological age. Although the mean age of the obese children was almost 1 yr younger than the controls, they were somewhat taller and their mean weight was almost 80% greater (Table 1). Thus, the shorter mean LTL of the obese children compared with the nonobese children may simply be due to an increased rate of hematopoietic stem cell turnover. Consistent with this is the finding that no association was observed in either cohort between BMI Z-score and telomere length, indicating that the correlation is with absolute body size, rather than size relative to age- and gender-matched peers.\nAlthough the difference in body size is the most striking phenotypic difference between our case and control cohorts, the observed association with telomere length may nonetheless reflect some other aspect of obesity, such as circulating lipid levels, or inflammation. There is evidence from a longitudinal study in adults that telomere length is positively associated with HDL-cholesterol levels (3). The authors propose that this association might be explained by the antioxidant and antiinflammatory effects of HDL-cholesterol, both of which might slow leukocyte telomere attrition rates. In our study, levels of HDL-cholesterol were significantly lower in the obese, compared with the nonobese children, and LTL in obese children showed a discernible (although not quite significant) positive correlation with HDL-cholesterol, each of which is consistent with this hypothesis. Conversely, another potential explanation for the shorter telomeres in obese children is inflammation (7), which would increase leukocyte progenitor cell turnover.\nAlthough shorter telomeres are associated with hypertension in adults (16), we found no association between blood pressure and LTL in either the nonobese or obese cohort. This suggests that this trait is only a significant predictor of telomere length in adulthood and not childhood. However, because ours was a case-control study designed to test for differences between normal-weight and obese children, it may have been underpowered to detect associations between LTL and continuous traits such as blood pressure and circulating lipids.\nIn conclusion, we have demonstrated that obese children have telomeres that are substantially shorter than those of nonobese controls of comparable age. Further population-based studies in young cohorts are required to identify the obesity-related traits that explain this observation and additionally to investigate the extent to which this difference in telomere length extends into adulthood. The fact that obese children have an apparent biological age that is significantly greater than their chronological age highlights the importance of intervention and also support for these individuals at the earliest opportunity to minimize their risk of future disease." ]

[ "methods", null, null, null, "results", "discussion" ]

[]

[ "Age Factors", "Aged", "Aged, 80 and over", "Carcinoma, Hepatocellular", "Female", "Humans", "Liver Neoplasms", "Male", "Middle Aged", "Multivariate Analysis", "Retrospective Studies", "Survival Analysis", "Treatment Outcome" ]

[ "Introduction", "Patient Selection", "Statistical Analysis", "Results", "Patients Characteristics", "Comorbidities", "Overall Survival", "HCC-Specific Survival", "HCC-Specific Survival Adjusted for Competing Risk", "Multivariate Analysis for Survival", "Discussion", "Limitation" ]

[ "The incidence and mortality of hepatocellular carcinoma (HCC) have been dramatically increasing in the United States [1]. Age is a known risk factor for HCC [2–4]. In the United States, the average age of HCC diagnosis is 65 years [5]. The incidence rate of HCC among the elderly is progressively increasing [6]. This trend may be partially attributable to screening, improved management of chronic liver disease, the rising incidence of nonalcoholic steatohepatitis-related cirrhosis, and an epidemic of hepatitis C virus (HCV) infection in the late 1970s [7]. As the population ages, it is becoming increasingly important to recognize age-specific cancer epidemiologic trends and differences in cancer outcomes to aid in prevention strategies, diagnostic approaches, and treatment decisions.\nHistorically, the elderly represent an important population particularly vulnerable to disparities in cancer treatment [8, 9]. There is a paucity of information on the possible differences in the clinical presentation, treatment patterns, and outcomes of HCC based on age. The limited available data are mainly from Asia, where HCC has a distinct underlying etiology and different treatment pattern [10–17]. Additional studies from Europe reflect disease trends inherent to a less demographically diverse population and particular patterns of clinical practice, thus making it difficult to extrapolate to U.S. patients [18, 19].\nThe increasing HCC incidence, aging population, development of new HCC treatment modalities, and limited age-related data available prompted the current study. By using a multi-institutional database, we designed the present study to examine the clinical characteristics, treatment patterns, and outcomes in elderly patients with HCC.", "Retrospective analysis of HCC patients treated at two U.S. tertiary institutions from January 1998 to December 2008 was performed. A diagnosis of HCC was confirmed histopathologically or according to the noninvasive European Association for the Study of the Liver (EASL) diagnostic criteria [20]. The age of 70 years was chosen as the cutoff point for the analysis because the 70 years of age landmark has been described as the lower boundary of senescence and the incidence of age-related organ dysfunction and the development of comorbid conditions start to increase after the age of 70 [21]. Additionally, 70 years of age has been the most frequently used cutoff age for comparisons between younger and older patients in previous analyses of HCC patients, thus allowing for some comparability with these other studies. The demographics, etiology of liver disease, tumor parameters, comorbidities, performance status, body mass index, detailed treatment information, length of survival from diagnosis, and cause of death were captured. Race/ethnicity was categorized as White, Black, Asians, and Hispanic. The etiology of the liver disease was categorized as infection with hepatitis B (HBV), HCV, alcoholic liver disease, nonalcoholic fatty liver disease, hemochromatosis, autoimmune hepatitis, and primary biliary cirrhosis. The etiology of cirrhosis was considered to be cryptogenic if no cause for chronic liver disease was identified after exhaustive testing. No quantitative analysis of alcohol consumption was performed; assessment of habitual alcohol use was based on patient self-reporting of habitual drinking.\nUnderlying cirrhosis was confirmed histologically or based on clinical and/or radiological criteria including nodular liver contour, presence of ascites, varices, enlargement of the caudate lobe, splenomegaly, and collateral portal-venous anastomoses. We also assessed whether the diagnosis of HCC was made during evaluation for possible HCC based on presenting signs or symptoms or as a result of surveillance.\nMultimodality treatment was defined as therapy that combines more than one method (e.g., combination of chemotherapy and embolization) of treatment. Death was categorized as attributable to HCC based on review of death certificate, discharge summary, or hospice documentation. Institutional Review Board approval was granted from each institution for this retrospective study.", "The patient characteristics were summarized using mean ± SD, median and interquartile range for continuous variables, and percentages for categorical variables. Chi-square tests and Wilcoxon rank sum tests were used to compare the patient characteristics between the two age groups. Three types of survival analysis were performed to compare the clinical outcome between the elderly and younger patients: (a) Overall survival (OS) defined as the interval between the date of HCC diagnosis and the date of death due to any cause; (b) HCC-specific survival defined as the time from HCC diagnosis to death from HCC (deaths from non-HCC causes were treated as censoring); (c) HCC-specific survival, but non-HCC death was treated as a competing risk. The Kaplan-Meier method was used to estimate survival distribution for OS and HCC-specific survival. Cumulative incidence rate for HCC death was estimated for the two age groups when non-HCC death was treated as a competing risk.\nA Cox proportional hazard regression model was developed to evaluate the survival difference between the elderly and nonelderly groups while other prognostic factors were controlled. All factors listed in Table 1, but not age, were included in the original model, and a stepwise method was used for variable selection. Eastern Cooperative Oncology Group (ECOG) performance status (PS), cancer stage, presence of symptoms, Child-Turcotte-Pugh (CTP) class, portal vein thrombosis (PVT), BMI, and HCV were identified as the significant prognostic factors for OS and HCC-specific survival and included in the final model. Analyses were performed separately for patients with and without orthotopic liver transplantation (OLT) because transplanted HCC patients have substantial long-term survival and the majority of transplanted patients were in the younger group.\nPatient characteristics\n(Continued)\nAbbreviations: AFP, alpha-fetoprotein; CAD, coronary artery disease; CLIP, Cancer of the Liver Italian Program; COPD, chronic obstructive pulmonary disease; DM, diabetes mellitus; ECOG, Eastern Cooperative Oncology Group; HBV, hepatitis B virus; HCV, hepatitis C virus; HTN, hypertension; IQR, interquartile range.\nAll tests were two-sided with significance level p < .05. All statistical analyses were conducted using SAS 9.2 statistical software (SAS Institute, Cary, NC).", "[SUBTITLE] Patients Characteristics [SUBSECTION] A total of 335 patients with HCC were included. Ninety-five patients ≥70 years old (median age of 74 years) at diagnosis of HCC were defined as the elderly group. Two hundred forty patients aged <70 years (median age of 56) were regarded as the younger group. The patient characteristics are summarized in Table 1. The male (M)-to-female (F) ratio was 1.7:1 in the elderly and 5.8:1 in the younger group, showing that there was a higher proportion of women in the elderly group (p < .0001). The proportion of patients with HCV infection was significantly lower in the elderly group (21.1% versus 48.3%, p < .0001). There were no significant differences in HBV infection rates between younger and elderly patients (21.2% and 14.7%, respectively). Alcohol-induced liver disease was found in 37.5% of younger patients and 29.5% of the elderly group; the difference was not statistically significant (p = .1657). There was a trend toward more advanced CTP class in the younger group: CTP class B and C cirrhosis 35.8% versus 25.3% (p = .063). younger patients were more frequently diagnosed as a result of HCC screening or surveillance (61.3% versus 32.6%, p < .0001). More patients in the elderly group had ECOG PS > 1 than in the younger group (24.2% versus 7.9%, p < .0001). There were no significant differences in the following: race distribution, stage, number of lesions, symptomatic at presentation, PVT, alpha-fetoprotein (AFP) level, Cancer of the Liver Italian Program score, diabetes mellitus, or BMI.\nThe types of treatments received are summarized in Table 2. Forty-seven (19.6%) of the younger patients underwent liver transplant compared to 5.3% of the elderly patients (p = .0002). Conversely, a higher proportion of elderly patients received only symptomatic treatment (36.8% versus 22.9%, p = .01). There was no difference in the number of liver resections or liver directed treatments, including radiofrequency ablation, transarterial chemoembolization, radiation, and chemotherapy. The rate of utilization of multimodality treatment was 28.3% in the younger group and 23.2% in the elderly group; the difference was not statistically significant.\nTreatments received\nA total of 335 patients with HCC were included. Ninety-five patients ≥70 years old (median age of 74 years) at diagnosis of HCC were defined as the elderly group. Two hundred forty patients aged <70 years (median age of 56) were regarded as the younger group. The patient characteristics are summarized in Table 1. The male (M)-to-female (F) ratio was 1.7:1 in the elderly and 5.8:1 in the younger group, showing that there was a higher proportion of women in the elderly group (p < .0001). The proportion of patients with HCV infection was significantly lower in the elderly group (21.1% versus 48.3%, p < .0001). There were no significant differences in HBV infection rates between younger and elderly patients (21.2% and 14.7%, respectively). Alcohol-induced liver disease was found in 37.5% of younger patients and 29.5% of the elderly group; the difference was not statistically significant (p = .1657). There was a trend toward more advanced CTP class in the younger group: CTP class B and C cirrhosis 35.8% versus 25.3% (p = .063). younger patients were more frequently diagnosed as a result of HCC screening or surveillance (61.3% versus 32.6%, p < .0001). More patients in the elderly group had ECOG PS > 1 than in the younger group (24.2% versus 7.9%, p < .0001). There were no significant differences in the following: race distribution, stage, number of lesions, symptomatic at presentation, PVT, alpha-fetoprotein (AFP) level, Cancer of the Liver Italian Program score, diabetes mellitus, or BMI.\nThe types of treatments received are summarized in Table 2. Forty-seven (19.6%) of the younger patients underwent liver transplant compared to 5.3% of the elderly patients (p = .0002). Conversely, a higher proportion of elderly patients received only symptomatic treatment (36.8% versus 22.9%, p = .01). There was no difference in the number of liver resections or liver directed treatments, including radiofrequency ablation, transarterial chemoembolization, radiation, and chemotherapy. The rate of utilization of multimodality treatment was 28.3% in the younger group and 23.2% in the elderly group; the difference was not statistically significant.\nTreatments received\n[SUBTITLE] Comorbidities [SUBSECTION] Eighty nine of 95 patients (93%) in the elderly group had comorbid conditions compared with 102 of 240 patients (42.5%) in the younger group (p < .0001). Cardiovascular and cerebrovascular disease, chronic lung conditions, second primary malignancies, chronic renal disease, or cognitive disorders constituted the most common comorbid conditions. The following statistically significant differences between the two groups were observed: 36.8% elderly patients had coronary artery disease compared to 10.4% in the younger group (p < .0001); similar trends were observed for hypertension (37.9% versus 24.2%, p = .0117) and second primary malignancy (21.1% versus 3.8%, p < .0001).\nEighty nine of 95 patients (93%) in the elderly group had comorbid conditions compared with 102 of 240 patients (42.5%) in the younger group (p < .0001). Cardiovascular and cerebrovascular disease, chronic lung conditions, second primary malignancies, chronic renal disease, or cognitive disorders constituted the most common comorbid conditions. The following statistically significant differences between the two groups were observed: 36.8% elderly patients had coronary artery disease compared to 10.4% in the younger group (p < .0001); similar trends were observed for hypertension (37.9% versus 24.2%, p = .0117) and second primary malignancy (21.1% versus 3.8%, p < .0001).\n[SUBTITLE] Overall Survival [SUBSECTION] Median follow-up was 15 months for the younger group and 11 months for the elderly group. For subjects who were alive at the time of this analysis, the median follow-up time was 27 months for the elderly and 31 months for the younger.\nFor all patients, the 1-, 2-, and 3-year survival rates were 52.9 ± 5.3%, 38.3 ± 5.2%, and 27.1 ± 5.2%, respectively, for the elderly group and 63.1 ± 3.2%, 46.8 ± 3.4%, and 35.9 ± 3.4%, respectively, for the younger group. Figure 1A shows the overall survival curves for the elderly and younger group. This difference was borderline significant (log-rank test, p = .06). One hundred forty-eight of the 190 patients in the younger group and 66 of the 90 patients in the elderly group died from all causes. Figure 1B shows the overall survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. The median survival time was 12.5 months (95% CI, 7.9–18.7 months) for the elderly group and 13.9 months (95% CI, 10.2–16.4 months) for the younger group. After exclusion of transplanted patients, the 1-, 2-, and 3-year survival rates were 51.4 ± 5.4%, 35.7 ± 5.3%, and 23.0 ± 5.2%, respectively, for the elderly group and 54.4 ± 3.7%, 35.9 ± 3.7%, and 23.5 ± 3.5%, respectively, for the younger group. This difference was not statistically significant (log-rank test, p = .47).\nKaplan-Meier overall survival curves. (A): Elderly (n = 95) and younger (n =240) patients. (B): Overall survival curves of elderly (n = 90) and younger (n = 193) patients after exclusion of patients who underwent orthotic liver transplantation. p values were calculated with the use of the log-rank test.\nMedian follow-up was 15 months for the younger group and 11 months for the elderly group. For subjects who were alive at the time of this analysis, the median follow-up time was 27 months for the elderly and 31 months for the younger.\nFor all patients, the 1-, 2-, and 3-year survival rates were 52.9 ± 5.3%, 38.3 ± 5.2%, and 27.1 ± 5.2%, respectively, for the elderly group and 63.1 ± 3.2%, 46.8 ± 3.4%, and 35.9 ± 3.4%, respectively, for the younger group. Figure 1A shows the overall survival curves for the elderly and younger group. This difference was borderline significant (log-rank test, p = .06). One hundred forty-eight of the 190 patients in the younger group and 66 of the 90 patients in the elderly group died from all causes. Figure 1B shows the overall survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. The median survival time was 12.5 months (95% CI, 7.9–18.7 months) for the elderly group and 13.9 months (95% CI, 10.2–16.4 months) for the younger group. After exclusion of transplanted patients, the 1-, 2-, and 3-year survival rates were 51.4 ± 5.4%, 35.7 ± 5.3%, and 23.0 ± 5.2%, respectively, for the elderly group and 54.4 ± 3.7%, 35.9 ± 3.7%, and 23.5 ± 3.5%, respectively, for the younger group. This difference was not statistically significant (log-rank test, p = .47).\nKaplan-Meier overall survival curves. (A): Elderly (n = 95) and younger (n =240) patients. (B): Overall survival curves of elderly (n = 90) and younger (n = 193) patients after exclusion of patients who underwent orthotic liver transplantation. p values were calculated with the use of the log-rank test.\n[SUBTITLE] HCC-Specific Survival [SUBSECTION] One hundred eighty-six HCC-related deaths were documented: 46 in the elderly group and 140 in younger group. The median HCC specific survival time was 27.8 months (95% CI, 13.4–36.9 months) for the elderly group and 24.5 months (95% CI, 18.7–31.3 months) for the younger group.\nFor all patients, the 1-, 2-, and 3-year HCC-specific survival rates were 66.5 ± 5.3%, 52.3 ± 5.9%, and 39.1 ± 6.5%, respectively, for the elderly group and 66.7 ± 3.2%, 51.7 ± 3.5%, and 40.3 ± 3.6%, respectively, for the younger group. Figure 2A shows the HCC-specific survival curves for the elderly and younger group. This difference was not statistically significant (log-rank test, p = .89).\nKaplan-Meier HCC-specific survival curves. (A): All elderly (n = 95) and younger (n = 240) patients. (B): HCC-specific survival for elderly (n = 90) and younger (n =193) patients after exclusion of patients who underwent orthotic liver transplantation. p-values were calculated with the use of the log-rank test.\nFor the 280 patients who did not receive OLT, 45 (50%) of the 90 elderly patients died from HCC, versus 131 patients (69%) of 190 patients in the younger group. The median disease-specific survival time was 24.9 months (95% CI, 12.9–32.9 months) for the elderly group and 14.9 months (95% CI, 1–21.0 months) for the younger group. The estimated 1-, 2-, and 3-year HCC-specific survival rates were 65.9 ± 5.4%, 50.3 ± 6.2% and 34.9 ± 6.9%, respectively, for the elderly group and 58.4 ± 3.8%, 40.9 ± 3.9%, and 27.6 ± 3.9%, respectively, for the younger group. Figure 2B shows the HCC-specific survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. This difference was not statistically significant (log-rank test, p = .38).\nOne hundred eighty-six HCC-related deaths were documented: 46 in the elderly group and 140 in younger group. The median HCC specific survival time was 27.8 months (95% CI, 13.4–36.9 months) for the elderly group and 24.5 months (95% CI, 18.7–31.3 months) for the younger group.\nFor all patients, the 1-, 2-, and 3-year HCC-specific survival rates were 66.5 ± 5.3%, 52.3 ± 5.9%, and 39.1 ± 6.5%, respectively, for the elderly group and 66.7 ± 3.2%, 51.7 ± 3.5%, and 40.3 ± 3.6%, respectively, for the younger group. Figure 2A shows the HCC-specific survival curves for the elderly and younger group. This difference was not statistically significant (log-rank test, p = .89).\nKaplan-Meier HCC-specific survival curves. (A): All elderly (n = 95) and younger (n = 240) patients. (B): HCC-specific survival for elderly (n = 90) and younger (n =193) patients after exclusion of patients who underwent orthotic liver transplantation. p-values were calculated with the use of the log-rank test.\nFor the 280 patients who did not receive OLT, 45 (50%) of the 90 elderly patients died from HCC, versus 131 patients (69%) of 190 patients in the younger group. The median disease-specific survival time was 24.9 months (95% CI, 12.9–32.9 months) for the elderly group and 14.9 months (95% CI, 1–21.0 months) for the younger group. The estimated 1-, 2-, and 3-year HCC-specific survival rates were 65.9 ± 5.4%, 50.3 ± 6.2% and 34.9 ± 6.9%, respectively, for the elderly group and 58.4 ± 3.8%, 40.9 ± 3.9%, and 27.6 ± 3.9%, respectively, for the younger group. Figure 2B shows the HCC-specific survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. This difference was not statistically significant (log-rank test, p = .38).\n[SUBTITLE] HCC-Specific Survival Adjusted for Competing Risk [SUBSECTION] Forty-one non-HCC-related deaths were documented: 21 in the elderly group and 20 in younger group. Although non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death between the two patients groups, hazard ratio of 1.02 (95% CI, 0.73–1.43, p = .89). There were a total of 38 non-HCC deaths after transplanted patients were excluded: 21 in the elderly group and 17 in the younger group. The non-HCC-related deaths included infections, myocardial infarction, stroke, renal failure, suicide, and trauma.\nAlthough non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death among the two patients groups, hazard ratio of 0.86 (95% CI, 0.61–1.21, p = .39). However, comparing with younger patients, elderly patients had a significantly worse survival for non-HCC related deaths with a hazard ratio of 3.27 (95% CI, 1.76–6.05, p = .0002).\nForty-one non-HCC-related deaths were documented: 21 in the elderly group and 20 in younger group. Although non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death between the two patients groups, hazard ratio of 1.02 (95% CI, 0.73–1.43, p = .89). There were a total of 38 non-HCC deaths after transplanted patients were excluded: 21 in the elderly group and 17 in the younger group. The non-HCC-related deaths included infections, myocardial infarction, stroke, renal failure, suicide, and trauma.\nAlthough non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death among the two patients groups, hazard ratio of 0.86 (95% CI, 0.61–1.21, p = .39). However, comparing with younger patients, elderly patients had a significantly worse survival for non-HCC related deaths with a hazard ratio of 3.27 (95% CI, 1.76–6.05, p = .0002).\n[SUBTITLE] Multivariate Analysis for Survival [SUBSECTION] Multivariate analysis did not show a significant difference in clinical outcomes between the two age groups. When the effects of the important prognostic factors were adjusted, the estimated hazard ratio was 1.33 (95% CI, 0.97–1.82, p = .07) for OS; 0.98 (95% CI, 0.68–1.41, p = .92) for HCC-specific survival; and 0.98 (95% CI, 0.66–1.46, p = .92) for HCC-specific survival adjusted for competing risk.\nThe results of multivariate analysis for patients without liver transplant are summarized in Table 3. No significant difference between the two age groups was found. The estimated hazard ratio was 1.23 (95% CI, 0.89–1.71, p = .21) for OS; 0.90 (95% CI, 0.62–1.30, p = .57) for HCC-specific survival; and 0.90 (95% CI, 0.59–1.36, p = .61) for HCC-specific survival adjusted for competing risk.\nEstimated hazard ratios from Cox proportional hazards analysis for patients without liver transplant\nAbbreviations: ECOG, Eastern Cooperative Oncology Group; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; PVT, portal vein thrombosis.\naHigh if BMI >27 (median); low otherwise.\nMultivariate analysis did not show a significant difference in clinical outcomes between the two age groups. When the effects of the important prognostic factors were adjusted, the estimated hazard ratio was 1.33 (95% CI, 0.97–1.82, p = .07) for OS; 0.98 (95% CI, 0.68–1.41, p = .92) for HCC-specific survival; and 0.98 (95% CI, 0.66–1.46, p = .92) for HCC-specific survival adjusted for competing risk.\nThe results of multivariate analysis for patients without liver transplant are summarized in Table 3. No significant difference between the two age groups was found. The estimated hazard ratio was 1.23 (95% CI, 0.89–1.71, p = .21) for OS; 0.90 (95% CI, 0.62–1.30, p = .57) for HCC-specific survival; and 0.90 (95% CI, 0.59–1.36, p = .61) for HCC-specific survival adjusted for competing risk.\nEstimated hazard ratios from Cox proportional hazards analysis for patients without liver transplant\nAbbreviations: ECOG, Eastern Cooperative Oncology Group; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; PVT, portal vein thrombosis.\naHigh if BMI >27 (median); low otherwise.", "A total of 335 patients with HCC were included. Ninety-five patients ≥70 years old (median age of 74 years) at diagnosis of HCC were defined as the elderly group. Two hundred forty patients aged <70 years (median age of 56) were regarded as the younger group. The patient characteristics are summarized in Table 1. The male (M)-to-female (F) ratio was 1.7:1 in the elderly and 5.8:1 in the younger group, showing that there was a higher proportion of women in the elderly group (p < .0001). The proportion of patients with HCV infection was significantly lower in the elderly group (21.1% versus 48.3%, p < .0001). There were no significant differences in HBV infection rates between younger and elderly patients (21.2% and 14.7%, respectively). Alcohol-induced liver disease was found in 37.5% of younger patients and 29.5% of the elderly group; the difference was not statistically significant (p = .1657). There was a trend toward more advanced CTP class in the younger group: CTP class B and C cirrhosis 35.8% versus 25.3% (p = .063). younger patients were more frequently diagnosed as a result of HCC screening or surveillance (61.3% versus 32.6%, p < .0001). More patients in the elderly group had ECOG PS > 1 than in the younger group (24.2% versus 7.9%, p < .0001). There were no significant differences in the following: race distribution, stage, number of lesions, symptomatic at presentation, PVT, alpha-fetoprotein (AFP) level, Cancer of the Liver Italian Program score, diabetes mellitus, or BMI.\nThe types of treatments received are summarized in Table 2. Forty-seven (19.6%) of the younger patients underwent liver transplant compared to 5.3% of the elderly patients (p = .0002). Conversely, a higher proportion of elderly patients received only symptomatic treatment (36.8% versus 22.9%, p = .01). There was no difference in the number of liver resections or liver directed treatments, including radiofrequency ablation, transarterial chemoembolization, radiation, and chemotherapy. The rate of utilization of multimodality treatment was 28.3% in the younger group and 23.2% in the elderly group; the difference was not statistically significant.\nTreatments received", "Eighty nine of 95 patients (93%) in the elderly group had comorbid conditions compared with 102 of 240 patients (42.5%) in the younger group (p < .0001). Cardiovascular and cerebrovascular disease, chronic lung conditions, second primary malignancies, chronic renal disease, or cognitive disorders constituted the most common comorbid conditions. The following statistically significant differences between the two groups were observed: 36.8% elderly patients had coronary artery disease compared to 10.4% in the younger group (p < .0001); similar trends were observed for hypertension (37.9% versus 24.2%, p = .0117) and second primary malignancy (21.1% versus 3.8%, p < .0001).", "Median follow-up was 15 months for the younger group and 11 months for the elderly group. For subjects who were alive at the time of this analysis, the median follow-up time was 27 months for the elderly and 31 months for the younger.\nFor all patients, the 1-, 2-, and 3-year survival rates were 52.9 ± 5.3%, 38.3 ± 5.2%, and 27.1 ± 5.2%, respectively, for the elderly group and 63.1 ± 3.2%, 46.8 ± 3.4%, and 35.9 ± 3.4%, respectively, for the younger group. Figure 1A shows the overall survival curves for the elderly and younger group. This difference was borderline significant (log-rank test, p = .06). One hundred forty-eight of the 190 patients in the younger group and 66 of the 90 patients in the elderly group died from all causes. Figure 1B shows the overall survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. The median survival time was 12.5 months (95% CI, 7.9–18.7 months) for the elderly group and 13.9 months (95% CI, 10.2–16.4 months) for the younger group. After exclusion of transplanted patients, the 1-, 2-, and 3-year survival rates were 51.4 ± 5.4%, 35.7 ± 5.3%, and 23.0 ± 5.2%, respectively, for the elderly group and 54.4 ± 3.7%, 35.9 ± 3.7%, and 23.5 ± 3.5%, respectively, for the younger group. This difference was not statistically significant (log-rank test, p = .47).\nKaplan-Meier overall survival curves. (A): Elderly (n = 95) and younger (n =240) patients. (B): Overall survival curves of elderly (n = 90) and younger (n = 193) patients after exclusion of patients who underwent orthotic liver transplantation. p values were calculated with the use of the log-rank test.", "One hundred eighty-six HCC-related deaths were documented: 46 in the elderly group and 140 in younger group. The median HCC specific survival time was 27.8 months (95% CI, 13.4–36.9 months) for the elderly group and 24.5 months (95% CI, 18.7–31.3 months) for the younger group.\nFor all patients, the 1-, 2-, and 3-year HCC-specific survival rates were 66.5 ± 5.3%, 52.3 ± 5.9%, and 39.1 ± 6.5%, respectively, for the elderly group and 66.7 ± 3.2%, 51.7 ± 3.5%, and 40.3 ± 3.6%, respectively, for the younger group. Figure 2A shows the HCC-specific survival curves for the elderly and younger group. This difference was not statistically significant (log-rank test, p = .89).\nKaplan-Meier HCC-specific survival curves. (A): All elderly (n = 95) and younger (n = 240) patients. (B): HCC-specific survival for elderly (n = 90) and younger (n =193) patients after exclusion of patients who underwent orthotic liver transplantation. p-values were calculated with the use of the log-rank test.\nFor the 280 patients who did not receive OLT, 45 (50%) of the 90 elderly patients died from HCC, versus 131 patients (69%) of 190 patients in the younger group. The median disease-specific survival time was 24.9 months (95% CI, 12.9–32.9 months) for the elderly group and 14.9 months (95% CI, 1–21.0 months) for the younger group. The estimated 1-, 2-, and 3-year HCC-specific survival rates were 65.9 ± 5.4%, 50.3 ± 6.2% and 34.9 ± 6.9%, respectively, for the elderly group and 58.4 ± 3.8%, 40.9 ± 3.9%, and 27.6 ± 3.9%, respectively, for the younger group. Figure 2B shows the HCC-specific survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. This difference was not statistically significant (log-rank test, p = .38).", "Forty-one non-HCC-related deaths were documented: 21 in the elderly group and 20 in younger group. Although non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death between the two patients groups, hazard ratio of 1.02 (95% CI, 0.73–1.43, p = .89). There were a total of 38 non-HCC deaths after transplanted patients were excluded: 21 in the elderly group and 17 in the younger group. The non-HCC-related deaths included infections, myocardial infarction, stroke, renal failure, suicide, and trauma.\nAlthough non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death among the two patients groups, hazard ratio of 0.86 (95% CI, 0.61–1.21, p = .39). However, comparing with younger patients, elderly patients had a significantly worse survival for non-HCC related deaths with a hazard ratio of 3.27 (95% CI, 1.76–6.05, p = .0002).", "Multivariate analysis did not show a significant difference in clinical outcomes between the two age groups. When the effects of the important prognostic factors were adjusted, the estimated hazard ratio was 1.33 (95% CI, 0.97–1.82, p = .07) for OS; 0.98 (95% CI, 0.68–1.41, p = .92) for HCC-specific survival; and 0.98 (95% CI, 0.66–1.46, p = .92) for HCC-specific survival adjusted for competing risk.\nThe results of multivariate analysis for patients without liver transplant are summarized in Table 3. No significant difference between the two age groups was found. The estimated hazard ratio was 1.23 (95% CI, 0.89–1.71, p = .21) for OS; 0.90 (95% CI, 0.62–1.30, p = .57) for HCC-specific survival; and 0.90 (95% CI, 0.59–1.36, p = .61) for HCC-specific survival adjusted for competing risk.\nEstimated hazard ratios from Cox proportional hazards analysis for patients without liver transplant\nAbbreviations: ECOG, Eastern Cooperative Oncology Group; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; PVT, portal vein thrombosis.\naHigh if BMI >27 (median); low otherwise.", "In this retrospective multicenter study, we compared the clinical characteristics, treatment modalities, and outcome data for elderly (age ≥70 years) compared to younger patients (age <70 years) with HCC. Characteristics that distinguished the elderly group included lower M/F ratio, fewer patients with HCV infections, and less advanced underlying liver disease than the younger group. Several previous studies have shown differences in clinico-pathological characteristics of elderly patients with HCC including a lower rate of viral hepatitis positivity, lower M/F ratio, lower AFP values, smaller tumor diameter, and a higher proportion of stage I-II patients in elderly group. However, most of these studies were conducted in Asia where HCC has a distinct epidemiologic pattern [22–24]. Additional data available from small-scale retrospective European studies reflect a relatively homogenous patient population and particular patterns of clinical practice, thus making it difficult to extrapolate to U.S. patients. Both of these studies found shorter survival in elderly patients with HCC that could not be accounted for just by differences in tumor extent or liver failure with the conclusion reached that this difference might be due to a lower rate of therapeutic intervention in elderly patients [18, 19].\nA recent retrospective report from a U.S. institution illustrated differences in clinicopathological characteristics between young and elderly HCC patients. This cohort of patients differed from ours in that they were selected for transarterial embolization and therefore had relatively preserved hepatic function and less severe medical comorbidities. Similar to our findings, they found a lower M/F ratio and a lower rate of HCV/HBV co-infection in the elderly group. As opposed to our finding where no difference in stage distribution between the two age groups was observed, they found that the younger group had higher clinical tumor-node-metastasis stage [25]. The differences found in stage between the elderly and younger patients in the two studies may in part be due to the fact that elderly patients selected for transarterial embolization might have lower stage on average than younger patients.\nIt is interesting that a higher proportion of women in elderly HCC patients were observed in a recent epidemiologic study that analyzed age-adjusted trends in HCC incidence using U.S. Surveillance, Epidemiology, and End Results (SEER) registry data [1]. Although the reason for this difference has not been defined, it is possible that behavioral risk factors in younger males leading to earlier infection with HCV or HBV, and alcohol abuse might lead to a disproportionate increase in HCC incidence in younger males [26]. However, further studies are needed to confirm this finding, which could have implications for public health measures to develop strategies to decrease the higher rate of HCC in younger males. Moreover, SEER registry data demonstrated increased HCC incidence rates among Hispanic and black middle-aged men over the last 3 decades [1]. In our study, a higher proportion of Hispanic and black patients in the younger group were seen, although the numbers were small and not statistically different.\nHCC screening and surveillance in a well-defined risk population is considered the standard of care in the United States. Nevertheless, in our study only 32.6% of patients in the elderly group were diagnosed via active screening and surveillance, as compared to 62% of the younger patient group. Despite this, there was no difference in stage distribution between the two groups. This lack of difference may be partially explained by accidental discovery of HCC in an elderly patient undergoing state-of-the-art medical care for comorbidities. It also reflects the relative inefficiency of current surveillance and screening approaches in detecting early HCC.\nIn contrast to other solid tumors, the presence of underlying liver cirrhosis in patients with HCC adds an important dimension that cannot be overemphasized when the prognosis and treatment of HCC are assessed. Interestingly, elderly patients in our study had less advanced liver disease, with CTP class B/C in 25.3% of patients compared to 35.8% in younger counterparts, although it did not reach statistical significance (p = .063). A higher proportion of younger patients (92.5%) had underlying chronic liver disease, based on histological and/or clinical and radiological studies as opposed to 83.2% of elderly patients. This may be partially explained by the fact that a proportion of patients with poor liver function might succumb to liver failure and not survive into elderly age. It is also possible that fewer HCV infections in the elderly group may partially contribute to these findings, although genetic, environmental, and other less clearly defined factors of hepatocarcinogenesis may play a role.\nThe prognosis of HCC patients is largely determined by tumor burden, hepatic function, and performance status [27, 28]. In our study, no difference was found in tumor burden, hepatic function, and symptomatic presentation in the two groups, although performance status was worse in the elderly group. Given that HCC disease characteristics (number of lesions, AFP levels, and stage) were similar between the elderly and younger patients, this poor performance status is likely a reflection of the higher number of comorbidities in the elderly, rather than directly related to HCC.\nSeveral trends in selection of treatment modalities were noted between the two groups in our study. Because the majority of patients with HCC have underlying cirrhosis, it is the degree of hepatic dysfunction that dictates the choice of treatment [29, 30]. Not surprisingly, liver transplantation was uncommon in the elderly group, as only 10% of liver transplantation across the United States is performed in patients over 65 years old [29]. Supportive care alone was a more common course of action in the elderly, despite equal stage distribution and more preserved liver function on average in the elderly group. This discrepancy likely reflects decisions based on a poor performance status and competing medical conditions. Age-based patterns of care, as defined by practitioner and the patient, are likely implicated in this complex decision process. Our study complements the findings from the largest retrospective Western HCC database presented at ASCO 2009. Although this work illuminates differences in the pattern of care in different age groups, it does not provide details of the degree of liver dysfunction, comorbidities, and performance status, which are important for HCC treatment decisions [30].\nIn recent years, the chemotherapeutic armamentarium for HCC has expanded with introduction of the targeted therapy. Clinical trials are a major source of information for clinical decision making. However, historically, elderly patients have been underrepresented in clinical trials [31]. In contrast to lung cancer, lymphoma, breast cancer, and myeloma, there are few prospective studies designed for elderly patients with liver cancer. Elderly subgroup analyses have not been reported from the largest clinical trials involving patients with HCC. As a consequence, limited data are available about toxicity and tolerability of systemic therapy in this population. In our study, 27% of elderly patients received chemotherapy at some point of their treatment, which was not statistically different from the younger patient group. Moreover, 15% of elderly patients in our study were enrolled in clinical trials, a number similar to that of their younger counterparts. Advanced liver disease commonly presents a major limitation, when considering enrolling a HCC patient into a clinical trial. Given our results suggesting that elderly patients, on average, have less advanced liver disease, they should be offered the chance to participate in appropriate clinical trials. On the basis of our observations, elderly patients as a group are as willing to participate in clinical research as younger patients, when presented with an opportunity to enroll into a clinical trial.\nDespite a higher prevalence of comorbidities and a significantly higher mean age, overall and HCC-specific survival of elderly patients was comparable to that of their younger counterparts. Low survival rate in both groups indicating morbid nature of HCC could potentially explain this result as HCC mortality surpasses the impact of both comorbidity and age-adjusted life expectancy. These findings are in line with results previously reported focusing on liver-directed therapy and surgical intervention, which include a healthier subset of the HCC population [12, 25, 32, 33]. Given the high mortality for HCC in both the younger and the elderly populations, clinical trials that have proportionate representation of the elderly would permit extrapolation of the results to the elderly and allow for direct comparisons of outcomes for younger and older patients. The complexity of HCC diagnosis and management argues for development of a prospective multi-institutional HCC registration database. Prospective data collection is required to optimize evidence-based treatment choices appropriate for elderly patients with HCC and ensure meaningful gains in survival and quality of life in the elderly population.", "An advantage of this analysis is that there was access to demographic and clinical variables not available in many population-based databases, providing the ability to analyze associations that have not been well characterized in the past. However, the retrospective nature of this study is its biggest limitation. All patients in this study were managed at tertiary care hospitals with liver transplant programs, interventional radiology experts, and clinical trials available. Thus, our findings may not be applicable to the patients managed in settings without similar resources available." ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Patients and Methods", "Patient Selection", "Statistical Analysis", "Results", "Patients Characteristics", "Comorbidities", "Overall Survival", "HCC-Specific Survival", "HCC-Specific Survival Adjusted for Competing Risk", "Multivariate Analysis for Survival", "Discussion", "Limitation" ]

[ "The incidence and mortality of hepatocellular carcinoma (HCC) have been dramatically increasing in the United States [1]. Age is a known risk factor for HCC [2–4]. In the United States, the average age of HCC diagnosis is 65 years [5]. The incidence rate of HCC among the elderly is progressively increasing [6]. This trend may be partially attributable to screening, improved management of chronic liver disease, the rising incidence of nonalcoholic steatohepatitis-related cirrhosis, and an epidemic of hepatitis C virus (HCV) infection in the late 1970s [7]. As the population ages, it is becoming increasingly important to recognize age-specific cancer epidemiologic trends and differences in cancer outcomes to aid in prevention strategies, diagnostic approaches, and treatment decisions.\nHistorically, the elderly represent an important population particularly vulnerable to disparities in cancer treatment [8, 9]. There is a paucity of information on the possible differences in the clinical presentation, treatment patterns, and outcomes of HCC based on age. The limited available data are mainly from Asia, where HCC has a distinct underlying etiology and different treatment pattern [10–17]. Additional studies from Europe reflect disease trends inherent to a less demographically diverse population and particular patterns of clinical practice, thus making it difficult to extrapolate to U.S. patients [18, 19].\nThe increasing HCC incidence, aging population, development of new HCC treatment modalities, and limited age-related data available prompted the current study. By using a multi-institutional database, we designed the present study to examine the clinical characteristics, treatment patterns, and outcomes in elderly patients with HCC.", "[SUBTITLE] Patient Selection [SUBSECTION] Retrospective analysis of HCC patients treated at two U.S. tertiary institutions from January 1998 to December 2008 was performed. A diagnosis of HCC was confirmed histopathologically or according to the noninvasive European Association for the Study of the Liver (EASL) diagnostic criteria [20]. The age of 70 years was chosen as the cutoff point for the analysis because the 70 years of age landmark has been described as the lower boundary of senescence and the incidence of age-related organ dysfunction and the development of comorbid conditions start to increase after the age of 70 [21]. Additionally, 70 years of age has been the most frequently used cutoff age for comparisons between younger and older patients in previous analyses of HCC patients, thus allowing for some comparability with these other studies. The demographics, etiology of liver disease, tumor parameters, comorbidities, performance status, body mass index, detailed treatment information, length of survival from diagnosis, and cause of death were captured. Race/ethnicity was categorized as White, Black, Asians, and Hispanic. The etiology of the liver disease was categorized as infection with hepatitis B (HBV), HCV, alcoholic liver disease, nonalcoholic fatty liver disease, hemochromatosis, autoimmune hepatitis, and primary biliary cirrhosis. The etiology of cirrhosis was considered to be cryptogenic if no cause for chronic liver disease was identified after exhaustive testing. No quantitative analysis of alcohol consumption was performed; assessment of habitual alcohol use was based on patient self-reporting of habitual drinking.\nUnderlying cirrhosis was confirmed histologically or based on clinical and/or radiological criteria including nodular liver contour, presence of ascites, varices, enlargement of the caudate lobe, splenomegaly, and collateral portal-venous anastomoses. We also assessed whether the diagnosis of HCC was made during evaluation for possible HCC based on presenting signs or symptoms or as a result of surveillance.\nMultimodality treatment was defined as therapy that combines more than one method (e.g., combination of chemotherapy and embolization) of treatment. Death was categorized as attributable to HCC based on review of death certificate, discharge summary, or hospice documentation. Institutional Review Board approval was granted from each institution for this retrospective study.\nRetrospective analysis of HCC patients treated at two U.S. tertiary institutions from January 1998 to December 2008 was performed. A diagnosis of HCC was confirmed histopathologically or according to the noninvasive European Association for the Study of the Liver (EASL) diagnostic criteria [20]. The age of 70 years was chosen as the cutoff point for the analysis because the 70 years of age landmark has been described as the lower boundary of senescence and the incidence of age-related organ dysfunction and the development of comorbid conditions start to increase after the age of 70 [21]. Additionally, 70 years of age has been the most frequently used cutoff age for comparisons between younger and older patients in previous analyses of HCC patients, thus allowing for some comparability with these other studies. The demographics, etiology of liver disease, tumor parameters, comorbidities, performance status, body mass index, detailed treatment information, length of survival from diagnosis, and cause of death were captured. Race/ethnicity was categorized as White, Black, Asians, and Hispanic. The etiology of the liver disease was categorized as infection with hepatitis B (HBV), HCV, alcoholic liver disease, nonalcoholic fatty liver disease, hemochromatosis, autoimmune hepatitis, and primary biliary cirrhosis. The etiology of cirrhosis was considered to be cryptogenic if no cause for chronic liver disease was identified after exhaustive testing. No quantitative analysis of alcohol consumption was performed; assessment of habitual alcohol use was based on patient self-reporting of habitual drinking.\nUnderlying cirrhosis was confirmed histologically or based on clinical and/or radiological criteria including nodular liver contour, presence of ascites, varices, enlargement of the caudate lobe, splenomegaly, and collateral portal-venous anastomoses. We also assessed whether the diagnosis of HCC was made during evaluation for possible HCC based on presenting signs or symptoms or as a result of surveillance.\nMultimodality treatment was defined as therapy that combines more than one method (e.g., combination of chemotherapy and embolization) of treatment. Death was categorized as attributable to HCC based on review of death certificate, discharge summary, or hospice documentation. Institutional Review Board approval was granted from each institution for this retrospective study.\n[SUBTITLE] Statistical Analysis [SUBSECTION] The patient characteristics were summarized using mean ± SD, median and interquartile range for continuous variables, and percentages for categorical variables. Chi-square tests and Wilcoxon rank sum tests were used to compare the patient characteristics between the two age groups. Three types of survival analysis were performed to compare the clinical outcome between the elderly and younger patients: (a) Overall survival (OS) defined as the interval between the date of HCC diagnosis and the date of death due to any cause; (b) HCC-specific survival defined as the time from HCC diagnosis to death from HCC (deaths from non-HCC causes were treated as censoring); (c) HCC-specific survival, but non-HCC death was treated as a competing risk. The Kaplan-Meier method was used to estimate survival distribution for OS and HCC-specific survival. Cumulative incidence rate for HCC death was estimated for the two age groups when non-HCC death was treated as a competing risk.\nA Cox proportional hazard regression model was developed to evaluate the survival difference between the elderly and nonelderly groups while other prognostic factors were controlled. All factors listed in Table 1, but not age, were included in the original model, and a stepwise method was used for variable selection. Eastern Cooperative Oncology Group (ECOG) performance status (PS), cancer stage, presence of symptoms, Child-Turcotte-Pugh (CTP) class, portal vein thrombosis (PVT), BMI, and HCV were identified as the significant prognostic factors for OS and HCC-specific survival and included in the final model. Analyses were performed separately for patients with and without orthotopic liver transplantation (OLT) because transplanted HCC patients have substantial long-term survival and the majority of transplanted patients were in the younger group.\nPatient characteristics\n(Continued)\nAbbreviations: AFP, alpha-fetoprotein; CAD, coronary artery disease; CLIP, Cancer of the Liver Italian Program; COPD, chronic obstructive pulmonary disease; DM, diabetes mellitus; ECOG, Eastern Cooperative Oncology Group; HBV, hepatitis B virus; HCV, hepatitis C virus; HTN, hypertension; IQR, interquartile range.\nAll tests were two-sided with significance level p < .05. All statistical analyses were conducted using SAS 9.2 statistical software (SAS Institute, Cary, NC).\nThe patient characteristics were summarized using mean ± SD, median and interquartile range for continuous variables, and percentages for categorical variables. Chi-square tests and Wilcoxon rank sum tests were used to compare the patient characteristics between the two age groups. Three types of survival analysis were performed to compare the clinical outcome between the elderly and younger patients: (a) Overall survival (OS) defined as the interval between the date of HCC diagnosis and the date of death due to any cause; (b) HCC-specific survival defined as the time from HCC diagnosis to death from HCC (deaths from non-HCC causes were treated as censoring); (c) HCC-specific survival, but non-HCC death was treated as a competing risk. The Kaplan-Meier method was used to estimate survival distribution for OS and HCC-specific survival. Cumulative incidence rate for HCC death was estimated for the two age groups when non-HCC death was treated as a competing risk.\nA Cox proportional hazard regression model was developed to evaluate the survival difference between the elderly and nonelderly groups while other prognostic factors were controlled. All factors listed in Table 1, but not age, were included in the original model, and a stepwise method was used for variable selection. Eastern Cooperative Oncology Group (ECOG) performance status (PS), cancer stage, presence of symptoms, Child-Turcotte-Pugh (CTP) class, portal vein thrombosis (PVT), BMI, and HCV were identified as the significant prognostic factors for OS and HCC-specific survival and included in the final model. Analyses were performed separately for patients with and without orthotopic liver transplantation (OLT) because transplanted HCC patients have substantial long-term survival and the majority of transplanted patients were in the younger group.\nPatient characteristics\n(Continued)\nAbbreviations: AFP, alpha-fetoprotein; CAD, coronary artery disease; CLIP, Cancer of the Liver Italian Program; COPD, chronic obstructive pulmonary disease; DM, diabetes mellitus; ECOG, Eastern Cooperative Oncology Group; HBV, hepatitis B virus; HCV, hepatitis C virus; HTN, hypertension; IQR, interquartile range.\nAll tests were two-sided with significance level p < .05. All statistical analyses were conducted using SAS 9.2 statistical software (SAS Institute, Cary, NC).", "Retrospective analysis of HCC patients treated at two U.S. tertiary institutions from January 1998 to December 2008 was performed. A diagnosis of HCC was confirmed histopathologically or according to the noninvasive European Association for the Study of the Liver (EASL) diagnostic criteria [20]. The age of 70 years was chosen as the cutoff point for the analysis because the 70 years of age landmark has been described as the lower boundary of senescence and the incidence of age-related organ dysfunction and the development of comorbid conditions start to increase after the age of 70 [21]. Additionally, 70 years of age has been the most frequently used cutoff age for comparisons between younger and older patients in previous analyses of HCC patients, thus allowing for some comparability with these other studies. The demographics, etiology of liver disease, tumor parameters, comorbidities, performance status, body mass index, detailed treatment information, length of survival from diagnosis, and cause of death were captured. Race/ethnicity was categorized as White, Black, Asians, and Hispanic. The etiology of the liver disease was categorized as infection with hepatitis B (HBV), HCV, alcoholic liver disease, nonalcoholic fatty liver disease, hemochromatosis, autoimmune hepatitis, and primary biliary cirrhosis. The etiology of cirrhosis was considered to be cryptogenic if no cause for chronic liver disease was identified after exhaustive testing. No quantitative analysis of alcohol consumption was performed; assessment of habitual alcohol use was based on patient self-reporting of habitual drinking.\nUnderlying cirrhosis was confirmed histologically or based on clinical and/or radiological criteria including nodular liver contour, presence of ascites, varices, enlargement of the caudate lobe, splenomegaly, and collateral portal-venous anastomoses. We also assessed whether the diagnosis of HCC was made during evaluation for possible HCC based on presenting signs or symptoms or as a result of surveillance.\nMultimodality treatment was defined as therapy that combines more than one method (e.g., combination of chemotherapy and embolization) of treatment. Death was categorized as attributable to HCC based on review of death certificate, discharge summary, or hospice documentation. Institutional Review Board approval was granted from each institution for this retrospective study.", "The patient characteristics were summarized using mean ± SD, median and interquartile range for continuous variables, and percentages for categorical variables. Chi-square tests and Wilcoxon rank sum tests were used to compare the patient characteristics between the two age groups. Three types of survival analysis were performed to compare the clinical outcome between the elderly and younger patients: (a) Overall survival (OS) defined as the interval between the date of HCC diagnosis and the date of death due to any cause; (b) HCC-specific survival defined as the time from HCC diagnosis to death from HCC (deaths from non-HCC causes were treated as censoring); (c) HCC-specific survival, but non-HCC death was treated as a competing risk. The Kaplan-Meier method was used to estimate survival distribution for OS and HCC-specific survival. Cumulative incidence rate for HCC death was estimated for the two age groups when non-HCC death was treated as a competing risk.\nA Cox proportional hazard regression model was developed to evaluate the survival difference between the elderly and nonelderly groups while other prognostic factors were controlled. All factors listed in Table 1, but not age, were included in the original model, and a stepwise method was used for variable selection. Eastern Cooperative Oncology Group (ECOG) performance status (PS), cancer stage, presence of symptoms, Child-Turcotte-Pugh (CTP) class, portal vein thrombosis (PVT), BMI, and HCV were identified as the significant prognostic factors for OS and HCC-specific survival and included in the final model. Analyses were performed separately for patients with and without orthotopic liver transplantation (OLT) because transplanted HCC patients have substantial long-term survival and the majority of transplanted patients were in the younger group.\nPatient characteristics\n(Continued)\nAbbreviations: AFP, alpha-fetoprotein; CAD, coronary artery disease; CLIP, Cancer of the Liver Italian Program; COPD, chronic obstructive pulmonary disease; DM, diabetes mellitus; ECOG, Eastern Cooperative Oncology Group; HBV, hepatitis B virus; HCV, hepatitis C virus; HTN, hypertension; IQR, interquartile range.\nAll tests were two-sided with significance level p < .05. All statistical analyses were conducted using SAS 9.2 statistical software (SAS Institute, Cary, NC).", "[SUBTITLE] Patients Characteristics [SUBSECTION] A total of 335 patients with HCC were included. Ninety-five patients ≥70 years old (median age of 74 years) at diagnosis of HCC were defined as the elderly group. Two hundred forty patients aged <70 years (median age of 56) were regarded as the younger group. The patient characteristics are summarized in Table 1. The male (M)-to-female (F) ratio was 1.7:1 in the elderly and 5.8:1 in the younger group, showing that there was a higher proportion of women in the elderly group (p < .0001). The proportion of patients with HCV infection was significantly lower in the elderly group (21.1% versus 48.3%, p < .0001). There were no significant differences in HBV infection rates between younger and elderly patients (21.2% and 14.7%, respectively). Alcohol-induced liver disease was found in 37.5% of younger patients and 29.5% of the elderly group; the difference was not statistically significant (p = .1657). There was a trend toward more advanced CTP class in the younger group: CTP class B and C cirrhosis 35.8% versus 25.3% (p = .063). younger patients were more frequently diagnosed as a result of HCC screening or surveillance (61.3% versus 32.6%, p < .0001). More patients in the elderly group had ECOG PS > 1 than in the younger group (24.2% versus 7.9%, p < .0001). There were no significant differences in the following: race distribution, stage, number of lesions, symptomatic at presentation, PVT, alpha-fetoprotein (AFP) level, Cancer of the Liver Italian Program score, diabetes mellitus, or BMI.\nThe types of treatments received are summarized in Table 2. Forty-seven (19.6%) of the younger patients underwent liver transplant compared to 5.3% of the elderly patients (p = .0002). Conversely, a higher proportion of elderly patients received only symptomatic treatment (36.8% versus 22.9%, p = .01). There was no difference in the number of liver resections or liver directed treatments, including radiofrequency ablation, transarterial chemoembolization, radiation, and chemotherapy. The rate of utilization of multimodality treatment was 28.3% in the younger group and 23.2% in the elderly group; the difference was not statistically significant.\nTreatments received\nA total of 335 patients with HCC were included. Ninety-five patients ≥70 years old (median age of 74 years) at diagnosis of HCC were defined as the elderly group. Two hundred forty patients aged <70 years (median age of 56) were regarded as the younger group. The patient characteristics are summarized in Table 1. The male (M)-to-female (F) ratio was 1.7:1 in the elderly and 5.8:1 in the younger group, showing that there was a higher proportion of women in the elderly group (p < .0001). The proportion of patients with HCV infection was significantly lower in the elderly group (21.1% versus 48.3%, p < .0001). There were no significant differences in HBV infection rates between younger and elderly patients (21.2% and 14.7%, respectively). Alcohol-induced liver disease was found in 37.5% of younger patients and 29.5% of the elderly group; the difference was not statistically significant (p = .1657). There was a trend toward more advanced CTP class in the younger group: CTP class B and C cirrhosis 35.8% versus 25.3% (p = .063). younger patients were more frequently diagnosed as a result of HCC screening or surveillance (61.3% versus 32.6%, p < .0001). More patients in the elderly group had ECOG PS > 1 than in the younger group (24.2% versus 7.9%, p < .0001). There were no significant differences in the following: race distribution, stage, number of lesions, symptomatic at presentation, PVT, alpha-fetoprotein (AFP) level, Cancer of the Liver Italian Program score, diabetes mellitus, or BMI.\nThe types of treatments received are summarized in Table 2. Forty-seven (19.6%) of the younger patients underwent liver transplant compared to 5.3% of the elderly patients (p = .0002). Conversely, a higher proportion of elderly patients received only symptomatic treatment (36.8% versus 22.9%, p = .01). There was no difference in the number of liver resections or liver directed treatments, including radiofrequency ablation, transarterial chemoembolization, radiation, and chemotherapy. The rate of utilization of multimodality treatment was 28.3% in the younger group and 23.2% in the elderly group; the difference was not statistically significant.\nTreatments received\n[SUBTITLE] Comorbidities [SUBSECTION] Eighty nine of 95 patients (93%) in the elderly group had comorbid conditions compared with 102 of 240 patients (42.5%) in the younger group (p < .0001). Cardiovascular and cerebrovascular disease, chronic lung conditions, second primary malignancies, chronic renal disease, or cognitive disorders constituted the most common comorbid conditions. The following statistically significant differences between the two groups were observed: 36.8% elderly patients had coronary artery disease compared to 10.4% in the younger group (p < .0001); similar trends were observed for hypertension (37.9% versus 24.2%, p = .0117) and second primary malignancy (21.1% versus 3.8%, p < .0001).\nEighty nine of 95 patients (93%) in the elderly group had comorbid conditions compared with 102 of 240 patients (42.5%) in the younger group (p < .0001). Cardiovascular and cerebrovascular disease, chronic lung conditions, second primary malignancies, chronic renal disease, or cognitive disorders constituted the most common comorbid conditions. The following statistically significant differences between the two groups were observed: 36.8% elderly patients had coronary artery disease compared to 10.4% in the younger group (p < .0001); similar trends were observed for hypertension (37.9% versus 24.2%, p = .0117) and second primary malignancy (21.1% versus 3.8%, p < .0001).\n[SUBTITLE] Overall Survival [SUBSECTION] Median follow-up was 15 months for the younger group and 11 months for the elderly group. For subjects who were alive at the time of this analysis, the median follow-up time was 27 months for the elderly and 31 months for the younger.\nFor all patients, the 1-, 2-, and 3-year survival rates were 52.9 ± 5.3%, 38.3 ± 5.2%, and 27.1 ± 5.2%, respectively, for the elderly group and 63.1 ± 3.2%, 46.8 ± 3.4%, and 35.9 ± 3.4%, respectively, for the younger group. Figure 1A shows the overall survival curves for the elderly and younger group. This difference was borderline significant (log-rank test, p = .06). One hundred forty-eight of the 190 patients in the younger group and 66 of the 90 patients in the elderly group died from all causes. Figure 1B shows the overall survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. The median survival time was 12.5 months (95% CI, 7.9–18.7 months) for the elderly group and 13.9 months (95% CI, 10.2–16.4 months) for the younger group. After exclusion of transplanted patients, the 1-, 2-, and 3-year survival rates were 51.4 ± 5.4%, 35.7 ± 5.3%, and 23.0 ± 5.2%, respectively, for the elderly group and 54.4 ± 3.7%, 35.9 ± 3.7%, and 23.5 ± 3.5%, respectively, for the younger group. This difference was not statistically significant (log-rank test, p = .47).\nKaplan-Meier overall survival curves. (A): Elderly (n = 95) and younger (n =240) patients. (B): Overall survival curves of elderly (n = 90) and younger (n = 193) patients after exclusion of patients who underwent orthotic liver transplantation. p values were calculated with the use of the log-rank test.\nMedian follow-up was 15 months for the younger group and 11 months for the elderly group. For subjects who were alive at the time of this analysis, the median follow-up time was 27 months for the elderly and 31 months for the younger.\nFor all patients, the 1-, 2-, and 3-year survival rates were 52.9 ± 5.3%, 38.3 ± 5.2%, and 27.1 ± 5.2%, respectively, for the elderly group and 63.1 ± 3.2%, 46.8 ± 3.4%, and 35.9 ± 3.4%, respectively, for the younger group. Figure 1A shows the overall survival curves for the elderly and younger group. This difference was borderline significant (log-rank test, p = .06). One hundred forty-eight of the 190 patients in the younger group and 66 of the 90 patients in the elderly group died from all causes. Figure 1B shows the overall survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. The median survival time was 12.5 months (95% CI, 7.9–18.7 months) for the elderly group and 13.9 months (95% CI, 10.2–16.4 months) for the younger group. After exclusion of transplanted patients, the 1-, 2-, and 3-year survival rates were 51.4 ± 5.4%, 35.7 ± 5.3%, and 23.0 ± 5.2%, respectively, for the elderly group and 54.4 ± 3.7%, 35.9 ± 3.7%, and 23.5 ± 3.5%, respectively, for the younger group. This difference was not statistically significant (log-rank test, p = .47).\nKaplan-Meier overall survival curves. (A): Elderly (n = 95) and younger (n =240) patients. (B): Overall survival curves of elderly (n = 90) and younger (n = 193) patients after exclusion of patients who underwent orthotic liver transplantation. p values were calculated with the use of the log-rank test.\n[SUBTITLE] HCC-Specific Survival [SUBSECTION] One hundred eighty-six HCC-related deaths were documented: 46 in the elderly group and 140 in younger group. The median HCC specific survival time was 27.8 months (95% CI, 13.4–36.9 months) for the elderly group and 24.5 months (95% CI, 18.7–31.3 months) for the younger group.\nFor all patients, the 1-, 2-, and 3-year HCC-specific survival rates were 66.5 ± 5.3%, 52.3 ± 5.9%, and 39.1 ± 6.5%, respectively, for the elderly group and 66.7 ± 3.2%, 51.7 ± 3.5%, and 40.3 ± 3.6%, respectively, for the younger group. Figure 2A shows the HCC-specific survival curves for the elderly and younger group. This difference was not statistically significant (log-rank test, p = .89).\nKaplan-Meier HCC-specific survival curves. (A): All elderly (n = 95) and younger (n = 240) patients. (B): HCC-specific survival for elderly (n = 90) and younger (n =193) patients after exclusion of patients who underwent orthotic liver transplantation. p-values were calculated with the use of the log-rank test.\nFor the 280 patients who did not receive OLT, 45 (50%) of the 90 elderly patients died from HCC, versus 131 patients (69%) of 190 patients in the younger group. The median disease-specific survival time was 24.9 months (95% CI, 12.9–32.9 months) for the elderly group and 14.9 months (95% CI, 1–21.0 months) for the younger group. The estimated 1-, 2-, and 3-year HCC-specific survival rates were 65.9 ± 5.4%, 50.3 ± 6.2% and 34.9 ± 6.9%, respectively, for the elderly group and 58.4 ± 3.8%, 40.9 ± 3.9%, and 27.6 ± 3.9%, respectively, for the younger group. Figure 2B shows the HCC-specific survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. This difference was not statistically significant (log-rank test, p = .38).\nOne hundred eighty-six HCC-related deaths were documented: 46 in the elderly group and 140 in younger group. The median HCC specific survival time was 27.8 months (95% CI, 13.4–36.9 months) for the elderly group and 24.5 months (95% CI, 18.7–31.3 months) for the younger group.\nFor all patients, the 1-, 2-, and 3-year HCC-specific survival rates were 66.5 ± 5.3%, 52.3 ± 5.9%, and 39.1 ± 6.5%, respectively, for the elderly group and 66.7 ± 3.2%, 51.7 ± 3.5%, and 40.3 ± 3.6%, respectively, for the younger group. Figure 2A shows the HCC-specific survival curves for the elderly and younger group. This difference was not statistically significant (log-rank test, p = .89).\nKaplan-Meier HCC-specific survival curves. (A): All elderly (n = 95) and younger (n = 240) patients. (B): HCC-specific survival for elderly (n = 90) and younger (n =193) patients after exclusion of patients who underwent orthotic liver transplantation. p-values were calculated with the use of the log-rank test.\nFor the 280 patients who did not receive OLT, 45 (50%) of the 90 elderly patients died from HCC, versus 131 patients (69%) of 190 patients in the younger group. The median disease-specific survival time was 24.9 months (95% CI, 12.9–32.9 months) for the elderly group and 14.9 months (95% CI, 1–21.0 months) for the younger group. The estimated 1-, 2-, and 3-year HCC-specific survival rates were 65.9 ± 5.4%, 50.3 ± 6.2% and 34.9 ± 6.9%, respectively, for the elderly group and 58.4 ± 3.8%, 40.9 ± 3.9%, and 27.6 ± 3.9%, respectively, for the younger group. Figure 2B shows the HCC-specific survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. This difference was not statistically significant (log-rank test, p = .38).\n[SUBTITLE] HCC-Specific Survival Adjusted for Competing Risk [SUBSECTION] Forty-one non-HCC-related deaths were documented: 21 in the elderly group and 20 in younger group. Although non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death between the two patients groups, hazard ratio of 1.02 (95% CI, 0.73–1.43, p = .89). There were a total of 38 non-HCC deaths after transplanted patients were excluded: 21 in the elderly group and 17 in the younger group. The non-HCC-related deaths included infections, myocardial infarction, stroke, renal failure, suicide, and trauma.\nAlthough non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death among the two patients groups, hazard ratio of 0.86 (95% CI, 0.61–1.21, p = .39). However, comparing with younger patients, elderly patients had a significantly worse survival for non-HCC related deaths with a hazard ratio of 3.27 (95% CI, 1.76–6.05, p = .0002).\nForty-one non-HCC-related deaths were documented: 21 in the elderly group and 20 in younger group. Although non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death between the two patients groups, hazard ratio of 1.02 (95% CI, 0.73–1.43, p = .89). There were a total of 38 non-HCC deaths after transplanted patients were excluded: 21 in the elderly group and 17 in the younger group. The non-HCC-related deaths included infections, myocardial infarction, stroke, renal failure, suicide, and trauma.\nAlthough non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death among the two patients groups, hazard ratio of 0.86 (95% CI, 0.61–1.21, p = .39). However, comparing with younger patients, elderly patients had a significantly worse survival for non-HCC related deaths with a hazard ratio of 3.27 (95% CI, 1.76–6.05, p = .0002).\n[SUBTITLE] Multivariate Analysis for Survival [SUBSECTION] Multivariate analysis did not show a significant difference in clinical outcomes between the two age groups. When the effects of the important prognostic factors were adjusted, the estimated hazard ratio was 1.33 (95% CI, 0.97–1.82, p = .07) for OS; 0.98 (95% CI, 0.68–1.41, p = .92) for HCC-specific survival; and 0.98 (95% CI, 0.66–1.46, p = .92) for HCC-specific survival adjusted for competing risk.\nThe results of multivariate analysis for patients without liver transplant are summarized in Table 3. No significant difference between the two age groups was found. The estimated hazard ratio was 1.23 (95% CI, 0.89–1.71, p = .21) for OS; 0.90 (95% CI, 0.62–1.30, p = .57) for HCC-specific survival; and 0.90 (95% CI, 0.59–1.36, p = .61) for HCC-specific survival adjusted for competing risk.\nEstimated hazard ratios from Cox proportional hazards analysis for patients without liver transplant\nAbbreviations: ECOG, Eastern Cooperative Oncology Group; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; PVT, portal vein thrombosis.\naHigh if BMI >27 (median); low otherwise.\nMultivariate analysis did not show a significant difference in clinical outcomes between the two age groups. When the effects of the important prognostic factors were adjusted, the estimated hazard ratio was 1.33 (95% CI, 0.97–1.82, p = .07) for OS; 0.98 (95% CI, 0.68–1.41, p = .92) for HCC-specific survival; and 0.98 (95% CI, 0.66–1.46, p = .92) for HCC-specific survival adjusted for competing risk.\nThe results of multivariate analysis for patients without liver transplant are summarized in Table 3. No significant difference between the two age groups was found. The estimated hazard ratio was 1.23 (95% CI, 0.89–1.71, p = .21) for OS; 0.90 (95% CI, 0.62–1.30, p = .57) for HCC-specific survival; and 0.90 (95% CI, 0.59–1.36, p = .61) for HCC-specific survival adjusted for competing risk.\nEstimated hazard ratios from Cox proportional hazards analysis for patients without liver transplant\nAbbreviations: ECOG, Eastern Cooperative Oncology Group; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; PVT, portal vein thrombosis.\naHigh if BMI >27 (median); low otherwise.", "A total of 335 patients with HCC were included. Ninety-five patients ≥70 years old (median age of 74 years) at diagnosis of HCC were defined as the elderly group. Two hundred forty patients aged <70 years (median age of 56) were regarded as the younger group. The patient characteristics are summarized in Table 1. The male (M)-to-female (F) ratio was 1.7:1 in the elderly and 5.8:1 in the younger group, showing that there was a higher proportion of women in the elderly group (p < .0001). The proportion of patients with HCV infection was significantly lower in the elderly group (21.1% versus 48.3%, p < .0001). There were no significant differences in HBV infection rates between younger and elderly patients (21.2% and 14.7%, respectively). Alcohol-induced liver disease was found in 37.5% of younger patients and 29.5% of the elderly group; the difference was not statistically significant (p = .1657). There was a trend toward more advanced CTP class in the younger group: CTP class B and C cirrhosis 35.8% versus 25.3% (p = .063). younger patients were more frequently diagnosed as a result of HCC screening or surveillance (61.3% versus 32.6%, p < .0001). More patients in the elderly group had ECOG PS > 1 than in the younger group (24.2% versus 7.9%, p < .0001). There were no significant differences in the following: race distribution, stage, number of lesions, symptomatic at presentation, PVT, alpha-fetoprotein (AFP) level, Cancer of the Liver Italian Program score, diabetes mellitus, or BMI.\nThe types of treatments received are summarized in Table 2. Forty-seven (19.6%) of the younger patients underwent liver transplant compared to 5.3% of the elderly patients (p = .0002). Conversely, a higher proportion of elderly patients received only symptomatic treatment (36.8% versus 22.9%, p = .01). There was no difference in the number of liver resections or liver directed treatments, including radiofrequency ablation, transarterial chemoembolization, radiation, and chemotherapy. The rate of utilization of multimodality treatment was 28.3% in the younger group and 23.2% in the elderly group; the difference was not statistically significant.\nTreatments received", "Eighty nine of 95 patients (93%) in the elderly group had comorbid conditions compared with 102 of 240 patients (42.5%) in the younger group (p < .0001). Cardiovascular and cerebrovascular disease, chronic lung conditions, second primary malignancies, chronic renal disease, or cognitive disorders constituted the most common comorbid conditions. The following statistically significant differences between the two groups were observed: 36.8% elderly patients had coronary artery disease compared to 10.4% in the younger group (p < .0001); similar trends were observed for hypertension (37.9% versus 24.2%, p = .0117) and second primary malignancy (21.1% versus 3.8%, p < .0001).", "Median follow-up was 15 months for the younger group and 11 months for the elderly group. For subjects who were alive at the time of this analysis, the median follow-up time was 27 months for the elderly and 31 months for the younger.\nFor all patients, the 1-, 2-, and 3-year survival rates were 52.9 ± 5.3%, 38.3 ± 5.2%, and 27.1 ± 5.2%, respectively, for the elderly group and 63.1 ± 3.2%, 46.8 ± 3.4%, and 35.9 ± 3.4%, respectively, for the younger group. Figure 1A shows the overall survival curves for the elderly and younger group. This difference was borderline significant (log-rank test, p = .06). One hundred forty-eight of the 190 patients in the younger group and 66 of the 90 patients in the elderly group died from all causes. Figure 1B shows the overall survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. The median survival time was 12.5 months (95% CI, 7.9–18.7 months) for the elderly group and 13.9 months (95% CI, 10.2–16.4 months) for the younger group. After exclusion of transplanted patients, the 1-, 2-, and 3-year survival rates were 51.4 ± 5.4%, 35.7 ± 5.3%, and 23.0 ± 5.2%, respectively, for the elderly group and 54.4 ± 3.7%, 35.9 ± 3.7%, and 23.5 ± 3.5%, respectively, for the younger group. This difference was not statistically significant (log-rank test, p = .47).\nKaplan-Meier overall survival curves. (A): Elderly (n = 95) and younger (n =240) patients. (B): Overall survival curves of elderly (n = 90) and younger (n = 193) patients after exclusion of patients who underwent orthotic liver transplantation. p values were calculated with the use of the log-rank test.", "One hundred eighty-six HCC-related deaths were documented: 46 in the elderly group and 140 in younger group. The median HCC specific survival time was 27.8 months (95% CI, 13.4–36.9 months) for the elderly group and 24.5 months (95% CI, 18.7–31.3 months) for the younger group.\nFor all patients, the 1-, 2-, and 3-year HCC-specific survival rates were 66.5 ± 5.3%, 52.3 ± 5.9%, and 39.1 ± 6.5%, respectively, for the elderly group and 66.7 ± 3.2%, 51.7 ± 3.5%, and 40.3 ± 3.6%, respectively, for the younger group. Figure 2A shows the HCC-specific survival curves for the elderly and younger group. This difference was not statistically significant (log-rank test, p = .89).\nKaplan-Meier HCC-specific survival curves. (A): All elderly (n = 95) and younger (n = 240) patients. (B): HCC-specific survival for elderly (n = 90) and younger (n =193) patients after exclusion of patients who underwent orthotic liver transplantation. p-values were calculated with the use of the log-rank test.\nFor the 280 patients who did not receive OLT, 45 (50%) of the 90 elderly patients died from HCC, versus 131 patients (69%) of 190 patients in the younger group. The median disease-specific survival time was 24.9 months (95% CI, 12.9–32.9 months) for the elderly group and 14.9 months (95% CI, 1–21.0 months) for the younger group. The estimated 1-, 2-, and 3-year HCC-specific survival rates were 65.9 ± 5.4%, 50.3 ± 6.2% and 34.9 ± 6.9%, respectively, for the elderly group and 58.4 ± 3.8%, 40.9 ± 3.9%, and 27.6 ± 3.9%, respectively, for the younger group. Figure 2B shows the HCC-specific survival curves of the elderly group and the younger group after exclusion of patients who underwent OLT. This difference was not statistically significant (log-rank test, p = .38).", "Forty-one non-HCC-related deaths were documented: 21 in the elderly group and 20 in younger group. Although non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death between the two patients groups, hazard ratio of 1.02 (95% CI, 0.73–1.43, p = .89). There were a total of 38 non-HCC deaths after transplanted patients were excluded: 21 in the elderly group and 17 in the younger group. The non-HCC-related deaths included infections, myocardial infarction, stroke, renal failure, suicide, and trauma.\nAlthough non-HCC death was considered as a competing risk, there was no statistically significant difference in the cumulative incidence of the HCC-related death among the two patients groups, hazard ratio of 0.86 (95% CI, 0.61–1.21, p = .39). However, comparing with younger patients, elderly patients had a significantly worse survival for non-HCC related deaths with a hazard ratio of 3.27 (95% CI, 1.76–6.05, p = .0002).", "Multivariate analysis did not show a significant difference in clinical outcomes between the two age groups. When the effects of the important prognostic factors were adjusted, the estimated hazard ratio was 1.33 (95% CI, 0.97–1.82, p = .07) for OS; 0.98 (95% CI, 0.68–1.41, p = .92) for HCC-specific survival; and 0.98 (95% CI, 0.66–1.46, p = .92) for HCC-specific survival adjusted for competing risk.\nThe results of multivariate analysis for patients without liver transplant are summarized in Table 3. No significant difference between the two age groups was found. The estimated hazard ratio was 1.23 (95% CI, 0.89–1.71, p = .21) for OS; 0.90 (95% CI, 0.62–1.30, p = .57) for HCC-specific survival; and 0.90 (95% CI, 0.59–1.36, p = .61) for HCC-specific survival adjusted for competing risk.\nEstimated hazard ratios from Cox proportional hazards analysis for patients without liver transplant\nAbbreviations: ECOG, Eastern Cooperative Oncology Group; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; PVT, portal vein thrombosis.\naHigh if BMI >27 (median); low otherwise.", "In this retrospective multicenter study, we compared the clinical characteristics, treatment modalities, and outcome data for elderly (age ≥70 years) compared to younger patients (age <70 years) with HCC. Characteristics that distinguished the elderly group included lower M/F ratio, fewer patients with HCV infections, and less advanced underlying liver disease than the younger group. Several previous studies have shown differences in clinico-pathological characteristics of elderly patients with HCC including a lower rate of viral hepatitis positivity, lower M/F ratio, lower AFP values, smaller tumor diameter, and a higher proportion of stage I-II patients in elderly group. However, most of these studies were conducted in Asia where HCC has a distinct epidemiologic pattern [22–24]. Additional data available from small-scale retrospective European studies reflect a relatively homogenous patient population and particular patterns of clinical practice, thus making it difficult to extrapolate to U.S. patients. Both of these studies found shorter survival in elderly patients with HCC that could not be accounted for just by differences in tumor extent or liver failure with the conclusion reached that this difference might be due to a lower rate of therapeutic intervention in elderly patients [18, 19].\nA recent retrospective report from a U.S. institution illustrated differences in clinicopathological characteristics between young and elderly HCC patients. This cohort of patients differed from ours in that they were selected for transarterial embolization and therefore had relatively preserved hepatic function and less severe medical comorbidities. Similar to our findings, they found a lower M/F ratio and a lower rate of HCV/HBV co-infection in the elderly group. As opposed to our finding where no difference in stage distribution between the two age groups was observed, they found that the younger group had higher clinical tumor-node-metastasis stage [25]. The differences found in stage between the elderly and younger patients in the two studies may in part be due to the fact that elderly patients selected for transarterial embolization might have lower stage on average than younger patients.\nIt is interesting that a higher proportion of women in elderly HCC patients were observed in a recent epidemiologic study that analyzed age-adjusted trends in HCC incidence using U.S. Surveillance, Epidemiology, and End Results (SEER) registry data [1]. Although the reason for this difference has not been defined, it is possible that behavioral risk factors in younger males leading to earlier infection with HCV or HBV, and alcohol abuse might lead to a disproportionate increase in HCC incidence in younger males [26]. However, further studies are needed to confirm this finding, which could have implications for public health measures to develop strategies to decrease the higher rate of HCC in younger males. Moreover, SEER registry data demonstrated increased HCC incidence rates among Hispanic and black middle-aged men over the last 3 decades [1]. In our study, a higher proportion of Hispanic and black patients in the younger group were seen, although the numbers were small and not statistically different.\nHCC screening and surveillance in a well-defined risk population is considered the standard of care in the United States. Nevertheless, in our study only 32.6% of patients in the elderly group were diagnosed via active screening and surveillance, as compared to 62% of the younger patient group. Despite this, there was no difference in stage distribution between the two groups. This lack of difference may be partially explained by accidental discovery of HCC in an elderly patient undergoing state-of-the-art medical care for comorbidities. It also reflects the relative inefficiency of current surveillance and screening approaches in detecting early HCC.\nIn contrast to other solid tumors, the presence of underlying liver cirrhosis in patients with HCC adds an important dimension that cannot be overemphasized when the prognosis and treatment of HCC are assessed. Interestingly, elderly patients in our study had less advanced liver disease, with CTP class B/C in 25.3% of patients compared to 35.8% in younger counterparts, although it did not reach statistical significance (p = .063). A higher proportion of younger patients (92.5%) had underlying chronic liver disease, based on histological and/or clinical and radiological studies as opposed to 83.2% of elderly patients. This may be partially explained by the fact that a proportion of patients with poor liver function might succumb to liver failure and not survive into elderly age. It is also possible that fewer HCV infections in the elderly group may partially contribute to these findings, although genetic, environmental, and other less clearly defined factors of hepatocarcinogenesis may play a role.\nThe prognosis of HCC patients is largely determined by tumor burden, hepatic function, and performance status [27, 28]. In our study, no difference was found in tumor burden, hepatic function, and symptomatic presentation in the two groups, although performance status was worse in the elderly group. Given that HCC disease characteristics (number of lesions, AFP levels, and stage) were similar between the elderly and younger patients, this poor performance status is likely a reflection of the higher number of comorbidities in the elderly, rather than directly related to HCC.\nSeveral trends in selection of treatment modalities were noted between the two groups in our study. Because the majority of patients with HCC have underlying cirrhosis, it is the degree of hepatic dysfunction that dictates the choice of treatment [29, 30]. Not surprisingly, liver transplantation was uncommon in the elderly group, as only 10% of liver transplantation across the United States is performed in patients over 65 years old [29]. Supportive care alone was a more common course of action in the elderly, despite equal stage distribution and more preserved liver function on average in the elderly group. This discrepancy likely reflects decisions based on a poor performance status and competing medical conditions. Age-based patterns of care, as defined by practitioner and the patient, are likely implicated in this complex decision process. Our study complements the findings from the largest retrospective Western HCC database presented at ASCO 2009. Although this work illuminates differences in the pattern of care in different age groups, it does not provide details of the degree of liver dysfunction, comorbidities, and performance status, which are important for HCC treatment decisions [30].\nIn recent years, the chemotherapeutic armamentarium for HCC has expanded with introduction of the targeted therapy. Clinical trials are a major source of information for clinical decision making. However, historically, elderly patients have been underrepresented in clinical trials [31]. In contrast to lung cancer, lymphoma, breast cancer, and myeloma, there are few prospective studies designed for elderly patients with liver cancer. Elderly subgroup analyses have not been reported from the largest clinical trials involving patients with HCC. As a consequence, limited data are available about toxicity and tolerability of systemic therapy in this population. In our study, 27% of elderly patients received chemotherapy at some point of their treatment, which was not statistically different from the younger patient group. Moreover, 15% of elderly patients in our study were enrolled in clinical trials, a number similar to that of their younger counterparts. Advanced liver disease commonly presents a major limitation, when considering enrolling a HCC patient into a clinical trial. Given our results suggesting that elderly patients, on average, have less advanced liver disease, they should be offered the chance to participate in appropriate clinical trials. On the basis of our observations, elderly patients as a group are as willing to participate in clinical research as younger patients, when presented with an opportunity to enroll into a clinical trial.\nDespite a higher prevalence of comorbidities and a significantly higher mean age, overall and HCC-specific survival of elderly patients was comparable to that of their younger counterparts. Low survival rate in both groups indicating morbid nature of HCC could potentially explain this result as HCC mortality surpasses the impact of both comorbidity and age-adjusted life expectancy. These findings are in line with results previously reported focusing on liver-directed therapy and surgical intervention, which include a healthier subset of the HCC population [12, 25, 32, 33]. Given the high mortality for HCC in both the younger and the elderly populations, clinical trials that have proportionate representation of the elderly would permit extrapolation of the results to the elderly and allow for direct comparisons of outcomes for younger and older patients. The complexity of HCC diagnosis and management argues for development of a prospective multi-institutional HCC registration database. Prospective data collection is required to optimize evidence-based treatment choices appropriate for elderly patients with HCC and ensure meaningful gains in survival and quality of life in the elderly population.", "An advantage of this analysis is that there was access to demographic and clinical variables not available in many population-based databases, providing the ability to analyze associations that have not been well characterized in the past. However, the retrospective nature of this study is its biggest limitation. All patients in this study were managed at tertiary care hospitals with liver transplant programs, interventional radiology experts, and clinical trials available. Thus, our findings may not be applicable to the patients managed in settings without similar resources available." ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null ]

[ "Hepatocellular carcinoma", "Epidemiology", "Elderly", "Treatment pattern", "Survival" ]

[ "Animals", "Down-Regulation", "Electrophoresis", "Female", "Infrared Rays", "Lens, Crystalline", "Male", "Rabbits", "Sodium-Potassium-Exchanging ATPase", "Spectroscopy, Fourier Transform Infrared", "Time Factors" ]

[]

[]

[]

[ "Materials and Methods", "Results", "Discussion" ]

[ "Fifteen healthy New Zealand rabbits of either sex, weighing 2–2.5 kg, were used for this study. The animals were divided into three groups; one of them served as control. The other two groups were exposed to IR for 5 or 10 minutes. Animals from these two irradiated groups were subdivided into two subgroups; one of them was decapitated directly after IR exposure, while the other subgroup was decapitated 1 hour post exposure.\nIR was delivered from a General Electric Lamp, model 250R 50/10, placed 20 cm from the rabbit and aimed at each eye. The lamp was calibrated at the Photometry Department, National Institute of standards, Giza, Egypt. The wavelengths emitted from the anterior surface of the IR lamp as provided by General Electric Lighting Division (Cleveland, OH, USA)[3] were 0.34–0.4 µm (UV light), 0.4–0.76 µm (visible light), 0.76–3.0 µm (IR-A, IR-B light 83%) and 3.0–7.0 µm (IR-C 10%). The total IR percentage emitted was 93%. The irradiance of the IR lamp detected at 20 cm was 0.2 W/cm2. The heat flux reaching the rabbit cornea after 5 minute IR exposure was 44 J/cm2; consequently, the heat flux reaching the cornea after 10 minutes was calculated and found to be 88 J/cm2.\nThe lenses were removed from the eye and their capsules were removed carefully. Each lens capsule was weighed in a separate container, and then homogenized in extraction medium [0.32 M sucrose, 1 mM ethylenediaminetetraacetic acid (EDTA) and 0.15% deoxycholic acid]. Na+-K+ ATPase measurement was carried out on the lens membrane by the method of Bowler and Tirri.[13]\nThe lenses without their capsules were weighed, homogenized separately in de-ionized water and centrifuged at 16,000 rpm to extract soluble lens proteins and then stored at –20°C for the following measurements.\nTotal proteins in the soluble part of crystalline lens were determined by the method of Lowry et al.[14] This method depends on preliminary treatment of proteins with an alkaline copper reagent followed by foline-phenol reagent. The developing color (measured spectrophotometrically at 750 nm) for different proteins depends on tyrosine, tryptophan content and the sequence of various amino acids with functional side groups, especially, histidine, arginine and glutamic acid.\nSoluble lens proteins were separated according to their molecular weights by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) according to Laemmli,[15] using 5% stacking gel and 12% separating gel. The data were represented graphically with an automatic scanner (model R-112, manufactured by Beckman Coulter, CA, USA).\nFourier transformation infrared (FTIR) spectra of lyophilized soluble lens proteins were recorded on JASCO FTIR 430 spectrometer (JASCO Corporation, Tokyo, Japan) in the range 4000–400 cm–1. Measurements were made with an IR cell (KBr windows) according to Lamba et al.[16] The resolution was 2.0 cm–1 and all the spectra were recorded at physiological temperature. The instrument was operated under continuous N2 gas to reduce the effects of atmospheric CO2 and water vapor.\nData were presented as the mean ± SD. To determine the significance difference between the groups, analysis of variance (ANOVA) procedure was used followed by student’s t-test, where commercially available statistical software package, SPSS-11 for windows, was used. The significance level was set at P < 0.001.[17]", "Table 1 gives the total soluble lens proteins of control rabbits and those exposed to IR radiation. The mean value of the control lens was 290.8 ± 3.8 mg/g wet wt. The exposed groups showed significant decrease, with an average value of 256.3 ± 15.7 mg/g wet wt. for both the groups exposed to IR radiation for 5 minutes. On the other hand, the average value of the groups exposed to IR for 10 minutes was 222.6 ± 18.5 mg/g wet wt. Na+-K+ ATPase activity [Table 1] shows a similar behavior as the total lens proteins but with different significant levels. For normal lens membrane, the enzyme activity was 48.2 ± 3.2 µMpi/hour/g wet wt. After exposure of rabbit eyes to IR radiation for 5 minutes, the enzyme activity was decreased for both the groups (P < 0.01), with an average value of 35.8 ± 1.7 µMpi/hour/g wet wt. After exposure to IR for 10 minutes also, the enzyme activity for both the groups decreased, with an average value of 25.9 ± 1.5 µMpi/hour/g wet wt.\nTotal soluble protein content of rabbit lens and activity of Na+-K+ ATPase of rabbit lens membrane for the different studied groups\nStatistically significant\nThe electrophoretic patterns of the lens proteins for control [Fig. 1, panel a], 5 minutes exposure to IR direct and 1 hour post exposure [Fig. 1, panel b] and 10 minutes exposure to IR direct and 1 hour post exposure [Fig. 1, panel c] are shown.\nElectrophoretic profile of rabbit’s lens protein for (panel a) control group, (panel b) 5 minutes exposed groups to IR, directly decapitated and after 1 hour, and (panel c) 10 minutes exposed groups to IR, directly decapitated and after 1 hour\nIn panel a of Fig. 1, the control was characterized by the presence of 10 peaks representing different soluble protein fractions with specific intensities and broadening, which cover the molecular weight range 177–47 kDa. In panel b of Fig. 1, the direct effect revealed the reduction of the soluble protein fractions to nine peaks which cover the molecular weight range 198–53 kDa. The maximum intensity was noticed for the last protein fraction (53 kDa). The reduction in the number of soluble protein fractions was pronounced after 1 hour of IR exposure (eight peaks). These peaks cover the molecular weight range 199–46 kDa. The maximum intensities were noticed for the last two fractions (60 and 46 kDa). Panel c of Fig. 1 shows the same result as that of panel b. The molecular weight range for 10 minute exposure-direct decapitation was 206–44 kDa, whereas after 1 hour of IR irradiation the molecular weight range was 205–42 kDa. The intensity of the peaks in the molecular weight range 126–44 kDa for the direct effect is higher relative to the corresponding peaks in 1 hour post exposure subgroup.\nFig. 2 shows the IR spectra of soluble lens proteins for control, 5 minute and 10 minute groups in the range 4000–3000 cm–1. The NH-OH region of the control pattern indicates the presence of six bands that centered nearly at 3850 cm–1 (strOH), 3739 cm–1 (strOH), 3616 cm–1 (strOH), 3288 cm–1 (symOH), 3144 cm–1 (symNH) and 3072 cm–1 (CHring), respectively, as previously mentioned by Dovbeshko et al.[18] There were dramatic changes in the NH-OH pattern of all IR irradiated groups with different characteristics that differed according to the exposure period. For 5 minute irradiated groups, the changes can be noticed in the strOH region as reduced number of bands or as change in the contour; also, there was a distortion in the rest of bands where the contour was broad but not for the CHring where its intensities were increased. On the other hand, as the exposure time was increased to 10 minutes, the strOH was greatly affected where the number of bands was increased concomitant with the presence of asymOH vibration mode. Again, the intensity of CHring was increased. The asymNH mode can be noticed at 5 minutes-1 hour group as well as 10 minutes-direct one.\nFTIR of soluble lens proteins for all studied groups in the NH-OH region (4000–3000 cm–1)\nFig. 3 shows IR spectra of soluble lens proteins for control, 5 minute and 10 minute groups in the range 3000-2800 cm–1. This range characterizes the CH stretching region. The mean band position of control asymCH3 was 2967 ± 3 cm–1 and its bandwidth was 25 ± 3 cm–1, while for irradiated groups it was 2965 ± 3 cm–1 and 26 ± 2 cm-1, respectively. The same phenomenon was noticed for the other two bands: asymCH2 (for control: 2930 ± 2 cm–1 and 43 ± 6 cm–1; for irradiated groups: 2929 ± 3 cm–1 and 40 ± 7 cm–1) and CHsym (for control: 2973 ± 2 cm–1 and 32 ± 5 cm–1; for irradiated groups: 2971 ± 2 cm–1 and 31 ± 6 cm–1), i.e., no changes in either the band position or the bandwidth.\nCH stretching region (3000–2800 cm–1) of soluble lens proteins studied by Fourier transform infrared for all the groups\nFig. 4 shows the fingerprint region for all groups involved in this study. The interesting observation in this figure is the splitting of amide I band and amide II band into two components each. The mean vibrational frequency of control amide I was 1668 ± 1 cm–1. In the same context, the two components that were noticed in amide I region of irradiated groups had a mean frequency of 1680 ± 1 cm–1 and 1650 ± 2 cm–1, respectively. On the other hand, for amide II region, the mean vibrational frequency of the two bands in irradiated groups was 1543 ± 2 cm–1 and 1618 ± 2 cm–1, while in the control pattern this band was centered on 1533 ± 2 cm–1. The intensity of CH2 bending mode was increased for irradiated groups for 10 minute rather than 5 minute groups. COOsym band was characterized by change in its contour at the 10 minutes-1 hour group. The frequency of amide III band was increased from 1237 ± 1 cm–1 in the control to 1241 ± 1 cm–1 in the IR irradiated groups. No change in the band intensity or band width was found for 5 minute irradiated groups. As the exposure time was increased to 10 minutes, the intensity of amide III band fluctuated, i.e., increased for the direct effect and reduced for the delayed 1 hour group. The bandwidth was reduced relative to the control in the direct studied group. The previous band assignment was based on that described by Toyran et al.[19]\nFTIR of soluble lens proteins for all the studied groups in the range 1700–1200 cm–1 (fingerprint region)", "Exposure to IR radiation may cause the corneal opacity, burns on the retina, miosis, breakdown of blood–aqueous barrier and delayed cataract.[20–23] The present study is an attempt to investigate the effect of IR radiation with different exposure times (5 and 10 minutes) on the molecular structure of the soluble lens proteins.\nThe lens proteins are markedly decreased after exposure to IR radiation for all the studied groups. This decrease is directly proportional to the time of exposure. Also, the decrease of total lens proteins is more pronounced in the group consisting of animals which were decapitated after 1 hour of exposure. This decrease may be due to cataract formation. When IR radiation is incident on the eye, it is absorbed by the cornea and converted into heat which is then conducted to the lens and induces cataract.[4] The changes in the lens crystalline, evidenced by SDS-PAGE, are given in Fig. 1. It is clear that there were changes in the molecular weight, electrophoretic mobility and intensity of different peaks representing different crystalline fractions. The obtained changes in the molecular weight of lens proteins are responsible for the decrease in the concentration of soluble lens proteins of all the studied groups and this may be a characteristic of cataract. Michael et al.[24] concluded that during cataract formation, the decrease of biosynthesis of lens crystalline was followed by their aggregations which then led to opacity of the lens.\nNa+-K+ ATPase is an enzyme that has been found to play a major role in ionic transport through cell membrane. Decreased levels of Na+-K+ ATPase activity have been detected after exposure to IR radiation. This decrease may be due to the disturbance of lens cell membrane function, which leads to a decrease in active transport of nutrients and electrolytes from the aqueous into the lens. Kantorow et al.[25] and Delamere et al.[26] found that Na+-K+ ATPase has a lower activity in certain types of human cataract and also in a number of experimental cataracts.\nThe decrease in the total soluble proteins and the change in the molecular weight of lens proteins result from the conformational changes in protein secondary structure, which were noticed in the FTIR data. Chen et al.[27] recorded similar findings as ours, where compositional and conformational changes of lens proteins were associated with cataract formation. In the present study, the dramatic changes noticed in the NH-OH region [Fig. 2] indicate that the soluble lens protein was greatly influenced by IR radiation and that the magnitude of response was proportional to the exposure time. These changes indicate the presence of different interaction mechanisms between the soluble lens crystallines that function in the exposure time, i.e., there are two different environments. The vibrational motion of hydrocarbon chains of proteins is unaffected by these IR exposure periods [Fig. 3]. The soluble protein part of the rabbit’s lens was greatly affected by the IR irradiation periods of 5–10 minutes [Fig. 4]. Amide I band indicates that the soluble protein secondary structure was modified with the domination of β-turns structure (band at 1680 cm–1) and the α-helix one (band at 1652 cm–1). The presence of β-turns structure implies that protein becomes more folded and may tend to be aggregated. These findings were clearly confirmed by the observation of splitting of amide II band.\nIn conclusion, our FTIR findings provide information about the mechanism of protein structural changes induced by IR exposure periods of 5 or 10 minutes. IR exposure may participate in cataract formation. Furthermore, the alteration of the soluble lens protein secondary structure resulted in a reduction of the Na+-K+ ATPase activity." ]

[ "materials|methods", "results", "discussion" ]

[ "Electrophoresis", "Fourier transform infrared spectroscopy", "infrared", "lens", "protein", "rabbit" ]

[ "Decision Making", "Evidence-Based Medicine", "Humans", "Practice Guidelines as Topic", "Review Literature as Topic", "Terminology as Topic", "Urologic Neoplasms" ]

[ "Challenges in urological cancer treatment decision-making", "The strengths and limitations of systematic review methodology", "Defining the question", "Development of urological cancer care pathways", "What are care pathways?", "Methodology", "Engagement with clinical content experts and patient groups", "Use of urological cancer care pathways", "Standardisation of terminology (treatment options and definitions)", "Informing search strategy", "Prioritisation (review scope, guideline scope and development process)", "Economic evaluation and care pathways" ]

[ "In making healthcare decisions regarding treatment, decision makers are confronted with several fundamental issues. At the individual level, patients and clinicians are primarily concerned with the balance between the perceived benefits and harms of treatment discussed, whilst for healthcare systems, policy makers need to know which treatments should be provided. This societal aspect will be informed by information on costs, cost-effectiveness and arguably, some notion of fairness such as equal access for equal need. In conditions such as localised prostate cancer for which multiple management options exist, the situation becomes even more complex. One approach to facilitate individual decision-making and decisions regarding treatment provision is evidence-based medicine [1], in which choices are made on the basis of the best available evidence obtained from robust methodology; evidence that is valid, reliable and of high quality. To be useful, this evidence must be readily accessible. Systematic reviews are one method of identifying and synthesising research evidence on a particular subject [2]. The review process, review findings and subsequent guideline recommendations can also be used to identify gaps in knowledge about treatment effects (uncertainties) and inform future work to address important gaps.", "Reasons for undertaking a systematic review include resolution of conflicting evidence or clinical uncertainties, explanation of variations in practice, or to confirm the appropriateness of current practice. To achieve these outcomes, a systematic review requires a transparent and replicable synthesis process [3] with effect estimates obtained by meta-analysis using appropriate statistical techniques. Typically, a systematic review involves a well-formulated question or questions, comprehensive searches of the major electronic databases, pre-defined study eligibility criteria, an unbiased study selection process and data extraction against a set of pre-determined outcomes using standard forms. These data are critically appraised, including rating of the quality of the evidence and quantitative synthesis with meta-analysis where appropriate. This transparent and replicable process differentiates systematic from narrative reviews, which are more prone to bias.\nA key limitation of any systematic review is that it cannot overcome problems inherent in the design, conduct and reporting of the included primary studies [3–5]. Their authority can also be undermined by errors in review methodology or reporting leading to variation in conclusions drawn by separate systematic reviews attempting to answer the same question. As a consequence, widely accepted guidelines for the conduct and reporting of primary studies and systematic reviews, such as CONSORT, Cochrane Handbook and PRISMA, [6–8] have been established.", "Defining the review question for a systematic review is the first and most important stage of the process as this provides the direction for all subsequent stages [9]. Paradoxically, while methodology has advanced for the actual process of a systematic review, there has been little work done to establish the best way of identifying areas of clinical uncertainty and prioritising a list of questions that maintains relevance for all interested groups. One approach is to perform a scoping study [10]; this involves an initial literature search to assess whether a full systematic review is both feasible, that is there are sufficient primary studies available for synthesis, and relevant, that is there is no existing equivalent review document. Another approach is evidence mapping [11], which was used by the Global Evidence Mapping Initiative in Australia [12]. Here, the number and quality of relevant studies retrieved from literature searches and their summary outcomes are tabulated for each condition or treatment of interest. A potential drawback of this approach is that it does not map the entirety of the research that could be conducted within a given clinical subject area or indeed the importance of any gaps within the evidence.\nThe urological cancer care pathways (UCAN care pathways from here on) being developed by our group are an attempt to address these issues. In September 2004, we facilitated plenary discussions involving urological clinicians, patients and their partners including expertise and experience of the five main cancers: kidney, bladder, prostate, testis and penis. The purpose was to better understand the needs of individuals with urological cancer in the context of their cultural setting and health system. The principal messages from patients and their families were that the following improvements are needed: (1) Better and more accessible information backed up by evidence, which could help them make decisions about their care, (2) Better care of for those that suffer unwanted effects of cancer treatment, and (3) Better support throughout their journey of care both in the clinical setting and at home. In response, a Scottish Charity (UCAN) committed funding of £2.6 million to address these gaps in urological cancer care focussed on people living in north-east Scotland but also reflecting an international perspective. Discussion with the research group on how to address the first principal message, the requirement for better and more accessible evidenced-based information for patients led to the development methodology to formulate UCAN care pathways, which we discuss in the present paper.\nOur main objective is to map all plausible treatment options for each of the five urological cancers to allow collection of an appropriate range of existing and new systematic evidence reviews of effectiveness and cost-effectiveness of alternative treatment options including the magnitude of risk of short- and long-term adverse effects. We will also engage key clinical experts and patients to prioritise unanswered questions and identify significant evidence gaps in the evidence to direct future research priorities. The ultimate aim of the UCAN care pathways is to build an evidence base for the major urological cancers and develop a framework to inform future systematic reviews, clinical practice guidelines, care algorithms, integrated care pathways and research priorities [13]. We will also work to establish a framework for wider engagement of clinicians, patients, researchers, healthcare policy makers and healthcare funders in priority setting within this clinical specialty and inform the development of a set of core outcomes for both research and clinical practices. The strength of the UCAN care pathways will be to harness and rationalise all our efforts to answering the key questions for universal benefit.", "[SUBTITLE] What are care pathways? [SUBSECTION] A care pathway is a tool constructed with multidisciplinary input and used by healthcare professionals and/or researchers to map a patient’s journey in terms of which treatments should be given, by whom, when, and to what outcome [13–16]. This broad characterisation includes many different terminologies including care profile, care protocol, critical care pathway, care map, integrated care pathway, and different uses dependent on the context, user and the stage in the research process at which it is considered. Systematic reviewers, for example, may use a care pathway to identify all plausible treatment options for a particular patient group with a view to identifying gaps in current knowledge, thereby informing possible review questions. Policy makers may use a care pathway to establish a standardised treatment algorithm for a specific clinical problem in clinical practice that can then be implemented in a particular healthcare organisation [13, 17].\nThese examples can be seen as occupying two ends of a process: driving the research agenda on the one hand and providing a clinical useable output on the other. We see the UCAN care pathways more as being drivers of the research agenda but also providing structure for guideline developers. They are designed to inform the scope, search strategy, development and prioritisation of focussed clinical questions for relevant systematic reviews. They lend themselves well to informing development (including scope) of clinical practice guidelines rather than informing actual practice as is the purpose of clinical practice guidelines. They provide a framework for systematic review teams and guidelines panel members to identify and prioritise topic areas, to highlight areas where guidance is currently lacking or the evidence is unclear and to reduce redundancy and resource wasting.\nA care pathway is a tool constructed with multidisciplinary input and used by healthcare professionals and/or researchers to map a patient’s journey in terms of which treatments should be given, by whom, when, and to what outcome [13–16]. This broad characterisation includes many different terminologies including care profile, care protocol, critical care pathway, care map, integrated care pathway, and different uses dependent on the context, user and the stage in the research process at which it is considered. Systematic reviewers, for example, may use a care pathway to identify all plausible treatment options for a particular patient group with a view to identifying gaps in current knowledge, thereby informing possible review questions. Policy makers may use a care pathway to establish a standardised treatment algorithm for a specific clinical problem in clinical practice that can then be implemented in a particular healthcare organisation [13, 17].\nThese examples can be seen as occupying two ends of a process: driving the research agenda on the one hand and providing a clinical useable output on the other. We see the UCAN care pathways more as being drivers of the research agenda but also providing structure for guideline developers. They are designed to inform the scope, search strategy, development and prioritisation of focussed clinical questions for relevant systematic reviews. They lend themselves well to informing development (including scope) of clinical practice guidelines rather than informing actual practice as is the purpose of clinical practice guidelines. They provide a framework for systematic review teams and guidelines panel members to identify and prioritise topic areas, to highlight areas where guidance is currently lacking or the evidence is unclear and to reduce redundancy and resource wasting.\n[SUBTITLE] Methodology [SUBSECTION] Steps in the development of the UCAN care pathwaysDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.Establishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.Draughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.Iterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\n\nDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.\nEstablishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.\nDraughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.\nIterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\nLocalised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nLocally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nMetastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nHormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nLocalised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nAdvanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nTesticular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nPenile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nMuscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nNon-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\nSteps in the development of the UCAN care pathwaysDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.Establishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.Draughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.Iterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\n\nDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.\nEstablishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.\nDraughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.\nIterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\nLocalised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nLocally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nMetastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nHormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nLocalised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nAdvanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nTesticular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nPenile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nMuscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nNon-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n[SUBTITLE] Engagement with clinical content experts and patient groups [SUBSECTION] The reasons for engaging with content experts are threefold. First, to ensure that the UCAN care pathways are comprehensive and reflect contemporary clinical practice within the pre-defined societal location. Second, to achieve engagement with the process and to give ownership of the UCAN care pathways, the systematic reviews and any subsequent clinical practice guidelines driven by the systematic reviews to key figures (opinion formers) within the discipline. This is important to facilitate any required behaviour change and the adoption of evidence-based practice across the discipline. Third, to develop an international collaboration to advance the practice of evidence-based medicine within urology. Patient involvement is also a necessary step in identifying appropriate patient-reported outcomes, understanding which approaches to delivering care and which outcomes are most important to patients, facilitating shared decision-making and increasing patient satisfaction.\nThe reasons for engaging with content experts are threefold. First, to ensure that the UCAN care pathways are comprehensive and reflect contemporary clinical practice within the pre-defined societal location. Second, to achieve engagement with the process and to give ownership of the UCAN care pathways, the systematic reviews and any subsequent clinical practice guidelines driven by the systematic reviews to key figures (opinion formers) within the discipline. This is important to facilitate any required behaviour change and the adoption of evidence-based practice across the discipline. Third, to develop an international collaboration to advance the practice of evidence-based medicine within urology. Patient involvement is also a necessary step in identifying appropriate patient-reported outcomes, understanding which approaches to delivering care and which outcomes are most important to patients, facilitating shared decision-making and increasing patient satisfaction.", "A care pathway is a tool constructed with multidisciplinary input and used by healthcare professionals and/or researchers to map a patient’s journey in terms of which treatments should be given, by whom, when, and to what outcome [13–16]. This broad characterisation includes many different terminologies including care profile, care protocol, critical care pathway, care map, integrated care pathway, and different uses dependent on the context, user and the stage in the research process at which it is considered. Systematic reviewers, for example, may use a care pathway to identify all plausible treatment options for a particular patient group with a view to identifying gaps in current knowledge, thereby informing possible review questions. Policy makers may use a care pathway to establish a standardised treatment algorithm for a specific clinical problem in clinical practice that can then be implemented in a particular healthcare organisation [13, 17].\nThese examples can be seen as occupying two ends of a process: driving the research agenda on the one hand and providing a clinical useable output on the other. We see the UCAN care pathways more as being drivers of the research agenda but also providing structure for guideline developers. They are designed to inform the scope, search strategy, development and prioritisation of focussed clinical questions for relevant systematic reviews. They lend themselves well to informing development (including scope) of clinical practice guidelines rather than informing actual practice as is the purpose of clinical practice guidelines. They provide a framework for systematic review teams and guidelines panel members to identify and prioritise topic areas, to highlight areas where guidance is currently lacking or the evidence is unclear and to reduce redundancy and resource wasting.", "Steps in the development of the UCAN care pathwaysDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.Establishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.Draughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.Iterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\n\nDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.\nEstablishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.\nDraughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.\nIterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\nLocalised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nLocally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nMetastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nHormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nLocalised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nAdvanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nTesticular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nPenile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nMuscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nNon-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n", "The reasons for engaging with content experts are threefold. First, to ensure that the UCAN care pathways are comprehensive and reflect contemporary clinical practice within the pre-defined societal location. Second, to achieve engagement with the process and to give ownership of the UCAN care pathways, the systematic reviews and any subsequent clinical practice guidelines driven by the systematic reviews to key figures (opinion formers) within the discipline. This is important to facilitate any required behaviour change and the adoption of evidence-based practice across the discipline. Third, to develop an international collaboration to advance the practice of evidence-based medicine within urology. Patient involvement is also a necessary step in identifying appropriate patient-reported outcomes, understanding which approaches to delivering care and which outcomes are most important to patients, facilitating shared decision-making and increasing patient satisfaction.", "[SUBTITLE] Standardisation of terminology (treatment options and definitions) [SUBSECTION] Our methodology gives a clear opportunity to gain clear consensus regarding definitions of treatment options. This precision in terminology is vital to ensure that the conduct and reporting of research, the systematic review process and the clinical practice guidelines development process are relevant to all.\nOur methodology gives a clear opportunity to gain clear consensus regarding definitions of treatment options. This precision in terminology is vital to ensure that the conduct and reporting of research, the systematic review process and the clinical practice guidelines development process are relevant to all.\n[SUBTITLE] Informing search strategy [SUBSECTION] The aim of the search strategy for a systematic review is to maximise both the sensitivity and the specificity of the search results, that is, to retrieve all the relevant articles and exclude the irrelevant ones. In order to design the strategy, the question first needs to be translated into searchable terms using the PICO (Patient, Intervention, Comparison, Outcome) framework. The UCAN care pathways greatly facilitate this step, as all plausible interventions that can be compared are clearly shown for each patient group at every stage in the disease process. Also, the format of the pathway allows the reviewer to see where the search question is focussed in the context of the entire pathway and to see which interventions and stages of the disease are being excluded from the search. This is particularly useful when refining and reviewing the results of the search.\nTherefore, as the UCAN care pathways succinctly represent the required detailed and complex information, the reviewer is readily able to understand the specific patient groups and interventions being reviewed and focus the search accordingly. Hence, the pathways are a key element in the design and execution of searches to inform systematic reviews for specific stages of the pathway. The same process can be employed in the definition of focused clinical questions by clinical practice guideline panels.\nThe aim of the search strategy for a systematic review is to maximise both the sensitivity and the specificity of the search results, that is, to retrieve all the relevant articles and exclude the irrelevant ones. In order to design the strategy, the question first needs to be translated into searchable terms using the PICO (Patient, Intervention, Comparison, Outcome) framework. The UCAN care pathways greatly facilitate this step, as all plausible interventions that can be compared are clearly shown for each patient group at every stage in the disease process. Also, the format of the pathway allows the reviewer to see where the search question is focussed in the context of the entire pathway and to see which interventions and stages of the disease are being excluded from the search. This is particularly useful when refining and reviewing the results of the search.\nTherefore, as the UCAN care pathways succinctly represent the required detailed and complex information, the reviewer is readily able to understand the specific patient groups and interventions being reviewed and focus the search accordingly. Hence, the pathways are a key element in the design and execution of searches to inform systematic reviews for specific stages of the pathway. The same process can be employed in the definition of focused clinical questions by clinical practice guideline panels.\n[SUBTITLE] Prioritisation (review scope, guideline scope and development process) [SUBSECTION] Systematic reviews can be used to provide more reliable and precise estimates of relative effectiveness of competing treatments and highlight areas where further primary research is needed. Resources are however always limited, and therefore, the questions requiring performance of a systematic review need to be prioritised. There is also a linked need to prioritise topics for clinical practice guidelines based on evidence from systematic reviews. Prioritisation should be a process that is systematic and transparent and that includes consultation with key experts [18].\nThe UCAN care pathways inform the prioritisation process by providing a comprehensive framework for discussion that is applicable to multiple settings both nationally and internationally. They also inform the prioritisation process by highlighting where in the process of care the uncertainty is and the magnitude of the problem in terms of the number of people affected and its consequences. As such, they provide a concise summary of the various treatment options and inform the scope of systematic reviews and clinical practice guidelines. The UCAN care pathway model also ensures that the systematic review activity is firmly embedded in the context of the overall management of patients.\nSystematic reviews can be used to provide more reliable and precise estimates of relative effectiveness of competing treatments and highlight areas where further primary research is needed. Resources are however always limited, and therefore, the questions requiring performance of a systematic review need to be prioritised. There is also a linked need to prioritise topics for clinical practice guidelines based on evidence from systematic reviews. Prioritisation should be a process that is systematic and transparent and that includes consultation with key experts [18].\nThe UCAN care pathways inform the prioritisation process by providing a comprehensive framework for discussion that is applicable to multiple settings both nationally and internationally. They also inform the prioritisation process by highlighting where in the process of care the uncertainty is and the magnitude of the problem in terms of the number of people affected and its consequences. As such, they provide a concise summary of the various treatment options and inform the scope of systematic reviews and clinical practice guidelines. The UCAN care pathway model also ensures that the systematic review activity is firmly embedded in the context of the overall management of patients.\n[SUBTITLE] Economic evaluation and care pathways [SUBSECTION] Economic evaluation involves the comparative analysis of alternative interventions in terms of benefits such as improvement in health, the value of those improvements to individual patients and costs [19]. An economic evaluation has to be based upon a care pathway, as an understanding is needed of what the sequence of events is from initiation of the treatment under study. The UCAN care pathways can be used to identify events that will influence a patient’s well-being and events that use or save resources. Once these events have been established, they can then be measured and valued [20].\nEstimates of benefits and costs can be based on data from randomised controlled trials or from a mathematical model. The models can be constructed using decision analytic methods to provide a mathematical representation of a UCAN care pathway. A model is a further level of evidence synthesis as it may incorporate the results from a number of systematic reviews.\nIn addition to information on cost-effectiveness, an economic model can show who stands to benefit most from use of a particular healthcare intervention and who is most likely to bear the cost. These analyses can inform judgements about equity of provision. Economic models can also be used to highlight areas for further research either by showing precisely what evidence is missing or more formally by quantifying the value of further research in terms of balancing the cost of gaining the new evidence against the subsequent benefits that are likely to accrue by calculation of value of information [21].\nEconomic evaluation involves the comparative analysis of alternative interventions in terms of benefits such as improvement in health, the value of those improvements to individual patients and costs [19]. An economic evaluation has to be based upon a care pathway, as an understanding is needed of what the sequence of events is from initiation of the treatment under study. The UCAN care pathways can be used to identify events that will influence a patient’s well-being and events that use or save resources. Once these events have been established, they can then be measured and valued [20].\nEstimates of benefits and costs can be based on data from randomised controlled trials or from a mathematical model. The models can be constructed using decision analytic methods to provide a mathematical representation of a UCAN care pathway. A model is a further level of evidence synthesis as it may incorporate the results from a number of systematic reviews.\nIn addition to information on cost-effectiveness, an economic model can show who stands to benefit most from use of a particular healthcare intervention and who is most likely to bear the cost. These analyses can inform judgements about equity of provision. Economic models can also be used to highlight areas for further research either by showing precisely what evidence is missing or more formally by quantifying the value of further research in terms of balancing the cost of gaining the new evidence against the subsequent benefits that are likely to accrue by calculation of value of information [21].", "Our methodology gives a clear opportunity to gain clear consensus regarding definitions of treatment options. This precision in terminology is vital to ensure that the conduct and reporting of research, the systematic review process and the clinical practice guidelines development process are relevant to all.", "The aim of the search strategy for a systematic review is to maximise both the sensitivity and the specificity of the search results, that is, to retrieve all the relevant articles and exclude the irrelevant ones. In order to design the strategy, the question first needs to be translated into searchable terms using the PICO (Patient, Intervention, Comparison, Outcome) framework. The UCAN care pathways greatly facilitate this step, as all plausible interventions that can be compared are clearly shown for each patient group at every stage in the disease process. Also, the format of the pathway allows the reviewer to see where the search question is focussed in the context of the entire pathway and to see which interventions and stages of the disease are being excluded from the search. This is particularly useful when refining and reviewing the results of the search.\nTherefore, as the UCAN care pathways succinctly represent the required detailed and complex information, the reviewer is readily able to understand the specific patient groups and interventions being reviewed and focus the search accordingly. Hence, the pathways are a key element in the design and execution of searches to inform systematic reviews for specific stages of the pathway. The same process can be employed in the definition of focused clinical questions by clinical practice guideline panels.", "Systematic reviews can be used to provide more reliable and precise estimates of relative effectiveness of competing treatments and highlight areas where further primary research is needed. Resources are however always limited, and therefore, the questions requiring performance of a systematic review need to be prioritised. There is also a linked need to prioritise topics for clinical practice guidelines based on evidence from systematic reviews. Prioritisation should be a process that is systematic and transparent and that includes consultation with key experts [18].\nThe UCAN care pathways inform the prioritisation process by providing a comprehensive framework for discussion that is applicable to multiple settings both nationally and internationally. They also inform the prioritisation process by highlighting where in the process of care the uncertainty is and the magnitude of the problem in terms of the number of people affected and its consequences. As such, they provide a concise summary of the various treatment options and inform the scope of systematic reviews and clinical practice guidelines. The UCAN care pathway model also ensures that the systematic review activity is firmly embedded in the context of the overall management of patients.", "Economic evaluation involves the comparative analysis of alternative interventions in terms of benefits such as improvement in health, the value of those improvements to individual patients and costs [19]. An economic evaluation has to be based upon a care pathway, as an understanding is needed of what the sequence of events is from initiation of the treatment under study. The UCAN care pathways can be used to identify events that will influence a patient’s well-being and events that use or save resources. Once these events have been established, they can then be measured and valued [20].\nEstimates of benefits and costs can be based on data from randomised controlled trials or from a mathematical model. The models can be constructed using decision analytic methods to provide a mathematical representation of a UCAN care pathway. A model is a further level of evidence synthesis as it may incorporate the results from a number of systematic reviews.\nIn addition to information on cost-effectiveness, an economic model can show who stands to benefit most from use of a particular healthcare intervention and who is most likely to bear the cost. These analyses can inform judgements about equity of provision. Economic models can also be used to highlight areas for further research either by showing precisely what evidence is missing or more formally by quantifying the value of further research in terms of balancing the cost of gaining the new evidence against the subsequent benefits that are likely to accrue by calculation of value of information [21]." ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Challenges in urological cancer treatment decision-making", "The strengths and limitations of systematic review methodology", "Defining the question", "Development of urological cancer care pathways", "What are care pathways?", "Methodology", "Engagement with clinical content experts and patient groups", "Use of urological cancer care pathways", "Standardisation of terminology (treatment options and definitions)", "Informing search strategy", "Prioritisation (review scope, guideline scope and development process)", "Economic evaluation and care pathways", "Conclusions" ]

[ "[SUBTITLE] Challenges in urological cancer treatment decision-making [SUBSECTION] In making healthcare decisions regarding treatment, decision makers are confronted with several fundamental issues. At the individual level, patients and clinicians are primarily concerned with the balance between the perceived benefits and harms of treatment discussed, whilst for healthcare systems, policy makers need to know which treatments should be provided. This societal aspect will be informed by information on costs, cost-effectiveness and arguably, some notion of fairness such as equal access for equal need. In conditions such as localised prostate cancer for which multiple management options exist, the situation becomes even more complex. One approach to facilitate individual decision-making and decisions regarding treatment provision is evidence-based medicine [1], in which choices are made on the basis of the best available evidence obtained from robust methodology; evidence that is valid, reliable and of high quality. To be useful, this evidence must be readily accessible. Systematic reviews are one method of identifying and synthesising research evidence on a particular subject [2]. The review process, review findings and subsequent guideline recommendations can also be used to identify gaps in knowledge about treatment effects (uncertainties) and inform future work to address important gaps.\nIn making healthcare decisions regarding treatment, decision makers are confronted with several fundamental issues. At the individual level, patients and clinicians are primarily concerned with the balance between the perceived benefits and harms of treatment discussed, whilst for healthcare systems, policy makers need to know which treatments should be provided. This societal aspect will be informed by information on costs, cost-effectiveness and arguably, some notion of fairness such as equal access for equal need. In conditions such as localised prostate cancer for which multiple management options exist, the situation becomes even more complex. One approach to facilitate individual decision-making and decisions regarding treatment provision is evidence-based medicine [1], in which choices are made on the basis of the best available evidence obtained from robust methodology; evidence that is valid, reliable and of high quality. To be useful, this evidence must be readily accessible. Systematic reviews are one method of identifying and synthesising research evidence on a particular subject [2]. The review process, review findings and subsequent guideline recommendations can also be used to identify gaps in knowledge about treatment effects (uncertainties) and inform future work to address important gaps.\n[SUBTITLE] The strengths and limitations of systematic review methodology [SUBSECTION] Reasons for undertaking a systematic review include resolution of conflicting evidence or clinical uncertainties, explanation of variations in practice, or to confirm the appropriateness of current practice. To achieve these outcomes, a systematic review requires a transparent and replicable synthesis process [3] with effect estimates obtained by meta-analysis using appropriate statistical techniques. Typically, a systematic review involves a well-formulated question or questions, comprehensive searches of the major electronic databases, pre-defined study eligibility criteria, an unbiased study selection process and data extraction against a set of pre-determined outcomes using standard forms. These data are critically appraised, including rating of the quality of the evidence and quantitative synthesis with meta-analysis where appropriate. This transparent and replicable process differentiates systematic from narrative reviews, which are more prone to bias.\nA key limitation of any systematic review is that it cannot overcome problems inherent in the design, conduct and reporting of the included primary studies [3–5]. Their authority can also be undermined by errors in review methodology or reporting leading to variation in conclusions drawn by separate systematic reviews attempting to answer the same question. As a consequence, widely accepted guidelines for the conduct and reporting of primary studies and systematic reviews, such as CONSORT, Cochrane Handbook and PRISMA, [6–8] have been established.\nReasons for undertaking a systematic review include resolution of conflicting evidence or clinical uncertainties, explanation of variations in practice, or to confirm the appropriateness of current practice. To achieve these outcomes, a systematic review requires a transparent and replicable synthesis process [3] with effect estimates obtained by meta-analysis using appropriate statistical techniques. Typically, a systematic review involves a well-formulated question or questions, comprehensive searches of the major electronic databases, pre-defined study eligibility criteria, an unbiased study selection process and data extraction against a set of pre-determined outcomes using standard forms. These data are critically appraised, including rating of the quality of the evidence and quantitative synthesis with meta-analysis where appropriate. This transparent and replicable process differentiates systematic from narrative reviews, which are more prone to bias.\nA key limitation of any systematic review is that it cannot overcome problems inherent in the design, conduct and reporting of the included primary studies [3–5]. Their authority can also be undermined by errors in review methodology or reporting leading to variation in conclusions drawn by separate systematic reviews attempting to answer the same question. As a consequence, widely accepted guidelines for the conduct and reporting of primary studies and systematic reviews, such as CONSORT, Cochrane Handbook and PRISMA, [6–8] have been established.\n[SUBTITLE] Defining the question [SUBSECTION] Defining the review question for a systematic review is the first and most important stage of the process as this provides the direction for all subsequent stages [9]. Paradoxically, while methodology has advanced for the actual process of a systematic review, there has been little work done to establish the best way of identifying areas of clinical uncertainty and prioritising a list of questions that maintains relevance for all interested groups. One approach is to perform a scoping study [10]; this involves an initial literature search to assess whether a full systematic review is both feasible, that is there are sufficient primary studies available for synthesis, and relevant, that is there is no existing equivalent review document. Another approach is evidence mapping [11], which was used by the Global Evidence Mapping Initiative in Australia [12]. Here, the number and quality of relevant studies retrieved from literature searches and their summary outcomes are tabulated for each condition or treatment of interest. A potential drawback of this approach is that it does not map the entirety of the research that could be conducted within a given clinical subject area or indeed the importance of any gaps within the evidence.\nThe urological cancer care pathways (UCAN care pathways from here on) being developed by our group are an attempt to address these issues. In September 2004, we facilitated plenary discussions involving urological clinicians, patients and their partners including expertise and experience of the five main cancers: kidney, bladder, prostate, testis and penis. The purpose was to better understand the needs of individuals with urological cancer in the context of their cultural setting and health system. The principal messages from patients and their families were that the following improvements are needed: (1) Better and more accessible information backed up by evidence, which could help them make decisions about their care, (2) Better care of for those that suffer unwanted effects of cancer treatment, and (3) Better support throughout their journey of care both in the clinical setting and at home. In response, a Scottish Charity (UCAN) committed funding of £2.6 million to address these gaps in urological cancer care focussed on people living in north-east Scotland but also reflecting an international perspective. Discussion with the research group on how to address the first principal message, the requirement for better and more accessible evidenced-based information for patients led to the development methodology to formulate UCAN care pathways, which we discuss in the present paper.\nOur main objective is to map all plausible treatment options for each of the five urological cancers to allow collection of an appropriate range of existing and new systematic evidence reviews of effectiveness and cost-effectiveness of alternative treatment options including the magnitude of risk of short- and long-term adverse effects. We will also engage key clinical experts and patients to prioritise unanswered questions and identify significant evidence gaps in the evidence to direct future research priorities. The ultimate aim of the UCAN care pathways is to build an evidence base for the major urological cancers and develop a framework to inform future systematic reviews, clinical practice guidelines, care algorithms, integrated care pathways and research priorities [13]. We will also work to establish a framework for wider engagement of clinicians, patients, researchers, healthcare policy makers and healthcare funders in priority setting within this clinical specialty and inform the development of a set of core outcomes for both research and clinical practices. The strength of the UCAN care pathways will be to harness and rationalise all our efforts to answering the key questions for universal benefit.\nDefining the review question for a systematic review is the first and most important stage of the process as this provides the direction for all subsequent stages [9]. Paradoxically, while methodology has advanced for the actual process of a systematic review, there has been little work done to establish the best way of identifying areas of clinical uncertainty and prioritising a list of questions that maintains relevance for all interested groups. One approach is to perform a scoping study [10]; this involves an initial literature search to assess whether a full systematic review is both feasible, that is there are sufficient primary studies available for synthesis, and relevant, that is there is no existing equivalent review document. Another approach is evidence mapping [11], which was used by the Global Evidence Mapping Initiative in Australia [12]. Here, the number and quality of relevant studies retrieved from literature searches and their summary outcomes are tabulated for each condition or treatment of interest. A potential drawback of this approach is that it does not map the entirety of the research that could be conducted within a given clinical subject area or indeed the importance of any gaps within the evidence.\nThe urological cancer care pathways (UCAN care pathways from here on) being developed by our group are an attempt to address these issues. In September 2004, we facilitated plenary discussions involving urological clinicians, patients and their partners including expertise and experience of the five main cancers: kidney, bladder, prostate, testis and penis. The purpose was to better understand the needs of individuals with urological cancer in the context of their cultural setting and health system. The principal messages from patients and their families were that the following improvements are needed: (1) Better and more accessible information backed up by evidence, which could help them make decisions about their care, (2) Better care of for those that suffer unwanted effects of cancer treatment, and (3) Better support throughout their journey of care both in the clinical setting and at home. In response, a Scottish Charity (UCAN) committed funding of £2.6 million to address these gaps in urological cancer care focussed on people living in north-east Scotland but also reflecting an international perspective. Discussion with the research group on how to address the first principal message, the requirement for better and more accessible evidenced-based information for patients led to the development methodology to formulate UCAN care pathways, which we discuss in the present paper.\nOur main objective is to map all plausible treatment options for each of the five urological cancers to allow collection of an appropriate range of existing and new systematic evidence reviews of effectiveness and cost-effectiveness of alternative treatment options including the magnitude of risk of short- and long-term adverse effects. We will also engage key clinical experts and patients to prioritise unanswered questions and identify significant evidence gaps in the evidence to direct future research priorities. The ultimate aim of the UCAN care pathways is to build an evidence base for the major urological cancers and develop a framework to inform future systematic reviews, clinical practice guidelines, care algorithms, integrated care pathways and research priorities [13]. We will also work to establish a framework for wider engagement of clinicians, patients, researchers, healthcare policy makers and healthcare funders in priority setting within this clinical specialty and inform the development of a set of core outcomes for both research and clinical practices. The strength of the UCAN care pathways will be to harness and rationalise all our efforts to answering the key questions for universal benefit.", "In making healthcare decisions regarding treatment, decision makers are confronted with several fundamental issues. At the individual level, patients and clinicians are primarily concerned with the balance between the perceived benefits and harms of treatment discussed, whilst for healthcare systems, policy makers need to know which treatments should be provided. This societal aspect will be informed by information on costs, cost-effectiveness and arguably, some notion of fairness such as equal access for equal need. In conditions such as localised prostate cancer for which multiple management options exist, the situation becomes even more complex. One approach to facilitate individual decision-making and decisions regarding treatment provision is evidence-based medicine [1], in which choices are made on the basis of the best available evidence obtained from robust methodology; evidence that is valid, reliable and of high quality. To be useful, this evidence must be readily accessible. Systematic reviews are one method of identifying and synthesising research evidence on a particular subject [2]. The review process, review findings and subsequent guideline recommendations can also be used to identify gaps in knowledge about treatment effects (uncertainties) and inform future work to address important gaps.", "Reasons for undertaking a systematic review include resolution of conflicting evidence or clinical uncertainties, explanation of variations in practice, or to confirm the appropriateness of current practice. To achieve these outcomes, a systematic review requires a transparent and replicable synthesis process [3] with effect estimates obtained by meta-analysis using appropriate statistical techniques. Typically, a systematic review involves a well-formulated question or questions, comprehensive searches of the major electronic databases, pre-defined study eligibility criteria, an unbiased study selection process and data extraction against a set of pre-determined outcomes using standard forms. These data are critically appraised, including rating of the quality of the evidence and quantitative synthesis with meta-analysis where appropriate. This transparent and replicable process differentiates systematic from narrative reviews, which are more prone to bias.\nA key limitation of any systematic review is that it cannot overcome problems inherent in the design, conduct and reporting of the included primary studies [3–5]. Their authority can also be undermined by errors in review methodology or reporting leading to variation in conclusions drawn by separate systematic reviews attempting to answer the same question. As a consequence, widely accepted guidelines for the conduct and reporting of primary studies and systematic reviews, such as CONSORT, Cochrane Handbook and PRISMA, [6–8] have been established.", "Defining the review question for a systematic review is the first and most important stage of the process as this provides the direction for all subsequent stages [9]. Paradoxically, while methodology has advanced for the actual process of a systematic review, there has been little work done to establish the best way of identifying areas of clinical uncertainty and prioritising a list of questions that maintains relevance for all interested groups. One approach is to perform a scoping study [10]; this involves an initial literature search to assess whether a full systematic review is both feasible, that is there are sufficient primary studies available for synthesis, and relevant, that is there is no existing equivalent review document. Another approach is evidence mapping [11], which was used by the Global Evidence Mapping Initiative in Australia [12]. Here, the number and quality of relevant studies retrieved from literature searches and their summary outcomes are tabulated for each condition or treatment of interest. A potential drawback of this approach is that it does not map the entirety of the research that could be conducted within a given clinical subject area or indeed the importance of any gaps within the evidence.\nThe urological cancer care pathways (UCAN care pathways from here on) being developed by our group are an attempt to address these issues. In September 2004, we facilitated plenary discussions involving urological clinicians, patients and their partners including expertise and experience of the five main cancers: kidney, bladder, prostate, testis and penis. The purpose was to better understand the needs of individuals with urological cancer in the context of their cultural setting and health system. The principal messages from patients and their families were that the following improvements are needed: (1) Better and more accessible information backed up by evidence, which could help them make decisions about their care, (2) Better care of for those that suffer unwanted effects of cancer treatment, and (3) Better support throughout their journey of care both in the clinical setting and at home. In response, a Scottish Charity (UCAN) committed funding of £2.6 million to address these gaps in urological cancer care focussed on people living in north-east Scotland but also reflecting an international perspective. Discussion with the research group on how to address the first principal message, the requirement for better and more accessible evidenced-based information for patients led to the development methodology to formulate UCAN care pathways, which we discuss in the present paper.\nOur main objective is to map all plausible treatment options for each of the five urological cancers to allow collection of an appropriate range of existing and new systematic evidence reviews of effectiveness and cost-effectiveness of alternative treatment options including the magnitude of risk of short- and long-term adverse effects. We will also engage key clinical experts and patients to prioritise unanswered questions and identify significant evidence gaps in the evidence to direct future research priorities. The ultimate aim of the UCAN care pathways is to build an evidence base for the major urological cancers and develop a framework to inform future systematic reviews, clinical practice guidelines, care algorithms, integrated care pathways and research priorities [13]. We will also work to establish a framework for wider engagement of clinicians, patients, researchers, healthcare policy makers and healthcare funders in priority setting within this clinical specialty and inform the development of a set of core outcomes for both research and clinical practices. The strength of the UCAN care pathways will be to harness and rationalise all our efforts to answering the key questions for universal benefit.", "[SUBTITLE] What are care pathways? [SUBSECTION] A care pathway is a tool constructed with multidisciplinary input and used by healthcare professionals and/or researchers to map a patient’s journey in terms of which treatments should be given, by whom, when, and to what outcome [13–16]. This broad characterisation includes many different terminologies including care profile, care protocol, critical care pathway, care map, integrated care pathway, and different uses dependent on the context, user and the stage in the research process at which it is considered. Systematic reviewers, for example, may use a care pathway to identify all plausible treatment options for a particular patient group with a view to identifying gaps in current knowledge, thereby informing possible review questions. Policy makers may use a care pathway to establish a standardised treatment algorithm for a specific clinical problem in clinical practice that can then be implemented in a particular healthcare organisation [13, 17].\nThese examples can be seen as occupying two ends of a process: driving the research agenda on the one hand and providing a clinical useable output on the other. We see the UCAN care pathways more as being drivers of the research agenda but also providing structure for guideline developers. They are designed to inform the scope, search strategy, development and prioritisation of focussed clinical questions for relevant systematic reviews. They lend themselves well to informing development (including scope) of clinical practice guidelines rather than informing actual practice as is the purpose of clinical practice guidelines. They provide a framework for systematic review teams and guidelines panel members to identify and prioritise topic areas, to highlight areas where guidance is currently lacking or the evidence is unclear and to reduce redundancy and resource wasting.\nA care pathway is a tool constructed with multidisciplinary input and used by healthcare professionals and/or researchers to map a patient’s journey in terms of which treatments should be given, by whom, when, and to what outcome [13–16]. This broad characterisation includes many different terminologies including care profile, care protocol, critical care pathway, care map, integrated care pathway, and different uses dependent on the context, user and the stage in the research process at which it is considered. Systematic reviewers, for example, may use a care pathway to identify all plausible treatment options for a particular patient group with a view to identifying gaps in current knowledge, thereby informing possible review questions. Policy makers may use a care pathway to establish a standardised treatment algorithm for a specific clinical problem in clinical practice that can then be implemented in a particular healthcare organisation [13, 17].\nThese examples can be seen as occupying two ends of a process: driving the research agenda on the one hand and providing a clinical useable output on the other. We see the UCAN care pathways more as being drivers of the research agenda but also providing structure for guideline developers. They are designed to inform the scope, search strategy, development and prioritisation of focussed clinical questions for relevant systematic reviews. They lend themselves well to informing development (including scope) of clinical practice guidelines rather than informing actual practice as is the purpose of clinical practice guidelines. They provide a framework for systematic review teams and guidelines panel members to identify and prioritise topic areas, to highlight areas where guidance is currently lacking or the evidence is unclear and to reduce redundancy and resource wasting.\n[SUBTITLE] Methodology [SUBSECTION] Steps in the development of the UCAN care pathwaysDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.Establishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.Draughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.Iterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\n\nDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.\nEstablishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.\nDraughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.\nIterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\nLocalised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nLocally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nMetastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nHormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nLocalised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nAdvanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nTesticular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nPenile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nMuscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nNon-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\nSteps in the development of the UCAN care pathwaysDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.Establishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.Draughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.Iterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\n\nDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.\nEstablishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.\nDraughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.\nIterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\nLocalised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nLocally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nMetastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nHormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nLocalised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nAdvanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nTesticular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nPenile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nMuscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nNon-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n[SUBTITLE] Engagement with clinical content experts and patient groups [SUBSECTION] The reasons for engaging with content experts are threefold. First, to ensure that the UCAN care pathways are comprehensive and reflect contemporary clinical practice within the pre-defined societal location. Second, to achieve engagement with the process and to give ownership of the UCAN care pathways, the systematic reviews and any subsequent clinical practice guidelines driven by the systematic reviews to key figures (opinion formers) within the discipline. This is important to facilitate any required behaviour change and the adoption of evidence-based practice across the discipline. Third, to develop an international collaboration to advance the practice of evidence-based medicine within urology. Patient involvement is also a necessary step in identifying appropriate patient-reported outcomes, understanding which approaches to delivering care and which outcomes are most important to patients, facilitating shared decision-making and increasing patient satisfaction.\nThe reasons for engaging with content experts are threefold. First, to ensure that the UCAN care pathways are comprehensive and reflect contemporary clinical practice within the pre-defined societal location. Second, to achieve engagement with the process and to give ownership of the UCAN care pathways, the systematic reviews and any subsequent clinical practice guidelines driven by the systematic reviews to key figures (opinion formers) within the discipline. This is important to facilitate any required behaviour change and the adoption of evidence-based practice across the discipline. Third, to develop an international collaboration to advance the practice of evidence-based medicine within urology. Patient involvement is also a necessary step in identifying appropriate patient-reported outcomes, understanding which approaches to delivering care and which outcomes are most important to patients, facilitating shared decision-making and increasing patient satisfaction.", "A care pathway is a tool constructed with multidisciplinary input and used by healthcare professionals and/or researchers to map a patient’s journey in terms of which treatments should be given, by whom, when, and to what outcome [13–16]. This broad characterisation includes many different terminologies including care profile, care protocol, critical care pathway, care map, integrated care pathway, and different uses dependent on the context, user and the stage in the research process at which it is considered. Systematic reviewers, for example, may use a care pathway to identify all plausible treatment options for a particular patient group with a view to identifying gaps in current knowledge, thereby informing possible review questions. Policy makers may use a care pathway to establish a standardised treatment algorithm for a specific clinical problem in clinical practice that can then be implemented in a particular healthcare organisation [13, 17].\nThese examples can be seen as occupying two ends of a process: driving the research agenda on the one hand and providing a clinical useable output on the other. We see the UCAN care pathways more as being drivers of the research agenda but also providing structure for guideline developers. They are designed to inform the scope, search strategy, development and prioritisation of focussed clinical questions for relevant systematic reviews. They lend themselves well to informing development (including scope) of clinical practice guidelines rather than informing actual practice as is the purpose of clinical practice guidelines. They provide a framework for systematic review teams and guidelines panel members to identify and prioritise topic areas, to highlight areas where guidance is currently lacking or the evidence is unclear and to reduce redundancy and resource wasting.", "Steps in the development of the UCAN care pathwaysDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.Establishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.Draughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.Iterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\n\nDefinition of the aim of an individual UCAN care pathway. For example, what are the treatment options for localised prostate cancer? This will require a structured list of options and the eligibility characteristics to enter the pathway.\nEstablishment of an advisory group. This should include both clinical and methodological experts drawn from national and international professional bodies, together with representation from patient groups as lay experts and those who have experienced the disease. The purpose of the advisory group is to provide a range of user perspectives necessary to inform the development of the pathway.\nDraughting of all plausible treatment options from the point of diagnosis moving forward to the end of treatment, follow-up and death in the form of a flow chart to producing a comprehensive but concise one-page summary.\nIterative development of the pathway by consensus clearly describing key areas of uncertainty to create a final version for dissemination (Figs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10).Fig. 1Localised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nFig. 2Locally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 3Metastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 4Hormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nFig. 5Localised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nFig. 6Advanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nFig. 7Testicular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nFig. 8Penile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nFig. 9Muscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nFig. 10Non-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n\n\nLocalised prostate cancer care pathway. Abbreviations: PSA prostate-specific antigen, EBRT electron beam radiotherapy, IMRT intensity-modulated radiation therapy, HRPC hormone refractory prostate cancer, LHRH luteinizing hormone-releasing hormone, DXT deep X-ray therapy, TURP transurethral resection of the prostate, GVAX granulocyte–macrophage colony-stimulating factor [or GM-CSF] vaccine, ZD4054 Zibotentan\nLocally advanced prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nMetastatic prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nHormone-resistant prostate cancer care pathway. For abbreviations, refer Fig. 1\n\nLocalised renal cell cancer care pathway. Abbreviations: HIFU high-intensity focussed ultrasound, CT computed tomography, IFA interferon-alpha, IL2 interleukin 2, 5FU 5 Fluorouracil\nAdvanced renal cell cancer care pathway. For abbreviations, refer Fig. 5\n\nTesticular cancer care pathway. Abbreviations: AFP alpha-fetoprotein, bhCG beta-human chorionic gonadotropin, BEP bleomycin/etoposide/cisplatin, C/A/P chest/abdomen/pelvis, CBOP carboplatin, bleomycin, vincristine, and cisplatin, CT computed tomography, EP etoposide/cisplatin, FSH Follicle-stimulating hormone, HIV human immunodeficiency virus, HBV hepatitis B virus, HCV hepatitis C virus, i-s inguinoscrotal, LDH lactate dehydrogenase, LH luteinizing hormone, med 1°  medistinal primary, NSGCT non-seminoma germ cell tumour, PET positron emission tomography, RP 1° retroperitoneal primary, RPLND retroperitoneal lymph node dissection, RT radiotherapy, Te21 trial to compare Taxol-BEP vs BEP alone in intermediate prognosis disease, Te23 trial comparing CBOP BEP combinations in poor prognosis with BEP, TRISST TRial of imaging and schedule in seminoma testis, UK United Kingdom, USA United States of America, Yrs years\nPenile cancer care pathway. Abbreviations: 5-FU 5 fluorouracil, CIS carcinoma in situ, DSNB dynamic sentinel node biopsy, FNAC Fine needle aspiration cytology, MDT multidisciplinary team, SCC squamous cell carcinoma, USS ultrasound scanning\nMuscle invasive metastatic bladder cancer care pathway. Abbreviations: BCG bacillus calmette-guérin, CIS carcinoma in situ, CXR chest X-ray, INFα interferon-alpha, IVU intravenous urogram, LUTS lower urinary tract symptoms, LND lymph node dissection, MIBC muscle invasive bladder cancer, NMIBC non-muscle invasive bladder cancer, Prostatic TCC transitional cell carcinoma of the prostate, TB tuberculosis\nNon-muscle invasive bladder cancer care pathway. For abbreviations, refer Fig. 9\n", "The reasons for engaging with content experts are threefold. First, to ensure that the UCAN care pathways are comprehensive and reflect contemporary clinical practice within the pre-defined societal location. Second, to achieve engagement with the process and to give ownership of the UCAN care pathways, the systematic reviews and any subsequent clinical practice guidelines driven by the systematic reviews to key figures (opinion formers) within the discipline. This is important to facilitate any required behaviour change and the adoption of evidence-based practice across the discipline. Third, to develop an international collaboration to advance the practice of evidence-based medicine within urology. Patient involvement is also a necessary step in identifying appropriate patient-reported outcomes, understanding which approaches to delivering care and which outcomes are most important to patients, facilitating shared decision-making and increasing patient satisfaction.", "[SUBTITLE] Standardisation of terminology (treatment options and definitions) [SUBSECTION] Our methodology gives a clear opportunity to gain clear consensus regarding definitions of treatment options. This precision in terminology is vital to ensure that the conduct and reporting of research, the systematic review process and the clinical practice guidelines development process are relevant to all.\nOur methodology gives a clear opportunity to gain clear consensus regarding definitions of treatment options. This precision in terminology is vital to ensure that the conduct and reporting of research, the systematic review process and the clinical practice guidelines development process are relevant to all.\n[SUBTITLE] Informing search strategy [SUBSECTION] The aim of the search strategy for a systematic review is to maximise both the sensitivity and the specificity of the search results, that is, to retrieve all the relevant articles and exclude the irrelevant ones. In order to design the strategy, the question first needs to be translated into searchable terms using the PICO (Patient, Intervention, Comparison, Outcome) framework. The UCAN care pathways greatly facilitate this step, as all plausible interventions that can be compared are clearly shown for each patient group at every stage in the disease process. Also, the format of the pathway allows the reviewer to see where the search question is focussed in the context of the entire pathway and to see which interventions and stages of the disease are being excluded from the search. This is particularly useful when refining and reviewing the results of the search.\nTherefore, as the UCAN care pathways succinctly represent the required detailed and complex information, the reviewer is readily able to understand the specific patient groups and interventions being reviewed and focus the search accordingly. Hence, the pathways are a key element in the design and execution of searches to inform systematic reviews for specific stages of the pathway. The same process can be employed in the definition of focused clinical questions by clinical practice guideline panels.\nThe aim of the search strategy for a systematic review is to maximise both the sensitivity and the specificity of the search results, that is, to retrieve all the relevant articles and exclude the irrelevant ones. In order to design the strategy, the question first needs to be translated into searchable terms using the PICO (Patient, Intervention, Comparison, Outcome) framework. The UCAN care pathways greatly facilitate this step, as all plausible interventions that can be compared are clearly shown for each patient group at every stage in the disease process. Also, the format of the pathway allows the reviewer to see where the search question is focussed in the context of the entire pathway and to see which interventions and stages of the disease are being excluded from the search. This is particularly useful when refining and reviewing the results of the search.\nTherefore, as the UCAN care pathways succinctly represent the required detailed and complex information, the reviewer is readily able to understand the specific patient groups and interventions being reviewed and focus the search accordingly. Hence, the pathways are a key element in the design and execution of searches to inform systematic reviews for specific stages of the pathway. The same process can be employed in the definition of focused clinical questions by clinical practice guideline panels.\n[SUBTITLE] Prioritisation (review scope, guideline scope and development process) [SUBSECTION] Systematic reviews can be used to provide more reliable and precise estimates of relative effectiveness of competing treatments and highlight areas where further primary research is needed. Resources are however always limited, and therefore, the questions requiring performance of a systematic review need to be prioritised. There is also a linked need to prioritise topics for clinical practice guidelines based on evidence from systematic reviews. Prioritisation should be a process that is systematic and transparent and that includes consultation with key experts [18].\nThe UCAN care pathways inform the prioritisation process by providing a comprehensive framework for discussion that is applicable to multiple settings both nationally and internationally. They also inform the prioritisation process by highlighting where in the process of care the uncertainty is and the magnitude of the problem in terms of the number of people affected and its consequences. As such, they provide a concise summary of the various treatment options and inform the scope of systematic reviews and clinical practice guidelines. The UCAN care pathway model also ensures that the systematic review activity is firmly embedded in the context of the overall management of patients.\nSystematic reviews can be used to provide more reliable and precise estimates of relative effectiveness of competing treatments and highlight areas where further primary research is needed. Resources are however always limited, and therefore, the questions requiring performance of a systematic review need to be prioritised. There is also a linked need to prioritise topics for clinical practice guidelines based on evidence from systematic reviews. Prioritisation should be a process that is systematic and transparent and that includes consultation with key experts [18].\nThe UCAN care pathways inform the prioritisation process by providing a comprehensive framework for discussion that is applicable to multiple settings both nationally and internationally. They also inform the prioritisation process by highlighting where in the process of care the uncertainty is and the magnitude of the problem in terms of the number of people affected and its consequences. As such, they provide a concise summary of the various treatment options and inform the scope of systematic reviews and clinical practice guidelines. The UCAN care pathway model also ensures that the systematic review activity is firmly embedded in the context of the overall management of patients.\n[SUBTITLE] Economic evaluation and care pathways [SUBSECTION] Economic evaluation involves the comparative analysis of alternative interventions in terms of benefits such as improvement in health, the value of those improvements to individual patients and costs [19]. An economic evaluation has to be based upon a care pathway, as an understanding is needed of what the sequence of events is from initiation of the treatment under study. The UCAN care pathways can be used to identify events that will influence a patient’s well-being and events that use or save resources. Once these events have been established, they can then be measured and valued [20].\nEstimates of benefits and costs can be based on data from randomised controlled trials or from a mathematical model. The models can be constructed using decision analytic methods to provide a mathematical representation of a UCAN care pathway. A model is a further level of evidence synthesis as it may incorporate the results from a number of systematic reviews.\nIn addition to information on cost-effectiveness, an economic model can show who stands to benefit most from use of a particular healthcare intervention and who is most likely to bear the cost. These analyses can inform judgements about equity of provision. Economic models can also be used to highlight areas for further research either by showing precisely what evidence is missing or more formally by quantifying the value of further research in terms of balancing the cost of gaining the new evidence against the subsequent benefits that are likely to accrue by calculation of value of information [21].\nEconomic evaluation involves the comparative analysis of alternative interventions in terms of benefits such as improvement in health, the value of those improvements to individual patients and costs [19]. An economic evaluation has to be based upon a care pathway, as an understanding is needed of what the sequence of events is from initiation of the treatment under study. The UCAN care pathways can be used to identify events that will influence a patient’s well-being and events that use or save resources. Once these events have been established, they can then be measured and valued [20].\nEstimates of benefits and costs can be based on data from randomised controlled trials or from a mathematical model. The models can be constructed using decision analytic methods to provide a mathematical representation of a UCAN care pathway. A model is a further level of evidence synthesis as it may incorporate the results from a number of systematic reviews.\nIn addition to information on cost-effectiveness, an economic model can show who stands to benefit most from use of a particular healthcare intervention and who is most likely to bear the cost. These analyses can inform judgements about equity of provision. Economic models can also be used to highlight areas for further research either by showing precisely what evidence is missing or more formally by quantifying the value of further research in terms of balancing the cost of gaining the new evidence against the subsequent benefits that are likely to accrue by calculation of value of information [21].", "Our methodology gives a clear opportunity to gain clear consensus regarding definitions of treatment options. This precision in terminology is vital to ensure that the conduct and reporting of research, the systematic review process and the clinical practice guidelines development process are relevant to all.", "The aim of the search strategy for a systematic review is to maximise both the sensitivity and the specificity of the search results, that is, to retrieve all the relevant articles and exclude the irrelevant ones. In order to design the strategy, the question first needs to be translated into searchable terms using the PICO (Patient, Intervention, Comparison, Outcome) framework. The UCAN care pathways greatly facilitate this step, as all plausible interventions that can be compared are clearly shown for each patient group at every stage in the disease process. Also, the format of the pathway allows the reviewer to see where the search question is focussed in the context of the entire pathway and to see which interventions and stages of the disease are being excluded from the search. This is particularly useful when refining and reviewing the results of the search.\nTherefore, as the UCAN care pathways succinctly represent the required detailed and complex information, the reviewer is readily able to understand the specific patient groups and interventions being reviewed and focus the search accordingly. Hence, the pathways are a key element in the design and execution of searches to inform systematic reviews for specific stages of the pathway. The same process can be employed in the definition of focused clinical questions by clinical practice guideline panels.", "Systematic reviews can be used to provide more reliable and precise estimates of relative effectiveness of competing treatments and highlight areas where further primary research is needed. Resources are however always limited, and therefore, the questions requiring performance of a systematic review need to be prioritised. There is also a linked need to prioritise topics for clinical practice guidelines based on evidence from systematic reviews. Prioritisation should be a process that is systematic and transparent and that includes consultation with key experts [18].\nThe UCAN care pathways inform the prioritisation process by providing a comprehensive framework for discussion that is applicable to multiple settings both nationally and internationally. They also inform the prioritisation process by highlighting where in the process of care the uncertainty is and the magnitude of the problem in terms of the number of people affected and its consequences. As such, they provide a concise summary of the various treatment options and inform the scope of systematic reviews and clinical practice guidelines. The UCAN care pathway model also ensures that the systematic review activity is firmly embedded in the context of the overall management of patients.", "Economic evaluation involves the comparative analysis of alternative interventions in terms of benefits such as improvement in health, the value of those improvements to individual patients and costs [19]. An economic evaluation has to be based upon a care pathway, as an understanding is needed of what the sequence of events is from initiation of the treatment under study. The UCAN care pathways can be used to identify events that will influence a patient’s well-being and events that use or save resources. Once these events have been established, they can then be measured and valued [20].\nEstimates of benefits and costs can be based on data from randomised controlled trials or from a mathematical model. The models can be constructed using decision analytic methods to provide a mathematical representation of a UCAN care pathway. A model is a further level of evidence synthesis as it may incorporate the results from a number of systematic reviews.\nIn addition to information on cost-effectiveness, an economic model can show who stands to benefit most from use of a particular healthcare intervention and who is most likely to bear the cost. These analyses can inform judgements about equity of provision. Economic models can also be used to highlight areas for further research either by showing precisely what evidence is missing or more formally by quantifying the value of further research in terms of balancing the cost of gaining the new evidence against the subsequent benefits that are likely to accrue by calculation of value of information [21].", "We believe that the UCAN care pathways provide a crucial framework for improving urological cancer care through evidence synthesis, research prioritisation, stakeholder involvement and international collaboration. The process of developing these pathways is an important vehicle for influencing the discipline of urology by facilitating engagement with the principles of evidence-based medicine through international collaboration. They also hold the promise of advancing the goal of standardising terminology within urology and improving communication between healthcare professionals, researchers, patients, policy makers and funders, in different geographical locations. Finally, they inform the development and conduct of systematic reviews, the development of clinical practice guidelines and economic evaluation of interventions within urology." ]

[ "introduction", null, null, null, null, null, null, null, null, null, null, null, null, "conclusion" ]

[ "Urological cancer", "Care pathways", "Systematic review", "Clinical practice guidelines" ]

[ "Bone Density", "Bone Remodeling", "Bone and Bones", "Europe", "Evidence-Based Medicine", "Food Industry", "Functional Food", "Guidelines as Topic", "Humans", "Legislation, Food", "Nutritional Physiological Phenomena", "Research Design" ]

[ "Pre-clinical models", "Acceptable health claims in human bone health", "Design of clinical studies" ]

[ "A variety of invasive and non-invasive techniques can be used to provide relevant endpoints [4, 7], including bioavailability studies, microarray or PCR analysis of modulated genes, histomorphometry, culture of bone forming or bone resorbing cells ex vivo, exposure to primary cell cultures to plasma harvested from treated animals, the chemistry and biochemistry of bone tissue, the assessment of biochemical indices of skeletal turnover in blood and urine, metabolic balance of calcium combined with radioactive calcium kinetics, radiogrammetry of bone radiographs, neutron activation for whole body calcium, dual x-ray absorptiometry (DXA), and the assessment of bone strength [8]. The latter endpoint is considered to be the most relevant in the field of bone health claims. Bone strength reflects both bone density and bone quality. Bone quality depends on bone architecture, mineralization, turnover, and accumulation of microdamage. Therefore, the assessment of bone health would benefit from the measurement of bone strength in vivo. No validated non-invasive tools capable of measuring bone strength in vivo are available to date. However, biomechanical tests of resistance to fracture provide an objective measure of overall bone strength. The three main types of biomechanical tests for bone strength are bending, torsional, and compression tests [9]. It is considered that both long bones and vertebral bodies should be tested since they may be differentially affected by food products.\nBesides the assessment of a direct effect of the food product on bone strength, two other aims of animal data could be to better understand the mechanism of action of the food product or to validate surrogate variables used in human animal data to see if these variables reflect bone strength.\nKey criteria of suitable/acceptable animal studies are:\n➢to deliver the food product in the manner in which it will be delivered in a human setting;➢to utilize a site of delivery and/or assessment site that is as closely matched as possible to the settings in which it will be used;➢to utilize an animal that provides a metabolic background and physiological responsiveness comparable to humans;➢to utilize a formulation of active agent that has the same composition, release, retention, and degradation properties as the formulation that will be used in humans.\n\n➢to deliver the food product in the manner in which it will be delivered in a human setting;\n➢to utilize a site of delivery and/or assessment site that is as closely matched as possible to the settings in which it will be used;\n➢to utilize an animal that provides a metabolic background and physiological responsiveness comparable to humans;\n➢to utilize a formulation of active agent that has the same composition, release, retention, and degradation properties as the formulation that will be used in humans.", "The GREES panel considers that six different health claims could be accepted for an effect of food products on bone health. However, as already used by the European Food and Safety Authority, different wording to reflect the level of evidence of the effect could be used depending on the effect that is (always), may (demonstrated only under certain circumstances) or might be (logically expected benefit from physiology but yet not demonstrated) beneficial for bone health.\nImprovement of calcium bioavailabilityCalcium bioavailability may be defined as the proportion of calcium in foods which is absorbed and utilized for normal metabolic functions. In addition to the amount of calcium in the diet, the fractional absorption of dietary calcium in food and its retention in the body are also a factor that determines the availability of calcium for bone development and maintenance of bone health [10, 11]. Many methods can be used to assess bioavailability (i.e., classical and isotopic balances, urinary excretion, isotope labeling in the urine, plasma, and bones) [12]. The group considers that an increase in bioavailability is not beneficial if not accompanied by calcium retention in the body. A food product with an effect on calcium bioavailability with or without calcium retention data, unless associated with appropriate animal studies would not fulfill a claim related to article 14. However, food products that show an effect on bioavailability and calcium retention could have an article 13 claim: “X increases calcium absorption” or “X increases calcium bioavailability”.Maintenance of bone metabolism (through an effect on osteoclast regulatory proteins)The transition of osteoclast precursors to mature osteoclasts that are capable of resorbing bone is tightly regulated by osteoclast regulatory proteins that either affect the differentiation and proliferation of osteoclast precursors into mature osteoclasts or are involved in the coupling between osteoblasts and osteoclasts [13]. Markers of osteoclastogenesis include receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, whereas markers of osteoclast number include tartrate-resistant acid phosphatase and cathepsin K. These proteins contribute to bone metabolism but are not yet strongly associated with elements of bone strength. A food product with an effect on osteoclast regulatory proteins, unless supported by animal studies (see next BMD and BTM sections) would not fulfill a claim related to article 14. The product, however, might have the label under the article 13: “X contributes to the maintenance of bone metabolism”.Maintenance or changes in bone turnover markerA determinant of bone strength that is not assessed by bone mineral density (BMD) is the rate of bone remodeling. Depending upon their origin, bone turnover markers (BTMs) are classified as indices of bone resorption or formation [14–17]. The rate of bone resorption and formation can be estimated by assays that measure the serum concentration or urinary excretion of different target molecules specific to these cellular processes. GREES panel recommends the inclusion of reference markers of bone formation (serum procollagen type I N propeptide, s-PINP) and resorption (serum C-terminal cross-linking telopeptide of type I collagen, s-CTX) in keeping with the recommendations of the International Osteoporosis Foundation [18]. A food product with a positive BTM balance might have the claim: “X maintains normal bone remodeling that could contribute to the normal structure and function of bones” or “X increases markers of bone formation that could contribute to the normal structure and function of bones” or “X decreases markers of bone resorption that could contribute to the normal structure and function of bones”.As in the case of BMD, BTMs are only indicators of fracture risk, but the change in BTM induced by a product is not necessarily associated with a change in fracture risk or bone strength. In this regard, animal models are useful to assess if changes in BTMs due to the intake of the food product are associated with an increase in bone strength. A food product with an effect on BTMs together with animal studies that showed improved bone strength or a relationship between changes in BTMs induced by the food product and bone strength could have the claim: “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that is associated with bone strength” or “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that increases bone strength” or “X increases bone strength”.Maintenance or improvement in bone structureThe key role of bone microarchitecture in bone health was suggested by the classic definition of osteoporosis adopted in 1993 [19]. Methods for investigating 3-D bone microarchitecture and bone strength include in vitro μCT, in vitro μMRI, in vivo pQCT, and in vivo high-resolution MRI [20]. They aim to quantify various determinants of bone strength such as bone geometry, volumetric bone density, microarchitecture, and properties of the bone matrix [21]. A food product with an effect on bone microarchitecture could have the claim: “X improves (or maintains) bone microarchitecture that could contribute to the normal structure and function of bones”.It is considered that the assessment of bone structure with the tools currently available in man is not sufficiently validated to be a reliable surrogate of bone strength. For this reason, animal models are needed to assess the relationship between changes in bone microarchitecture induced by the food product and any increase in bone strength. A food product with an effect on microarchitecture of the human bone and animals studies that show improvement in bone strength or show the relationship between change in bone structure induced by the food product and bone strength could have the claim: “X improves bone microarchitecture that increases bone strength” or “X increases bone strength”Maintenance or increase in bone mineral densityBone strength is determined by many factors, including bone mass. Bone mass is estimated in clinical practice by the measurement of BMD. BMD, as measured by DXA, represents an estimate of the quantity of mineral (grams of calcium) divided by the two-dimensional area of the bone [22]. There is a strong relationship between the risk of fracture and BMD but there is a wide overlap in the bone densities of patients who develop a fracture and those who do not. Since BMD is only a surrogate marker for bone strength or fracture risk, and since product-induced changes in BMD are not clearly associated with changes in bone strength or fracture risk, an increase in BMD may not be associated with an increased bone strength or decreased fracture risk [23]. A food product with a positive effect on BMD could have the claim: “X increases BMD. A low BMD is associated with an increased risk of fracture” or “X maintains BMD. A low BMD is associated with an increased risk of fracture”.Animal models are appropriate to determine whether an increase in BMD associated with a food product is accompanied by an increase in bone strength. A food product with a positive effect on BMD, together with animal studies showing an improvement in bone strength or showing a relationship between BMD changes induced by the food product and bone strength, could have the claim: “X increases (or maintains) BMD that could reduce the risk of fracture” or “X increases (or maintains) BMD that increases bone strength” or “X increases bone strength”.Reduction of the risk of fractureA reduction of the incidence of fracture is a major aim of food products beneficial to skeletal health, but according to the regulation cannot be claimed as such without mentioning the effect on a risk factor. However, a reduction in the fracture risk is obviously supportive for a claim on the reduction of an identified risk factor. For non-spinal fractures, either femoral (hip) or major non-vertebral (pelvis, distal femur, proximal tibia, ribs, proximal humerus forearm, and hip) fractures should be assessed.\n\nImprovement of calcium bioavailability\nCalcium bioavailability may be defined as the proportion of calcium in foods which is absorbed and utilized for normal metabolic functions. In addition to the amount of calcium in the diet, the fractional absorption of dietary calcium in food and its retention in the body are also a factor that determines the availability of calcium for bone development and maintenance of bone health [10, 11]. Many methods can be used to assess bioavailability (i.e., classical and isotopic balances, urinary excretion, isotope labeling in the urine, plasma, and bones) [12]. The group considers that an increase in bioavailability is not beneficial if not accompanied by calcium retention in the body. A food product with an effect on calcium bioavailability with or without calcium retention data, unless associated with appropriate animal studies would not fulfill a claim related to article 14. However, food products that show an effect on bioavailability and calcium retention could have an article 13 claim: “X increases calcium absorption” or “X increases calcium bioavailability”.\nMaintenance of bone metabolism (through an effect on osteoclast regulatory proteins)\nThe transition of osteoclast precursors to mature osteoclasts that are capable of resorbing bone is tightly regulated by osteoclast regulatory proteins that either affect the differentiation and proliferation of osteoclast precursors into mature osteoclasts or are involved in the coupling between osteoblasts and osteoclasts [13]. Markers of osteoclastogenesis include receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, whereas markers of osteoclast number include tartrate-resistant acid phosphatase and cathepsin K. These proteins contribute to bone metabolism but are not yet strongly associated with elements of bone strength. A food product with an effect on osteoclast regulatory proteins, unless supported by animal studies (see next BMD and BTM sections) would not fulfill a claim related to article 14. The product, however, might have the label under the article 13: “X contributes to the maintenance of bone metabolism”.\nMaintenance or changes in bone turnover marker\nA determinant of bone strength that is not assessed by bone mineral density (BMD) is the rate of bone remodeling. Depending upon their origin, bone turnover markers (BTMs) are classified as indices of bone resorption or formation [14–17]. The rate of bone resorption and formation can be estimated by assays that measure the serum concentration or urinary excretion of different target molecules specific to these cellular processes. GREES panel recommends the inclusion of reference markers of bone formation (serum procollagen type I N propeptide, s-PINP) and resorption (serum C-terminal cross-linking telopeptide of type I collagen, s-CTX) in keeping with the recommendations of the International Osteoporosis Foundation [18]. A food product with a positive BTM balance might have the claim: “X maintains normal bone remodeling that could contribute to the normal structure and function of bones” or “X increases markers of bone formation that could contribute to the normal structure and function of bones” or “X decreases markers of bone resorption that could contribute to the normal structure and function of bones”.\nAs in the case of BMD, BTMs are only indicators of fracture risk, but the change in BTM induced by a product is not necessarily associated with a change in fracture risk or bone strength. In this regard, animal models are useful to assess if changes in BTMs due to the intake of the food product are associated with an increase in bone strength. A food product with an effect on BTMs together with animal studies that showed improved bone strength or a relationship between changes in BTMs induced by the food product and bone strength could have the claim: “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that is associated with bone strength” or “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that increases bone strength” or “X increases bone strength”.\nMaintenance or improvement in bone structure\nThe key role of bone microarchitecture in bone health was suggested by the classic definition of osteoporosis adopted in 1993 [19]. Methods for investigating 3-D bone microarchitecture and bone strength include in vitro μCT, in vitro μMRI, in vivo pQCT, and in vivo high-resolution MRI [20]. They aim to quantify various determinants of bone strength such as bone geometry, volumetric bone density, microarchitecture, and properties of the bone matrix [21]. A food product with an effect on bone microarchitecture could have the claim: “X improves (or maintains) bone microarchitecture that could contribute to the normal structure and function of bones”.\nIt is considered that the assessment of bone structure with the tools currently available in man is not sufficiently validated to be a reliable surrogate of bone strength. For this reason, animal models are needed to assess the relationship between changes in bone microarchitecture induced by the food product and any increase in bone strength. A food product with an effect on microarchitecture of the human bone and animals studies that show improvement in bone strength or show the relationship between change in bone structure induced by the food product and bone strength could have the claim: “X improves bone microarchitecture that increases bone strength” or “X increases bone strength”\nMaintenance or increase in bone mineral density\nBone strength is determined by many factors, including bone mass. Bone mass is estimated in clinical practice by the measurement of BMD. BMD, as measured by DXA, represents an estimate of the quantity of mineral (grams of calcium) divided by the two-dimensional area of the bone [22]. There is a strong relationship between the risk of fracture and BMD but there is a wide overlap in the bone densities of patients who develop a fracture and those who do not. Since BMD is only a surrogate marker for bone strength or fracture risk, and since product-induced changes in BMD are not clearly associated with changes in bone strength or fracture risk, an increase in BMD may not be associated with an increased bone strength or decreased fracture risk [23]. A food product with a positive effect on BMD could have the claim: “X increases BMD. A low BMD is associated with an increased risk of fracture” or “X maintains BMD. A low BMD is associated with an increased risk of fracture”.\nAnimal models are appropriate to determine whether an increase in BMD associated with a food product is accompanied by an increase in bone strength. A food product with a positive effect on BMD, together with animal studies showing an improvement in bone strength or showing a relationship between BMD changes induced by the food product and bone strength, could have the claim: “X increases (or maintains) BMD that could reduce the risk of fracture” or “X increases (or maintains) BMD that increases bone strength” or “X increases bone strength”.\nReduction of the risk of fracture\nA reduction of the incidence of fracture is a major aim of food products beneficial to skeletal health, but according to the regulation cannot be claimed as such without mentioning the effect on a risk factor. However, a reduction in the fracture risk is obviously supportive for a claim on the reduction of an identified risk factor. For non-spinal fractures, either femoral (hip) or major non-vertebral (pelvis, distal femur, proximal tibia, ribs, proximal humerus forearm, and hip) fractures should be assessed.", "\nPopulationThe subjects studied should be representative of the population targeted for the food product. The applicant should take all necessary precautions to make sure that the tested population is equivalent to the user population with respect of ethnicity, age, physiological status (such as menopause for example), life habits (such as exercise) and diet. No densitometric criteria are required for inclusion. However, the experimental and the control group must show no significant differences in term of baseline BMD.DesignThe ideal design would be a multicentre randomized controlled study (RCT). The control could be a placebo, another active product or nothing, depending on the tested food. When possible, subjects and/or investigators should be blinded of the intervention. Treatment and control groups should be balanced with respect to gender, age, menopausal status, dietary habits, or underlying diseases. The GREES panel recognizes that a RCT is not always possible in practice or from an ethical point of view. Since the totality of the evidence should be weighed for the substantiation of a claim, well-designed prospective cohort studies, case–control studies and/or observational studies of high quality could be acceptable if accompanied by other data (e.g., animal data, effect on multiple surrogate endpoints). Cross-over studies design can also be considered. All the efforts should be made to eliminate potential confounders.Duration of studyThe duration of the trial should be predetermined and should depend on the outcome. For BMD, duration of at least 1 year seems necessary. For BTMs, a 3-month study is the minimum. The primary efficacy endpoint should be assessed at the end of the predetermined treatment period in comparison with the measurement at baseline. Intermediate measurements are also recommended.Statistical analysisIntention-to-treat analysis should be the primary method of evaluation. Statistical significance will be inferred if a P value is equal to or less than 0.05. The beta risk will be equal to or less than 20%. The sample size of the study must be calculated prior to the start of the study. Possible confounding variables should be managed using appropriate statistical analysis. Within group (end vs. baseline) and between groups comparisons should be made.Diet habit and lifestyleThe control of critical effect modifiers such as physical activity, synergies with a multitude of other nutrients and the influence of nutrigenomic relationships must be taken into account.Intakes of other nutrients or foods, on which the tested nutrient is dependent, must be optimized. Any supplementation with other food products known to have an effect on bone (e.g., calcium and/or vitamin D) should be consistent within all patient groups.ObservanceObservance of food product intake should be monitored during the study to be able to perform pre-planned analyses on individuals with high and poor compliance rates or analyses of dose–response.SafetyAll adverse experiences occurring during the course of clinical trials should be fully documented with separate analysis of adverse events, dropouts, and patients who died while being on the study.\n\nPopulation\nThe subjects studied should be representative of the population targeted for the food product. The applicant should take all necessary precautions to make sure that the tested population is equivalent to the user population with respect of ethnicity, age, physiological status (such as menopause for example), life habits (such as exercise) and diet. No densitometric criteria are required for inclusion. However, the experimental and the control group must show no significant differences in term of baseline BMD.\nDesign\nThe ideal design would be a multicentre randomized controlled study (RCT). The control could be a placebo, another active product or nothing, depending on the tested food. When possible, subjects and/or investigators should be blinded of the intervention. Treatment and control groups should be balanced with respect to gender, age, menopausal status, dietary habits, or underlying diseases. The GREES panel recognizes that a RCT is not always possible in practice or from an ethical point of view. Since the totality of the evidence should be weighed for the substantiation of a claim, well-designed prospective cohort studies, case–control studies and/or observational studies of high quality could be acceptable if accompanied by other data (e.g., animal data, effect on multiple surrogate endpoints). Cross-over studies design can also be considered. All the efforts should be made to eliminate potential confounders.\nDuration of study\nThe duration of the trial should be predetermined and should depend on the outcome. For BMD, duration of at least 1 year seems necessary. For BTMs, a 3-month study is the minimum. The primary efficacy endpoint should be assessed at the end of the predetermined treatment period in comparison with the measurement at baseline. Intermediate measurements are also recommended.\nStatistical analysis\nIntention-to-treat analysis should be the primary method of evaluation. Statistical significance will be inferred if a P value is equal to or less than 0.05. The beta risk will be equal to or less than 20%. The sample size of the study must be calculated prior to the start of the study. Possible confounding variables should be managed using appropriate statistical analysis. Within group (end vs. baseline) and between groups comparisons should be made.\nDiet habit and lifestyle\nThe control of critical effect modifiers such as physical activity, synergies with a multitude of other nutrients and the influence of nutrigenomic relationships must be taken into account.\nIntakes of other nutrients or foods, on which the tested nutrient is dependent, must be optimized. Any supplementation with other food products known to have an effect on bone (e.g., calcium and/or vitamin D) should be consistent within all patient groups.\nObservance\nObservance of food product intake should be monitored during the study to be able to perform pre-planned analyses on individuals with high and poor compliance rates or analyses of dose–response.\nSafety\nAll adverse experiences occurring during the course of clinical trials should be fully documented with separate analysis of adverse events, dropouts, and patients who died while being on the study." ]

[ null, null, null ]

[ "Introduction", "Methods", "Results", "Pre-clinical models", "Acceptable health claims in human bone health", "Design of clinical studies", "Conclusion" ]

[ "More and more food products bear health claims. The skepticism of consumers regarding functional foods is mainly due to doubts over the veracity of health claims and in the poor and often inadequate control of their claimed properties. It is important that health claims should provide genuine information to help consumers choose healthy diets. Consequently, claims should be supported by a sound and sufficient body of scientific evidence to substantiate them and be reinforced by specific consumer education.\nSince health claims on food products are increasingly recognized to be important, they are being legally regulated in more and more countries around the world [1]. Although there is a general scientific consensus on how to substantiate health claims on food [2], there is no agreement on the specific approaches and indicators that can be used in different fields. Various jurisdictions have developed systematic approaches for reviewing scientific data linking food products and health with the objective of identifying the threshold of scientific evidence needed to substantiate an authoritative statement to the general public in the form of a label claim for a given marketed food product. In Europe, a regulation on nutrition and health claims made on foods was introduced in 2007. This regulation provides opportunities for the use of health claims on foods in Europe, including reduction of disease risk [3]. According to Regulation EC 1924/2006, the use of nutrition and health claims shall only be permitted if the substance in respect of which the claim is made has been shown to have a beneficial nutritional or physiological effect. A community list of permitted and rejected claims has been established and made available to the public (http://ec.europa.eu/food/food/labellingnutrition/claims/community_register/health_claims_en.htm).\nThe regulation defines a health claim in general as “any claim that states, suggests or implies that a relationship exists between a food category, a food or one of its constituents and health.” All claims are addressed in Articles 13 and 14 of the Regulation EC 1924/2006 (Table 1).\nTable 1Claims addressed in articles 13 and 14 of the Regulation EC 1924/2006Article 13Article 14Article 13.1Article 13.5Referring tothe role of a nutrient or other substance in growth, development and the functions of the bodythe role of a nutrient or other substance in growth, development and the functions of the body based on newly developed scientific evidence and/or which include a request for the protection of proprietary data.the reduction of disease risk and claims relating to children's development and healthApplication based ongenerally accepted scientific evidencesubmission of an extensive scientific dossiersubmission of an extensive scientific dossier\n\nClaims addressed in articles 13 and 14 of the Regulation EC 1924/2006\nIn the context of health claims in foods, bone health is of potential interest as it is a major public health problem, at least in Western countries [4]. Up to 60% of the variance in bone mass is determined by genetic factors. Environmental factors account for the remainder, including nutritional intake and lifestyle habits throughout life [5, 6].\nIn the field of bone health, there are no scientifically based definitions of health claims and no uniform recommendations of the preferred study and/or methodology, even though some preparatory work had been done before the introduction of the European regulations [4]. The objective of this paper was to define the relevant biomarker for bone health and to provide recommendations for the design and the methodology of clinical studies which need to be fulfilled to assert claims related to bone health. The intent was to aid regulatory authorities in defining claims and assessing scientific evidence used to support those that relate to bone health. By establishing common criteria for these assessments, it is hoped that these recommendations will lead to harmonization of the requirements for scientific substantiation of claims worldwide.", "Two 1-day meetings were organized by the Group for the Respect of Ethics and Excellence in Science (GREES). This non-for-profit organization has expertise in literature research and consensus meetings. The meetings were attended by academic scientists with expertise in the field of bone health or nutrition, members of regulatory authorities as well as industrialists with interests in health claims relating to bone. The objective of the first day of the meeting was to critically review the current literature in the field of health claims related to bone and to discuss the needs and problems to assert such claims. The objective of the second day was to reach consensus on scientifically acceptable health claims related to bone and to provide guidelines for the design and the methodology of clinical studies which need to be adopted to assert such health claims.\nA literature search, using Medline database up to August 2010, was performed using keywords including health claims, nutrition, bone, osteoporosis, clinical study methodology, surrogate endpoint. A selection of relevant papers was made by OB, RR, and JYR.", "The GREES panel considers that clinical data in humans are indispensable, and that health claims cannot be accepted solely on the basis of animal data. However, as discussed below, animal studies can give important information not available in humans and can provide data for the generalization of results obtained in a specific tested population to a larger group. Thus, different levels of heath claims should be considered based both on the endpoint used and on the information provided by animal studies.\n[SUBTITLE] Pre-clinical models [SUBSECTION] A variety of invasive and non-invasive techniques can be used to provide relevant endpoints [4, 7], including bioavailability studies, microarray or PCR analysis of modulated genes, histomorphometry, culture of bone forming or bone resorbing cells ex vivo, exposure to primary cell cultures to plasma harvested from treated animals, the chemistry and biochemistry of bone tissue, the assessment of biochemical indices of skeletal turnover in blood and urine, metabolic balance of calcium combined with radioactive calcium kinetics, radiogrammetry of bone radiographs, neutron activation for whole body calcium, dual x-ray absorptiometry (DXA), and the assessment of bone strength [8]. The latter endpoint is considered to be the most relevant in the field of bone health claims. Bone strength reflects both bone density and bone quality. Bone quality depends on bone architecture, mineralization, turnover, and accumulation of microdamage. Therefore, the assessment of bone health would benefit from the measurement of bone strength in vivo. No validated non-invasive tools capable of measuring bone strength in vivo are available to date. However, biomechanical tests of resistance to fracture provide an objective measure of overall bone strength. The three main types of biomechanical tests for bone strength are bending, torsional, and compression tests [9]. It is considered that both long bones and vertebral bodies should be tested since they may be differentially affected by food products.\nBesides the assessment of a direct effect of the food product on bone strength, two other aims of animal data could be to better understand the mechanism of action of the food product or to validate surrogate variables used in human animal data to see if these variables reflect bone strength.\nKey criteria of suitable/acceptable animal studies are:\n➢to deliver the food product in the manner in which it will be delivered in a human setting;➢to utilize a site of delivery and/or assessment site that is as closely matched as possible to the settings in which it will be used;➢to utilize an animal that provides a metabolic background and physiological responsiveness comparable to humans;➢to utilize a formulation of active agent that has the same composition, release, retention, and degradation properties as the formulation that will be used in humans.\n\n➢to deliver the food product in the manner in which it will be delivered in a human setting;\n➢to utilize a site of delivery and/or assessment site that is as closely matched as possible to the settings in which it will be used;\n➢to utilize an animal that provides a metabolic background and physiological responsiveness comparable to humans;\n➢to utilize a formulation of active agent that has the same composition, release, retention, and degradation properties as the formulation that will be used in humans.\nA variety of invasive and non-invasive techniques can be used to provide relevant endpoints [4, 7], including bioavailability studies, microarray or PCR analysis of modulated genes, histomorphometry, culture of bone forming or bone resorbing cells ex vivo, exposure to primary cell cultures to plasma harvested from treated animals, the chemistry and biochemistry of bone tissue, the assessment of biochemical indices of skeletal turnover in blood and urine, metabolic balance of calcium combined with radioactive calcium kinetics, radiogrammetry of bone radiographs, neutron activation for whole body calcium, dual x-ray absorptiometry (DXA), and the assessment of bone strength [8]. The latter endpoint is considered to be the most relevant in the field of bone health claims. Bone strength reflects both bone density and bone quality. Bone quality depends on bone architecture, mineralization, turnover, and accumulation of microdamage. Therefore, the assessment of bone health would benefit from the measurement of bone strength in vivo. No validated non-invasive tools capable of measuring bone strength in vivo are available to date. However, biomechanical tests of resistance to fracture provide an objective measure of overall bone strength. The three main types of biomechanical tests for bone strength are bending, torsional, and compression tests [9]. It is considered that both long bones and vertebral bodies should be tested since they may be differentially affected by food products.\nBesides the assessment of a direct effect of the food product on bone strength, two other aims of animal data could be to better understand the mechanism of action of the food product or to validate surrogate variables used in human animal data to see if these variables reflect bone strength.\nKey criteria of suitable/acceptable animal studies are:\n➢to deliver the food product in the manner in which it will be delivered in a human setting;➢to utilize a site of delivery and/or assessment site that is as closely matched as possible to the settings in which it will be used;➢to utilize an animal that provides a metabolic background and physiological responsiveness comparable to humans;➢to utilize a formulation of active agent that has the same composition, release, retention, and degradation properties as the formulation that will be used in humans.\n\n➢to deliver the food product in the manner in which it will be delivered in a human setting;\n➢to utilize a site of delivery and/or assessment site that is as closely matched as possible to the settings in which it will be used;\n➢to utilize an animal that provides a metabolic background and physiological responsiveness comparable to humans;\n➢to utilize a formulation of active agent that has the same composition, release, retention, and degradation properties as the formulation that will be used in humans.\n[SUBTITLE] Acceptable health claims in human bone health [SUBSECTION] The GREES panel considers that six different health claims could be accepted for an effect of food products on bone health. However, as already used by the European Food and Safety Authority, different wording to reflect the level of evidence of the effect could be used depending on the effect that is (always), may (demonstrated only under certain circumstances) or might be (logically expected benefit from physiology but yet not demonstrated) beneficial for bone health.\nImprovement of calcium bioavailabilityCalcium bioavailability may be defined as the proportion of calcium in foods which is absorbed and utilized for normal metabolic functions. In addition to the amount of calcium in the diet, the fractional absorption of dietary calcium in food and its retention in the body are also a factor that determines the availability of calcium for bone development and maintenance of bone health [10, 11]. Many methods can be used to assess bioavailability (i.e., classical and isotopic balances, urinary excretion, isotope labeling in the urine, plasma, and bones) [12]. The group considers that an increase in bioavailability is not beneficial if not accompanied by calcium retention in the body. A food product with an effect on calcium bioavailability with or without calcium retention data, unless associated with appropriate animal studies would not fulfill a claim related to article 14. However, food products that show an effect on bioavailability and calcium retention could have an article 13 claim: “X increases calcium absorption” or “X increases calcium bioavailability”.Maintenance of bone metabolism (through an effect on osteoclast regulatory proteins)The transition of osteoclast precursors to mature osteoclasts that are capable of resorbing bone is tightly regulated by osteoclast regulatory proteins that either affect the differentiation and proliferation of osteoclast precursors into mature osteoclasts or are involved in the coupling between osteoblasts and osteoclasts [13]. Markers of osteoclastogenesis include receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, whereas markers of osteoclast number include tartrate-resistant acid phosphatase and cathepsin K. These proteins contribute to bone metabolism but are not yet strongly associated with elements of bone strength. A food product with an effect on osteoclast regulatory proteins, unless supported by animal studies (see next BMD and BTM sections) would not fulfill a claim related to article 14. The product, however, might have the label under the article 13: “X contributes to the maintenance of bone metabolism”.Maintenance or changes in bone turnover markerA determinant of bone strength that is not assessed by bone mineral density (BMD) is the rate of bone remodeling. Depending upon their origin, bone turnover markers (BTMs) are classified as indices of bone resorption or formation [14–17]. The rate of bone resorption and formation can be estimated by assays that measure the serum concentration or urinary excretion of different target molecules specific to these cellular processes. GREES panel recommends the inclusion of reference markers of bone formation (serum procollagen type I N propeptide, s-PINP) and resorption (serum C-terminal cross-linking telopeptide of type I collagen, s-CTX) in keeping with the recommendations of the International Osteoporosis Foundation [18]. A food product with a positive BTM balance might have the claim: “X maintains normal bone remodeling that could contribute to the normal structure and function of bones” or “X increases markers of bone formation that could contribute to the normal structure and function of bones” or “X decreases markers of bone resorption that could contribute to the normal structure and function of bones”.As in the case of BMD, BTMs are only indicators of fracture risk, but the change in BTM induced by a product is not necessarily associated with a change in fracture risk or bone strength. In this regard, animal models are useful to assess if changes in BTMs due to the intake of the food product are associated with an increase in bone strength. A food product with an effect on BTMs together with animal studies that showed improved bone strength or a relationship between changes in BTMs induced by the food product and bone strength could have the claim: “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that is associated with bone strength” or “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that increases bone strength” or “X increases bone strength”.Maintenance or improvement in bone structureThe key role of bone microarchitecture in bone health was suggested by the classic definition of osteoporosis adopted in 1993 [19]. Methods for investigating 3-D bone microarchitecture and bone strength include in vitro μCT, in vitro μMRI, in vivo pQCT, and in vivo high-resolution MRI [20]. They aim to quantify various determinants of bone strength such as bone geometry, volumetric bone density, microarchitecture, and properties of the bone matrix [21]. A food product with an effect on bone microarchitecture could have the claim: “X improves (or maintains) bone microarchitecture that could contribute to the normal structure and function of bones”.It is considered that the assessment of bone structure with the tools currently available in man is not sufficiently validated to be a reliable surrogate of bone strength. For this reason, animal models are needed to assess the relationship between changes in bone microarchitecture induced by the food product and any increase in bone strength. A food product with an effect on microarchitecture of the human bone and animals studies that show improvement in bone strength or show the relationship between change in bone structure induced by the food product and bone strength could have the claim: “X improves bone microarchitecture that increases bone strength” or “X increases bone strength”Maintenance or increase in bone mineral densityBone strength is determined by many factors, including bone mass. Bone mass is estimated in clinical practice by the measurement of BMD. BMD, as measured by DXA, represents an estimate of the quantity of mineral (grams of calcium) divided by the two-dimensional area of the bone [22]. There is a strong relationship between the risk of fracture and BMD but there is a wide overlap in the bone densities of patients who develop a fracture and those who do not. Since BMD is only a surrogate marker for bone strength or fracture risk, and since product-induced changes in BMD are not clearly associated with changes in bone strength or fracture risk, an increase in BMD may not be associated with an increased bone strength or decreased fracture risk [23]. A food product with a positive effect on BMD could have the claim: “X increases BMD. A low BMD is associated with an increased risk of fracture” or “X maintains BMD. A low BMD is associated with an increased risk of fracture”.Animal models are appropriate to determine whether an increase in BMD associated with a food product is accompanied by an increase in bone strength. A food product with a positive effect on BMD, together with animal studies showing an improvement in bone strength or showing a relationship between BMD changes induced by the food product and bone strength, could have the claim: “X increases (or maintains) BMD that could reduce the risk of fracture” or “X increases (or maintains) BMD that increases bone strength” or “X increases bone strength”.Reduction of the risk of fractureA reduction of the incidence of fracture is a major aim of food products beneficial to skeletal health, but according to the regulation cannot be claimed as such without mentioning the effect on a risk factor. However, a reduction in the fracture risk is obviously supportive for a claim on the reduction of an identified risk factor. For non-spinal fractures, either femoral (hip) or major non-vertebral (pelvis, distal femur, proximal tibia, ribs, proximal humerus forearm, and hip) fractures should be assessed.\n\nImprovement of calcium bioavailability\nCalcium bioavailability may be defined as the proportion of calcium in foods which is absorbed and utilized for normal metabolic functions. In addition to the amount of calcium in the diet, the fractional absorption of dietary calcium in food and its retention in the body are also a factor that determines the availability of calcium for bone development and maintenance of bone health [10, 11]. Many methods can be used to assess bioavailability (i.e., classical and isotopic balances, urinary excretion, isotope labeling in the urine, plasma, and bones) [12]. The group considers that an increase in bioavailability is not beneficial if not accompanied by calcium retention in the body. A food product with an effect on calcium bioavailability with or without calcium retention data, unless associated with appropriate animal studies would not fulfill a claim related to article 14. However, food products that show an effect on bioavailability and calcium retention could have an article 13 claim: “X increases calcium absorption” or “X increases calcium bioavailability”.\nMaintenance of bone metabolism (through an effect on osteoclast regulatory proteins)\nThe transition of osteoclast precursors to mature osteoclasts that are capable of resorbing bone is tightly regulated by osteoclast regulatory proteins that either affect the differentiation and proliferation of osteoclast precursors into mature osteoclasts or are involved in the coupling between osteoblasts and osteoclasts [13]. Markers of osteoclastogenesis include receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, whereas markers of osteoclast number include tartrate-resistant acid phosphatase and cathepsin K. These proteins contribute to bone metabolism but are not yet strongly associated with elements of bone strength. A food product with an effect on osteoclast regulatory proteins, unless supported by animal studies (see next BMD and BTM sections) would not fulfill a claim related to article 14. The product, however, might have the label under the article 13: “X contributes to the maintenance of bone metabolism”.\nMaintenance or changes in bone turnover marker\nA determinant of bone strength that is not assessed by bone mineral density (BMD) is the rate of bone remodeling. Depending upon their origin, bone turnover markers (BTMs) are classified as indices of bone resorption or formation [14–17]. The rate of bone resorption and formation can be estimated by assays that measure the serum concentration or urinary excretion of different target molecules specific to these cellular processes. GREES panel recommends the inclusion of reference markers of bone formation (serum procollagen type I N propeptide, s-PINP) and resorption (serum C-terminal cross-linking telopeptide of type I collagen, s-CTX) in keeping with the recommendations of the International Osteoporosis Foundation [18]. A food product with a positive BTM balance might have the claim: “X maintains normal bone remodeling that could contribute to the normal structure and function of bones” or “X increases markers of bone formation that could contribute to the normal structure and function of bones” or “X decreases markers of bone resorption that could contribute to the normal structure and function of bones”.\nAs in the case of BMD, BTMs are only indicators of fracture risk, but the change in BTM induced by a product is not necessarily associated with a change in fracture risk or bone strength. In this regard, animal models are useful to assess if changes in BTMs due to the intake of the food product are associated with an increase in bone strength. A food product with an effect on BTMs together with animal studies that showed improved bone strength or a relationship between changes in BTMs induced by the food product and bone strength could have the claim: “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that is associated with bone strength” or “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that increases bone strength” or “X increases bone strength”.\nMaintenance or improvement in bone structure\nThe key role of bone microarchitecture in bone health was suggested by the classic definition of osteoporosis adopted in 1993 [19]. Methods for investigating 3-D bone microarchitecture and bone strength include in vitro μCT, in vitro μMRI, in vivo pQCT, and in vivo high-resolution MRI [20]. They aim to quantify various determinants of bone strength such as bone geometry, volumetric bone density, microarchitecture, and properties of the bone matrix [21]. A food product with an effect on bone microarchitecture could have the claim: “X improves (or maintains) bone microarchitecture that could contribute to the normal structure and function of bones”.\nIt is considered that the assessment of bone structure with the tools currently available in man is not sufficiently validated to be a reliable surrogate of bone strength. For this reason, animal models are needed to assess the relationship between changes in bone microarchitecture induced by the food product and any increase in bone strength. A food product with an effect on microarchitecture of the human bone and animals studies that show improvement in bone strength or show the relationship between change in bone structure induced by the food product and bone strength could have the claim: “X improves bone microarchitecture that increases bone strength” or “X increases bone strength”\nMaintenance or increase in bone mineral density\nBone strength is determined by many factors, including bone mass. Bone mass is estimated in clinical practice by the measurement of BMD. BMD, as measured by DXA, represents an estimate of the quantity of mineral (grams of calcium) divided by the two-dimensional area of the bone [22]. There is a strong relationship between the risk of fracture and BMD but there is a wide overlap in the bone densities of patients who develop a fracture and those who do not. Since BMD is only a surrogate marker for bone strength or fracture risk, and since product-induced changes in BMD are not clearly associated with changes in bone strength or fracture risk, an increase in BMD may not be associated with an increased bone strength or decreased fracture risk [23]. A food product with a positive effect on BMD could have the claim: “X increases BMD. A low BMD is associated with an increased risk of fracture” or “X maintains BMD. A low BMD is associated with an increased risk of fracture”.\nAnimal models are appropriate to determine whether an increase in BMD associated with a food product is accompanied by an increase in bone strength. A food product with a positive effect on BMD, together with animal studies showing an improvement in bone strength or showing a relationship between BMD changes induced by the food product and bone strength, could have the claim: “X increases (or maintains) BMD that could reduce the risk of fracture” or “X increases (or maintains) BMD that increases bone strength” or “X increases bone strength”.\nReduction of the risk of fracture\nA reduction of the incidence of fracture is a major aim of food products beneficial to skeletal health, but according to the regulation cannot be claimed as such without mentioning the effect on a risk factor. However, a reduction in the fracture risk is obviously supportive for a claim on the reduction of an identified risk factor. For non-spinal fractures, either femoral (hip) or major non-vertebral (pelvis, distal femur, proximal tibia, ribs, proximal humerus forearm, and hip) fractures should be assessed.\nThe GREES panel considers that six different health claims could be accepted for an effect of food products on bone health. However, as already used by the European Food and Safety Authority, different wording to reflect the level of evidence of the effect could be used depending on the effect that is (always), may (demonstrated only under certain circumstances) or might be (logically expected benefit from physiology but yet not demonstrated) beneficial for bone health.\nImprovement of calcium bioavailabilityCalcium bioavailability may be defined as the proportion of calcium in foods which is absorbed and utilized for normal metabolic functions. In addition to the amount of calcium in the diet, the fractional absorption of dietary calcium in food and its retention in the body are also a factor that determines the availability of calcium for bone development and maintenance of bone health [10, 11]. Many methods can be used to assess bioavailability (i.e., classical and isotopic balances, urinary excretion, isotope labeling in the urine, plasma, and bones) [12]. The group considers that an increase in bioavailability is not beneficial if not accompanied by calcium retention in the body. A food product with an effect on calcium bioavailability with or without calcium retention data, unless associated with appropriate animal studies would not fulfill a claim related to article 14. However, food products that show an effect on bioavailability and calcium retention could have an article 13 claim: “X increases calcium absorption” or “X increases calcium bioavailability”.Maintenance of bone metabolism (through an effect on osteoclast regulatory proteins)The transition of osteoclast precursors to mature osteoclasts that are capable of resorbing bone is tightly regulated by osteoclast regulatory proteins that either affect the differentiation and proliferation of osteoclast precursors into mature osteoclasts or are involved in the coupling between osteoblasts and osteoclasts [13]. Markers of osteoclastogenesis include receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, whereas markers of osteoclast number include tartrate-resistant acid phosphatase and cathepsin K. These proteins contribute to bone metabolism but are not yet strongly associated with elements of bone strength. A food product with an effect on osteoclast regulatory proteins, unless supported by animal studies (see next BMD and BTM sections) would not fulfill a claim related to article 14. The product, however, might have the label under the article 13: “X contributes to the maintenance of bone metabolism”.Maintenance or changes in bone turnover markerA determinant of bone strength that is not assessed by bone mineral density (BMD) is the rate of bone remodeling. Depending upon their origin, bone turnover markers (BTMs) are classified as indices of bone resorption or formation [14–17]. The rate of bone resorption and formation can be estimated by assays that measure the serum concentration or urinary excretion of different target molecules specific to these cellular processes. GREES panel recommends the inclusion of reference markers of bone formation (serum procollagen type I N propeptide, s-PINP) and resorption (serum C-terminal cross-linking telopeptide of type I collagen, s-CTX) in keeping with the recommendations of the International Osteoporosis Foundation [18]. A food product with a positive BTM balance might have the claim: “X maintains normal bone remodeling that could contribute to the normal structure and function of bones” or “X increases markers of bone formation that could contribute to the normal structure and function of bones” or “X decreases markers of bone resorption that could contribute to the normal structure and function of bones”.As in the case of BMD, BTMs are only indicators of fracture risk, but the change in BTM induced by a product is not necessarily associated with a change in fracture risk or bone strength. In this regard, animal models are useful to assess if changes in BTMs due to the intake of the food product are associated with an increase in bone strength. A food product with an effect on BTMs together with animal studies that showed improved bone strength or a relationship between changes in BTMs induced by the food product and bone strength could have the claim: “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that is associated with bone strength” or “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that increases bone strength” or “X increases bone strength”.Maintenance or improvement in bone structureThe key role of bone microarchitecture in bone health was suggested by the classic definition of osteoporosis adopted in 1993 [19]. Methods for investigating 3-D bone microarchitecture and bone strength include in vitro μCT, in vitro μMRI, in vivo pQCT, and in vivo high-resolution MRI [20]. They aim to quantify various determinants of bone strength such as bone geometry, volumetric bone density, microarchitecture, and properties of the bone matrix [21]. A food product with an effect on bone microarchitecture could have the claim: “X improves (or maintains) bone microarchitecture that could contribute to the normal structure and function of bones”.It is considered that the assessment of bone structure with the tools currently available in man is not sufficiently validated to be a reliable surrogate of bone strength. For this reason, animal models are needed to assess the relationship between changes in bone microarchitecture induced by the food product and any increase in bone strength. A food product with an effect on microarchitecture of the human bone and animals studies that show improvement in bone strength or show the relationship between change in bone structure induced by the food product and bone strength could have the claim: “X improves bone microarchitecture that increases bone strength” or “X increases bone strength”Maintenance or increase in bone mineral densityBone strength is determined by many factors, including bone mass. Bone mass is estimated in clinical practice by the measurement of BMD. BMD, as measured by DXA, represents an estimate of the quantity of mineral (grams of calcium) divided by the two-dimensional area of the bone [22]. There is a strong relationship between the risk of fracture and BMD but there is a wide overlap in the bone densities of patients who develop a fracture and those who do not. Since BMD is only a surrogate marker for bone strength or fracture risk, and since product-induced changes in BMD are not clearly associated with changes in bone strength or fracture risk, an increase in BMD may not be associated with an increased bone strength or decreased fracture risk [23]. A food product with a positive effect on BMD could have the claim: “X increases BMD. A low BMD is associated with an increased risk of fracture” or “X maintains BMD. A low BMD is associated with an increased risk of fracture”.Animal models are appropriate to determine whether an increase in BMD associated with a food product is accompanied by an increase in bone strength. A food product with a positive effect on BMD, together with animal studies showing an improvement in bone strength or showing a relationship between BMD changes induced by the food product and bone strength, could have the claim: “X increases (or maintains) BMD that could reduce the risk of fracture” or “X increases (or maintains) BMD that increases bone strength” or “X increases bone strength”.Reduction of the risk of fractureA reduction of the incidence of fracture is a major aim of food products beneficial to skeletal health, but according to the regulation cannot be claimed as such without mentioning the effect on a risk factor. However, a reduction in the fracture risk is obviously supportive for a claim on the reduction of an identified risk factor. For non-spinal fractures, either femoral (hip) or major non-vertebral (pelvis, distal femur, proximal tibia, ribs, proximal humerus forearm, and hip) fractures should be assessed.\n\nImprovement of calcium bioavailability\nCalcium bioavailability may be defined as the proportion of calcium in foods which is absorbed and utilized for normal metabolic functions. In addition to the amount of calcium in the diet, the fractional absorption of dietary calcium in food and its retention in the body are also a factor that determines the availability of calcium for bone development and maintenance of bone health [10, 11]. Many methods can be used to assess bioavailability (i.e., classical and isotopic balances, urinary excretion, isotope labeling in the urine, plasma, and bones) [12]. The group considers that an increase in bioavailability is not beneficial if not accompanied by calcium retention in the body. A food product with an effect on calcium bioavailability with or without calcium retention data, unless associated with appropriate animal studies would not fulfill a claim related to article 14. However, food products that show an effect on bioavailability and calcium retention could have an article 13 claim: “X increases calcium absorption” or “X increases calcium bioavailability”.\nMaintenance of bone metabolism (through an effect on osteoclast regulatory proteins)\nThe transition of osteoclast precursors to mature osteoclasts that are capable of resorbing bone is tightly regulated by osteoclast regulatory proteins that either affect the differentiation and proliferation of osteoclast precursors into mature osteoclasts or are involved in the coupling between osteoblasts and osteoclasts [13]. Markers of osteoclastogenesis include receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, whereas markers of osteoclast number include tartrate-resistant acid phosphatase and cathepsin K. These proteins contribute to bone metabolism but are not yet strongly associated with elements of bone strength. A food product with an effect on osteoclast regulatory proteins, unless supported by animal studies (see next BMD and BTM sections) would not fulfill a claim related to article 14. The product, however, might have the label under the article 13: “X contributes to the maintenance of bone metabolism”.\nMaintenance or changes in bone turnover marker\nA determinant of bone strength that is not assessed by bone mineral density (BMD) is the rate of bone remodeling. Depending upon their origin, bone turnover markers (BTMs) are classified as indices of bone resorption or formation [14–17]. The rate of bone resorption and formation can be estimated by assays that measure the serum concentration or urinary excretion of different target molecules specific to these cellular processes. GREES panel recommends the inclusion of reference markers of bone formation (serum procollagen type I N propeptide, s-PINP) and resorption (serum C-terminal cross-linking telopeptide of type I collagen, s-CTX) in keeping with the recommendations of the International Osteoporosis Foundation [18]. A food product with a positive BTM balance might have the claim: “X maintains normal bone remodeling that could contribute to the normal structure and function of bones” or “X increases markers of bone formation that could contribute to the normal structure and function of bones” or “X decreases markers of bone resorption that could contribute to the normal structure and function of bones”.\nAs in the case of BMD, BTMs are only indicators of fracture risk, but the change in BTM induced by a product is not necessarily associated with a change in fracture risk or bone strength. In this regard, animal models are useful to assess if changes in BTMs due to the intake of the food product are associated with an increase in bone strength. A food product with an effect on BTMs together with animal studies that showed improved bone strength or a relationship between changes in BTMs induced by the food product and bone strength could have the claim: “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that is associated with bone strength” or “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that increases bone strength” or “X increases bone strength”.\nMaintenance or improvement in bone structure\nThe key role of bone microarchitecture in bone health was suggested by the classic definition of osteoporosis adopted in 1993 [19]. Methods for investigating 3-D bone microarchitecture and bone strength include in vitro μCT, in vitro μMRI, in vivo pQCT, and in vivo high-resolution MRI [20]. They aim to quantify various determinants of bone strength such as bone geometry, volumetric bone density, microarchitecture, and properties of the bone matrix [21]. A food product with an effect on bone microarchitecture could have the claim: “X improves (or maintains) bone microarchitecture that could contribute to the normal structure and function of bones”.\nIt is considered that the assessment of bone structure with the tools currently available in man is not sufficiently validated to be a reliable surrogate of bone strength. For this reason, animal models are needed to assess the relationship between changes in bone microarchitecture induced by the food product and any increase in bone strength. A food product with an effect on microarchitecture of the human bone and animals studies that show improvement in bone strength or show the relationship between change in bone structure induced by the food product and bone strength could have the claim: “X improves bone microarchitecture that increases bone strength” or “X increases bone strength”\nMaintenance or increase in bone mineral density\nBone strength is determined by many factors, including bone mass. Bone mass is estimated in clinical practice by the measurement of BMD. BMD, as measured by DXA, represents an estimate of the quantity of mineral (grams of calcium) divided by the two-dimensional area of the bone [22]. There is a strong relationship between the risk of fracture and BMD but there is a wide overlap in the bone densities of patients who develop a fracture and those who do not. Since BMD is only a surrogate marker for bone strength or fracture risk, and since product-induced changes in BMD are not clearly associated with changes in bone strength or fracture risk, an increase in BMD may not be associated with an increased bone strength or decreased fracture risk [23]. A food product with a positive effect on BMD could have the claim: “X increases BMD. A low BMD is associated with an increased risk of fracture” or “X maintains BMD. A low BMD is associated with an increased risk of fracture”.\nAnimal models are appropriate to determine whether an increase in BMD associated with a food product is accompanied by an increase in bone strength. A food product with a positive effect on BMD, together with animal studies showing an improvement in bone strength or showing a relationship between BMD changes induced by the food product and bone strength, could have the claim: “X increases (or maintains) BMD that could reduce the risk of fracture” or “X increases (or maintains) BMD that increases bone strength” or “X increases bone strength”.\nReduction of the risk of fracture\nA reduction of the incidence of fracture is a major aim of food products beneficial to skeletal health, but according to the regulation cannot be claimed as such without mentioning the effect on a risk factor. However, a reduction in the fracture risk is obviously supportive for a claim on the reduction of an identified risk factor. For non-spinal fractures, either femoral (hip) or major non-vertebral (pelvis, distal femur, proximal tibia, ribs, proximal humerus forearm, and hip) fractures should be assessed.\n[SUBTITLE] Design of clinical studies [SUBSECTION] \nPopulationThe subjects studied should be representative of the population targeted for the food product. The applicant should take all necessary precautions to make sure that the tested population is equivalent to the user population with respect of ethnicity, age, physiological status (such as menopause for example), life habits (such as exercise) and diet. No densitometric criteria are required for inclusion. However, the experimental and the control group must show no significant differences in term of baseline BMD.DesignThe ideal design would be a multicentre randomized controlled study (RCT). The control could be a placebo, another active product or nothing, depending on the tested food. When possible, subjects and/or investigators should be blinded of the intervention. Treatment and control groups should be balanced with respect to gender, age, menopausal status, dietary habits, or underlying diseases. The GREES panel recognizes that a RCT is not always possible in practice or from an ethical point of view. Since the totality of the evidence should be weighed for the substantiation of a claim, well-designed prospective cohort studies, case–control studies and/or observational studies of high quality could be acceptable if accompanied by other data (e.g., animal data, effect on multiple surrogate endpoints). Cross-over studies design can also be considered. All the efforts should be made to eliminate potential confounders.Duration of studyThe duration of the trial should be predetermined and should depend on the outcome. For BMD, duration of at least 1 year seems necessary. For BTMs, a 3-month study is the minimum. The primary efficacy endpoint should be assessed at the end of the predetermined treatment period in comparison with the measurement at baseline. Intermediate measurements are also recommended.Statistical analysisIntention-to-treat analysis should be the primary method of evaluation. Statistical significance will be inferred if a P value is equal to or less than 0.05. The beta risk will be equal to or less than 20%. The sample size of the study must be calculated prior to the start of the study. Possible confounding variables should be managed using appropriate statistical analysis. Within group (end vs. baseline) and between groups comparisons should be made.Diet habit and lifestyleThe control of critical effect modifiers such as physical activity, synergies with a multitude of other nutrients and the influence of nutrigenomic relationships must be taken into account.Intakes of other nutrients or foods, on which the tested nutrient is dependent, must be optimized. Any supplementation with other food products known to have an effect on bone (e.g., calcium and/or vitamin D) should be consistent within all patient groups.ObservanceObservance of food product intake should be monitored during the study to be able to perform pre-planned analyses on individuals with high and poor compliance rates or analyses of dose–response.SafetyAll adverse experiences occurring during the course of clinical trials should be fully documented with separate analysis of adverse events, dropouts, and patients who died while being on the study.\n\nPopulation\nThe subjects studied should be representative of the population targeted for the food product. The applicant should take all necessary precautions to make sure that the tested population is equivalent to the user population with respect of ethnicity, age, physiological status (such as menopause for example), life habits (such as exercise) and diet. No densitometric criteria are required for inclusion. However, the experimental and the control group must show no significant differences in term of baseline BMD.\nDesign\nThe ideal design would be a multicentre randomized controlled study (RCT). The control could be a placebo, another active product or nothing, depending on the tested food. When possible, subjects and/or investigators should be blinded of the intervention. Treatment and control groups should be balanced with respect to gender, age, menopausal status, dietary habits, or underlying diseases. The GREES panel recognizes that a RCT is not always possible in practice or from an ethical point of view. Since the totality of the evidence should be weighed for the substantiation of a claim, well-designed prospective cohort studies, case–control studies and/or observational studies of high quality could be acceptable if accompanied by other data (e.g., animal data, effect on multiple surrogate endpoints). Cross-over studies design can also be considered. All the efforts should be made to eliminate potential confounders.\nDuration of study\nThe duration of the trial should be predetermined and should depend on the outcome. For BMD, duration of at least 1 year seems necessary. For BTMs, a 3-month study is the minimum. The primary efficacy endpoint should be assessed at the end of the predetermined treatment period in comparison with the measurement at baseline. Intermediate measurements are also recommended.\nStatistical analysis\nIntention-to-treat analysis should be the primary method of evaluation. Statistical significance will be inferred if a P value is equal to or less than 0.05. The beta risk will be equal to or less than 20%. The sample size of the study must be calculated prior to the start of the study. Possible confounding variables should be managed using appropriate statistical analysis. Within group (end vs. baseline) and between groups comparisons should be made.\nDiet habit and lifestyle\nThe control of critical effect modifiers such as physical activity, synergies with a multitude of other nutrients and the influence of nutrigenomic relationships must be taken into account.\nIntakes of other nutrients or foods, on which the tested nutrient is dependent, must be optimized. Any supplementation with other food products known to have an effect on bone (e.g., calcium and/or vitamin D) should be consistent within all patient groups.\nObservance\nObservance of food product intake should be monitored during the study to be able to perform pre-planned analyses on individuals with high and poor compliance rates or analyses of dose–response.\nSafety\nAll adverse experiences occurring during the course of clinical trials should be fully documented with separate analysis of adverse events, dropouts, and patients who died while being on the study.\n\nPopulationThe subjects studied should be representative of the population targeted for the food product. The applicant should take all necessary precautions to make sure that the tested population is equivalent to the user population with respect of ethnicity, age, physiological status (such as menopause for example), life habits (such as exercise) and diet. No densitometric criteria are required for inclusion. However, the experimental and the control group must show no significant differences in term of baseline BMD.DesignThe ideal design would be a multicentre randomized controlled study (RCT). The control could be a placebo, another active product or nothing, depending on the tested food. When possible, subjects and/or investigators should be blinded of the intervention. Treatment and control groups should be balanced with respect to gender, age, menopausal status, dietary habits, or underlying diseases. The GREES panel recognizes that a RCT is not always possible in practice or from an ethical point of view. Since the totality of the evidence should be weighed for the substantiation of a claim, well-designed prospective cohort studies, case–control studies and/or observational studies of high quality could be acceptable if accompanied by other data (e.g., animal data, effect on multiple surrogate endpoints). Cross-over studies design can also be considered. All the efforts should be made to eliminate potential confounders.Duration of studyThe duration of the trial should be predetermined and should depend on the outcome. For BMD, duration of at least 1 year seems necessary. For BTMs, a 3-month study is the minimum. The primary efficacy endpoint should be assessed at the end of the predetermined treatment period in comparison with the measurement at baseline. Intermediate measurements are also recommended.Statistical analysisIntention-to-treat analysis should be the primary method of evaluation. Statistical significance will be inferred if a P value is equal to or less than 0.05. The beta risk will be equal to or less than 20%. The sample size of the study must be calculated prior to the start of the study. Possible confounding variables should be managed using appropriate statistical analysis. Within group (end vs. baseline) and between groups comparisons should be made.Diet habit and lifestyleThe control of critical effect modifiers such as physical activity, synergies with a multitude of other nutrients and the influence of nutrigenomic relationships must be taken into account.Intakes of other nutrients or foods, on which the tested nutrient is dependent, must be optimized. Any supplementation with other food products known to have an effect on bone (e.g., calcium and/or vitamin D) should be consistent within all patient groups.ObservanceObservance of food product intake should be monitored during the study to be able to perform pre-planned analyses on individuals with high and poor compliance rates or analyses of dose–response.SafetyAll adverse experiences occurring during the course of clinical trials should be fully documented with separate analysis of adverse events, dropouts, and patients who died while being on the study.\n\nPopulation\nThe subjects studied should be representative of the population targeted for the food product. The applicant should take all necessary precautions to make sure that the tested population is equivalent to the user population with respect of ethnicity, age, physiological status (such as menopause for example), life habits (such as exercise) and diet. No densitometric criteria are required for inclusion. However, the experimental and the control group must show no significant differences in term of baseline BMD.\nDesign\nThe ideal design would be a multicentre randomized controlled study (RCT). The control could be a placebo, another active product or nothing, depending on the tested food. When possible, subjects and/or investigators should be blinded of the intervention. Treatment and control groups should be balanced with respect to gender, age, menopausal status, dietary habits, or underlying diseases. The GREES panel recognizes that a RCT is not always possible in practice or from an ethical point of view. Since the totality of the evidence should be weighed for the substantiation of a claim, well-designed prospective cohort studies, case–control studies and/or observational studies of high quality could be acceptable if accompanied by other data (e.g., animal data, effect on multiple surrogate endpoints). Cross-over studies design can also be considered. All the efforts should be made to eliminate potential confounders.\nDuration of study\nThe duration of the trial should be predetermined and should depend on the outcome. For BMD, duration of at least 1 year seems necessary. For BTMs, a 3-month study is the minimum. The primary efficacy endpoint should be assessed at the end of the predetermined treatment period in comparison with the measurement at baseline. Intermediate measurements are also recommended.\nStatistical analysis\nIntention-to-treat analysis should be the primary method of evaluation. Statistical significance will be inferred if a P value is equal to or less than 0.05. The beta risk will be equal to or less than 20%. The sample size of the study must be calculated prior to the start of the study. Possible confounding variables should be managed using appropriate statistical analysis. Within group (end vs. baseline) and between groups comparisons should be made.\nDiet habit and lifestyle\nThe control of critical effect modifiers such as physical activity, synergies with a multitude of other nutrients and the influence of nutrigenomic relationships must be taken into account.\nIntakes of other nutrients or foods, on which the tested nutrient is dependent, must be optimized. Any supplementation with other food products known to have an effect on bone (e.g., calcium and/or vitamin D) should be consistent within all patient groups.\nObservance\nObservance of food product intake should be monitored during the study to be able to perform pre-planned analyses on individuals with high and poor compliance rates or analyses of dose–response.\nSafety\nAll adverse experiences occurring during the course of clinical trials should be fully documented with separate analysis of adverse events, dropouts, and patients who died while being on the study.", "A variety of invasive and non-invasive techniques can be used to provide relevant endpoints [4, 7], including bioavailability studies, microarray or PCR analysis of modulated genes, histomorphometry, culture of bone forming or bone resorbing cells ex vivo, exposure to primary cell cultures to plasma harvested from treated animals, the chemistry and biochemistry of bone tissue, the assessment of biochemical indices of skeletal turnover in blood and urine, metabolic balance of calcium combined with radioactive calcium kinetics, radiogrammetry of bone radiographs, neutron activation for whole body calcium, dual x-ray absorptiometry (DXA), and the assessment of bone strength [8]. The latter endpoint is considered to be the most relevant in the field of bone health claims. Bone strength reflects both bone density and bone quality. Bone quality depends on bone architecture, mineralization, turnover, and accumulation of microdamage. Therefore, the assessment of bone health would benefit from the measurement of bone strength in vivo. No validated non-invasive tools capable of measuring bone strength in vivo are available to date. However, biomechanical tests of resistance to fracture provide an objective measure of overall bone strength. The three main types of biomechanical tests for bone strength are bending, torsional, and compression tests [9]. It is considered that both long bones and vertebral bodies should be tested since they may be differentially affected by food products.\nBesides the assessment of a direct effect of the food product on bone strength, two other aims of animal data could be to better understand the mechanism of action of the food product or to validate surrogate variables used in human animal data to see if these variables reflect bone strength.\nKey criteria of suitable/acceptable animal studies are:\n➢to deliver the food product in the manner in which it will be delivered in a human setting;➢to utilize a site of delivery and/or assessment site that is as closely matched as possible to the settings in which it will be used;➢to utilize an animal that provides a metabolic background and physiological responsiveness comparable to humans;➢to utilize a formulation of active agent that has the same composition, release, retention, and degradation properties as the formulation that will be used in humans.\n\n➢to deliver the food product in the manner in which it will be delivered in a human setting;\n➢to utilize a site of delivery and/or assessment site that is as closely matched as possible to the settings in which it will be used;\n➢to utilize an animal that provides a metabolic background and physiological responsiveness comparable to humans;\n➢to utilize a formulation of active agent that has the same composition, release, retention, and degradation properties as the formulation that will be used in humans.", "The GREES panel considers that six different health claims could be accepted for an effect of food products on bone health. However, as already used by the European Food and Safety Authority, different wording to reflect the level of evidence of the effect could be used depending on the effect that is (always), may (demonstrated only under certain circumstances) or might be (logically expected benefit from physiology but yet not demonstrated) beneficial for bone health.\nImprovement of calcium bioavailabilityCalcium bioavailability may be defined as the proportion of calcium in foods which is absorbed and utilized for normal metabolic functions. In addition to the amount of calcium in the diet, the fractional absorption of dietary calcium in food and its retention in the body are also a factor that determines the availability of calcium for bone development and maintenance of bone health [10, 11]. Many methods can be used to assess bioavailability (i.e., classical and isotopic balances, urinary excretion, isotope labeling in the urine, plasma, and bones) [12]. The group considers that an increase in bioavailability is not beneficial if not accompanied by calcium retention in the body. A food product with an effect on calcium bioavailability with or without calcium retention data, unless associated with appropriate animal studies would not fulfill a claim related to article 14. However, food products that show an effect on bioavailability and calcium retention could have an article 13 claim: “X increases calcium absorption” or “X increases calcium bioavailability”.Maintenance of bone metabolism (through an effect on osteoclast regulatory proteins)The transition of osteoclast precursors to mature osteoclasts that are capable of resorbing bone is tightly regulated by osteoclast regulatory proteins that either affect the differentiation and proliferation of osteoclast precursors into mature osteoclasts or are involved in the coupling between osteoblasts and osteoclasts [13]. Markers of osteoclastogenesis include receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, whereas markers of osteoclast number include tartrate-resistant acid phosphatase and cathepsin K. These proteins contribute to bone metabolism but are not yet strongly associated with elements of bone strength. A food product with an effect on osteoclast regulatory proteins, unless supported by animal studies (see next BMD and BTM sections) would not fulfill a claim related to article 14. The product, however, might have the label under the article 13: “X contributes to the maintenance of bone metabolism”.Maintenance or changes in bone turnover markerA determinant of bone strength that is not assessed by bone mineral density (BMD) is the rate of bone remodeling. Depending upon their origin, bone turnover markers (BTMs) are classified as indices of bone resorption or formation [14–17]. The rate of bone resorption and formation can be estimated by assays that measure the serum concentration or urinary excretion of different target molecules specific to these cellular processes. GREES panel recommends the inclusion of reference markers of bone formation (serum procollagen type I N propeptide, s-PINP) and resorption (serum C-terminal cross-linking telopeptide of type I collagen, s-CTX) in keeping with the recommendations of the International Osteoporosis Foundation [18]. A food product with a positive BTM balance might have the claim: “X maintains normal bone remodeling that could contribute to the normal structure and function of bones” or “X increases markers of bone formation that could contribute to the normal structure and function of bones” or “X decreases markers of bone resorption that could contribute to the normal structure and function of bones”.As in the case of BMD, BTMs are only indicators of fracture risk, but the change in BTM induced by a product is not necessarily associated with a change in fracture risk or bone strength. In this regard, animal models are useful to assess if changes in BTMs due to the intake of the food product are associated with an increase in bone strength. A food product with an effect on BTMs together with animal studies that showed improved bone strength or a relationship between changes in BTMs induced by the food product and bone strength could have the claim: “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that is associated with bone strength” or “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that increases bone strength” or “X increases bone strength”.Maintenance or improvement in bone structureThe key role of bone microarchitecture in bone health was suggested by the classic definition of osteoporosis adopted in 1993 [19]. Methods for investigating 3-D bone microarchitecture and bone strength include in vitro μCT, in vitro μMRI, in vivo pQCT, and in vivo high-resolution MRI [20]. They aim to quantify various determinants of bone strength such as bone geometry, volumetric bone density, microarchitecture, and properties of the bone matrix [21]. A food product with an effect on bone microarchitecture could have the claim: “X improves (or maintains) bone microarchitecture that could contribute to the normal structure and function of bones”.It is considered that the assessment of bone structure with the tools currently available in man is not sufficiently validated to be a reliable surrogate of bone strength. For this reason, animal models are needed to assess the relationship between changes in bone microarchitecture induced by the food product and any increase in bone strength. A food product with an effect on microarchitecture of the human bone and animals studies that show improvement in bone strength or show the relationship between change in bone structure induced by the food product and bone strength could have the claim: “X improves bone microarchitecture that increases bone strength” or “X increases bone strength”Maintenance or increase in bone mineral densityBone strength is determined by many factors, including bone mass. Bone mass is estimated in clinical practice by the measurement of BMD. BMD, as measured by DXA, represents an estimate of the quantity of mineral (grams of calcium) divided by the two-dimensional area of the bone [22]. There is a strong relationship between the risk of fracture and BMD but there is a wide overlap in the bone densities of patients who develop a fracture and those who do not. Since BMD is only a surrogate marker for bone strength or fracture risk, and since product-induced changes in BMD are not clearly associated with changes in bone strength or fracture risk, an increase in BMD may not be associated with an increased bone strength or decreased fracture risk [23]. A food product with a positive effect on BMD could have the claim: “X increases BMD. A low BMD is associated with an increased risk of fracture” or “X maintains BMD. A low BMD is associated with an increased risk of fracture”.Animal models are appropriate to determine whether an increase in BMD associated with a food product is accompanied by an increase in bone strength. A food product with a positive effect on BMD, together with animal studies showing an improvement in bone strength or showing a relationship between BMD changes induced by the food product and bone strength, could have the claim: “X increases (or maintains) BMD that could reduce the risk of fracture” or “X increases (or maintains) BMD that increases bone strength” or “X increases bone strength”.Reduction of the risk of fractureA reduction of the incidence of fracture is a major aim of food products beneficial to skeletal health, but according to the regulation cannot be claimed as such without mentioning the effect on a risk factor. However, a reduction in the fracture risk is obviously supportive for a claim on the reduction of an identified risk factor. For non-spinal fractures, either femoral (hip) or major non-vertebral (pelvis, distal femur, proximal tibia, ribs, proximal humerus forearm, and hip) fractures should be assessed.\n\nImprovement of calcium bioavailability\nCalcium bioavailability may be defined as the proportion of calcium in foods which is absorbed and utilized for normal metabolic functions. In addition to the amount of calcium in the diet, the fractional absorption of dietary calcium in food and its retention in the body are also a factor that determines the availability of calcium for bone development and maintenance of bone health [10, 11]. Many methods can be used to assess bioavailability (i.e., classical and isotopic balances, urinary excretion, isotope labeling in the urine, plasma, and bones) [12]. The group considers that an increase in bioavailability is not beneficial if not accompanied by calcium retention in the body. A food product with an effect on calcium bioavailability with or without calcium retention data, unless associated with appropriate animal studies would not fulfill a claim related to article 14. However, food products that show an effect on bioavailability and calcium retention could have an article 13 claim: “X increases calcium absorption” or “X increases calcium bioavailability”.\nMaintenance of bone metabolism (through an effect on osteoclast regulatory proteins)\nThe transition of osteoclast precursors to mature osteoclasts that are capable of resorbing bone is tightly regulated by osteoclast regulatory proteins that either affect the differentiation and proliferation of osteoclast precursors into mature osteoclasts or are involved in the coupling between osteoblasts and osteoclasts [13]. Markers of osteoclastogenesis include receptor activator of nuclear factor kappa-B ligand and osteoprotegerin, whereas markers of osteoclast number include tartrate-resistant acid phosphatase and cathepsin K. These proteins contribute to bone metabolism but are not yet strongly associated with elements of bone strength. A food product with an effect on osteoclast regulatory proteins, unless supported by animal studies (see next BMD and BTM sections) would not fulfill a claim related to article 14. The product, however, might have the label under the article 13: “X contributes to the maintenance of bone metabolism”.\nMaintenance or changes in bone turnover marker\nA determinant of bone strength that is not assessed by bone mineral density (BMD) is the rate of bone remodeling. Depending upon their origin, bone turnover markers (BTMs) are classified as indices of bone resorption or formation [14–17]. The rate of bone resorption and formation can be estimated by assays that measure the serum concentration or urinary excretion of different target molecules specific to these cellular processes. GREES panel recommends the inclusion of reference markers of bone formation (serum procollagen type I N propeptide, s-PINP) and resorption (serum C-terminal cross-linking telopeptide of type I collagen, s-CTX) in keeping with the recommendations of the International Osteoporosis Foundation [18]. A food product with a positive BTM balance might have the claim: “X maintains normal bone remodeling that could contribute to the normal structure and function of bones” or “X increases markers of bone formation that could contribute to the normal structure and function of bones” or “X decreases markers of bone resorption that could contribute to the normal structure and function of bones”.\nAs in the case of BMD, BTMs are only indicators of fracture risk, but the change in BTM induced by a product is not necessarily associated with a change in fracture risk or bone strength. In this regard, animal models are useful to assess if changes in BTMs due to the intake of the food product are associated with an increase in bone strength. A food product with an effect on BTMs together with animal studies that showed improved bone strength or a relationship between changes in BTMs induced by the food product and bone strength could have the claim: “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that is associated with bone strength” or “X contributes to the maintenance of normal bone remodeling (or increases bone formation or decreases bone resorption) that increases bone strength” or “X increases bone strength”.\nMaintenance or improvement in bone structure\nThe key role of bone microarchitecture in bone health was suggested by the classic definition of osteoporosis adopted in 1993 [19]. Methods for investigating 3-D bone microarchitecture and bone strength include in vitro μCT, in vitro μMRI, in vivo pQCT, and in vivo high-resolution MRI [20]. They aim to quantify various determinants of bone strength such as bone geometry, volumetric bone density, microarchitecture, and properties of the bone matrix [21]. A food product with an effect on bone microarchitecture could have the claim: “X improves (or maintains) bone microarchitecture that could contribute to the normal structure and function of bones”.\nIt is considered that the assessment of bone structure with the tools currently available in man is not sufficiently validated to be a reliable surrogate of bone strength. For this reason, animal models are needed to assess the relationship between changes in bone microarchitecture induced by the food product and any increase in bone strength. A food product with an effect on microarchitecture of the human bone and animals studies that show improvement in bone strength or show the relationship between change in bone structure induced by the food product and bone strength could have the claim: “X improves bone microarchitecture that increases bone strength” or “X increases bone strength”\nMaintenance or increase in bone mineral density\nBone strength is determined by many factors, including bone mass. Bone mass is estimated in clinical practice by the measurement of BMD. BMD, as measured by DXA, represents an estimate of the quantity of mineral (grams of calcium) divided by the two-dimensional area of the bone [22]. There is a strong relationship between the risk of fracture and BMD but there is a wide overlap in the bone densities of patients who develop a fracture and those who do not. Since BMD is only a surrogate marker for bone strength or fracture risk, and since product-induced changes in BMD are not clearly associated with changes in bone strength or fracture risk, an increase in BMD may not be associated with an increased bone strength or decreased fracture risk [23]. A food product with a positive effect on BMD could have the claim: “X increases BMD. A low BMD is associated with an increased risk of fracture” or “X maintains BMD. A low BMD is associated with an increased risk of fracture”.\nAnimal models are appropriate to determine whether an increase in BMD associated with a food product is accompanied by an increase in bone strength. A food product with a positive effect on BMD, together with animal studies showing an improvement in bone strength or showing a relationship between BMD changes induced by the food product and bone strength, could have the claim: “X increases (or maintains) BMD that could reduce the risk of fracture” or “X increases (or maintains) BMD that increases bone strength” or “X increases bone strength”.\nReduction of the risk of fracture\nA reduction of the incidence of fracture is a major aim of food products beneficial to skeletal health, but according to the regulation cannot be claimed as such without mentioning the effect on a risk factor. However, a reduction in the fracture risk is obviously supportive for a claim on the reduction of an identified risk factor. For non-spinal fractures, either femoral (hip) or major non-vertebral (pelvis, distal femur, proximal tibia, ribs, proximal humerus forearm, and hip) fractures should be assessed.", "\nPopulationThe subjects studied should be representative of the population targeted for the food product. The applicant should take all necessary precautions to make sure that the tested population is equivalent to the user population with respect of ethnicity, age, physiological status (such as menopause for example), life habits (such as exercise) and diet. No densitometric criteria are required for inclusion. However, the experimental and the control group must show no significant differences in term of baseline BMD.DesignThe ideal design would be a multicentre randomized controlled study (RCT). The control could be a placebo, another active product or nothing, depending on the tested food. When possible, subjects and/or investigators should be blinded of the intervention. Treatment and control groups should be balanced with respect to gender, age, menopausal status, dietary habits, or underlying diseases. The GREES panel recognizes that a RCT is not always possible in practice or from an ethical point of view. Since the totality of the evidence should be weighed for the substantiation of a claim, well-designed prospective cohort studies, case–control studies and/or observational studies of high quality could be acceptable if accompanied by other data (e.g., animal data, effect on multiple surrogate endpoints). Cross-over studies design can also be considered. All the efforts should be made to eliminate potential confounders.Duration of studyThe duration of the trial should be predetermined and should depend on the outcome. For BMD, duration of at least 1 year seems necessary. For BTMs, a 3-month study is the minimum. The primary efficacy endpoint should be assessed at the end of the predetermined treatment period in comparison with the measurement at baseline. Intermediate measurements are also recommended.Statistical analysisIntention-to-treat analysis should be the primary method of evaluation. Statistical significance will be inferred if a P value is equal to or less than 0.05. The beta risk will be equal to or less than 20%. The sample size of the study must be calculated prior to the start of the study. Possible confounding variables should be managed using appropriate statistical analysis. Within group (end vs. baseline) and between groups comparisons should be made.Diet habit and lifestyleThe control of critical effect modifiers such as physical activity, synergies with a multitude of other nutrients and the influence of nutrigenomic relationships must be taken into account.Intakes of other nutrients or foods, on which the tested nutrient is dependent, must be optimized. Any supplementation with other food products known to have an effect on bone (e.g., calcium and/or vitamin D) should be consistent within all patient groups.ObservanceObservance of food product intake should be monitored during the study to be able to perform pre-planned analyses on individuals with high and poor compliance rates or analyses of dose–response.SafetyAll adverse experiences occurring during the course of clinical trials should be fully documented with separate analysis of adverse events, dropouts, and patients who died while being on the study.\n\nPopulation\nThe subjects studied should be representative of the population targeted for the food product. The applicant should take all necessary precautions to make sure that the tested population is equivalent to the user population with respect of ethnicity, age, physiological status (such as menopause for example), life habits (such as exercise) and diet. No densitometric criteria are required for inclusion. However, the experimental and the control group must show no significant differences in term of baseline BMD.\nDesign\nThe ideal design would be a multicentre randomized controlled study (RCT). The control could be a placebo, another active product or nothing, depending on the tested food. When possible, subjects and/or investigators should be blinded of the intervention. Treatment and control groups should be balanced with respect to gender, age, menopausal status, dietary habits, or underlying diseases. The GREES panel recognizes that a RCT is not always possible in practice or from an ethical point of view. Since the totality of the evidence should be weighed for the substantiation of a claim, well-designed prospective cohort studies, case–control studies and/or observational studies of high quality could be acceptable if accompanied by other data (e.g., animal data, effect on multiple surrogate endpoints). Cross-over studies design can also be considered. All the efforts should be made to eliminate potential confounders.\nDuration of study\nThe duration of the trial should be predetermined and should depend on the outcome. For BMD, duration of at least 1 year seems necessary. For BTMs, a 3-month study is the minimum. The primary efficacy endpoint should be assessed at the end of the predetermined treatment period in comparison with the measurement at baseline. Intermediate measurements are also recommended.\nStatistical analysis\nIntention-to-treat analysis should be the primary method of evaluation. Statistical significance will be inferred if a P value is equal to or less than 0.05. The beta risk will be equal to or less than 20%. The sample size of the study must be calculated prior to the start of the study. Possible confounding variables should be managed using appropriate statistical analysis. Within group (end vs. baseline) and between groups comparisons should be made.\nDiet habit and lifestyle\nThe control of critical effect modifiers such as physical activity, synergies with a multitude of other nutrients and the influence of nutrigenomic relationships must be taken into account.\nIntakes of other nutrients or foods, on which the tested nutrient is dependent, must be optimized. Any supplementation with other food products known to have an effect on bone (e.g., calcium and/or vitamin D) should be consistent within all patient groups.\nObservance\nObservance of food product intake should be monitored during the study to be able to perform pre-planned analyses on individuals with high and poor compliance rates or analyses of dose–response.\nSafety\nAll adverse experiences occurring during the course of clinical trials should be fully documented with separate analysis of adverse events, dropouts, and patients who died while being on the study.", "According to the European regulation, the use of nutrition and health claims shall only be permitted if the food product has been shown to have a beneficial nutritional or physiological effect in agreement with the health claim. However, it must also be pointed out that during the evaluation of the health claim, besides the characterization of the effect, important elements will be taken into account, such as the characterization of the food and the substantiation of the effect. In the field of bone health, claimed effects are not sufficiently defined and there are no standardized recommendations for the design and the methodology of clinical studies needed to reach such health claims. The consensus reached by the GREES is that the level of health claim may differ according to the surrogate endpoint used and on additional animal studies provided to support the claim. The ideal study design is a RCT but, is some particular cases, prospective cohort, case-control, or observational studies can be acceptable. In our opinion, general principles of the consensus reached are in line with the principles adopted in the EFSA's published opinions. This consensus is subject to future modifications when new validated surrogate markers will be available." ]

[ "introduction", "materials|methods", "results", null, null, null, "conclusion" ]

[ "Bone", "Health claim", "Nutrition", "Surrogate" ]

[ "Adult", "Aged", "Aged, 80 and over", "Antineoplastic Combined Chemotherapy Protocols", "Benzenesulfonates", "Carcinoma, Renal Cell", "Female", "Humans", "Interferon-alpha", "Kidney Neoplasms", "Male", "Middle Aged", "Niacinamide", "Phenylurea Compounds", "Prognosis", "Pyridines", "RNA, Messenger", "Receptor, Interferon alpha-beta", "Reverse Transcriptase Polymerase Chain Reaction", "Sorafenib" ]

[ "Patients, blood samples, and tissue specimens", "Real-time RT-PCR", "Western blotting", "Statistical analysis", "Serum IFNAR mRNA and characteristics of RCC", "Tumor expression of IFNAR mRNA and characteristics of RCC", "Tumor expression of phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) and characteristics of RCC", "Relationship between IFNAR2 mRNA, phosphorylated Akt (Ser-473), Akt, or phosphorylated S6 ribosomal protein (Ser-235/236) and the response to therapy", "Prognostic value of IFNAR2 mRNA, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236)", "Combination therapy with IFN-α and sorafenib", "Role of the IFNAR2 and mTOR pathways in progression of RCC", "Role of serum IFNAR2 in progression of RCC" ]

[ "We studied 66 consecutive Japanese patients (39 men and 27 women) aged 32–82 years (mean age: 62.9 years), who were newly diagnosed with clear cell RCC from 2008 to 2010. All patients routinely underwent imaging with CT and/or MRI for preoperative staging prior to radical nephrectomy. The postoperative follow-up period ranged from 2 to 35 months (median: 19 months). Surgery was done before the patients received any other therapy.\nA peripheral whole blood sample (10 ml) was collected from each patient and was diluted in PBS with 2 mM EDTA. Then, peripheral blood mononuclear cells (PBMC) were separated using Ficol-Hypaque medium (Biocoll, Berlin, Germany) and gradient centrifugation and stored at −80°C for future use. Total RNA was isolated with an RNeasy kit (Qiagen, Hamburg, Germany), including a DNA digestion step using the RNase-free DNase set.\nIn every patient, three different tumor tissue specimens and various parts of the non-neoplastic kidney were harvested for this study and stored at −80°C, as described previously [13–15]. The tumor grade and clinical stage were determined according to the Fuhrman grading system and the TNM classification, respectively [16, 17]. This study was conducted in accordance with the Helsinki Declaration, and approval of the Dokkyo Medical University Hospital institutional review board was obtained. In addition, each patient signed a consent form that was approved by the Committee on Human Rights in Research of our institution.\nPostoperative immunotherapy with IFN-α was given to 26 patients with extra-renal involvement. These patients received 3, 5, or 6 million units of natural human IFN-α intravenously or intramuscularly two or three times a week until tumor progression occurred. If the tumor was refractory to IFN-α monotherapy (progressive disease; PD), concomitant treatment with IFN-α and sorafenib (400 mg/day) was performed [18].\nThe doses of IFN-α and sorafenib were decreased if grade 3/4 toxicity occurred. Peripheral whole blood samples (10 ml) were obtained every 2 or 3 months from each metastatic RCC patient on treatment with IFN-α ± sorafenib. Tumor response was assessed according to RECIST criteria [19].", "Total RNA was purified from all 66 sets of serum, tumor tissue samples, and non-tumor tissue samples with an RNA preparation kit (“High Pure RNA Kit”; Roche Diagnostic Ltd., Germany) and was used as the template for cDNA synthesis. A 100-μl reaction mixture containing 1 mg of random hexamers and 100 units of MMLV reverse transcriptase was incubated at 25°C for 10 min, at 42°C for 30 min, and then at 99°C for 5 min. The IFNAR gene expression profile was analyzed with an ABI PRISM 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) using the SYBR Green method. The following primers were employed to amplify the β-actin gene after confirming their specificity: sense; 5′-CTGGCATCGTGATGGACTCCGG-3′; anti-sense, 5′-GTGGATGCCACAGGACTCCATG-3′. The primers used for IFNAR1 and IFNAR2 have been previously reported [14]. Real-time RT-PCR was performed in a 25-μl reaction mixture containing 20 ng of sample cDNA, 100 nM sense primer, 100 nM anti-sense primer, and 12.5 μl of SYBR Green PCR Master Mix (Applied Biosystems), with 45 cycles of 95°C for 15 s and 60°C for 1 min. A standard curve for each mRNA was generated using fivefold dilutions of a control RNA sample (25×, 5×, 1×, 0.2×, and 0.04×). The expression of each IFNAR mRNA was calculated as a ratio to that of β-actin in the serum, tumor tissue, and corresponding normal tissue samples to determine the relative level of expression [13, 14]. Individual variations of mRNA expression considered to be important [13]. To investigate the influence of such individual variations in the expression of IFNAR1 mRNA and IFNAR2 mRNA in the blood and tissues, we compared the expression of these mRNAs in serum and paired samples of tumor and non-tumor tissues. For quantification of mRNA expression, the relative amounts of IFNAR1 and IFNAR2 mRNAs in the tumors were calculated as a ratio of the optical density of the bands for the tumor specimens to the density of the bands for the corresponding normal tissue specimens (T/N ratio) by densitometric analysis, as described previously [13, 14]. The mean value for three tissue samples was used, as described previously [13, 14]. We also measured the serum C-reactive protein (CRP) level (normal < 0.3 mg/dl) in each patient.", "We could only perform Western blotting for 15 M1 tumors and four M0 tumors. Samples of tumor tissue and normal tissue were carefully dissected free of stromal tissue. Western blotting for phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) was carried out as described previously [15, 20, 21]. Briefly, 50 μg of cytosolic protein was separated by SDS–PAGE on 12.5% gel and electrotransferred to a polyvinylidene difluoride membrane (Immobilon-P membrane; Millipore, Bedford, MA). After the membrane was blocked, the bound proteins were probed with an anti-rabbit monoclonal antibody for phosphorylated Akt (Ser-473) and an anti-rabbit monoclonal antibody for Akt (Cell Signaling Technology, Inc; PhosphoPlus Akt (Ser-473) Antibody Kit; # 9270, Danvers, MA), as well as an anti-rabbit monoclonal antibody for phosphorylated S6 ribosomal protein (Ser-235/236) (2F9) (Cell Signaling Technology, Inc; # 4856) and a primary antibody for β-actin (Santa Cruz Biotechnology, Santa Cruz, CA). Hela cells were used as the positive control. Next, the membranes were washed and incubated with horseradish peroxidase-conjugated secondary antibodies. Bands of antibody-bound proteins were visualized by chemiluminescence; the blotted membrane was scanned for densitometry with a PDI imaging scanner (Agfa Japan, Tokyo), and the data were analyzed with NIH Image software. Expression of phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) was calculated relative to that of β-actin in the tumor tissue specimens and corresponding normal tissue specimens. For quantification of these proteins, the relative amount of phosphorylated Akt (Ser-473), Akt, or phosphorylated S6 ribosomal protein (Ser-235/236) in tumor tissue was expressed as a ratio of the optical density of the band for the tumor tissue specimen to that for the corresponding normal tissue specimen (set at 1.0) by densitometric analysis, as described previously [15, 20, 21]. The mean values for specimens of tumor and non-tumor tissue were calculated from three experiments [15, 21].", "Real-time RT-PCR and Western blotting data were analyzed by the Mann–Whitney U-test for comparisons between two groups (pT stage, metastasis, microscopic pathological vascular invasion, and serum CRP), while the Kruskal–Wallis test was used to compare three groups (histological grade and treatment effect) [13–15]. Spearman’s rank correlation coefficient analysis was performed to determine the relations between IFNAR2 mRNA expression and the serum CRP level or tumor size. The Kaplan–Meier method was used to estimate survival, and differences of survival were assessed by the log-rank test. In all analyses, a P value of less than 0.05 was considered significant. Data were analyzed with commercially available software.", "Although the preoperative serum IFNAR2 mRNA level was not related to the IFNAR2 mRNA level in tumor tissues (Fig. 1a) or the preoperative serum CRP level (Fig. 1b), there was a weak positive correlation between serum IFNAR2 and tumor size (r\n2 = 0.121, P = 0.0056, Fig. 1c).Fig. 1Spearman rank correlation coefficient relationship between mRNA expression levels of serum IFNAR2 mRNA and other factors. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels were associated with tumor size (c), but not with the IFNAR2 mRNAs in tumors (a) and serum CRP levels (b)\n\nSpearman rank correlation coefficient relationship between mRNA expression levels of serum IFNAR2 mRNA and other factors. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels were associated with tumor size (c), but not with the IFNAR2 mRNAs in tumors (a) and serum CRP levels (b)\nThe preoperative serum IFNAR2 mRNA level was not associated with the histological grade of RCC (mean ± S.D., grade 1, 1.38 ± 0.80; grade 2, 1.41 ± 0.71; grade 3, 1.36 ± 0.78, P = 0.9390, Fig. 2a), the pT stage (pT1–2, 1.35 ± 0.67; pT3–4, 1.47 ± 0.80, P = 0.7692, Fig. 2b), tumor metastasis (M0, 1.37 ± 0.74; M1, 1.49 ± 0.71, P = 0.5058, Fig. 2c), or microscopic vascular invasion (v(−), 1.36 ± 0.69; v(+), 1.55 ± 0.80, P = 0.3886, Fig. 2d).Fig. 2Serum levels for IFNAR2 mRNAs were not associated with tumor grade (a), pT stage (b), metastasis (c), and pathological microscopic vessel invasion (d). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled\nP values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nSerum levels for IFNAR2 mRNAs were not associated with tumor grade (a), pT stage (b), metastasis (c), and pathological microscopic vessel invasion (d). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled\nP values were obtained by comparing the three groups with the Kruskal–Wallis test\nThe preoperative serum level of IFNAR1 mRNA was not associated with the IFNAR1 mRNA level in tumor tissues, the preoperative serum CRP level, tumor size, histological grade, pT stage, tumor metastasis, or microscopic vascular invasion (data not shown).\nA higher preoperative serum CRP level was associated with local invasion (pT1–2, 0.24 ± 0.30; pT3–4, 3.50 ± 5.48, P < 0.0001), metastatic disease (M0, 0.39 ± 0.63; M1, 4.11 ± 6.09, P = 0.0002, and microscopic vascular invasion (v(−), 0.49 ± 0.73; v(+), 4.11 ± 6.41, P = 0.0217), but not with the histological grade of RCC (grade 1, 1.18 ± 3.03; grade 2, 0.83 ± 1.47; grade 3, 4.49 ± 7.74, P = 0.0771) (data not shown).", "The IFNAR2 mRNA level in tumor tissue specimens was correlated with the histological grade of RCC (grade 1, 0.82 ± 0.46; grade 2, 1.12 ± 0.86; grade 3, 1.97 ± 1.61, P = 0.0480, Fig. 3a), pT stage (pT1–2, 0.94 ± 0.59; pT3–4, 1.58 ± 1.37, P = 0.0111, Fig. 3b), tumor metastasis (M0, 0.89 ± 0.61; M1, 1.83 ± 1.39, P = 0.0009, Fig. 3c), microscopic vascular invasion (v(−), 0.97 ± 0.70; v(+), 1.54 ± 1.39, P = 0.0392, Fig. 3d), and serum CRP (normal, <0.30 mg/dl; high CRP, 1.63 ± 1.35; low CRP, 0.88 ± 0.61; P = 0.0154, Fig. 3e). In contrast, there was no relationship between IFNAR1 mRNA expression and the histological grade (grade 1, 1.07 ± 0.79; grade 2, 1.75 ± 3.79; grade 3, 0.90 ± 0.67, P = 0.7472), stage (pT1–2, 0.87 ± 0.58; pT3–4, 2.30 ± 4.61, P = 0.2439), tumor metastasis (M0, 0.74 ± 0.51; M1, 3.07 ± 5.18, P = 0.0735), microscopic vascular invasion (v(-), 1.59 ± 3.91; v(+), 1.54 ± 1.08, P = 0.9736), or serum CRP (normal, < 0.30 mg/dl; high CRP, 2.50 ± 4.77; low CRP, 0.91 ± 0.71; P = 0.2646).Fig. 3The IFNAR2 mRNA levels in tumors were associated with tumor grade (a), pT stage (b), metastasis (c), pathological microscopic vessel invasion (d), and serum CRP levels (e). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nThe IFNAR2 mRNA levels in tumors were associated with tumor grade (a), pT stage (b), metastasis (c), pathological microscopic vessel invasion (d), and serum CRP levels (e). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test", "We could only perform Western blotting for 15 M1 tumors and four M0 tumors. The levels of phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were significantly higher in M1 tumors than in M0 tumors (3.93 ± 3.13, vs. 0.89 ± 0.40, P = 0.0257 and 3.82 ± 2.28 vs. 0.96 ± 0.84, P = 0.0079, respectively, Fig. 4). In contrast, there was no difference of Akt expression between M1 tumors and M0 tumors (1.13 ± 0.26, vs. 1.48 ± 1.87, P = 0.3662).Fig. 4Expression of phosphorylated Akt (Ser-473) (60 kDa), Akt (60 kDa), phosphorylated S6 ribosomal protein (Ser-235/236) (32 kDa), and beta actin (42 kDa) proteins in the primary tumor tissues with metastatic lesions (M1) and without (M0) using Western blotting. N non-tumor tissue, T primary tumor tissue with metastatic lesions. Each number corresponds to a case number\n\nExpression of phosphorylated Akt (Ser-473) (60 kDa), Akt (60 kDa), phosphorylated S6 ribosomal protein (Ser-235/236) (32 kDa), and beta actin (42 kDa) proteins in the primary tumor tissues with metastatic lesions (M1) and without (M0) using Western blotting. N non-tumor tissue, T primary tumor tissue with metastatic lesions. Each number corresponds to a case number\nPreoperative serum level of IFNAR2 mRNA showed a weak negative correlation with tumor tissue levels of phosphorylated Akt (Ser-473) (r\n2 = 0.205, P = 0.0903, Fig. 5a), but not with tumor tissue levels of AKT (r\n2 = 0.017, P = 0.6587) or phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.072, P = 0.3327, Fig. 5b). On the other hand, tumor tissue levels of IFNAR2 mRNA were positively associated with the levels of phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.35, P = 0.0202, Fig. 5d) and also showed a weak positive correlation with phosphorylated Akt (Ser-473) (r\n2 = 0.199, P = 0.0959, Fig. 5c), but not with Akt (r\n2 = 0.001, P = 0.9154). Although there was a positive correlation between tumor tissue levels of phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.332, P = 0.0245, Fig. 5e), there was no relationship between either of these two phosphorylated proteins and Akt (data not shown).Fig. 5Spearman rank correlation coefficient relationship. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels inversely correlated with tumor phosphorylated Akt (Ser-473) (a), but not tumor phosphorylated S6 ribosomal protein (Ser-235/236) (b). Tumor IFNAR2 mRNA levels were associated with tumor phosphorylated S6 ribosomal protein (Ser-235/236) (c), and weak positive correlation with tumor phosphorylated Akt (Ser-473) (d). Tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were related each other (e)\n\nSpearman rank correlation coefficient relationship. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels inversely correlated with tumor phosphorylated Akt (Ser-473) (a), but not tumor phosphorylated S6 ribosomal protein (Ser-235/236) (b). Tumor IFNAR2 mRNA levels were associated with tumor phosphorylated S6 ribosomal protein (Ser-235/236) (c), and weak positive correlation with tumor phosphorylated Akt (Ser-473) (d). Tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were related each other (e)", "Kaplan–Meier survival plots for twenty-six metastatic RCCs treated with IFN-α monotherapy showed that the patients with a good response to this agent had better progression-free survival (Fig. 6a). Among the 21 patients refractory to IFN-α, IFN-α + Sor: CR-PR group had longer progression-free survival than IFN-α + Sor: SD-PD group (Fig. 6b). Thus, the patients with a good response to IFN-α with/without sorafenib had favorable overall survival (Fig. 6c).Fig. 6Survival curve in the patients with metastatic lesions (M1) based on the treatment effects. a Progression-free survival curve in M1 patients treated with IFN-α alone. One patient of IFN-α: CR-PR group progressed gradually into progressive disease (PD) after partial response (PR), but continued IFN-α monotherapy. Median survival time in IFN-α: SD-PD group is 16.7 months. Survival time of IFN-α: CR-PR group does not reach. b Progression-free survival curve in M1 patients after combination therapy with IFN-α plus sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 4.3 months. Survival time of IFN-α + Sor: CR-PR group does not reach. c Overall survival curve in M1 patients treated with IFN-α with/without sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 15.6 months. Survival time of IFN-α ± Sor: CR-PR group does not reach. P value was analyzed by log-rank test\n\nSurvival curve in the patients with metastatic lesions (M1) based on the treatment effects. a Progression-free survival curve in M1 patients treated with IFN-α alone. One patient of IFN-α: CR-PR group progressed gradually into progressive disease (PD) after partial response (PR), but continued IFN-α monotherapy. Median survival time in IFN-α: SD-PD group is 16.7 months. Survival time of IFN-α: CR-PR group does not reach. b Progression-free survival curve in M1 patients after combination therapy with IFN-α plus sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 4.3 months. Survival time of IFN-α + Sor: CR-PR group does not reach. c Overall survival curve in M1 patients treated with IFN-α with/without sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 15.6 months. Survival time of IFN-α ± Sor: CR-PR group does not reach. P value was analyzed by log-rank test\nTwenty-six patients with metastatic disease received IFN-α as first-line adjuvant therapy. If these patients showed a poor response to IFN-α monotherapy, they received concomitant treatment with IFN-α and sorafenib (Sor) as second-line therapy. Five of the 26 patients showed a complete or partial response to IFN-α alone (IFN-α: CR-PR), while the other 21 patients received concomitant IFN-α + sorafenib as second-line therapy. The five patients with a good response to IFN-α alone (IFN-α: CR-PR) were low risk according to the MSKCC criteria. There was no difference of the preoperative serum IFNAR2 level between the patients with a good response to IFN-α (IFN-α: CR-PR, 1.74 ± 0.53) and those with a poor response, including stable disease or progressive disease (IFN-α: SD-PD; 1.31 ± 0.59, P = 0.1263, Fig. 7a). Lower tumor expression of IFNAR2 mRNA was related to a good response (IFN-α: CR-PR, 0.76 ± 0.40; IFN-α: SD-PD, 1.89 ± 1.38, P = 0.0345, Fig. 7b).Fig. 7The relationship between treatment effect and mRNA levels of serum and tumor IFNAR2. The serum and tumor IFNAR2 mRNAs levels in the IFN-α monotherapy (a, b), and IFN-α ± sorafenib therapy (c–f). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nThe relationship between treatment effect and mRNA levels of serum and tumor IFNAR2. The serum and tumor IFNAR2 mRNAs levels in the IFN-α monotherapy (a, b), and IFN-α ± sorafenib therapy (c–f). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\nAmong the 21 patients treated with IFN-α + sorafenib, eleven patients showed a good response (IFN-α + Sor: CR-PR), while the other 10 patients had stable disease or progressive disease (IFN-α + Sor: SD-PD). The background characteristics and the outcomes of the patients receiving concomitant therapy with IFN-α + sorafenib are summarized in Table 1. A higher preoperative serum IFNAR2 level was correlated with a good outcome (IFN-α + Sor: CR-PR, 1.70 ± 0.39; IFN-α: CR-PR, 1.74 ± 0.53; IFN-α + Sor: SD-PD, 0.88 ± 0.47, P = 0.0013, Fig. 7e). In order to better assess the effect of combination therapy, we combined the IFN-α : CR-PR group and the IFN-α + Sor: CR-PR group into a single good response group (IFN-α ± Sor: CR-PR). The preoperative serum IFNAR2 level was significantly higher in this good response group than in the group with a poor response to IFN-α + sorafenib (INF-α ± Sor: CR-PR, 1.71 ± 0.42; IFN-α + Sor: SD-PD, 0.88 ± 0.47, P < 0.0001) (Fig. 7c). On the other hand, low tumor expression of IFNAR2 mRNA was related to a good response (IFN-α + Sor: CR-PR, 1.34 ± 1.09; IFN-α: CR-PR, 0.76 ± 0.40; IFN-α + Sor: SD-PD, 2.51 ± 1.46, P = 0.0064, Fig. 7f, IFN-α ± Sor: CR-PR, 1.16 ± 0.95; IFN-α + Sor: SD-PD, 2.51 ± 1.46, P = 0.0027, Fig. 7d).Table 1Background of 26 metastatic RCCs treated with IFN-α ± sorafenib (400 mg/day)OptionIFN-α aloneIFN-α + sorafenibCR-PR (n = 5)CR-PR (n = 11)SD-PD (n = 10)Sex (male/female) 19/7Years (median) 39–78 (63)ECOG PS* (0/1/2) 17/7/25/0/08/3/04/4/2MSKCC* (Fav/Int/Poor) 11/12/34/1/05/6/02/5/3Duration of IFN-α* (mean:months)3–29 (14.7)Duration of pre-IFN-α* (mean:months)1–27 (7.8)Duration of IFN-α + sorafenib (mean:months)1–23 (9.7)Metastatic lesions* (numbers) PUL  255173 PLE  2011 HEP  4013 OSS  5Radiation for bone lesions023 LYM  8044 Others  2002AE*: Grade 1/2 (numbers) HT014 Fatigue39 Alopecia013 Fever48 HFS06 Thrombocytopenia24 Elevation of amylase and/or lipase05 Others29AE*: Grade 3/4 (numbers) HT01 Fatigue11 Alopecia02 Fever03 Thrombocytopenia02 Elevation of ALT and AST02 Erythema multiforme 01\nECOG PS* eastern cooperative oncology group (ECOG) performance status, MSKCC* Memorial Sloen-Kettering Cancer Center, Fav favorable risk, Int intermediated risk, Duration of IFN-α* duration of IFN-α monotherapy, Duration of pre-IFN-α* duration of IFN-α monotherapy prior to IFN-α plus sorafenib, Metastatic lesions* PUL Lung, PLE Pleura, HEP Liver, OSS Bone, LYM lymph node, AE* adverse events, HT hypertension, HFS hand-foot syndrome\n\nBackground of 26 metastatic RCCs treated with IFN-α ± sorafenib (400 mg/day)\n\nECOG PS* eastern cooperative oncology group (ECOG) performance status, MSKCC* Memorial Sloen-Kettering Cancer Center, Fav favorable risk, Int intermediated risk, Duration of IFN-α* duration of IFN-α monotherapy, Duration of pre-IFN-α* duration of IFN-α monotherapy prior to IFN-α plus sorafenib, Metastatic lesions* PUL Lung, PLE Pleura, HEP Liver, OSS Bone, LYM lymph node, AE* adverse events, HT hypertension, HFS hand-foot syndrome\nInterestingly, chronological analysis of serum IFNAR2 mRNA levels showed that these levels remained relatively high in the IFN-α ± Sor: CR-PR group and remained lower in the IFN-α + Sor: SD-PD group (Fig. 8).Fig. 8Chronological analysis of serum IFNAR2 mRNA throughout IFN-α ± sorafenib treatment; IFN-α ± Sor: CR-PR group (n = 12), IFN-α + Sor: SD-PD group (n = 7)\n\nChronological analysis of serum IFNAR2 mRNA throughout IFN-α ± sorafenib treatment; IFN-α ± Sor: CR-PR group (n = 12), IFN-α + Sor: SD-PD group (n = 7)\nThe IFN-α + Sor: CR-PR group had lower phosphorylated Akt (Ser-473) levels than the IFN-α + Sor: SD-PD group (1.65 ± 1.08 vs. 5.54 ± 3.22, P = 0.0208, Fig. 9a). Lower expression of phosphorylated S6 ribosomal protein (Ser-235/236) tended to be associated with a better response (IFN-α ± Sor: CR-PR, 2.50 ± 1.09; IFN-α + Sor: SD-PD, 4.72 ± 2.59, P = 0.1052, Fig. 9b). On the other hand, expression of Akt was similar in both the IFN-α + Sor: CR-PR group and the IFN-α + Sor: SD-PD group (1.11 ± 0.19 vs. 1.16 ± 0.34, P = 0.9078).Fig. 9The relationship between treatment effect and protein expression levels of Akt, phosphorylated Akt (Ser-473) (a) and phosphorylated S6 ribosomal protein (Ser-235/236) (b). The relative expression levels of targeted proteins in the primary tumor to those in corresponding non-tumor portion, which was set to 1.0. Hela cell was used as the positive control. The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers\n\nThe relationship between treatment effect and protein expression levels of Akt, phosphorylated Akt (Ser-473) (a) and phosphorylated S6 ribosomal protein (Ser-235/236) (b). The relative expression levels of targeted proteins in the primary tumor to those in corresponding non-tumor portion, which was set to 1.0. Hela cell was used as the positive control. The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers\nUnlike IFNAR2 mRNA, serum and tissue levels of IFNAR1 mRNA were unrelated to the response to treatment (data not shown).", "The mean serum level of IFNAR2 mRNA and mean tumor tissue levels of IFNAR2 mRNA, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236) in M1 patients treated by IFN-α ± sorafenib were 1.39 (±0.60), 1.67 (±1.33), 3.72 (±3.12), and 3.63 (±2.24), respectively. Patients were divided into two groups based on these mean values (i.e., a high group and a low group), as described previously [13–15, 21]. Kaplan–Meier survival plots for patients with low versus high levels of these possible prognostic factors showed that a lower serum level of IFNAR2 mRNA (P < 0.05), a higher tumor tissue levels of IFNAR2 mRNA (P < 0.01), phosphorylated Akt (Ser-473) (P < 0.05), and phosphorylated S6 ribosomal protein (Ser-235/236) (P < 0.05) were correlated with shorter overall survival (OS) (Fig. 10a–d).Fig. 10Survival curve in the patients with metastatic lesions (M1) based on the mean values of mRNA levels of serum and tumor lFNAR2 s, and of protein levels of tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236), the cases were divided into two groups at this levels—high and low expression. Progression-free survival curve based on serum (a) and tumor IFNAR2 mRNA levels (b) in M1 patients treated with IFN-α ± sorafenib therapy. Progression-free survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (c) and phosphorylated S6 ribosomal protein (Ser-235/236) (d) in M1 patients treated with IFN-α + sorafenib therapy. Overall survival curve based on serum (e) and tumor IFNAR2 mRNA levels (f) in M1 patients treated with IFN-α ± sorafenib therapy. Overall survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (g) and phosphorylated S6 ribosomal protein (Ser-235/236) (h) in M1 patients treated with IFN-α + sorafenib therapy. P value was analyzed by log-rank test\n\nSurvival curve in the patients with metastatic lesions (M1) based on the mean values of mRNA levels of serum and tumor lFNAR2 s, and of protein levels of tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236), the cases were divided into two groups at this levels—high and low expression. Progression-free survival curve based on serum (a) and tumor IFNAR2 mRNA levels (b) in M1 patients treated with IFN-α ± sorafenib therapy. Progression-free survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (c) and phosphorylated S6 ribosomal protein (Ser-235/236) (d) in M1 patients treated with IFN-α + sorafenib therapy. Overall survival curve based on serum (e) and tumor IFNAR2 mRNA levels (f) in M1 patients treated with IFN-α ± sorafenib therapy. Overall survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (g) and phosphorylated S6 ribosomal protein (Ser-235/236) (h) in M1 patients treated with IFN-α + sorafenib therapy. P value was analyzed by log-rank test", "While the efficacy of combination therapy with IFN-α and sorafenib needs clarification, the adverse effects of this regimen are related to the doses of each agent and are not additive [22–25]. We previously reported a good response to combination therapy with IFN-α and half the usual dose of sorafenib (400 mg/day rather than 800 mg/day) in patients with IFN-α-resistant RCC, along with tolerable adverse events [18].\nIt has also been reported that combined treatment with IFN-α + sorafenib suppresses proliferation and vascular endothelial growth factor (VEGF) production by several RCC cell lines more strongly than either agent alone [26, 27]. RCC is considered to be an immunogenic tumor [3], since cytotoxic T lymphocytes recognize and selectively kill autologous RCC cells, while tumor-specific T cells can be detected in the blood of RCC patients [3]. Sorafenib is a multikinase inhibitor targeting VEGF receptors 1–3, PDGFβ receptor, and Raf kinase, and it has both direct antitumor activity and antiangiogenic activity [28]. In a randomized phase III trial comparing sorafenib with placebo as second-line therapy for RCC, the response rate to sorafenib was 10%, and the stable disease rate was 74%, with the median progression-free survival time being 5.5 months in the sorafenib group versus 2.8 months in the placebo group [29]. In addition to its immunomodulatory effects, IFN-α also has direct antitumor activity as well as antiangiogenic activity, including inhibition of VEGF [30]. Antitumor immunity is usually suppressed in tumor-bearing mice because of the influence of regulatory T cells and suppressive cytokines, such as TGF-β and IL-10 [31]. Takeuchi et al. recently reported that the synergistic effect of sorafenib and IFN-α on RCC both in vitro and in tumor-bearing mice was related to a combination of antitumor, antiangiogenic, and immunologic responses [27]. In their study, sorafenib had no effect on the levels of natural killer (NK) cells, T cells, and regulatory T cells in the spleens of tumor-bearing BALB/c mice, irrespective of the use of IFN-α, while IFN-α showed weaker direct antitumor activity than sorafenib but stimulated CTL, NK cells, and tumor-infiltrating lymphocytes (which sorafenib did not), so that a synergistic antiproliferative effect of these two agents was demonstrated in vitro [27].\nAlthough an additive effect of IFN-α to sorafenib therapy in patients with metastatic RCC was recently reported [24], the clinical efficacy of this combination remains to be confirmed. Jonasch et al. reported that the outcome after addition of IFN-α to sorafenib was comparable to that of sorafenib monotherapy, when patients were randomized to treatment with either sorafenib (400 mg twice daily) or the combination of sorafenib (400 mg twice daily) plus IFN-α (0.5 MU twice daily) [25]. However, their IFN-α dose of 1 million units daily was probably too low to assess its additive effect because the average dose is 3–9 million units daily. So far, the published studies on combination therapy with IFN-α and sorafenib have employed concomitant therapy with both agents. In contrast, our treatment strategy was to add sorafenib (400 mg/day) to IFN-α in RCC patients whose tumors were refractory to IFN-α alone, i.e., first-line IFN-α monotherapy and second-line combination therapy with IFN-α plus sorafenib [18]. In the present study, 26 patients had metastatic disease (M1) at diagnosis. Among them, five patients showed a good response to IFN-α alone (IFN-α: CR-PR), 11 patients showed a good response to IFN-α + sorafenib (IFN-α + Sor: CR-PR, Table 1), and the remaining 10 patients had stable disease or showed a poor response to IFN-α + sorafenib (IFN-α + Sor: SD-PD, Table 1). Although half-dose sorafenib (400 mg/day) caused grade 1/2 toxicity, grade 3/4 toxicity was rare in the present study and such toxicity resolved when patients suspended sorafenib therapy. The response rate to the combination of IFN-α plus half-dose sorafenib was 52.4% (11/21 patients resistant to previous IFN-α monotherapy achieved CR or PR), indicating that sorafenib may be synergistic with IFN-α, leading to an increase of antitumor activity. Furthermore, the patients with good response to this combination therapy had favorable prognosis (Fig. 6). Thus, this combination seems to be tolerable and could be a useful treatment option for advanced RCC resistant to IFN-α monotherapy [18].", "The preoperative serum level of IFNAR2 mRNA was not correlated with the effect of IFN-α monotherapy, but a lower tumor tissue level of IFNAR2 mRNA was related to a better response to IFN-α. The present finding that a higher tumor tissue level of IFNAR2 mRNA was associated with a poor response to IFN-α is consistent with our previous results [14]. In contrast, regarding the relationship between IFNAR2 mRNA levels and the effect of combination therapy with IFN-α plus sorafenib, the preoperative serum level of IFNAR2 mRNA was higher in the IFN-α + Sor: CR-PR group and the IFN-α: CR-PR group than in the IFN-α + Sor: SD-PD group (Fig. 6e), while the tumor tissue level of IFNAR2 mRNA was lower in the IFN-α + Sor: CR-PR group and the IFN-α: CR-PR group than in the IFN-α + Sor: SD-PD group (Fig. 6f). These findings suggested that a different molecular mechanism might be involved in IFN-α ± Sor: CR-PR group and IFN-α + Sor: SD-PD group.\nIFNs are pleiotropic cytokines that regulate antiviral, antitumor, apoptotic, antiangiogenic, and cellular immune responses via activation of multiple downstream signaling cascades, including the Janus tyrosine kinase (Jak)-signal transducer and activator of transcription (STAT) pathway, the p38 mitogen-activated protein kinase (MAPK) pathway, and the mTOR pathway [32, 33]. The Jak-STAT and p38-MAPK signaling pathways have been shown to be responsible for transcription of the genes encoding proteins related to the antiviral and/or antiproliferative effects of IFNs. On the other hand, it has been reported that activation of the mTOR pathway by IFNs has an important regulatory role in mRNA translation and induction of the interferon response [34, 35]. IFN-α-induced tumor cell apoptosis is also mediated via the mTOR pathway in a nucleus-independent manner [36, 37]. Moreover, the Jak-STAT and mTOR pathways act separately from each other after activation by IFN-α [37]. Thus, it is likely that mTOR signaling selectively mediates apoptosis and survival.\nThe rapamycin-sensitive mTOR-raptor (regulatory-associated protein of mTOR) complex controls cell growth by regulating protein synthesis, so mTOR-raptor signaling is a potential antitumor target, and mTOR inhibitors are currently under investigation for the treatment of various human cancers. On the other hand, mTOR also interacts with rictor (rapamycin-insensitive companion of mTOR), and recent findings have suggested that the rapamycin-insensitive effect of mTOR on cell survival is overactive in many cancers. Thus, mTOR has dual rapamycin-sensitive (mTOR-raptor complex: mTORC1) and insensitive (mTOR-rictor complex: mTORC2) functions, indicating that treatment with rapamycin will not completely inhibit mTOR activity [38, 39]. Phosphatidylinositol 3`kinase (PI3 K), serine/threonine kinase Akt, and the mTOR pathway are all overactive in human cancers. mTORC1 lies downstream of PI3 K and is part of a pathway that is frequently activated in human cancers, so mTORC1 represents a pivotal target for anticancer therapy. The best-characterized pathways regulated by mTORC1 are phosphorylation and activation of ribosomal S6 kinase-1 (S6K1) and phosphorylation and inactivation of 4E-BP1, the suppressor of mRNA cap-binding protein eIF4E, leading to effects on cell growth and metabolism by acting as a restriction point in cells subjected to stresses [40, 41] such as hypoxia [42–44].\nPhosphorylation at two sites is required for full activation of Akt, since it is phosphorylated by PI3 K-dependent kinase-1 (PDK1) at a threonine residue in the catalytic domain (Thr 308) and by PI3 K-dependent kinase-2 (PDK2) at a serine residue (Ser 473) in the carboxy-terminal hydrophobic motif [45]. It has been reported that mTORC2 regulates the actin cytoskeleton and also possesses PDK2 activity that phosphorylates Ser-473 in the carboxy-terminal of Akt, making it essential for Akt activity [46]. Importantly, activation of Akt may lead to cell survival when mTORC1 is inhibited or could potentially increase VEGF production because PI3 K/Akt signaling induces tumor angiogenesis by regulating VEGF via both HIF-1α-dependent and -independent mechanisms [47]. It has been reported that hypoxia-inducible factor (HIF) 1α expression is dependent on both raptor and rictor, whereas HIF2α expression only depends on rictor and HIF2α is more important in RCC [48]. These findings suggest that phosphorylation of Ser 473 in AKT is a key molecular step in the progression of RCC and could be a target for treating these tumors [25]. In agreement with this, our current study showed that the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated and that both were related to tumor metastatic potential, while a high phosphorylated Akt (Ser-473) level in the primary tumor was associated with a poor response to IFN-α ± sorafenib therapy (stable or progressive disease).\nCurrent efforts to achieve the clinical development of mTOR inhibitors are based on the role of mTOR signaling in promoting the proliferation and survival of tumor cells. It has been reported that treatment with mTOR inhibitors can improve the outcome of patients with metastatic RCC [8–10]. On the other hand, the mTOR pathway is also important for IFN-dependent translational responses, and IFN-α is widely used to treat advanced RCC. Although we could not exclude a possible detrimental effect of IFN-α treatment in the patients with metastatic RCC and higher IFNAR2 mRNA levels in their tumors, our findings suggested that there may be different molecular mechanisms of cancer progression in the patients with a good or poor response to IFN-α ± sorafenib. It is possible that IFNAR2 signaling has different biological effects from normal when upregulated in RCC.\nJonasch et al. recently reported that an increase of phosphorylated Akt (Ser-473) was associated with worse survival by microarray analysis of paraffin-embedded specimens [25]. In the present study, the tumors with higher phosphorylated Akt (Ser-473) levels, but not higher phospho-S6 ribosomal protein (Ser-235/236) levels, were resistant to IFN-α ± sorafenib therapy. In addition, the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated, and both were related to metastatic potential. Tumor levels of IFNAR2 mRNA also had a weak positive correlation with those of phosphorylated Akt (Ser-473). Moreover, the patients whose tumors had higher levels of phosphorylated Akt (Ser-473), phosphorylated S6 ribosomal protein (Ser-235/236), and IFNAR2 mRNA showed shorter overall survival. Taken together, it is possible that phosphorylation of Ser 473 on Akt is a key molecular step in the progression of RCC and a potential therapeutic target, so that the tumor level of phosphorylated Akt (Ser-473) may be useful for predicting the response to treatment. At present, it remains to be elucidated why upregulation of IFNAR2 expression is linked to the progression of RCC and to a poor response to treatment, and it is unclear how IFNAR2 interacts with mTORC1 and mTORC2, but our findings suggested that the IFNAR2-mTORC1 pathway via phosphorylated S6 ribosomal protein (Ser-235/236) may act locally within tumors to promote proliferation and metastasis by modifying mRNA translation, while the IFNAR2-mTORC2 pathway via phosphorylated Akt (Ser-473) may be associated with tumor resistance. So, these interactions should be elucidated in the future. As Lekmine et al. have indicated, therefore, caution should be exercised when designing clinical trials that combine an mTOR inhibitor and IFN-α due to possible antagonism of antitumor activity [34]. In fact, Huges et al. reported that patients treated with temsirolimus alone had better overall survival than those given IFN-α alone, while patients treated with temsirolimus plus IFN-α did not [49]. In the future, the downstream targets of IFNAR2 should be identified, and the expression or activity of one or two such targets should be studied in cell lines or tissue samples. A better understanding of the IFNAR2 pathway may help to elucidate its role in cancer.\nAlthough none of our patients were treated by sorafenib alone, it would be interesting to assess the expression of not only IFNAR2, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236), but also VEGF receptor and Raf, in tumor cells and the effects of sorafenib, sunitinib, or mTOR inhibitors. Such information could lead to elucidation of the role of the IFNAR2-mTOR pathway in the progression of RCC and the selection of patients who will benefit from treatment with IFN-α, sorafenib, sunitinib, or mTOR inhibitors.", "The serum CRP level is associated with the stage and outcome of RCC [50, 51]. Elevation of CRP is primarily determined by an increase of circulating IL-6 [52], and the IL-6 level is correlated with the serum CRP level as well as with tumor histological grade and tumor metastasis [53]. We previously reported that increased serum levels of CRP and IL-6 were associated with local tumor invasion and metastasis [54]. In the present study, a higher preoperative serum CRP level was associated with local invasion and metastasis of RCC, but not with the response to treatment (data not shown). These findings suggest that the serum CRP level is associated with tumor aggressiveness, so that elevation of CRP might reflect the poorer general condition of the patient rather than the response to therapy. On the other hand, IFN-α has immunomodulatory effects and direct antitumor activity as well as antiangiogenic activity, including inhibition of VEGF [30]. Therefore, it is unclear exactly what serum IFNAR2 reflects, but it is likely to be associated with the overall immune status. Accordingly, we expected to find a relation between the preoperative serum levels of IFNAR2 mRNA and CRP, but no relation was observed.\nAlthough we could not examine how IFN-α binds to IFNAR2 on peripheral blood cells and then exhibits antitumor and antiangiogenic activity, our findings showed that the preoperative serum level of IFNAR2 mRNA was positively correlated with tumor size and was higher in patients with metastatic RCC who showed a good response to IFN-α ± sorafenib therapy than in those with a poor response. Because obtaining blood samples from patients is easier than harvesting tissue samples, chronological analysis of serum IFNAR2 mRNA levels is preferable for evaluation of the role of IFNAR2. While the effect of IFN-α therapy on the serum level of IFNAR2 mRNA is still unclear, our chronological evaluation of IFNAR2 mRNA throughout treatment showed that its serum level remained higher in the IFN-α ± Sor: CR-PR group than in the IFN-α + Sor: SD-PD group. Taken together, these observations suggest that an increased serum level of IFNAR2 mRNA may represent a systemic immunologic and antitumor response to the tumor burden in RCC patients, as well as showing antimicrobial activity if infection occurs.\nRegarding the effect of genetic polymorphism on the response of metastatic RCC to IFN-α, it has been reported that STAT3 polymorphism is a useful diagnostic marker for predicting the response to IFN-α therapy in these patients [55]. An efficient marker of the response to IFN-α is needed to establish individualized optimal treatment strategies, especially when newer therapies are used as first-line treatment for metastatic RCC. Our study showed that patients with higher serum levels of IFNAR2 mRNA may be more likely to respond to IFN-α ± sorafenib therapy and to show a longer overall survival, while patients with higher tumor tissue levels of IFNAR2 mRNA may show poor response and unfavorable overall survival. Although we need a surgical specimen to examine tumor tissue levels of IFNAR2 mRNA and protein, the serum level of IFNAR2 mRNA can be conveniently measured, so it may be more useful for predicting the response to IFN-α ± sorafenib therapy and as a prognostic indicator." ]

[ null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Introduction", "Materials and methods", "Patients, blood samples, and tissue specimens", "Real-time RT-PCR", "Western blotting", "Statistical analysis", "Results", "Serum IFNAR mRNA and characteristics of RCC", "Tumor expression of IFNAR mRNA and characteristics of RCC", "Tumor expression of phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) and characteristics of RCC", "Relationship between IFNAR2 mRNA, phosphorylated Akt (Ser-473), Akt, or phosphorylated S6 ribosomal protein (Ser-235/236) and the response to therapy", "Prognostic value of IFNAR2 mRNA, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236)", "Discussion", "Combination therapy with IFN-α and sorafenib", "Role of the IFNAR2 and mTOR pathways in progression of RCC", "Role of serum IFNAR2 in progression of RCC", "Conclusion" ]

[ "Localized renal cell carcinoma (RCC) is generally considered to be a surgical disease. However, almost 30% of the patients who present with limited disease and undergo surgery develop metastasis during the next 3 years [1]. The incidence of RCC is steadily increasing, and it accounts for 2–3% of all adult malignancies. Around 40% of RCC patients die of metastasis, because of the high frequency of metastasis at diagnosis and relapse following nephrectomy [2]. Patients with distant metastases have a very poor prognosis, and their 5-year survival rate is less than 10% [2]. Clear cell RCC is considered to be an immunogenic tumor [3], with immunocytokine therapy being the mainstay of treatment [4], while it is notoriously resistant to chemotherapy and radiotherapy [2]. Targeting novel pathways associated with the evolution of malignancy may lead to an improved outcome for patients with RCC. Novel treatment options include immunotherapy [5], monoclonal antibodies [6], anti-angiogenesis therapy [7], and inhibitors of mammalian target of rapamycin (mTOR) [8–10].\nJapanese patients who receive immunocytokine therapy including interferon alpha (IFN-α) show a better response and survival compared with American or European patients [11]. With the advent of molecular-targeting therapy, the use of sorafenib and sunitinib for RCC has been covered by the Japanese national health insurance system since 2008. Recently, Motzer et al. [12] identified five prognostic factors for RCC, which are known as the Memorial Slone-Kettering Cancer Center (MSKCC) classification (Karnofsky performance status, time from diagnosis of RCC to treatment or recurrence, serum lactate dehydrogenase, corrected serum calcium, and hemoglobin). The MSKCC classification is correlated with the overall survival of patients with metastatic RCC receiving IFN-α as initial systemic therapy. It has the potential to be employed for predicting the efficacy of immunocytokine therapy, and it may also be useful for predicting the response to molecular-targeting therapy. Recently, a number of molecular markers have been investigated in RCC patients to assess both their predictive value and their potential as therapeutic targets. Identification of new targets may lead to an improvement in the outcome of RCC, but no biomarkers have been established so far.\nIndividual variations of mRNA expression can have an important influence [13]. We previously reported that increased expression of IFNAR2 mRNA in tumor tissue is associated with the metastatic potential and progression of RCC and with a poor response to IFN-α therapy [14]. Our previous study also showed that a 102-KDa IFNAR2c protein, a functional domain of IFNAR2, is more highly expressed in the primary tumors compared with the non-tumor tissues of patients with metastatic RCC, but not localized RCC, indicating that the main component of IFNAR2 is the long form of IFNAR2c and that this protein may be important in the progression of RCC [14]. However, the influence of circulating (serum) IFNAR2 mRNA in patients with RCC has not been elucidated. In the present study, we examined serum levels of IFNAR2 mRNA, as well as the IFNAR2 mRNA levels in corresponding tumor tissue and non-tumor tissue samples, by the real-time reverse transcription polymerase chain reaction (RT-PCR). The relations among serum and tumor tissue levels of IFNAR2 mRNA and various clinicopathologic features of RCC patients were also examined. Furthermore, we investigated whether IFNAR2 mRNA could be used to predict the response of metastatic RCC to IFN-α ± sorafenib therapy. It was hoped that the information obtained might contribute to elucidation of the role of IFNAR2 in RCC.", "[SUBTITLE] Patients, blood samples, and tissue specimens [SUBSECTION] We studied 66 consecutive Japanese patients (39 men and 27 women) aged 32–82 years (mean age: 62.9 years), who were newly diagnosed with clear cell RCC from 2008 to 2010. All patients routinely underwent imaging with CT and/or MRI for preoperative staging prior to radical nephrectomy. The postoperative follow-up period ranged from 2 to 35 months (median: 19 months). Surgery was done before the patients received any other therapy.\nA peripheral whole blood sample (10 ml) was collected from each patient and was diluted in PBS with 2 mM EDTA. Then, peripheral blood mononuclear cells (PBMC) were separated using Ficol-Hypaque medium (Biocoll, Berlin, Germany) and gradient centrifugation and stored at −80°C for future use. Total RNA was isolated with an RNeasy kit (Qiagen, Hamburg, Germany), including a DNA digestion step using the RNase-free DNase set.\nIn every patient, three different tumor tissue specimens and various parts of the non-neoplastic kidney were harvested for this study and stored at −80°C, as described previously [13–15]. The tumor grade and clinical stage were determined according to the Fuhrman grading system and the TNM classification, respectively [16, 17]. This study was conducted in accordance with the Helsinki Declaration, and approval of the Dokkyo Medical University Hospital institutional review board was obtained. In addition, each patient signed a consent form that was approved by the Committee on Human Rights in Research of our institution.\nPostoperative immunotherapy with IFN-α was given to 26 patients with extra-renal involvement. These patients received 3, 5, or 6 million units of natural human IFN-α intravenously or intramuscularly two or three times a week until tumor progression occurred. If the tumor was refractory to IFN-α monotherapy (progressive disease; PD), concomitant treatment with IFN-α and sorafenib (400 mg/day) was performed [18].\nThe doses of IFN-α and sorafenib were decreased if grade 3/4 toxicity occurred. Peripheral whole blood samples (10 ml) were obtained every 2 or 3 months from each metastatic RCC patient on treatment with IFN-α ± sorafenib. Tumor response was assessed according to RECIST criteria [19].\nWe studied 66 consecutive Japanese patients (39 men and 27 women) aged 32–82 years (mean age: 62.9 years), who were newly diagnosed with clear cell RCC from 2008 to 2010. All patients routinely underwent imaging with CT and/or MRI for preoperative staging prior to radical nephrectomy. The postoperative follow-up period ranged from 2 to 35 months (median: 19 months). Surgery was done before the patients received any other therapy.\nA peripheral whole blood sample (10 ml) was collected from each patient and was diluted in PBS with 2 mM EDTA. Then, peripheral blood mononuclear cells (PBMC) were separated using Ficol-Hypaque medium (Biocoll, Berlin, Germany) and gradient centrifugation and stored at −80°C for future use. Total RNA was isolated with an RNeasy kit (Qiagen, Hamburg, Germany), including a DNA digestion step using the RNase-free DNase set.\nIn every patient, three different tumor tissue specimens and various parts of the non-neoplastic kidney were harvested for this study and stored at −80°C, as described previously [13–15]. The tumor grade and clinical stage were determined according to the Fuhrman grading system and the TNM classification, respectively [16, 17]. This study was conducted in accordance with the Helsinki Declaration, and approval of the Dokkyo Medical University Hospital institutional review board was obtained. In addition, each patient signed a consent form that was approved by the Committee on Human Rights in Research of our institution.\nPostoperative immunotherapy with IFN-α was given to 26 patients with extra-renal involvement. These patients received 3, 5, or 6 million units of natural human IFN-α intravenously or intramuscularly two or three times a week until tumor progression occurred. If the tumor was refractory to IFN-α monotherapy (progressive disease; PD), concomitant treatment with IFN-α and sorafenib (400 mg/day) was performed [18].\nThe doses of IFN-α and sorafenib were decreased if grade 3/4 toxicity occurred. Peripheral whole blood samples (10 ml) were obtained every 2 or 3 months from each metastatic RCC patient on treatment with IFN-α ± sorafenib. Tumor response was assessed according to RECIST criteria [19].\n[SUBTITLE] Real-time RT-PCR [SUBSECTION] Total RNA was purified from all 66 sets of serum, tumor tissue samples, and non-tumor tissue samples with an RNA preparation kit (“High Pure RNA Kit”; Roche Diagnostic Ltd., Germany) and was used as the template for cDNA synthesis. A 100-μl reaction mixture containing 1 mg of random hexamers and 100 units of MMLV reverse transcriptase was incubated at 25°C for 10 min, at 42°C for 30 min, and then at 99°C for 5 min. The IFNAR gene expression profile was analyzed with an ABI PRISM 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) using the SYBR Green method. The following primers were employed to amplify the β-actin gene after confirming their specificity: sense; 5′-CTGGCATCGTGATGGACTCCGG-3′; anti-sense, 5′-GTGGATGCCACAGGACTCCATG-3′. The primers used for IFNAR1 and IFNAR2 have been previously reported [14]. Real-time RT-PCR was performed in a 25-μl reaction mixture containing 20 ng of sample cDNA, 100 nM sense primer, 100 nM anti-sense primer, and 12.5 μl of SYBR Green PCR Master Mix (Applied Biosystems), with 45 cycles of 95°C for 15 s and 60°C for 1 min. A standard curve for each mRNA was generated using fivefold dilutions of a control RNA sample (25×, 5×, 1×, 0.2×, and 0.04×). The expression of each IFNAR mRNA was calculated as a ratio to that of β-actin in the serum, tumor tissue, and corresponding normal tissue samples to determine the relative level of expression [13, 14]. Individual variations of mRNA expression considered to be important [13]. To investigate the influence of such individual variations in the expression of IFNAR1 mRNA and IFNAR2 mRNA in the blood and tissues, we compared the expression of these mRNAs in serum and paired samples of tumor and non-tumor tissues. For quantification of mRNA expression, the relative amounts of IFNAR1 and IFNAR2 mRNAs in the tumors were calculated as a ratio of the optical density of the bands for the tumor specimens to the density of the bands for the corresponding normal tissue specimens (T/N ratio) by densitometric analysis, as described previously [13, 14]. The mean value for three tissue samples was used, as described previously [13, 14]. We also measured the serum C-reactive protein (CRP) level (normal < 0.3 mg/dl) in each patient.\nTotal RNA was purified from all 66 sets of serum, tumor tissue samples, and non-tumor tissue samples with an RNA preparation kit (“High Pure RNA Kit”; Roche Diagnostic Ltd., Germany) and was used as the template for cDNA synthesis. A 100-μl reaction mixture containing 1 mg of random hexamers and 100 units of MMLV reverse transcriptase was incubated at 25°C for 10 min, at 42°C for 30 min, and then at 99°C for 5 min. The IFNAR gene expression profile was analyzed with an ABI PRISM 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) using the SYBR Green method. The following primers were employed to amplify the β-actin gene after confirming their specificity: sense; 5′-CTGGCATCGTGATGGACTCCGG-3′; anti-sense, 5′-GTGGATGCCACAGGACTCCATG-3′. The primers used for IFNAR1 and IFNAR2 have been previously reported [14]. Real-time RT-PCR was performed in a 25-μl reaction mixture containing 20 ng of sample cDNA, 100 nM sense primer, 100 nM anti-sense primer, and 12.5 μl of SYBR Green PCR Master Mix (Applied Biosystems), with 45 cycles of 95°C for 15 s and 60°C for 1 min. A standard curve for each mRNA was generated using fivefold dilutions of a control RNA sample (25×, 5×, 1×, 0.2×, and 0.04×). The expression of each IFNAR mRNA was calculated as a ratio to that of β-actin in the serum, tumor tissue, and corresponding normal tissue samples to determine the relative level of expression [13, 14]. Individual variations of mRNA expression considered to be important [13]. To investigate the influence of such individual variations in the expression of IFNAR1 mRNA and IFNAR2 mRNA in the blood and tissues, we compared the expression of these mRNAs in serum and paired samples of tumor and non-tumor tissues. For quantification of mRNA expression, the relative amounts of IFNAR1 and IFNAR2 mRNAs in the tumors were calculated as a ratio of the optical density of the bands for the tumor specimens to the density of the bands for the corresponding normal tissue specimens (T/N ratio) by densitometric analysis, as described previously [13, 14]. The mean value for three tissue samples was used, as described previously [13, 14]. We also measured the serum C-reactive protein (CRP) level (normal < 0.3 mg/dl) in each patient.\n[SUBTITLE] Western blotting [SUBSECTION] We could only perform Western blotting for 15 M1 tumors and four M0 tumors. Samples of tumor tissue and normal tissue were carefully dissected free of stromal tissue. Western blotting for phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) was carried out as described previously [15, 20, 21]. Briefly, 50 μg of cytosolic protein was separated by SDS–PAGE on 12.5% gel and electrotransferred to a polyvinylidene difluoride membrane (Immobilon-P membrane; Millipore, Bedford, MA). After the membrane was blocked, the bound proteins were probed with an anti-rabbit monoclonal antibody for phosphorylated Akt (Ser-473) and an anti-rabbit monoclonal antibody for Akt (Cell Signaling Technology, Inc; PhosphoPlus Akt (Ser-473) Antibody Kit; # 9270, Danvers, MA), as well as an anti-rabbit monoclonal antibody for phosphorylated S6 ribosomal protein (Ser-235/236) (2F9) (Cell Signaling Technology, Inc; # 4856) and a primary antibody for β-actin (Santa Cruz Biotechnology, Santa Cruz, CA). Hela cells were used as the positive control. Next, the membranes were washed and incubated with horseradish peroxidase-conjugated secondary antibodies. Bands of antibody-bound proteins were visualized by chemiluminescence; the blotted membrane was scanned for densitometry with a PDI imaging scanner (Agfa Japan, Tokyo), and the data were analyzed with NIH Image software. Expression of phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) was calculated relative to that of β-actin in the tumor tissue specimens and corresponding normal tissue specimens. For quantification of these proteins, the relative amount of phosphorylated Akt (Ser-473), Akt, or phosphorylated S6 ribosomal protein (Ser-235/236) in tumor tissue was expressed as a ratio of the optical density of the band for the tumor tissue specimen to that for the corresponding normal tissue specimen (set at 1.0) by densitometric analysis, as described previously [15, 20, 21]. The mean values for specimens of tumor and non-tumor tissue were calculated from three experiments [15, 21].\nWe could only perform Western blotting for 15 M1 tumors and four M0 tumors. Samples of tumor tissue and normal tissue were carefully dissected free of stromal tissue. Western blotting for phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) was carried out as described previously [15, 20, 21]. Briefly, 50 μg of cytosolic protein was separated by SDS–PAGE on 12.5% gel and electrotransferred to a polyvinylidene difluoride membrane (Immobilon-P membrane; Millipore, Bedford, MA). After the membrane was blocked, the bound proteins were probed with an anti-rabbit monoclonal antibody for phosphorylated Akt (Ser-473) and an anti-rabbit monoclonal antibody for Akt (Cell Signaling Technology, Inc; PhosphoPlus Akt (Ser-473) Antibody Kit; # 9270, Danvers, MA), as well as an anti-rabbit monoclonal antibody for phosphorylated S6 ribosomal protein (Ser-235/236) (2F9) (Cell Signaling Technology, Inc; # 4856) and a primary antibody for β-actin (Santa Cruz Biotechnology, Santa Cruz, CA). Hela cells were used as the positive control. Next, the membranes were washed and incubated with horseradish peroxidase-conjugated secondary antibodies. Bands of antibody-bound proteins were visualized by chemiluminescence; the blotted membrane was scanned for densitometry with a PDI imaging scanner (Agfa Japan, Tokyo), and the data were analyzed with NIH Image software. Expression of phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) was calculated relative to that of β-actin in the tumor tissue specimens and corresponding normal tissue specimens. For quantification of these proteins, the relative amount of phosphorylated Akt (Ser-473), Akt, or phosphorylated S6 ribosomal protein (Ser-235/236) in tumor tissue was expressed as a ratio of the optical density of the band for the tumor tissue specimen to that for the corresponding normal tissue specimen (set at 1.0) by densitometric analysis, as described previously [15, 20, 21]. The mean values for specimens of tumor and non-tumor tissue were calculated from three experiments [15, 21].\n[SUBTITLE] Statistical analysis [SUBSECTION] Real-time RT-PCR and Western blotting data were analyzed by the Mann–Whitney U-test for comparisons between two groups (pT stage, metastasis, microscopic pathological vascular invasion, and serum CRP), while the Kruskal–Wallis test was used to compare three groups (histological grade and treatment effect) [13–15]. Spearman’s rank correlation coefficient analysis was performed to determine the relations between IFNAR2 mRNA expression and the serum CRP level or tumor size. The Kaplan–Meier method was used to estimate survival, and differences of survival were assessed by the log-rank test. In all analyses, a P value of less than 0.05 was considered significant. Data were analyzed with commercially available software.\nReal-time RT-PCR and Western blotting data were analyzed by the Mann–Whitney U-test for comparisons between two groups (pT stage, metastasis, microscopic pathological vascular invasion, and serum CRP), while the Kruskal–Wallis test was used to compare three groups (histological grade and treatment effect) [13–15]. Spearman’s rank correlation coefficient analysis was performed to determine the relations between IFNAR2 mRNA expression and the serum CRP level or tumor size. The Kaplan–Meier method was used to estimate survival, and differences of survival were assessed by the log-rank test. In all analyses, a P value of less than 0.05 was considered significant. Data were analyzed with commercially available software.", "We studied 66 consecutive Japanese patients (39 men and 27 women) aged 32–82 years (mean age: 62.9 years), who were newly diagnosed with clear cell RCC from 2008 to 2010. All patients routinely underwent imaging with CT and/or MRI for preoperative staging prior to radical nephrectomy. The postoperative follow-up period ranged from 2 to 35 months (median: 19 months). Surgery was done before the patients received any other therapy.\nA peripheral whole blood sample (10 ml) was collected from each patient and was diluted in PBS with 2 mM EDTA. Then, peripheral blood mononuclear cells (PBMC) were separated using Ficol-Hypaque medium (Biocoll, Berlin, Germany) and gradient centrifugation and stored at −80°C for future use. Total RNA was isolated with an RNeasy kit (Qiagen, Hamburg, Germany), including a DNA digestion step using the RNase-free DNase set.\nIn every patient, three different tumor tissue specimens and various parts of the non-neoplastic kidney were harvested for this study and stored at −80°C, as described previously [13–15]. The tumor grade and clinical stage were determined according to the Fuhrman grading system and the TNM classification, respectively [16, 17]. This study was conducted in accordance with the Helsinki Declaration, and approval of the Dokkyo Medical University Hospital institutional review board was obtained. In addition, each patient signed a consent form that was approved by the Committee on Human Rights in Research of our institution.\nPostoperative immunotherapy with IFN-α was given to 26 patients with extra-renal involvement. These patients received 3, 5, or 6 million units of natural human IFN-α intravenously or intramuscularly two or three times a week until tumor progression occurred. If the tumor was refractory to IFN-α monotherapy (progressive disease; PD), concomitant treatment with IFN-α and sorafenib (400 mg/day) was performed [18].\nThe doses of IFN-α and sorafenib were decreased if grade 3/4 toxicity occurred. Peripheral whole blood samples (10 ml) were obtained every 2 or 3 months from each metastatic RCC patient on treatment with IFN-α ± sorafenib. Tumor response was assessed according to RECIST criteria [19].", "Total RNA was purified from all 66 sets of serum, tumor tissue samples, and non-tumor tissue samples with an RNA preparation kit (“High Pure RNA Kit”; Roche Diagnostic Ltd., Germany) and was used as the template for cDNA synthesis. A 100-μl reaction mixture containing 1 mg of random hexamers and 100 units of MMLV reverse transcriptase was incubated at 25°C for 10 min, at 42°C for 30 min, and then at 99°C for 5 min. The IFNAR gene expression profile was analyzed with an ABI PRISM 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) using the SYBR Green method. The following primers were employed to amplify the β-actin gene after confirming their specificity: sense; 5′-CTGGCATCGTGATGGACTCCGG-3′; anti-sense, 5′-GTGGATGCCACAGGACTCCATG-3′. The primers used for IFNAR1 and IFNAR2 have been previously reported [14]. Real-time RT-PCR was performed in a 25-μl reaction mixture containing 20 ng of sample cDNA, 100 nM sense primer, 100 nM anti-sense primer, and 12.5 μl of SYBR Green PCR Master Mix (Applied Biosystems), with 45 cycles of 95°C for 15 s and 60°C for 1 min. A standard curve for each mRNA was generated using fivefold dilutions of a control RNA sample (25×, 5×, 1×, 0.2×, and 0.04×). The expression of each IFNAR mRNA was calculated as a ratio to that of β-actin in the serum, tumor tissue, and corresponding normal tissue samples to determine the relative level of expression [13, 14]. Individual variations of mRNA expression considered to be important [13]. To investigate the influence of such individual variations in the expression of IFNAR1 mRNA and IFNAR2 mRNA in the blood and tissues, we compared the expression of these mRNAs in serum and paired samples of tumor and non-tumor tissues. For quantification of mRNA expression, the relative amounts of IFNAR1 and IFNAR2 mRNAs in the tumors were calculated as a ratio of the optical density of the bands for the tumor specimens to the density of the bands for the corresponding normal tissue specimens (T/N ratio) by densitometric analysis, as described previously [13, 14]. The mean value for three tissue samples was used, as described previously [13, 14]. We also measured the serum C-reactive protein (CRP) level (normal < 0.3 mg/dl) in each patient.", "We could only perform Western blotting for 15 M1 tumors and four M0 tumors. Samples of tumor tissue and normal tissue were carefully dissected free of stromal tissue. Western blotting for phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) was carried out as described previously [15, 20, 21]. Briefly, 50 μg of cytosolic protein was separated by SDS–PAGE on 12.5% gel and electrotransferred to a polyvinylidene difluoride membrane (Immobilon-P membrane; Millipore, Bedford, MA). After the membrane was blocked, the bound proteins were probed with an anti-rabbit monoclonal antibody for phosphorylated Akt (Ser-473) and an anti-rabbit monoclonal antibody for Akt (Cell Signaling Technology, Inc; PhosphoPlus Akt (Ser-473) Antibody Kit; # 9270, Danvers, MA), as well as an anti-rabbit monoclonal antibody for phosphorylated S6 ribosomal protein (Ser-235/236) (2F9) (Cell Signaling Technology, Inc; # 4856) and a primary antibody for β-actin (Santa Cruz Biotechnology, Santa Cruz, CA). Hela cells were used as the positive control. Next, the membranes were washed and incubated with horseradish peroxidase-conjugated secondary antibodies. Bands of antibody-bound proteins were visualized by chemiluminescence; the blotted membrane was scanned for densitometry with a PDI imaging scanner (Agfa Japan, Tokyo), and the data were analyzed with NIH Image software. Expression of phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) was calculated relative to that of β-actin in the tumor tissue specimens and corresponding normal tissue specimens. For quantification of these proteins, the relative amount of phosphorylated Akt (Ser-473), Akt, or phosphorylated S6 ribosomal protein (Ser-235/236) in tumor tissue was expressed as a ratio of the optical density of the band for the tumor tissue specimen to that for the corresponding normal tissue specimen (set at 1.0) by densitometric analysis, as described previously [15, 20, 21]. The mean values for specimens of tumor and non-tumor tissue were calculated from three experiments [15, 21].", "Real-time RT-PCR and Western blotting data were analyzed by the Mann–Whitney U-test for comparisons between two groups (pT stage, metastasis, microscopic pathological vascular invasion, and serum CRP), while the Kruskal–Wallis test was used to compare three groups (histological grade and treatment effect) [13–15]. Spearman’s rank correlation coefficient analysis was performed to determine the relations between IFNAR2 mRNA expression and the serum CRP level or tumor size. The Kaplan–Meier method was used to estimate survival, and differences of survival were assessed by the log-rank test. In all analyses, a P value of less than 0.05 was considered significant. Data were analyzed with commercially available software.", "[SUBTITLE] Serum IFNAR mRNA and characteristics of RCC [SUBSECTION] Although the preoperative serum IFNAR2 mRNA level was not related to the IFNAR2 mRNA level in tumor tissues (Fig. 1a) or the preoperative serum CRP level (Fig. 1b), there was a weak positive correlation between serum IFNAR2 and tumor size (r\n2 = 0.121, P = 0.0056, Fig. 1c).Fig. 1Spearman rank correlation coefficient relationship between mRNA expression levels of serum IFNAR2 mRNA and other factors. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels were associated with tumor size (c), but not with the IFNAR2 mRNAs in tumors (a) and serum CRP levels (b)\n\nSpearman rank correlation coefficient relationship between mRNA expression levels of serum IFNAR2 mRNA and other factors. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels were associated with tumor size (c), but not with the IFNAR2 mRNAs in tumors (a) and serum CRP levels (b)\nThe preoperative serum IFNAR2 mRNA level was not associated with the histological grade of RCC (mean ± S.D., grade 1, 1.38 ± 0.80; grade 2, 1.41 ± 0.71; grade 3, 1.36 ± 0.78, P = 0.9390, Fig. 2a), the pT stage (pT1–2, 1.35 ± 0.67; pT3–4, 1.47 ± 0.80, P = 0.7692, Fig. 2b), tumor metastasis (M0, 1.37 ± 0.74; M1, 1.49 ± 0.71, P = 0.5058, Fig. 2c), or microscopic vascular invasion (v(−), 1.36 ± 0.69; v(+), 1.55 ± 0.80, P = 0.3886, Fig. 2d).Fig. 2Serum levels for IFNAR2 mRNAs were not associated with tumor grade (a), pT stage (b), metastasis (c), and pathological microscopic vessel invasion (d). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled\nP values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nSerum levels for IFNAR2 mRNAs were not associated with tumor grade (a), pT stage (b), metastasis (c), and pathological microscopic vessel invasion (d). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled\nP values were obtained by comparing the three groups with the Kruskal–Wallis test\nThe preoperative serum level of IFNAR1 mRNA was not associated with the IFNAR1 mRNA level in tumor tissues, the preoperative serum CRP level, tumor size, histological grade, pT stage, tumor metastasis, or microscopic vascular invasion (data not shown).\nA higher preoperative serum CRP level was associated with local invasion (pT1–2, 0.24 ± 0.30; pT3–4, 3.50 ± 5.48, P < 0.0001), metastatic disease (M0, 0.39 ± 0.63; M1, 4.11 ± 6.09, P = 0.0002, and microscopic vascular invasion (v(−), 0.49 ± 0.73; v(+), 4.11 ± 6.41, P = 0.0217), but not with the histological grade of RCC (grade 1, 1.18 ± 3.03; grade 2, 0.83 ± 1.47; grade 3, 4.49 ± 7.74, P = 0.0771) (data not shown).\nAlthough the preoperative serum IFNAR2 mRNA level was not related to the IFNAR2 mRNA level in tumor tissues (Fig. 1a) or the preoperative serum CRP level (Fig. 1b), there was a weak positive correlation between serum IFNAR2 and tumor size (r\n2 = 0.121, P = 0.0056, Fig. 1c).Fig. 1Spearman rank correlation coefficient relationship between mRNA expression levels of serum IFNAR2 mRNA and other factors. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels were associated with tumor size (c), but not with the IFNAR2 mRNAs in tumors (a) and serum CRP levels (b)\n\nSpearman rank correlation coefficient relationship between mRNA expression levels of serum IFNAR2 mRNA and other factors. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels were associated with tumor size (c), but not with the IFNAR2 mRNAs in tumors (a) and serum CRP levels (b)\nThe preoperative serum IFNAR2 mRNA level was not associated with the histological grade of RCC (mean ± S.D., grade 1, 1.38 ± 0.80; grade 2, 1.41 ± 0.71; grade 3, 1.36 ± 0.78, P = 0.9390, Fig. 2a), the pT stage (pT1–2, 1.35 ± 0.67; pT3–4, 1.47 ± 0.80, P = 0.7692, Fig. 2b), tumor metastasis (M0, 1.37 ± 0.74; M1, 1.49 ± 0.71, P = 0.5058, Fig. 2c), or microscopic vascular invasion (v(−), 1.36 ± 0.69; v(+), 1.55 ± 0.80, P = 0.3886, Fig. 2d).Fig. 2Serum levels for IFNAR2 mRNAs were not associated with tumor grade (a), pT stage (b), metastasis (c), and pathological microscopic vessel invasion (d). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled\nP values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nSerum levels for IFNAR2 mRNAs were not associated with tumor grade (a), pT stage (b), metastasis (c), and pathological microscopic vessel invasion (d). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled\nP values were obtained by comparing the three groups with the Kruskal–Wallis test\nThe preoperative serum level of IFNAR1 mRNA was not associated with the IFNAR1 mRNA level in tumor tissues, the preoperative serum CRP level, tumor size, histological grade, pT stage, tumor metastasis, or microscopic vascular invasion (data not shown).\nA higher preoperative serum CRP level was associated with local invasion (pT1–2, 0.24 ± 0.30; pT3–4, 3.50 ± 5.48, P < 0.0001), metastatic disease (M0, 0.39 ± 0.63; M1, 4.11 ± 6.09, P = 0.0002, and microscopic vascular invasion (v(−), 0.49 ± 0.73; v(+), 4.11 ± 6.41, P = 0.0217), but not with the histological grade of RCC (grade 1, 1.18 ± 3.03; grade 2, 0.83 ± 1.47; grade 3, 4.49 ± 7.74, P = 0.0771) (data not shown).\n[SUBTITLE] Tumor expression of IFNAR mRNA and characteristics of RCC [SUBSECTION] The IFNAR2 mRNA level in tumor tissue specimens was correlated with the histological grade of RCC (grade 1, 0.82 ± 0.46; grade 2, 1.12 ± 0.86; grade 3, 1.97 ± 1.61, P = 0.0480, Fig. 3a), pT stage (pT1–2, 0.94 ± 0.59; pT3–4, 1.58 ± 1.37, P = 0.0111, Fig. 3b), tumor metastasis (M0, 0.89 ± 0.61; M1, 1.83 ± 1.39, P = 0.0009, Fig. 3c), microscopic vascular invasion (v(−), 0.97 ± 0.70; v(+), 1.54 ± 1.39, P = 0.0392, Fig. 3d), and serum CRP (normal, <0.30 mg/dl; high CRP, 1.63 ± 1.35; low CRP, 0.88 ± 0.61; P = 0.0154, Fig. 3e). In contrast, there was no relationship between IFNAR1 mRNA expression and the histological grade (grade 1, 1.07 ± 0.79; grade 2, 1.75 ± 3.79; grade 3, 0.90 ± 0.67, P = 0.7472), stage (pT1–2, 0.87 ± 0.58; pT3–4, 2.30 ± 4.61, P = 0.2439), tumor metastasis (M0, 0.74 ± 0.51; M1, 3.07 ± 5.18, P = 0.0735), microscopic vascular invasion (v(-), 1.59 ± 3.91; v(+), 1.54 ± 1.08, P = 0.9736), or serum CRP (normal, < 0.30 mg/dl; high CRP, 2.50 ± 4.77; low CRP, 0.91 ± 0.71; P = 0.2646).Fig. 3The IFNAR2 mRNA levels in tumors were associated with tumor grade (a), pT stage (b), metastasis (c), pathological microscopic vessel invasion (d), and serum CRP levels (e). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nThe IFNAR2 mRNA levels in tumors were associated with tumor grade (a), pT stage (b), metastasis (c), pathological microscopic vessel invasion (d), and serum CRP levels (e). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\nThe IFNAR2 mRNA level in tumor tissue specimens was correlated with the histological grade of RCC (grade 1, 0.82 ± 0.46; grade 2, 1.12 ± 0.86; grade 3, 1.97 ± 1.61, P = 0.0480, Fig. 3a), pT stage (pT1–2, 0.94 ± 0.59; pT3–4, 1.58 ± 1.37, P = 0.0111, Fig. 3b), tumor metastasis (M0, 0.89 ± 0.61; M1, 1.83 ± 1.39, P = 0.0009, Fig. 3c), microscopic vascular invasion (v(−), 0.97 ± 0.70; v(+), 1.54 ± 1.39, P = 0.0392, Fig. 3d), and serum CRP (normal, <0.30 mg/dl; high CRP, 1.63 ± 1.35; low CRP, 0.88 ± 0.61; P = 0.0154, Fig. 3e). In contrast, there was no relationship between IFNAR1 mRNA expression and the histological grade (grade 1, 1.07 ± 0.79; grade 2, 1.75 ± 3.79; grade 3, 0.90 ± 0.67, P = 0.7472), stage (pT1–2, 0.87 ± 0.58; pT3–4, 2.30 ± 4.61, P = 0.2439), tumor metastasis (M0, 0.74 ± 0.51; M1, 3.07 ± 5.18, P = 0.0735), microscopic vascular invasion (v(-), 1.59 ± 3.91; v(+), 1.54 ± 1.08, P = 0.9736), or serum CRP (normal, < 0.30 mg/dl; high CRP, 2.50 ± 4.77; low CRP, 0.91 ± 0.71; P = 0.2646).Fig. 3The IFNAR2 mRNA levels in tumors were associated with tumor grade (a), pT stage (b), metastasis (c), pathological microscopic vessel invasion (d), and serum CRP levels (e). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nThe IFNAR2 mRNA levels in tumors were associated with tumor grade (a), pT stage (b), metastasis (c), pathological microscopic vessel invasion (d), and serum CRP levels (e). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n[SUBTITLE] Tumor expression of phosphorylated Akt (Ser-473), Akt, and phosphorylated S6 ribosomal protein (Ser-235/236) and characteristics of RCC [SUBSECTION] We could only perform Western blotting for 15 M1 tumors and four M0 tumors. The levels of phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were significantly higher in M1 tumors than in M0 tumors (3.93 ± 3.13, vs. 0.89 ± 0.40, P = 0.0257 and 3.82 ± 2.28 vs. 0.96 ± 0.84, P = 0.0079, respectively, Fig. 4). In contrast, there was no difference of Akt expression between M1 tumors and M0 tumors (1.13 ± 0.26, vs. 1.48 ± 1.87, P = 0.3662).Fig. 4Expression of phosphorylated Akt (Ser-473) (60 kDa), Akt (60 kDa), phosphorylated S6 ribosomal protein (Ser-235/236) (32 kDa), and beta actin (42 kDa) proteins in the primary tumor tissues with metastatic lesions (M1) and without (M0) using Western blotting. N non-tumor tissue, T primary tumor tissue with metastatic lesions. Each number corresponds to a case number\n\nExpression of phosphorylated Akt (Ser-473) (60 kDa), Akt (60 kDa), phosphorylated S6 ribosomal protein (Ser-235/236) (32 kDa), and beta actin (42 kDa) proteins in the primary tumor tissues with metastatic lesions (M1) and without (M0) using Western blotting. N non-tumor tissue, T primary tumor tissue with metastatic lesions. Each number corresponds to a case number\nPreoperative serum level of IFNAR2 mRNA showed a weak negative correlation with tumor tissue levels of phosphorylated Akt (Ser-473) (r\n2 = 0.205, P = 0.0903, Fig. 5a), but not with tumor tissue levels of AKT (r\n2 = 0.017, P = 0.6587) or phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.072, P = 0.3327, Fig. 5b). On the other hand, tumor tissue levels of IFNAR2 mRNA were positively associated with the levels of phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.35, P = 0.0202, Fig. 5d) and also showed a weak positive correlation with phosphorylated Akt (Ser-473) (r\n2 = 0.199, P = 0.0959, Fig. 5c), but not with Akt (r\n2 = 0.001, P = 0.9154). Although there was a positive correlation between tumor tissue levels of phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.332, P = 0.0245, Fig. 5e), there was no relationship between either of these two phosphorylated proteins and Akt (data not shown).Fig. 5Spearman rank correlation coefficient relationship. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels inversely correlated with tumor phosphorylated Akt (Ser-473) (a), but not tumor phosphorylated S6 ribosomal protein (Ser-235/236) (b). Tumor IFNAR2 mRNA levels were associated with tumor phosphorylated S6 ribosomal protein (Ser-235/236) (c), and weak positive correlation with tumor phosphorylated Akt (Ser-473) (d). Tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were related each other (e)\n\nSpearman rank correlation coefficient relationship. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels inversely correlated with tumor phosphorylated Akt (Ser-473) (a), but not tumor phosphorylated S6 ribosomal protein (Ser-235/236) (b). Tumor IFNAR2 mRNA levels were associated with tumor phosphorylated S6 ribosomal protein (Ser-235/236) (c), and weak positive correlation with tumor phosphorylated Akt (Ser-473) (d). Tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were related each other (e)\nWe could only perform Western blotting for 15 M1 tumors and four M0 tumors. The levels of phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were significantly higher in M1 tumors than in M0 tumors (3.93 ± 3.13, vs. 0.89 ± 0.40, P = 0.0257 and 3.82 ± 2.28 vs. 0.96 ± 0.84, P = 0.0079, respectively, Fig. 4). In contrast, there was no difference of Akt expression between M1 tumors and M0 tumors (1.13 ± 0.26, vs. 1.48 ± 1.87, P = 0.3662).Fig. 4Expression of phosphorylated Akt (Ser-473) (60 kDa), Akt (60 kDa), phosphorylated S6 ribosomal protein (Ser-235/236) (32 kDa), and beta actin (42 kDa) proteins in the primary tumor tissues with metastatic lesions (M1) and without (M0) using Western blotting. N non-tumor tissue, T primary tumor tissue with metastatic lesions. Each number corresponds to a case number\n\nExpression of phosphorylated Akt (Ser-473) (60 kDa), Akt (60 kDa), phosphorylated S6 ribosomal protein (Ser-235/236) (32 kDa), and beta actin (42 kDa) proteins in the primary tumor tissues with metastatic lesions (M1) and without (M0) using Western blotting. N non-tumor tissue, T primary tumor tissue with metastatic lesions. Each number corresponds to a case number\nPreoperative serum level of IFNAR2 mRNA showed a weak negative correlation with tumor tissue levels of phosphorylated Akt (Ser-473) (r\n2 = 0.205, P = 0.0903, Fig. 5a), but not with tumor tissue levels of AKT (r\n2 = 0.017, P = 0.6587) or phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.072, P = 0.3327, Fig. 5b). On the other hand, tumor tissue levels of IFNAR2 mRNA were positively associated with the levels of phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.35, P = 0.0202, Fig. 5d) and also showed a weak positive correlation with phosphorylated Akt (Ser-473) (r\n2 = 0.199, P = 0.0959, Fig. 5c), but not with Akt (r\n2 = 0.001, P = 0.9154). Although there was a positive correlation between tumor tissue levels of phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.332, P = 0.0245, Fig. 5e), there was no relationship between either of these two phosphorylated proteins and Akt (data not shown).Fig. 5Spearman rank correlation coefficient relationship. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels inversely correlated with tumor phosphorylated Akt (Ser-473) (a), but not tumor phosphorylated S6 ribosomal protein (Ser-235/236) (b). Tumor IFNAR2 mRNA levels were associated with tumor phosphorylated S6 ribosomal protein (Ser-235/236) (c), and weak positive correlation with tumor phosphorylated Akt (Ser-473) (d). Tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were related each other (e)\n\nSpearman rank correlation coefficient relationship. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels inversely correlated with tumor phosphorylated Akt (Ser-473) (a), but not tumor phosphorylated S6 ribosomal protein (Ser-235/236) (b). Tumor IFNAR2 mRNA levels were associated with tumor phosphorylated S6 ribosomal protein (Ser-235/236) (c), and weak positive correlation with tumor phosphorylated Akt (Ser-473) (d). Tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were related each other (e)\n[SUBTITLE] Relationship between IFNAR2 mRNA, phosphorylated Akt (Ser-473), Akt, or phosphorylated S6 ribosomal protein (Ser-235/236) and the response to therapy [SUBSECTION] Kaplan–Meier survival plots for twenty-six metastatic RCCs treated with IFN-α monotherapy showed that the patients with a good response to this agent had better progression-free survival (Fig. 6a). Among the 21 patients refractory to IFN-α, IFN-α + Sor: CR-PR group had longer progression-free survival than IFN-α + Sor: SD-PD group (Fig. 6b). Thus, the patients with a good response to IFN-α with/without sorafenib had favorable overall survival (Fig. 6c).Fig. 6Survival curve in the patients with metastatic lesions (M1) based on the treatment effects. a Progression-free survival curve in M1 patients treated with IFN-α alone. One patient of IFN-α: CR-PR group progressed gradually into progressive disease (PD) after partial response (PR), but continued IFN-α monotherapy. Median survival time in IFN-α: SD-PD group is 16.7 months. Survival time of IFN-α: CR-PR group does not reach. b Progression-free survival curve in M1 patients after combination therapy with IFN-α plus sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 4.3 months. Survival time of IFN-α + Sor: CR-PR group does not reach. c Overall survival curve in M1 patients treated with IFN-α with/without sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 15.6 months. Survival time of IFN-α ± Sor: CR-PR group does not reach. P value was analyzed by log-rank test\n\nSurvival curve in the patients with metastatic lesions (M1) based on the treatment effects. a Progression-free survival curve in M1 patients treated with IFN-α alone. One patient of IFN-α: CR-PR group progressed gradually into progressive disease (PD) after partial response (PR), but continued IFN-α monotherapy. Median survival time in IFN-α: SD-PD group is 16.7 months. Survival time of IFN-α: CR-PR group does not reach. b Progression-free survival curve in M1 patients after combination therapy with IFN-α plus sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 4.3 months. Survival time of IFN-α + Sor: CR-PR group does not reach. c Overall survival curve in M1 patients treated with IFN-α with/without sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 15.6 months. Survival time of IFN-α ± Sor: CR-PR group does not reach. P value was analyzed by log-rank test\nTwenty-six patients with metastatic disease received IFN-α as first-line adjuvant therapy. If these patients showed a poor response to IFN-α monotherapy, they received concomitant treatment with IFN-α and sorafenib (Sor) as second-line therapy. Five of the 26 patients showed a complete or partial response to IFN-α alone (IFN-α: CR-PR), while the other 21 patients received concomitant IFN-α + sorafenib as second-line therapy. The five patients with a good response to IFN-α alone (IFN-α: CR-PR) were low risk according to the MSKCC criteria. There was no difference of the preoperative serum IFNAR2 level between the patients with a good response to IFN-α (IFN-α: CR-PR, 1.74 ± 0.53) and those with a poor response, including stable disease or progressive disease (IFN-α: SD-PD; 1.31 ± 0.59, P = 0.1263, Fig. 7a). Lower tumor expression of IFNAR2 mRNA was related to a good response (IFN-α: CR-PR, 0.76 ± 0.40; IFN-α: SD-PD, 1.89 ± 1.38, P = 0.0345, Fig. 7b).Fig. 7The relationship between treatment effect and mRNA levels of serum and tumor IFNAR2. The serum and tumor IFNAR2 mRNAs levels in the IFN-α monotherapy (a, b), and IFN-α ± sorafenib therapy (c–f). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nThe relationship between treatment effect and mRNA levels of serum and tumor IFNAR2. The serum and tumor IFNAR2 mRNAs levels in the IFN-α monotherapy (a, b), and IFN-α ± sorafenib therapy (c–f). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\nAmong the 21 patients treated with IFN-α + sorafenib, eleven patients showed a good response (IFN-α + Sor: CR-PR), while the other 10 patients had stable disease or progressive disease (IFN-α + Sor: SD-PD). The background characteristics and the outcomes of the patients receiving concomitant therapy with IFN-α + sorafenib are summarized in Table 1. A higher preoperative serum IFNAR2 level was correlated with a good outcome (IFN-α + Sor: CR-PR, 1.70 ± 0.39; IFN-α: CR-PR, 1.74 ± 0.53; IFN-α + Sor: SD-PD, 0.88 ± 0.47, P = 0.0013, Fig. 7e). In order to better assess the effect of combination therapy, we combined the IFN-α : CR-PR group and the IFN-α + Sor: CR-PR group into a single good response group (IFN-α ± Sor: CR-PR). The preoperative serum IFNAR2 level was significantly higher in this good response group than in the group with a poor response to IFN-α + sorafenib (INF-α ± Sor: CR-PR, 1.71 ± 0.42; IFN-α + Sor: SD-PD, 0.88 ± 0.47, P < 0.0001) (Fig. 7c). On the other hand, low tumor expression of IFNAR2 mRNA was related to a good response (IFN-α + Sor: CR-PR, 1.34 ± 1.09; IFN-α: CR-PR, 0.76 ± 0.40; IFN-α + Sor: SD-PD, 2.51 ± 1.46, P = 0.0064, Fig. 7f, IFN-α ± Sor: CR-PR, 1.16 ± 0.95; IFN-α + Sor: SD-PD, 2.51 ± 1.46, P = 0.0027, Fig. 7d).Table 1Background of 26 metastatic RCCs treated with IFN-α ± sorafenib (400 mg/day)OptionIFN-α aloneIFN-α + sorafenibCR-PR (n = 5)CR-PR (n = 11)SD-PD (n = 10)Sex (male/female) 19/7Years (median) 39–78 (63)ECOG PS* (0/1/2) 17/7/25/0/08/3/04/4/2MSKCC* (Fav/Int/Poor) 11/12/34/1/05/6/02/5/3Duration of IFN-α* (mean:months)3–29 (14.7)Duration of pre-IFN-α* (mean:months)1–27 (7.8)Duration of IFN-α + sorafenib (mean:months)1–23 (9.7)Metastatic lesions* (numbers) PUL  255173 PLE  2011 HEP  4013 OSS  5Radiation for bone lesions023 LYM  8044 Others  2002AE*: Grade 1/2 (numbers) HT014 Fatigue39 Alopecia013 Fever48 HFS06 Thrombocytopenia24 Elevation of amylase and/or lipase05 Others29AE*: Grade 3/4 (numbers) HT01 Fatigue11 Alopecia02 Fever03 Thrombocytopenia02 Elevation of ALT and AST02 Erythema multiforme 01\nECOG PS* eastern cooperative oncology group (ECOG) performance status, MSKCC* Memorial Sloen-Kettering Cancer Center, Fav favorable risk, Int intermediated risk, Duration of IFN-α* duration of IFN-α monotherapy, Duration of pre-IFN-α* duration of IFN-α monotherapy prior to IFN-α plus sorafenib, Metastatic lesions* PUL Lung, PLE Pleura, HEP Liver, OSS Bone, LYM lymph node, AE* adverse events, HT hypertension, HFS hand-foot syndrome\n\nBackground of 26 metastatic RCCs treated with IFN-α ± sorafenib (400 mg/day)\n\nECOG PS* eastern cooperative oncology group (ECOG) performance status, MSKCC* Memorial Sloen-Kettering Cancer Center, Fav favorable risk, Int intermediated risk, Duration of IFN-α* duration of IFN-α monotherapy, Duration of pre-IFN-α* duration of IFN-α monotherapy prior to IFN-α plus sorafenib, Metastatic lesions* PUL Lung, PLE Pleura, HEP Liver, OSS Bone, LYM lymph node, AE* adverse events, HT hypertension, HFS hand-foot syndrome\nInterestingly, chronological analysis of serum IFNAR2 mRNA levels showed that these levels remained relatively high in the IFN-α ± Sor: CR-PR group and remained lower in the IFN-α + Sor: SD-PD group (Fig. 8).Fig. 8Chronological analysis of serum IFNAR2 mRNA throughout IFN-α ± sorafenib treatment; IFN-α ± Sor: CR-PR group (n = 12), IFN-α + Sor: SD-PD group (n = 7)\n\nChronological analysis of serum IFNAR2 mRNA throughout IFN-α ± sorafenib treatment; IFN-α ± Sor: CR-PR group (n = 12), IFN-α + Sor: SD-PD group (n = 7)\nThe IFN-α + Sor: CR-PR group had lower phosphorylated Akt (Ser-473) levels than the IFN-α + Sor: SD-PD group (1.65 ± 1.08 vs. 5.54 ± 3.22, P = 0.0208, Fig. 9a). Lower expression of phosphorylated S6 ribosomal protein (Ser-235/236) tended to be associated with a better response (IFN-α ± Sor: CR-PR, 2.50 ± 1.09; IFN-α + Sor: SD-PD, 4.72 ± 2.59, P = 0.1052, Fig. 9b). On the other hand, expression of Akt was similar in both the IFN-α + Sor: CR-PR group and the IFN-α + Sor: SD-PD group (1.11 ± 0.19 vs. 1.16 ± 0.34, P = 0.9078).Fig. 9The relationship between treatment effect and protein expression levels of Akt, phosphorylated Akt (Ser-473) (a) and phosphorylated S6 ribosomal protein (Ser-235/236) (b). The relative expression levels of targeted proteins in the primary tumor to those in corresponding non-tumor portion, which was set to 1.0. Hela cell was used as the positive control. The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers\n\nThe relationship between treatment effect and protein expression levels of Akt, phosphorylated Akt (Ser-473) (a) and phosphorylated S6 ribosomal protein (Ser-235/236) (b). The relative expression levels of targeted proteins in the primary tumor to those in corresponding non-tumor portion, which was set to 1.0. Hela cell was used as the positive control. The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers\nUnlike IFNAR2 mRNA, serum and tissue levels of IFNAR1 mRNA were unrelated to the response to treatment (data not shown).\nKaplan–Meier survival plots for twenty-six metastatic RCCs treated with IFN-α monotherapy showed that the patients with a good response to this agent had better progression-free survival (Fig. 6a). Among the 21 patients refractory to IFN-α, IFN-α + Sor: CR-PR group had longer progression-free survival than IFN-α + Sor: SD-PD group (Fig. 6b). Thus, the patients with a good response to IFN-α with/without sorafenib had favorable overall survival (Fig. 6c).Fig. 6Survival curve in the patients with metastatic lesions (M1) based on the treatment effects. a Progression-free survival curve in M1 patients treated with IFN-α alone. One patient of IFN-α: CR-PR group progressed gradually into progressive disease (PD) after partial response (PR), but continued IFN-α monotherapy. Median survival time in IFN-α: SD-PD group is 16.7 months. Survival time of IFN-α: CR-PR group does not reach. b Progression-free survival curve in M1 patients after combination therapy with IFN-α plus sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 4.3 months. Survival time of IFN-α + Sor: CR-PR group does not reach. c Overall survival curve in M1 patients treated with IFN-α with/without sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 15.6 months. Survival time of IFN-α ± Sor: CR-PR group does not reach. P value was analyzed by log-rank test\n\nSurvival curve in the patients with metastatic lesions (M1) based on the treatment effects. a Progression-free survival curve in M1 patients treated with IFN-α alone. One patient of IFN-α: CR-PR group progressed gradually into progressive disease (PD) after partial response (PR), but continued IFN-α monotherapy. Median survival time in IFN-α: SD-PD group is 16.7 months. Survival time of IFN-α: CR-PR group does not reach. b Progression-free survival curve in M1 patients after combination therapy with IFN-α plus sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 4.3 months. Survival time of IFN-α + Sor: CR-PR group does not reach. c Overall survival curve in M1 patients treated with IFN-α with/without sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 15.6 months. Survival time of IFN-α ± Sor: CR-PR group does not reach. P value was analyzed by log-rank test\nTwenty-six patients with metastatic disease received IFN-α as first-line adjuvant therapy. If these patients showed a poor response to IFN-α monotherapy, they received concomitant treatment with IFN-α and sorafenib (Sor) as second-line therapy. Five of the 26 patients showed a complete or partial response to IFN-α alone (IFN-α: CR-PR), while the other 21 patients received concomitant IFN-α + sorafenib as second-line therapy. The five patients with a good response to IFN-α alone (IFN-α: CR-PR) were low risk according to the MSKCC criteria. There was no difference of the preoperative serum IFNAR2 level between the patients with a good response to IFN-α (IFN-α: CR-PR, 1.74 ± 0.53) and those with a poor response, including stable disease or progressive disease (IFN-α: SD-PD; 1.31 ± 0.59, P = 0.1263, Fig. 7a). Lower tumor expression of IFNAR2 mRNA was related to a good response (IFN-α: CR-PR, 0.76 ± 0.40; IFN-α: SD-PD, 1.89 ± 1.38, P = 0.0345, Fig. 7b).Fig. 7The relationship between treatment effect and mRNA levels of serum and tumor IFNAR2. The serum and tumor IFNAR2 mRNAs levels in the IFN-α monotherapy (a, b), and IFN-α ± sorafenib therapy (c–f). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nThe relationship between treatment effect and mRNA levels of serum and tumor IFNAR2. The serum and tumor IFNAR2 mRNAs levels in the IFN-α monotherapy (a, b), and IFN-α ± sorafenib therapy (c–f). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\nAmong the 21 patients treated with IFN-α + sorafenib, eleven patients showed a good response (IFN-α + Sor: CR-PR), while the other 10 patients had stable disease or progressive disease (IFN-α + Sor: SD-PD). The background characteristics and the outcomes of the patients receiving concomitant therapy with IFN-α + sorafenib are summarized in Table 1. A higher preoperative serum IFNAR2 level was correlated with a good outcome (IFN-α + Sor: CR-PR, 1.70 ± 0.39; IFN-α: CR-PR, 1.74 ± 0.53; IFN-α + Sor: SD-PD, 0.88 ± 0.47, P = 0.0013, Fig. 7e). In order to better assess the effect of combination therapy, we combined the IFN-α : CR-PR group and the IFN-α + Sor: CR-PR group into a single good response group (IFN-α ± Sor: CR-PR). The preoperative serum IFNAR2 level was significantly higher in this good response group than in the group with a poor response to IFN-α + sorafenib (INF-α ± Sor: CR-PR, 1.71 ± 0.42; IFN-α + Sor: SD-PD, 0.88 ± 0.47, P < 0.0001) (Fig. 7c). On the other hand, low tumor expression of IFNAR2 mRNA was related to a good response (IFN-α + Sor: CR-PR, 1.34 ± 1.09; IFN-α: CR-PR, 0.76 ± 0.40; IFN-α + Sor: SD-PD, 2.51 ± 1.46, P = 0.0064, Fig. 7f, IFN-α ± Sor: CR-PR, 1.16 ± 0.95; IFN-α + Sor: SD-PD, 2.51 ± 1.46, P = 0.0027, Fig. 7d).Table 1Background of 26 metastatic RCCs treated with IFN-α ± sorafenib (400 mg/day)OptionIFN-α aloneIFN-α + sorafenibCR-PR (n = 5)CR-PR (n = 11)SD-PD (n = 10)Sex (male/female) 19/7Years (median) 39–78 (63)ECOG PS* (0/1/2) 17/7/25/0/08/3/04/4/2MSKCC* (Fav/Int/Poor) 11/12/34/1/05/6/02/5/3Duration of IFN-α* (mean:months)3–29 (14.7)Duration of pre-IFN-α* (mean:months)1–27 (7.8)Duration of IFN-α + sorafenib (mean:months)1–23 (9.7)Metastatic lesions* (numbers) PUL  255173 PLE  2011 HEP  4013 OSS  5Radiation for bone lesions023 LYM  8044 Others  2002AE*: Grade 1/2 (numbers) HT014 Fatigue39 Alopecia013 Fever48 HFS06 Thrombocytopenia24 Elevation of amylase and/or lipase05 Others29AE*: Grade 3/4 (numbers) HT01 Fatigue11 Alopecia02 Fever03 Thrombocytopenia02 Elevation of ALT and AST02 Erythema multiforme 01\nECOG PS* eastern cooperative oncology group (ECOG) performance status, MSKCC* Memorial Sloen-Kettering Cancer Center, Fav favorable risk, Int intermediated risk, Duration of IFN-α* duration of IFN-α monotherapy, Duration of pre-IFN-α* duration of IFN-α monotherapy prior to IFN-α plus sorafenib, Metastatic lesions* PUL Lung, PLE Pleura, HEP Liver, OSS Bone, LYM lymph node, AE* adverse events, HT hypertension, HFS hand-foot syndrome\n\nBackground of 26 metastatic RCCs treated with IFN-α ± sorafenib (400 mg/day)\n\nECOG PS* eastern cooperative oncology group (ECOG) performance status, MSKCC* Memorial Sloen-Kettering Cancer Center, Fav favorable risk, Int intermediated risk, Duration of IFN-α* duration of IFN-α monotherapy, Duration of pre-IFN-α* duration of IFN-α monotherapy prior to IFN-α plus sorafenib, Metastatic lesions* PUL Lung, PLE Pleura, HEP Liver, OSS Bone, LYM lymph node, AE* adverse events, HT hypertension, HFS hand-foot syndrome\nInterestingly, chronological analysis of serum IFNAR2 mRNA levels showed that these levels remained relatively high in the IFN-α ± Sor: CR-PR group and remained lower in the IFN-α + Sor: SD-PD group (Fig. 8).Fig. 8Chronological analysis of serum IFNAR2 mRNA throughout IFN-α ± sorafenib treatment; IFN-α ± Sor: CR-PR group (n = 12), IFN-α + Sor: SD-PD group (n = 7)\n\nChronological analysis of serum IFNAR2 mRNA throughout IFN-α ± sorafenib treatment; IFN-α ± Sor: CR-PR group (n = 12), IFN-α + Sor: SD-PD group (n = 7)\nThe IFN-α + Sor: CR-PR group had lower phosphorylated Akt (Ser-473) levels than the IFN-α + Sor: SD-PD group (1.65 ± 1.08 vs. 5.54 ± 3.22, P = 0.0208, Fig. 9a). Lower expression of phosphorylated S6 ribosomal protein (Ser-235/236) tended to be associated with a better response (IFN-α ± Sor: CR-PR, 2.50 ± 1.09; IFN-α + Sor: SD-PD, 4.72 ± 2.59, P = 0.1052, Fig. 9b). On the other hand, expression of Akt was similar in both the IFN-α + Sor: CR-PR group and the IFN-α + Sor: SD-PD group (1.11 ± 0.19 vs. 1.16 ± 0.34, P = 0.9078).Fig. 9The relationship between treatment effect and protein expression levels of Akt, phosphorylated Akt (Ser-473) (a) and phosphorylated S6 ribosomal protein (Ser-235/236) (b). The relative expression levels of targeted proteins in the primary tumor to those in corresponding non-tumor portion, which was set to 1.0. Hela cell was used as the positive control. The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers\n\nThe relationship between treatment effect and protein expression levels of Akt, phosphorylated Akt (Ser-473) (a) and phosphorylated S6 ribosomal protein (Ser-235/236) (b). The relative expression levels of targeted proteins in the primary tumor to those in corresponding non-tumor portion, which was set to 1.0. Hela cell was used as the positive control. The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers\nUnlike IFNAR2 mRNA, serum and tissue levels of IFNAR1 mRNA were unrelated to the response to treatment (data not shown).\n[SUBTITLE] Prognostic value of IFNAR2 mRNA, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236) [SUBSECTION] The mean serum level of IFNAR2 mRNA and mean tumor tissue levels of IFNAR2 mRNA, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236) in M1 patients treated by IFN-α ± sorafenib were 1.39 (±0.60), 1.67 (±1.33), 3.72 (±3.12), and 3.63 (±2.24), respectively. Patients were divided into two groups based on these mean values (i.e., a high group and a low group), as described previously [13–15, 21]. Kaplan–Meier survival plots for patients with low versus high levels of these possible prognostic factors showed that a lower serum level of IFNAR2 mRNA (P < 0.05), a higher tumor tissue levels of IFNAR2 mRNA (P < 0.01), phosphorylated Akt (Ser-473) (P < 0.05), and phosphorylated S6 ribosomal protein (Ser-235/236) (P < 0.05) were correlated with shorter overall survival (OS) (Fig. 10a–d).Fig. 10Survival curve in the patients with metastatic lesions (M1) based on the mean values of mRNA levels of serum and tumor lFNAR2 s, and of protein levels of tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236), the cases were divided into two groups at this levels—high and low expression. Progression-free survival curve based on serum (a) and tumor IFNAR2 mRNA levels (b) in M1 patients treated with IFN-α ± sorafenib therapy. Progression-free survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (c) and phosphorylated S6 ribosomal protein (Ser-235/236) (d) in M1 patients treated with IFN-α + sorafenib therapy. Overall survival curve based on serum (e) and tumor IFNAR2 mRNA levels (f) in M1 patients treated with IFN-α ± sorafenib therapy. Overall survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (g) and phosphorylated S6 ribosomal protein (Ser-235/236) (h) in M1 patients treated with IFN-α + sorafenib therapy. P value was analyzed by log-rank test\n\nSurvival curve in the patients with metastatic lesions (M1) based on the mean values of mRNA levels of serum and tumor lFNAR2 s, and of protein levels of tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236), the cases were divided into two groups at this levels—high and low expression. Progression-free survival curve based on serum (a) and tumor IFNAR2 mRNA levels (b) in M1 patients treated with IFN-α ± sorafenib therapy. Progression-free survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (c) and phosphorylated S6 ribosomal protein (Ser-235/236) (d) in M1 patients treated with IFN-α + sorafenib therapy. Overall survival curve based on serum (e) and tumor IFNAR2 mRNA levels (f) in M1 patients treated with IFN-α ± sorafenib therapy. Overall survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (g) and phosphorylated S6 ribosomal protein (Ser-235/236) (h) in M1 patients treated with IFN-α + sorafenib therapy. P value was analyzed by log-rank test\nThe mean serum level of IFNAR2 mRNA and mean tumor tissue levels of IFNAR2 mRNA, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236) in M1 patients treated by IFN-α ± sorafenib were 1.39 (±0.60), 1.67 (±1.33), 3.72 (±3.12), and 3.63 (±2.24), respectively. Patients were divided into two groups based on these mean values (i.e., a high group and a low group), as described previously [13–15, 21]. Kaplan–Meier survival plots for patients with low versus high levels of these possible prognostic factors showed that a lower serum level of IFNAR2 mRNA (P < 0.05), a higher tumor tissue levels of IFNAR2 mRNA (P < 0.01), phosphorylated Akt (Ser-473) (P < 0.05), and phosphorylated S6 ribosomal protein (Ser-235/236) (P < 0.05) were correlated with shorter overall survival (OS) (Fig. 10a–d).Fig. 10Survival curve in the patients with metastatic lesions (M1) based on the mean values of mRNA levels of serum and tumor lFNAR2 s, and of protein levels of tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236), the cases were divided into two groups at this levels—high and low expression. Progression-free survival curve based on serum (a) and tumor IFNAR2 mRNA levels (b) in M1 patients treated with IFN-α ± sorafenib therapy. Progression-free survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (c) and phosphorylated S6 ribosomal protein (Ser-235/236) (d) in M1 patients treated with IFN-α + sorafenib therapy. Overall survival curve based on serum (e) and tumor IFNAR2 mRNA levels (f) in M1 patients treated with IFN-α ± sorafenib therapy. Overall survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (g) and phosphorylated S6 ribosomal protein (Ser-235/236) (h) in M1 patients treated with IFN-α + sorafenib therapy. P value was analyzed by log-rank test\n\nSurvival curve in the patients with metastatic lesions (M1) based on the mean values of mRNA levels of serum and tumor lFNAR2 s, and of protein levels of tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236), the cases were divided into two groups at this levels—high and low expression. Progression-free survival curve based on serum (a) and tumor IFNAR2 mRNA levels (b) in M1 patients treated with IFN-α ± sorafenib therapy. Progression-free survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (c) and phosphorylated S6 ribosomal protein (Ser-235/236) (d) in M1 patients treated with IFN-α + sorafenib therapy. Overall survival curve based on serum (e) and tumor IFNAR2 mRNA levels (f) in M1 patients treated with IFN-α ± sorafenib therapy. Overall survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (g) and phosphorylated S6 ribosomal protein (Ser-235/236) (h) in M1 patients treated with IFN-α + sorafenib therapy. P value was analyzed by log-rank test", "Although the preoperative serum IFNAR2 mRNA level was not related to the IFNAR2 mRNA level in tumor tissues (Fig. 1a) or the preoperative serum CRP level (Fig. 1b), there was a weak positive correlation between serum IFNAR2 and tumor size (r\n2 = 0.121, P = 0.0056, Fig. 1c).Fig. 1Spearman rank correlation coefficient relationship between mRNA expression levels of serum IFNAR2 mRNA and other factors. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels were associated with tumor size (c), but not with the IFNAR2 mRNAs in tumors (a) and serum CRP levels (b)\n\nSpearman rank correlation coefficient relationship between mRNA expression levels of serum IFNAR2 mRNA and other factors. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels were associated with tumor size (c), but not with the IFNAR2 mRNAs in tumors (a) and serum CRP levels (b)\nThe preoperative serum IFNAR2 mRNA level was not associated with the histological grade of RCC (mean ± S.D., grade 1, 1.38 ± 0.80; grade 2, 1.41 ± 0.71; grade 3, 1.36 ± 0.78, P = 0.9390, Fig. 2a), the pT stage (pT1–2, 1.35 ± 0.67; pT3–4, 1.47 ± 0.80, P = 0.7692, Fig. 2b), tumor metastasis (M0, 1.37 ± 0.74; M1, 1.49 ± 0.71, P = 0.5058, Fig. 2c), or microscopic vascular invasion (v(−), 1.36 ± 0.69; v(+), 1.55 ± 0.80, P = 0.3886, Fig. 2d).Fig. 2Serum levels for IFNAR2 mRNAs were not associated with tumor grade (a), pT stage (b), metastasis (c), and pathological microscopic vessel invasion (d). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled\nP values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nSerum levels for IFNAR2 mRNAs were not associated with tumor grade (a), pT stage (b), metastasis (c), and pathological microscopic vessel invasion (d). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled\nP values were obtained by comparing the three groups with the Kruskal–Wallis test\nThe preoperative serum level of IFNAR1 mRNA was not associated with the IFNAR1 mRNA level in tumor tissues, the preoperative serum CRP level, tumor size, histological grade, pT stage, tumor metastasis, or microscopic vascular invasion (data not shown).\nA higher preoperative serum CRP level was associated with local invasion (pT1–2, 0.24 ± 0.30; pT3–4, 3.50 ± 5.48, P < 0.0001), metastatic disease (M0, 0.39 ± 0.63; M1, 4.11 ± 6.09, P = 0.0002, and microscopic vascular invasion (v(−), 0.49 ± 0.73; v(+), 4.11 ± 6.41, P = 0.0217), but not with the histological grade of RCC (grade 1, 1.18 ± 3.03; grade 2, 0.83 ± 1.47; grade 3, 4.49 ± 7.74, P = 0.0771) (data not shown).", "The IFNAR2 mRNA level in tumor tissue specimens was correlated with the histological grade of RCC (grade 1, 0.82 ± 0.46; grade 2, 1.12 ± 0.86; grade 3, 1.97 ± 1.61, P = 0.0480, Fig. 3a), pT stage (pT1–2, 0.94 ± 0.59; pT3–4, 1.58 ± 1.37, P = 0.0111, Fig. 3b), tumor metastasis (M0, 0.89 ± 0.61; M1, 1.83 ± 1.39, P = 0.0009, Fig. 3c), microscopic vascular invasion (v(−), 0.97 ± 0.70; v(+), 1.54 ± 1.39, P = 0.0392, Fig. 3d), and serum CRP (normal, <0.30 mg/dl; high CRP, 1.63 ± 1.35; low CRP, 0.88 ± 0.61; P = 0.0154, Fig. 3e). In contrast, there was no relationship between IFNAR1 mRNA expression and the histological grade (grade 1, 1.07 ± 0.79; grade 2, 1.75 ± 3.79; grade 3, 0.90 ± 0.67, P = 0.7472), stage (pT1–2, 0.87 ± 0.58; pT3–4, 2.30 ± 4.61, P = 0.2439), tumor metastasis (M0, 0.74 ± 0.51; M1, 3.07 ± 5.18, P = 0.0735), microscopic vascular invasion (v(-), 1.59 ± 3.91; v(+), 1.54 ± 1.08, P = 0.9736), or serum CRP (normal, < 0.30 mg/dl; high CRP, 2.50 ± 4.77; low CRP, 0.91 ± 0.71; P = 0.2646).Fig. 3The IFNAR2 mRNA levels in tumors were associated with tumor grade (a), pT stage (b), metastasis (c), pathological microscopic vessel invasion (d), and serum CRP levels (e). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nThe IFNAR2 mRNA levels in tumors were associated with tumor grade (a), pT stage (b), metastasis (c), pathological microscopic vessel invasion (d), and serum CRP levels (e). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test", "We could only perform Western blotting for 15 M1 tumors and four M0 tumors. The levels of phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were significantly higher in M1 tumors than in M0 tumors (3.93 ± 3.13, vs. 0.89 ± 0.40, P = 0.0257 and 3.82 ± 2.28 vs. 0.96 ± 0.84, P = 0.0079, respectively, Fig. 4). In contrast, there was no difference of Akt expression between M1 tumors and M0 tumors (1.13 ± 0.26, vs. 1.48 ± 1.87, P = 0.3662).Fig. 4Expression of phosphorylated Akt (Ser-473) (60 kDa), Akt (60 kDa), phosphorylated S6 ribosomal protein (Ser-235/236) (32 kDa), and beta actin (42 kDa) proteins in the primary tumor tissues with metastatic lesions (M1) and without (M0) using Western blotting. N non-tumor tissue, T primary tumor tissue with metastatic lesions. Each number corresponds to a case number\n\nExpression of phosphorylated Akt (Ser-473) (60 kDa), Akt (60 kDa), phosphorylated S6 ribosomal protein (Ser-235/236) (32 kDa), and beta actin (42 kDa) proteins in the primary tumor tissues with metastatic lesions (M1) and without (M0) using Western blotting. N non-tumor tissue, T primary tumor tissue with metastatic lesions. Each number corresponds to a case number\nPreoperative serum level of IFNAR2 mRNA showed a weak negative correlation with tumor tissue levels of phosphorylated Akt (Ser-473) (r\n2 = 0.205, P = 0.0903, Fig. 5a), but not with tumor tissue levels of AKT (r\n2 = 0.017, P = 0.6587) or phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.072, P = 0.3327, Fig. 5b). On the other hand, tumor tissue levels of IFNAR2 mRNA were positively associated with the levels of phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.35, P = 0.0202, Fig. 5d) and also showed a weak positive correlation with phosphorylated Akt (Ser-473) (r\n2 = 0.199, P = 0.0959, Fig. 5c), but not with Akt (r\n2 = 0.001, P = 0.9154). Although there was a positive correlation between tumor tissue levels of phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) (r\n2 = 0.332, P = 0.0245, Fig. 5e), there was no relationship between either of these two phosphorylated proteins and Akt (data not shown).Fig. 5Spearman rank correlation coefficient relationship. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels inversely correlated with tumor phosphorylated Akt (Ser-473) (a), but not tumor phosphorylated S6 ribosomal protein (Ser-235/236) (b). Tumor IFNAR2 mRNA levels were associated with tumor phosphorylated S6 ribosomal protein (Ser-235/236) (c), and weak positive correlation with tumor phosphorylated Akt (Ser-473) (d). Tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were related each other (e)\n\nSpearman rank correlation coefficient relationship. X axis is an independent variable. Y axis is a dependent variable. Serum IFNAR2 mRNA levels inversely correlated with tumor phosphorylated Akt (Ser-473) (a), but not tumor phosphorylated S6 ribosomal protein (Ser-235/236) (b). Tumor IFNAR2 mRNA levels were associated with tumor phosphorylated S6 ribosomal protein (Ser-235/236) (c), and weak positive correlation with tumor phosphorylated Akt (Ser-473) (d). Tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236) were related each other (e)", "Kaplan–Meier survival plots for twenty-six metastatic RCCs treated with IFN-α monotherapy showed that the patients with a good response to this agent had better progression-free survival (Fig. 6a). Among the 21 patients refractory to IFN-α, IFN-α + Sor: CR-PR group had longer progression-free survival than IFN-α + Sor: SD-PD group (Fig. 6b). Thus, the patients with a good response to IFN-α with/without sorafenib had favorable overall survival (Fig. 6c).Fig. 6Survival curve in the patients with metastatic lesions (M1) based on the treatment effects. a Progression-free survival curve in M1 patients treated with IFN-α alone. One patient of IFN-α: CR-PR group progressed gradually into progressive disease (PD) after partial response (PR), but continued IFN-α monotherapy. Median survival time in IFN-α: SD-PD group is 16.7 months. Survival time of IFN-α: CR-PR group does not reach. b Progression-free survival curve in M1 patients after combination therapy with IFN-α plus sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 4.3 months. Survival time of IFN-α + Sor: CR-PR group does not reach. c Overall survival curve in M1 patients treated with IFN-α with/without sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 15.6 months. Survival time of IFN-α ± Sor: CR-PR group does not reach. P value was analyzed by log-rank test\n\nSurvival curve in the patients with metastatic lesions (M1) based on the treatment effects. a Progression-free survival curve in M1 patients treated with IFN-α alone. One patient of IFN-α: CR-PR group progressed gradually into progressive disease (PD) after partial response (PR), but continued IFN-α monotherapy. Median survival time in IFN-α: SD-PD group is 16.7 months. Survival time of IFN-α: CR-PR group does not reach. b Progression-free survival curve in M1 patients after combination therapy with IFN-α plus sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 4.3 months. Survival time of IFN-α + Sor: CR-PR group does not reach. c Overall survival curve in M1 patients treated with IFN-α with/without sorafenib. Median survival time in IFN-α + Sor: SD-PD group is 15.6 months. Survival time of IFN-α ± Sor: CR-PR group does not reach. P value was analyzed by log-rank test\nTwenty-six patients with metastatic disease received IFN-α as first-line adjuvant therapy. If these patients showed a poor response to IFN-α monotherapy, they received concomitant treatment with IFN-α and sorafenib (Sor) as second-line therapy. Five of the 26 patients showed a complete or partial response to IFN-α alone (IFN-α: CR-PR), while the other 21 patients received concomitant IFN-α + sorafenib as second-line therapy. The five patients with a good response to IFN-α alone (IFN-α: CR-PR) were low risk according to the MSKCC criteria. There was no difference of the preoperative serum IFNAR2 level between the patients with a good response to IFN-α (IFN-α: CR-PR, 1.74 ± 0.53) and those with a poor response, including stable disease or progressive disease (IFN-α: SD-PD; 1.31 ± 0.59, P = 0.1263, Fig. 7a). Lower tumor expression of IFNAR2 mRNA was related to a good response (IFN-α: CR-PR, 0.76 ± 0.40; IFN-α: SD-PD, 1.89 ± 1.38, P = 0.0345, Fig. 7b).Fig. 7The relationship between treatment effect and mRNA levels of serum and tumor IFNAR2. The serum and tumor IFNAR2 mRNAs levels in the IFN-α monotherapy (a, b), and IFN-α ± sorafenib therapy (c–f). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\n\nThe relationship between treatment effect and mRNA levels of serum and tumor IFNAR2. The serum and tumor IFNAR2 mRNAs levels in the IFN-α monotherapy (a, b), and IFN-α ± sorafenib therapy (c–f). The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers. Bold circled P values were obtained by comparing the three groups with the Kruskal–Wallis test\nAmong the 21 patients treated with IFN-α + sorafenib, eleven patients showed a good response (IFN-α + Sor: CR-PR), while the other 10 patients had stable disease or progressive disease (IFN-α + Sor: SD-PD). The background characteristics and the outcomes of the patients receiving concomitant therapy with IFN-α + sorafenib are summarized in Table 1. A higher preoperative serum IFNAR2 level was correlated with a good outcome (IFN-α + Sor: CR-PR, 1.70 ± 0.39; IFN-α: CR-PR, 1.74 ± 0.53; IFN-α + Sor: SD-PD, 0.88 ± 0.47, P = 0.0013, Fig. 7e). In order to better assess the effect of combination therapy, we combined the IFN-α : CR-PR group and the IFN-α + Sor: CR-PR group into a single good response group (IFN-α ± Sor: CR-PR). The preoperative serum IFNAR2 level was significantly higher in this good response group than in the group with a poor response to IFN-α + sorafenib (INF-α ± Sor: CR-PR, 1.71 ± 0.42; IFN-α + Sor: SD-PD, 0.88 ± 0.47, P < 0.0001) (Fig. 7c). On the other hand, low tumor expression of IFNAR2 mRNA was related to a good response (IFN-α + Sor: CR-PR, 1.34 ± 1.09; IFN-α: CR-PR, 0.76 ± 0.40; IFN-α + Sor: SD-PD, 2.51 ± 1.46, P = 0.0064, Fig. 7f, IFN-α ± Sor: CR-PR, 1.16 ± 0.95; IFN-α + Sor: SD-PD, 2.51 ± 1.46, P = 0.0027, Fig. 7d).Table 1Background of 26 metastatic RCCs treated with IFN-α ± sorafenib (400 mg/day)OptionIFN-α aloneIFN-α + sorafenibCR-PR (n = 5)CR-PR (n = 11)SD-PD (n = 10)Sex (male/female) 19/7Years (median) 39–78 (63)ECOG PS* (0/1/2) 17/7/25/0/08/3/04/4/2MSKCC* (Fav/Int/Poor) 11/12/34/1/05/6/02/5/3Duration of IFN-α* (mean:months)3–29 (14.7)Duration of pre-IFN-α* (mean:months)1–27 (7.8)Duration of IFN-α + sorafenib (mean:months)1–23 (9.7)Metastatic lesions* (numbers) PUL  255173 PLE  2011 HEP  4013 OSS  5Radiation for bone lesions023 LYM  8044 Others  2002AE*: Grade 1/2 (numbers) HT014 Fatigue39 Alopecia013 Fever48 HFS06 Thrombocytopenia24 Elevation of amylase and/or lipase05 Others29AE*: Grade 3/4 (numbers) HT01 Fatigue11 Alopecia02 Fever03 Thrombocytopenia02 Elevation of ALT and AST02 Erythema multiforme 01\nECOG PS* eastern cooperative oncology group (ECOG) performance status, MSKCC* Memorial Sloen-Kettering Cancer Center, Fav favorable risk, Int intermediated risk, Duration of IFN-α* duration of IFN-α monotherapy, Duration of pre-IFN-α* duration of IFN-α monotherapy prior to IFN-α plus sorafenib, Metastatic lesions* PUL Lung, PLE Pleura, HEP Liver, OSS Bone, LYM lymph node, AE* adverse events, HT hypertension, HFS hand-foot syndrome\n\nBackground of 26 metastatic RCCs treated with IFN-α ± sorafenib (400 mg/day)\n\nECOG PS* eastern cooperative oncology group (ECOG) performance status, MSKCC* Memorial Sloen-Kettering Cancer Center, Fav favorable risk, Int intermediated risk, Duration of IFN-α* duration of IFN-α monotherapy, Duration of pre-IFN-α* duration of IFN-α monotherapy prior to IFN-α plus sorafenib, Metastatic lesions* PUL Lung, PLE Pleura, HEP Liver, OSS Bone, LYM lymph node, AE* adverse events, HT hypertension, HFS hand-foot syndrome\nInterestingly, chronological analysis of serum IFNAR2 mRNA levels showed that these levels remained relatively high in the IFN-α ± Sor: CR-PR group and remained lower in the IFN-α + Sor: SD-PD group (Fig. 8).Fig. 8Chronological analysis of serum IFNAR2 mRNA throughout IFN-α ± sorafenib treatment; IFN-α ± Sor: CR-PR group (n = 12), IFN-α + Sor: SD-PD group (n = 7)\n\nChronological analysis of serum IFNAR2 mRNA throughout IFN-α ± sorafenib treatment; IFN-α ± Sor: CR-PR group (n = 12), IFN-α + Sor: SD-PD group (n = 7)\nThe IFN-α + Sor: CR-PR group had lower phosphorylated Akt (Ser-473) levels than the IFN-α + Sor: SD-PD group (1.65 ± 1.08 vs. 5.54 ± 3.22, P = 0.0208, Fig. 9a). Lower expression of phosphorylated S6 ribosomal protein (Ser-235/236) tended to be associated with a better response (IFN-α ± Sor: CR-PR, 2.50 ± 1.09; IFN-α + Sor: SD-PD, 4.72 ± 2.59, P = 0.1052, Fig. 9b). On the other hand, expression of Akt was similar in both the IFN-α + Sor: CR-PR group and the IFN-α + Sor: SD-PD group (1.11 ± 0.19 vs. 1.16 ± 0.34, P = 0.9078).Fig. 9The relationship between treatment effect and protein expression levels of Akt, phosphorylated Akt (Ser-473) (a) and phosphorylated S6 ribosomal protein (Ser-235/236) (b). The relative expression levels of targeted proteins in the primary tumor to those in corresponding non-tumor portion, which was set to 1.0. Hela cell was used as the positive control. The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers\n\nThe relationship between treatment effect and protein expression levels of Akt, phosphorylated Akt (Ser-473) (a) and phosphorylated S6 ribosomal protein (Ser-235/236) (b). The relative expression levels of targeted proteins in the primary tumor to those in corresponding non-tumor portion, which was set to 1.0. Hela cell was used as the positive control. The median value is the central line, the box is the interquartile range, the bars are the full range, and the points are the outliers\nUnlike IFNAR2 mRNA, serum and tissue levels of IFNAR1 mRNA were unrelated to the response to treatment (data not shown).", "The mean serum level of IFNAR2 mRNA and mean tumor tissue levels of IFNAR2 mRNA, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236) in M1 patients treated by IFN-α ± sorafenib were 1.39 (±0.60), 1.67 (±1.33), 3.72 (±3.12), and 3.63 (±2.24), respectively. Patients were divided into two groups based on these mean values (i.e., a high group and a low group), as described previously [13–15, 21]. Kaplan–Meier survival plots for patients with low versus high levels of these possible prognostic factors showed that a lower serum level of IFNAR2 mRNA (P < 0.05), a higher tumor tissue levels of IFNAR2 mRNA (P < 0.01), phosphorylated Akt (Ser-473) (P < 0.05), and phosphorylated S6 ribosomal protein (Ser-235/236) (P < 0.05) were correlated with shorter overall survival (OS) (Fig. 10a–d).Fig. 10Survival curve in the patients with metastatic lesions (M1) based on the mean values of mRNA levels of serum and tumor lFNAR2 s, and of protein levels of tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236), the cases were divided into two groups at this levels—high and low expression. Progression-free survival curve based on serum (a) and tumor IFNAR2 mRNA levels (b) in M1 patients treated with IFN-α ± sorafenib therapy. Progression-free survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (c) and phosphorylated S6 ribosomal protein (Ser-235/236) (d) in M1 patients treated with IFN-α + sorafenib therapy. Overall survival curve based on serum (e) and tumor IFNAR2 mRNA levels (f) in M1 patients treated with IFN-α ± sorafenib therapy. Overall survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (g) and phosphorylated S6 ribosomal protein (Ser-235/236) (h) in M1 patients treated with IFN-α + sorafenib therapy. P value was analyzed by log-rank test\n\nSurvival curve in the patients with metastatic lesions (M1) based on the mean values of mRNA levels of serum and tumor lFNAR2 s, and of protein levels of tumor phosphorylated Akt (Ser-473) and phosphorylated S6 ribosomal protein (Ser-235/236), the cases were divided into two groups at this levels—high and low expression. Progression-free survival curve based on serum (a) and tumor IFNAR2 mRNA levels (b) in M1 patients treated with IFN-α ± sorafenib therapy. Progression-free survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (c) and phosphorylated S6 ribosomal protein (Ser-235/236) (d) in M1 patients treated with IFN-α + sorafenib therapy. Overall survival curve based on serum (e) and tumor IFNAR2 mRNA levels (f) in M1 patients treated with IFN-α ± sorafenib therapy. Overall survival curve based on tumor proteins for phosphorylated Akt (Ser-473) (g) and phosphorylated S6 ribosomal protein (Ser-235/236) (h) in M1 patients treated with IFN-α + sorafenib therapy. P value was analyzed by log-rank test", "There were five main findings of this study. First, the preoperative serum level of IFNAR2 mRNA was correlated with tumor size. Second, the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated with each other, and both were related to tumor metastatic potential. Third, while the patients with a high preoperative serum level of IFNAR2 mRNA showed a good response to IFN-α ± sorafenib (IFN-α ± Sor; CR-PR) and a longer overall survival, the patients with a low serum level and a high tumor tissue level of IFNAR2 mRNA level showed a poor response (SD-PD) and a shorter overall survival. Fourth, the patients with a high serum level of IFNAR2 mRNA throughout treatment showed a good response to IFN-α ± Sor. Fifth, a high level of phosphorylated Akt (Ser-473), but not a high level of phospho-S6 ribosomal protein (Ser-235/236), in the primary tumor was related to a poor response to IFN-α ± sorafenib. These findings suggest that the serum level of IFNAR2 mRNA might be useful to predict the efficacy of IFN-α ± sorafenib therapy, while the tumor tissue level of IFNAR2 mRNA could be associated with metastatic potential and tumor resistance.\n[SUBTITLE] Combination therapy with IFN-α and sorafenib [SUBSECTION] While the efficacy of combination therapy with IFN-α and sorafenib needs clarification, the adverse effects of this regimen are related to the doses of each agent and are not additive [22–25]. We previously reported a good response to combination therapy with IFN-α and half the usual dose of sorafenib (400 mg/day rather than 800 mg/day) in patients with IFN-α-resistant RCC, along with tolerable adverse events [18].\nIt has also been reported that combined treatment with IFN-α + sorafenib suppresses proliferation and vascular endothelial growth factor (VEGF) production by several RCC cell lines more strongly than either agent alone [26, 27]. RCC is considered to be an immunogenic tumor [3], since cytotoxic T lymphocytes recognize and selectively kill autologous RCC cells, while tumor-specific T cells can be detected in the blood of RCC patients [3]. Sorafenib is a multikinase inhibitor targeting VEGF receptors 1–3, PDGFβ receptor, and Raf kinase, and it has both direct antitumor activity and antiangiogenic activity [28]. In a randomized phase III trial comparing sorafenib with placebo as second-line therapy for RCC, the response rate to sorafenib was 10%, and the stable disease rate was 74%, with the median progression-free survival time being 5.5 months in the sorafenib group versus 2.8 months in the placebo group [29]. In addition to its immunomodulatory effects, IFN-α also has direct antitumor activity as well as antiangiogenic activity, including inhibition of VEGF [30]. Antitumor immunity is usually suppressed in tumor-bearing mice because of the influence of regulatory T cells and suppressive cytokines, such as TGF-β and IL-10 [31]. Takeuchi et al. recently reported that the synergistic effect of sorafenib and IFN-α on RCC both in vitro and in tumor-bearing mice was related to a combination of antitumor, antiangiogenic, and immunologic responses [27]. In their study, sorafenib had no effect on the levels of natural killer (NK) cells, T cells, and regulatory T cells in the spleens of tumor-bearing BALB/c mice, irrespective of the use of IFN-α, while IFN-α showed weaker direct antitumor activity than sorafenib but stimulated CTL, NK cells, and tumor-infiltrating lymphocytes (which sorafenib did not), so that a synergistic antiproliferative effect of these two agents was demonstrated in vitro [27].\nAlthough an additive effect of IFN-α to sorafenib therapy in patients with metastatic RCC was recently reported [24], the clinical efficacy of this combination remains to be confirmed. Jonasch et al. reported that the outcome after addition of IFN-α to sorafenib was comparable to that of sorafenib monotherapy, when patients were randomized to treatment with either sorafenib (400 mg twice daily) or the combination of sorafenib (400 mg twice daily) plus IFN-α (0.5 MU twice daily) [25]. However, their IFN-α dose of 1 million units daily was probably too low to assess its additive effect because the average dose is 3–9 million units daily. So far, the published studies on combination therapy with IFN-α and sorafenib have employed concomitant therapy with both agents. In contrast, our treatment strategy was to add sorafenib (400 mg/day) to IFN-α in RCC patients whose tumors were refractory to IFN-α alone, i.e., first-line IFN-α monotherapy and second-line combination therapy with IFN-α plus sorafenib [18]. In the present study, 26 patients had metastatic disease (M1) at diagnosis. Among them, five patients showed a good response to IFN-α alone (IFN-α: CR-PR), 11 patients showed a good response to IFN-α + sorafenib (IFN-α + Sor: CR-PR, Table 1), and the remaining 10 patients had stable disease or showed a poor response to IFN-α + sorafenib (IFN-α + Sor: SD-PD, Table 1). Although half-dose sorafenib (400 mg/day) caused grade 1/2 toxicity, grade 3/4 toxicity was rare in the present study and such toxicity resolved when patients suspended sorafenib therapy. The response rate to the combination of IFN-α plus half-dose sorafenib was 52.4% (11/21 patients resistant to previous IFN-α monotherapy achieved CR or PR), indicating that sorafenib may be synergistic with IFN-α, leading to an increase of antitumor activity. Furthermore, the patients with good response to this combination therapy had favorable prognosis (Fig. 6). Thus, this combination seems to be tolerable and could be a useful treatment option for advanced RCC resistant to IFN-α monotherapy [18].\nWhile the efficacy of combination therapy with IFN-α and sorafenib needs clarification, the adverse effects of this regimen are related to the doses of each agent and are not additive [22–25]. We previously reported a good response to combination therapy with IFN-α and half the usual dose of sorafenib (400 mg/day rather than 800 mg/day) in patients with IFN-α-resistant RCC, along with tolerable adverse events [18].\nIt has also been reported that combined treatment with IFN-α + sorafenib suppresses proliferation and vascular endothelial growth factor (VEGF) production by several RCC cell lines more strongly than either agent alone [26, 27]. RCC is considered to be an immunogenic tumor [3], since cytotoxic T lymphocytes recognize and selectively kill autologous RCC cells, while tumor-specific T cells can be detected in the blood of RCC patients [3]. Sorafenib is a multikinase inhibitor targeting VEGF receptors 1–3, PDGFβ receptor, and Raf kinase, and it has both direct antitumor activity and antiangiogenic activity [28]. In a randomized phase III trial comparing sorafenib with placebo as second-line therapy for RCC, the response rate to sorafenib was 10%, and the stable disease rate was 74%, with the median progression-free survival time being 5.5 months in the sorafenib group versus 2.8 months in the placebo group [29]. In addition to its immunomodulatory effects, IFN-α also has direct antitumor activity as well as antiangiogenic activity, including inhibition of VEGF [30]. Antitumor immunity is usually suppressed in tumor-bearing mice because of the influence of regulatory T cells and suppressive cytokines, such as TGF-β and IL-10 [31]. Takeuchi et al. recently reported that the synergistic effect of sorafenib and IFN-α on RCC both in vitro and in tumor-bearing mice was related to a combination of antitumor, antiangiogenic, and immunologic responses [27]. In their study, sorafenib had no effect on the levels of natural killer (NK) cells, T cells, and regulatory T cells in the spleens of tumor-bearing BALB/c mice, irrespective of the use of IFN-α, while IFN-α showed weaker direct antitumor activity than sorafenib but stimulated CTL, NK cells, and tumor-infiltrating lymphocytes (which sorafenib did not), so that a synergistic antiproliferative effect of these two agents was demonstrated in vitro [27].\nAlthough an additive effect of IFN-α to sorafenib therapy in patients with metastatic RCC was recently reported [24], the clinical efficacy of this combination remains to be confirmed. Jonasch et al. reported that the outcome after addition of IFN-α to sorafenib was comparable to that of sorafenib monotherapy, when patients were randomized to treatment with either sorafenib (400 mg twice daily) or the combination of sorafenib (400 mg twice daily) plus IFN-α (0.5 MU twice daily) [25]. However, their IFN-α dose of 1 million units daily was probably too low to assess its additive effect because the average dose is 3–9 million units daily. So far, the published studies on combination therapy with IFN-α and sorafenib have employed concomitant therapy with both agents. In contrast, our treatment strategy was to add sorafenib (400 mg/day) to IFN-α in RCC patients whose tumors were refractory to IFN-α alone, i.e., first-line IFN-α monotherapy and second-line combination therapy with IFN-α plus sorafenib [18]. In the present study, 26 patients had metastatic disease (M1) at diagnosis. Among them, five patients showed a good response to IFN-α alone (IFN-α: CR-PR), 11 patients showed a good response to IFN-α + sorafenib (IFN-α + Sor: CR-PR, Table 1), and the remaining 10 patients had stable disease or showed a poor response to IFN-α + sorafenib (IFN-α + Sor: SD-PD, Table 1). Although half-dose sorafenib (400 mg/day) caused grade 1/2 toxicity, grade 3/4 toxicity was rare in the present study and such toxicity resolved when patients suspended sorafenib therapy. The response rate to the combination of IFN-α plus half-dose sorafenib was 52.4% (11/21 patients resistant to previous IFN-α monotherapy achieved CR or PR), indicating that sorafenib may be synergistic with IFN-α, leading to an increase of antitumor activity. Furthermore, the patients with good response to this combination therapy had favorable prognosis (Fig. 6). Thus, this combination seems to be tolerable and could be a useful treatment option for advanced RCC resistant to IFN-α monotherapy [18].\n[SUBTITLE] Role of the IFNAR2 and mTOR pathways in progression of RCC [SUBSECTION] The preoperative serum level of IFNAR2 mRNA was not correlated with the effect of IFN-α monotherapy, but a lower tumor tissue level of IFNAR2 mRNA was related to a better response to IFN-α. The present finding that a higher tumor tissue level of IFNAR2 mRNA was associated with a poor response to IFN-α is consistent with our previous results [14]. In contrast, regarding the relationship between IFNAR2 mRNA levels and the effect of combination therapy with IFN-α plus sorafenib, the preoperative serum level of IFNAR2 mRNA was higher in the IFN-α + Sor: CR-PR group and the IFN-α: CR-PR group than in the IFN-α + Sor: SD-PD group (Fig. 6e), while the tumor tissue level of IFNAR2 mRNA was lower in the IFN-α + Sor: CR-PR group and the IFN-α: CR-PR group than in the IFN-α + Sor: SD-PD group (Fig. 6f). These findings suggested that a different molecular mechanism might be involved in IFN-α ± Sor: CR-PR group and IFN-α + Sor: SD-PD group.\nIFNs are pleiotropic cytokines that regulate antiviral, antitumor, apoptotic, antiangiogenic, and cellular immune responses via activation of multiple downstream signaling cascades, including the Janus tyrosine kinase (Jak)-signal transducer and activator of transcription (STAT) pathway, the p38 mitogen-activated protein kinase (MAPK) pathway, and the mTOR pathway [32, 33]. The Jak-STAT and p38-MAPK signaling pathways have been shown to be responsible for transcription of the genes encoding proteins related to the antiviral and/or antiproliferative effects of IFNs. On the other hand, it has been reported that activation of the mTOR pathway by IFNs has an important regulatory role in mRNA translation and induction of the interferon response [34, 35]. IFN-α-induced tumor cell apoptosis is also mediated via the mTOR pathway in a nucleus-independent manner [36, 37]. Moreover, the Jak-STAT and mTOR pathways act separately from each other after activation by IFN-α [37]. Thus, it is likely that mTOR signaling selectively mediates apoptosis and survival.\nThe rapamycin-sensitive mTOR-raptor (regulatory-associated protein of mTOR) complex controls cell growth by regulating protein synthesis, so mTOR-raptor signaling is a potential antitumor target, and mTOR inhibitors are currently under investigation for the treatment of various human cancers. On the other hand, mTOR also interacts with rictor (rapamycin-insensitive companion of mTOR), and recent findings have suggested that the rapamycin-insensitive effect of mTOR on cell survival is overactive in many cancers. Thus, mTOR has dual rapamycin-sensitive (mTOR-raptor complex: mTORC1) and insensitive (mTOR-rictor complex: mTORC2) functions, indicating that treatment with rapamycin will not completely inhibit mTOR activity [38, 39]. Phosphatidylinositol 3`kinase (PI3 K), serine/threonine kinase Akt, and the mTOR pathway are all overactive in human cancers. mTORC1 lies downstream of PI3 K and is part of a pathway that is frequently activated in human cancers, so mTORC1 represents a pivotal target for anticancer therapy. The best-characterized pathways regulated by mTORC1 are phosphorylation and activation of ribosomal S6 kinase-1 (S6K1) and phosphorylation and inactivation of 4E-BP1, the suppressor of mRNA cap-binding protein eIF4E, leading to effects on cell growth and metabolism by acting as a restriction point in cells subjected to stresses [40, 41] such as hypoxia [42–44].\nPhosphorylation at two sites is required for full activation of Akt, since it is phosphorylated by PI3 K-dependent kinase-1 (PDK1) at a threonine residue in the catalytic domain (Thr 308) and by PI3 K-dependent kinase-2 (PDK2) at a serine residue (Ser 473) in the carboxy-terminal hydrophobic motif [45]. It has been reported that mTORC2 regulates the actin cytoskeleton and also possesses PDK2 activity that phosphorylates Ser-473 in the carboxy-terminal of Akt, making it essential for Akt activity [46]. Importantly, activation of Akt may lead to cell survival when mTORC1 is inhibited or could potentially increase VEGF production because PI3 K/Akt signaling induces tumor angiogenesis by regulating VEGF via both HIF-1α-dependent and -independent mechanisms [47]. It has been reported that hypoxia-inducible factor (HIF) 1α expression is dependent on both raptor and rictor, whereas HIF2α expression only depends on rictor and HIF2α is more important in RCC [48]. These findings suggest that phosphorylation of Ser 473 in AKT is a key molecular step in the progression of RCC and could be a target for treating these tumors [25]. In agreement with this, our current study showed that the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated and that both were related to tumor metastatic potential, while a high phosphorylated Akt (Ser-473) level in the primary tumor was associated with a poor response to IFN-α ± sorafenib therapy (stable or progressive disease).\nCurrent efforts to achieve the clinical development of mTOR inhibitors are based on the role of mTOR signaling in promoting the proliferation and survival of tumor cells. It has been reported that treatment with mTOR inhibitors can improve the outcome of patients with metastatic RCC [8–10]. On the other hand, the mTOR pathway is also important for IFN-dependent translational responses, and IFN-α is widely used to treat advanced RCC. Although we could not exclude a possible detrimental effect of IFN-α treatment in the patients with metastatic RCC and higher IFNAR2 mRNA levels in their tumors, our findings suggested that there may be different molecular mechanisms of cancer progression in the patients with a good or poor response to IFN-α ± sorafenib. It is possible that IFNAR2 signaling has different biological effects from normal when upregulated in RCC.\nJonasch et al. recently reported that an increase of phosphorylated Akt (Ser-473) was associated with worse survival by microarray analysis of paraffin-embedded specimens [25]. In the present study, the tumors with higher phosphorylated Akt (Ser-473) levels, but not higher phospho-S6 ribosomal protein (Ser-235/236) levels, were resistant to IFN-α ± sorafenib therapy. In addition, the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated, and both were related to metastatic potential. Tumor levels of IFNAR2 mRNA also had a weak positive correlation with those of phosphorylated Akt (Ser-473). Moreover, the patients whose tumors had higher levels of phosphorylated Akt (Ser-473), phosphorylated S6 ribosomal protein (Ser-235/236), and IFNAR2 mRNA showed shorter overall survival. Taken together, it is possible that phosphorylation of Ser 473 on Akt is a key molecular step in the progression of RCC and a potential therapeutic target, so that the tumor level of phosphorylated Akt (Ser-473) may be useful for predicting the response to treatment. At present, it remains to be elucidated why upregulation of IFNAR2 expression is linked to the progression of RCC and to a poor response to treatment, and it is unclear how IFNAR2 interacts with mTORC1 and mTORC2, but our findings suggested that the IFNAR2-mTORC1 pathway via phosphorylated S6 ribosomal protein (Ser-235/236) may act locally within tumors to promote proliferation and metastasis by modifying mRNA translation, while the IFNAR2-mTORC2 pathway via phosphorylated Akt (Ser-473) may be associated with tumor resistance. So, these interactions should be elucidated in the future. As Lekmine et al. have indicated, therefore, caution should be exercised when designing clinical trials that combine an mTOR inhibitor and IFN-α due to possible antagonism of antitumor activity [34]. In fact, Huges et al. reported that patients treated with temsirolimus alone had better overall survival than those given IFN-α alone, while patients treated with temsirolimus plus IFN-α did not [49]. In the future, the downstream targets of IFNAR2 should be identified, and the expression or activity of one or two such targets should be studied in cell lines or tissue samples. A better understanding of the IFNAR2 pathway may help to elucidate its role in cancer.\nAlthough none of our patients were treated by sorafenib alone, it would be interesting to assess the expression of not only IFNAR2, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236), but also VEGF receptor and Raf, in tumor cells and the effects of sorafenib, sunitinib, or mTOR inhibitors. Such information could lead to elucidation of the role of the IFNAR2-mTOR pathway in the progression of RCC and the selection of patients who will benefit from treatment with IFN-α, sorafenib, sunitinib, or mTOR inhibitors.\nThe preoperative serum level of IFNAR2 mRNA was not correlated with the effect of IFN-α monotherapy, but a lower tumor tissue level of IFNAR2 mRNA was related to a better response to IFN-α. The present finding that a higher tumor tissue level of IFNAR2 mRNA was associated with a poor response to IFN-α is consistent with our previous results [14]. In contrast, regarding the relationship between IFNAR2 mRNA levels and the effect of combination therapy with IFN-α plus sorafenib, the preoperative serum level of IFNAR2 mRNA was higher in the IFN-α + Sor: CR-PR group and the IFN-α: CR-PR group than in the IFN-α + Sor: SD-PD group (Fig. 6e), while the tumor tissue level of IFNAR2 mRNA was lower in the IFN-α + Sor: CR-PR group and the IFN-α: CR-PR group than in the IFN-α + Sor: SD-PD group (Fig. 6f). These findings suggested that a different molecular mechanism might be involved in IFN-α ± Sor: CR-PR group and IFN-α + Sor: SD-PD group.\nIFNs are pleiotropic cytokines that regulate antiviral, antitumor, apoptotic, antiangiogenic, and cellular immune responses via activation of multiple downstream signaling cascades, including the Janus tyrosine kinase (Jak)-signal transducer and activator of transcription (STAT) pathway, the p38 mitogen-activated protein kinase (MAPK) pathway, and the mTOR pathway [32, 33]. The Jak-STAT and p38-MAPK signaling pathways have been shown to be responsible for transcription of the genes encoding proteins related to the antiviral and/or antiproliferative effects of IFNs. On the other hand, it has been reported that activation of the mTOR pathway by IFNs has an important regulatory role in mRNA translation and induction of the interferon response [34, 35]. IFN-α-induced tumor cell apoptosis is also mediated via the mTOR pathway in a nucleus-independent manner [36, 37]. Moreover, the Jak-STAT and mTOR pathways act separately from each other after activation by IFN-α [37]. Thus, it is likely that mTOR signaling selectively mediates apoptosis and survival.\nThe rapamycin-sensitive mTOR-raptor (regulatory-associated protein of mTOR) complex controls cell growth by regulating protein synthesis, so mTOR-raptor signaling is a potential antitumor target, and mTOR inhibitors are currently under investigation for the treatment of various human cancers. On the other hand, mTOR also interacts with rictor (rapamycin-insensitive companion of mTOR), and recent findings have suggested that the rapamycin-insensitive effect of mTOR on cell survival is overactive in many cancers. Thus, mTOR has dual rapamycin-sensitive (mTOR-raptor complex: mTORC1) and insensitive (mTOR-rictor complex: mTORC2) functions, indicating that treatment with rapamycin will not completely inhibit mTOR activity [38, 39]. Phosphatidylinositol 3`kinase (PI3 K), serine/threonine kinase Akt, and the mTOR pathway are all overactive in human cancers. mTORC1 lies downstream of PI3 K and is part of a pathway that is frequently activated in human cancers, so mTORC1 represents a pivotal target for anticancer therapy. The best-characterized pathways regulated by mTORC1 are phosphorylation and activation of ribosomal S6 kinase-1 (S6K1) and phosphorylation and inactivation of 4E-BP1, the suppressor of mRNA cap-binding protein eIF4E, leading to effects on cell growth and metabolism by acting as a restriction point in cells subjected to stresses [40, 41] such as hypoxia [42–44].\nPhosphorylation at two sites is required for full activation of Akt, since it is phosphorylated by PI3 K-dependent kinase-1 (PDK1) at a threonine residue in the catalytic domain (Thr 308) and by PI3 K-dependent kinase-2 (PDK2) at a serine residue (Ser 473) in the carboxy-terminal hydrophobic motif [45]. It has been reported that mTORC2 regulates the actin cytoskeleton and also possesses PDK2 activity that phosphorylates Ser-473 in the carboxy-terminal of Akt, making it essential for Akt activity [46]. Importantly, activation of Akt may lead to cell survival when mTORC1 is inhibited or could potentially increase VEGF production because PI3 K/Akt signaling induces tumor angiogenesis by regulating VEGF via both HIF-1α-dependent and -independent mechanisms [47]. It has been reported that hypoxia-inducible factor (HIF) 1α expression is dependent on both raptor and rictor, whereas HIF2α expression only depends on rictor and HIF2α is more important in RCC [48]. These findings suggest that phosphorylation of Ser 473 in AKT is a key molecular step in the progression of RCC and could be a target for treating these tumors [25]. In agreement with this, our current study showed that the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated and that both were related to tumor metastatic potential, while a high phosphorylated Akt (Ser-473) level in the primary tumor was associated with a poor response to IFN-α ± sorafenib therapy (stable or progressive disease).\nCurrent efforts to achieve the clinical development of mTOR inhibitors are based on the role of mTOR signaling in promoting the proliferation and survival of tumor cells. It has been reported that treatment with mTOR inhibitors can improve the outcome of patients with metastatic RCC [8–10]. On the other hand, the mTOR pathway is also important for IFN-dependent translational responses, and IFN-α is widely used to treat advanced RCC. Although we could not exclude a possible detrimental effect of IFN-α treatment in the patients with metastatic RCC and higher IFNAR2 mRNA levels in their tumors, our findings suggested that there may be different molecular mechanisms of cancer progression in the patients with a good or poor response to IFN-α ± sorafenib. It is possible that IFNAR2 signaling has different biological effects from normal when upregulated in RCC.\nJonasch et al. recently reported that an increase of phosphorylated Akt (Ser-473) was associated with worse survival by microarray analysis of paraffin-embedded specimens [25]. In the present study, the tumors with higher phosphorylated Akt (Ser-473) levels, but not higher phospho-S6 ribosomal protein (Ser-235/236) levels, were resistant to IFN-α ± sorafenib therapy. In addition, the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated, and both were related to metastatic potential. Tumor levels of IFNAR2 mRNA also had a weak positive correlation with those of phosphorylated Akt (Ser-473). Moreover, the patients whose tumors had higher levels of phosphorylated Akt (Ser-473), phosphorylated S6 ribosomal protein (Ser-235/236), and IFNAR2 mRNA showed shorter overall survival. Taken together, it is possible that phosphorylation of Ser 473 on Akt is a key molecular step in the progression of RCC and a potential therapeutic target, so that the tumor level of phosphorylated Akt (Ser-473) may be useful for predicting the response to treatment. At present, it remains to be elucidated why upregulation of IFNAR2 expression is linked to the progression of RCC and to a poor response to treatment, and it is unclear how IFNAR2 interacts with mTORC1 and mTORC2, but our findings suggested that the IFNAR2-mTORC1 pathway via phosphorylated S6 ribosomal protein (Ser-235/236) may act locally within tumors to promote proliferation and metastasis by modifying mRNA translation, while the IFNAR2-mTORC2 pathway via phosphorylated Akt (Ser-473) may be associated with tumor resistance. So, these interactions should be elucidated in the future. As Lekmine et al. have indicated, therefore, caution should be exercised when designing clinical trials that combine an mTOR inhibitor and IFN-α due to possible antagonism of antitumor activity [34]. In fact, Huges et al. reported that patients treated with temsirolimus alone had better overall survival than those given IFN-α alone, while patients treated with temsirolimus plus IFN-α did not [49]. In the future, the downstream targets of IFNAR2 should be identified, and the expression or activity of one or two such targets should be studied in cell lines or tissue samples. A better understanding of the IFNAR2 pathway may help to elucidate its role in cancer.\nAlthough none of our patients were treated by sorafenib alone, it would be interesting to assess the expression of not only IFNAR2, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236), but also VEGF receptor and Raf, in tumor cells and the effects of sorafenib, sunitinib, or mTOR inhibitors. Such information could lead to elucidation of the role of the IFNAR2-mTOR pathway in the progression of RCC and the selection of patients who will benefit from treatment with IFN-α, sorafenib, sunitinib, or mTOR inhibitors.\n[SUBTITLE] Role of serum IFNAR2 in progression of RCC [SUBSECTION] The serum CRP level is associated with the stage and outcome of RCC [50, 51]. Elevation of CRP is primarily determined by an increase of circulating IL-6 [52], and the IL-6 level is correlated with the serum CRP level as well as with tumor histological grade and tumor metastasis [53]. We previously reported that increased serum levels of CRP and IL-6 were associated with local tumor invasion and metastasis [54]. In the present study, a higher preoperative serum CRP level was associated with local invasion and metastasis of RCC, but not with the response to treatment (data not shown). These findings suggest that the serum CRP level is associated with tumor aggressiveness, so that elevation of CRP might reflect the poorer general condition of the patient rather than the response to therapy. On the other hand, IFN-α has immunomodulatory effects and direct antitumor activity as well as antiangiogenic activity, including inhibition of VEGF [30]. Therefore, it is unclear exactly what serum IFNAR2 reflects, but it is likely to be associated with the overall immune status. Accordingly, we expected to find a relation between the preoperative serum levels of IFNAR2 mRNA and CRP, but no relation was observed.\nAlthough we could not examine how IFN-α binds to IFNAR2 on peripheral blood cells and then exhibits antitumor and antiangiogenic activity, our findings showed that the preoperative serum level of IFNAR2 mRNA was positively correlated with tumor size and was higher in patients with metastatic RCC who showed a good response to IFN-α ± sorafenib therapy than in those with a poor response. Because obtaining blood samples from patients is easier than harvesting tissue samples, chronological analysis of serum IFNAR2 mRNA levels is preferable for evaluation of the role of IFNAR2. While the effect of IFN-α therapy on the serum level of IFNAR2 mRNA is still unclear, our chronological evaluation of IFNAR2 mRNA throughout treatment showed that its serum level remained higher in the IFN-α ± Sor: CR-PR group than in the IFN-α + Sor: SD-PD group. Taken together, these observations suggest that an increased serum level of IFNAR2 mRNA may represent a systemic immunologic and antitumor response to the tumor burden in RCC patients, as well as showing antimicrobial activity if infection occurs.\nRegarding the effect of genetic polymorphism on the response of metastatic RCC to IFN-α, it has been reported that STAT3 polymorphism is a useful diagnostic marker for predicting the response to IFN-α therapy in these patients [55]. An efficient marker of the response to IFN-α is needed to establish individualized optimal treatment strategies, especially when newer therapies are used as first-line treatment for metastatic RCC. Our study showed that patients with higher serum levels of IFNAR2 mRNA may be more likely to respond to IFN-α ± sorafenib therapy and to show a longer overall survival, while patients with higher tumor tissue levels of IFNAR2 mRNA may show poor response and unfavorable overall survival. Although we need a surgical specimen to examine tumor tissue levels of IFNAR2 mRNA and protein, the serum level of IFNAR2 mRNA can be conveniently measured, so it may be more useful for predicting the response to IFN-α ± sorafenib therapy and as a prognostic indicator.\nThe serum CRP level is associated with the stage and outcome of RCC [50, 51]. Elevation of CRP is primarily determined by an increase of circulating IL-6 [52], and the IL-6 level is correlated with the serum CRP level as well as with tumor histological grade and tumor metastasis [53]. We previously reported that increased serum levels of CRP and IL-6 were associated with local tumor invasion and metastasis [54]. In the present study, a higher preoperative serum CRP level was associated with local invasion and metastasis of RCC, but not with the response to treatment (data not shown). These findings suggest that the serum CRP level is associated with tumor aggressiveness, so that elevation of CRP might reflect the poorer general condition of the patient rather than the response to therapy. On the other hand, IFN-α has immunomodulatory effects and direct antitumor activity as well as antiangiogenic activity, including inhibition of VEGF [30]. Therefore, it is unclear exactly what serum IFNAR2 reflects, but it is likely to be associated with the overall immune status. Accordingly, we expected to find a relation between the preoperative serum levels of IFNAR2 mRNA and CRP, but no relation was observed.\nAlthough we could not examine how IFN-α binds to IFNAR2 on peripheral blood cells and then exhibits antitumor and antiangiogenic activity, our findings showed that the preoperative serum level of IFNAR2 mRNA was positively correlated with tumor size and was higher in patients with metastatic RCC who showed a good response to IFN-α ± sorafenib therapy than in those with a poor response. Because obtaining blood samples from patients is easier than harvesting tissue samples, chronological analysis of serum IFNAR2 mRNA levels is preferable for evaluation of the role of IFNAR2. While the effect of IFN-α therapy on the serum level of IFNAR2 mRNA is still unclear, our chronological evaluation of IFNAR2 mRNA throughout treatment showed that its serum level remained higher in the IFN-α ± Sor: CR-PR group than in the IFN-α + Sor: SD-PD group. Taken together, these observations suggest that an increased serum level of IFNAR2 mRNA may represent a systemic immunologic and antitumor response to the tumor burden in RCC patients, as well as showing antimicrobial activity if infection occurs.\nRegarding the effect of genetic polymorphism on the response of metastatic RCC to IFN-α, it has been reported that STAT3 polymorphism is a useful diagnostic marker for predicting the response to IFN-α therapy in these patients [55]. An efficient marker of the response to IFN-α is needed to establish individualized optimal treatment strategies, especially when newer therapies are used as first-line treatment for metastatic RCC. Our study showed that patients with higher serum levels of IFNAR2 mRNA may be more likely to respond to IFN-α ± sorafenib therapy and to show a longer overall survival, while patients with higher tumor tissue levels of IFNAR2 mRNA may show poor response and unfavorable overall survival. Although we need a surgical specimen to examine tumor tissue levels of IFNAR2 mRNA and protein, the serum level of IFNAR2 mRNA can be conveniently measured, so it may be more useful for predicting the response to IFN-α ± sorafenib therapy and as a prognostic indicator.", "While the efficacy of combination therapy with IFN-α and sorafenib needs clarification, the adverse effects of this regimen are related to the doses of each agent and are not additive [22–25]. We previously reported a good response to combination therapy with IFN-α and half the usual dose of sorafenib (400 mg/day rather than 800 mg/day) in patients with IFN-α-resistant RCC, along with tolerable adverse events [18].\nIt has also been reported that combined treatment with IFN-α + sorafenib suppresses proliferation and vascular endothelial growth factor (VEGF) production by several RCC cell lines more strongly than either agent alone [26, 27]. RCC is considered to be an immunogenic tumor [3], since cytotoxic T lymphocytes recognize and selectively kill autologous RCC cells, while tumor-specific T cells can be detected in the blood of RCC patients [3]. Sorafenib is a multikinase inhibitor targeting VEGF receptors 1–3, PDGFβ receptor, and Raf kinase, and it has both direct antitumor activity and antiangiogenic activity [28]. In a randomized phase III trial comparing sorafenib with placebo as second-line therapy for RCC, the response rate to sorafenib was 10%, and the stable disease rate was 74%, with the median progression-free survival time being 5.5 months in the sorafenib group versus 2.8 months in the placebo group [29]. In addition to its immunomodulatory effects, IFN-α also has direct antitumor activity as well as antiangiogenic activity, including inhibition of VEGF [30]. Antitumor immunity is usually suppressed in tumor-bearing mice because of the influence of regulatory T cells and suppressive cytokines, such as TGF-β and IL-10 [31]. Takeuchi et al. recently reported that the synergistic effect of sorafenib and IFN-α on RCC both in vitro and in tumor-bearing mice was related to a combination of antitumor, antiangiogenic, and immunologic responses [27]. In their study, sorafenib had no effect on the levels of natural killer (NK) cells, T cells, and regulatory T cells in the spleens of tumor-bearing BALB/c mice, irrespective of the use of IFN-α, while IFN-α showed weaker direct antitumor activity than sorafenib but stimulated CTL, NK cells, and tumor-infiltrating lymphocytes (which sorafenib did not), so that a synergistic antiproliferative effect of these two agents was demonstrated in vitro [27].\nAlthough an additive effect of IFN-α to sorafenib therapy in patients with metastatic RCC was recently reported [24], the clinical efficacy of this combination remains to be confirmed. Jonasch et al. reported that the outcome after addition of IFN-α to sorafenib was comparable to that of sorafenib monotherapy, when patients were randomized to treatment with either sorafenib (400 mg twice daily) or the combination of sorafenib (400 mg twice daily) plus IFN-α (0.5 MU twice daily) [25]. However, their IFN-α dose of 1 million units daily was probably too low to assess its additive effect because the average dose is 3–9 million units daily. So far, the published studies on combination therapy with IFN-α and sorafenib have employed concomitant therapy with both agents. In contrast, our treatment strategy was to add sorafenib (400 mg/day) to IFN-α in RCC patients whose tumors were refractory to IFN-α alone, i.e., first-line IFN-α monotherapy and second-line combination therapy with IFN-α plus sorafenib [18]. In the present study, 26 patients had metastatic disease (M1) at diagnosis. Among them, five patients showed a good response to IFN-α alone (IFN-α: CR-PR), 11 patients showed a good response to IFN-α + sorafenib (IFN-α + Sor: CR-PR, Table 1), and the remaining 10 patients had stable disease or showed a poor response to IFN-α + sorafenib (IFN-α + Sor: SD-PD, Table 1). Although half-dose sorafenib (400 mg/day) caused grade 1/2 toxicity, grade 3/4 toxicity was rare in the present study and such toxicity resolved when patients suspended sorafenib therapy. The response rate to the combination of IFN-α plus half-dose sorafenib was 52.4% (11/21 patients resistant to previous IFN-α monotherapy achieved CR or PR), indicating that sorafenib may be synergistic with IFN-α, leading to an increase of antitumor activity. Furthermore, the patients with good response to this combination therapy had favorable prognosis (Fig. 6). Thus, this combination seems to be tolerable and could be a useful treatment option for advanced RCC resistant to IFN-α monotherapy [18].", "The preoperative serum level of IFNAR2 mRNA was not correlated with the effect of IFN-α monotherapy, but a lower tumor tissue level of IFNAR2 mRNA was related to a better response to IFN-α. The present finding that a higher tumor tissue level of IFNAR2 mRNA was associated with a poor response to IFN-α is consistent with our previous results [14]. In contrast, regarding the relationship between IFNAR2 mRNA levels and the effect of combination therapy with IFN-α plus sorafenib, the preoperative serum level of IFNAR2 mRNA was higher in the IFN-α + Sor: CR-PR group and the IFN-α: CR-PR group than in the IFN-α + Sor: SD-PD group (Fig. 6e), while the tumor tissue level of IFNAR2 mRNA was lower in the IFN-α + Sor: CR-PR group and the IFN-α: CR-PR group than in the IFN-α + Sor: SD-PD group (Fig. 6f). These findings suggested that a different molecular mechanism might be involved in IFN-α ± Sor: CR-PR group and IFN-α + Sor: SD-PD group.\nIFNs are pleiotropic cytokines that regulate antiviral, antitumor, apoptotic, antiangiogenic, and cellular immune responses via activation of multiple downstream signaling cascades, including the Janus tyrosine kinase (Jak)-signal transducer and activator of transcription (STAT) pathway, the p38 mitogen-activated protein kinase (MAPK) pathway, and the mTOR pathway [32, 33]. The Jak-STAT and p38-MAPK signaling pathways have been shown to be responsible for transcription of the genes encoding proteins related to the antiviral and/or antiproliferative effects of IFNs. On the other hand, it has been reported that activation of the mTOR pathway by IFNs has an important regulatory role in mRNA translation and induction of the interferon response [34, 35]. IFN-α-induced tumor cell apoptosis is also mediated via the mTOR pathway in a nucleus-independent manner [36, 37]. Moreover, the Jak-STAT and mTOR pathways act separately from each other after activation by IFN-α [37]. Thus, it is likely that mTOR signaling selectively mediates apoptosis and survival.\nThe rapamycin-sensitive mTOR-raptor (regulatory-associated protein of mTOR) complex controls cell growth by regulating protein synthesis, so mTOR-raptor signaling is a potential antitumor target, and mTOR inhibitors are currently under investigation for the treatment of various human cancers. On the other hand, mTOR also interacts with rictor (rapamycin-insensitive companion of mTOR), and recent findings have suggested that the rapamycin-insensitive effect of mTOR on cell survival is overactive in many cancers. Thus, mTOR has dual rapamycin-sensitive (mTOR-raptor complex: mTORC1) and insensitive (mTOR-rictor complex: mTORC2) functions, indicating that treatment with rapamycin will not completely inhibit mTOR activity [38, 39]. Phosphatidylinositol 3`kinase (PI3 K), serine/threonine kinase Akt, and the mTOR pathway are all overactive in human cancers. mTORC1 lies downstream of PI3 K and is part of a pathway that is frequently activated in human cancers, so mTORC1 represents a pivotal target for anticancer therapy. The best-characterized pathways regulated by mTORC1 are phosphorylation and activation of ribosomal S6 kinase-1 (S6K1) and phosphorylation and inactivation of 4E-BP1, the suppressor of mRNA cap-binding protein eIF4E, leading to effects on cell growth and metabolism by acting as a restriction point in cells subjected to stresses [40, 41] such as hypoxia [42–44].\nPhosphorylation at two sites is required for full activation of Akt, since it is phosphorylated by PI3 K-dependent kinase-1 (PDK1) at a threonine residue in the catalytic domain (Thr 308) and by PI3 K-dependent kinase-2 (PDK2) at a serine residue (Ser 473) in the carboxy-terminal hydrophobic motif [45]. It has been reported that mTORC2 regulates the actin cytoskeleton and also possesses PDK2 activity that phosphorylates Ser-473 in the carboxy-terminal of Akt, making it essential for Akt activity [46]. Importantly, activation of Akt may lead to cell survival when mTORC1 is inhibited or could potentially increase VEGF production because PI3 K/Akt signaling induces tumor angiogenesis by regulating VEGF via both HIF-1α-dependent and -independent mechanisms [47]. It has been reported that hypoxia-inducible factor (HIF) 1α expression is dependent on both raptor and rictor, whereas HIF2α expression only depends on rictor and HIF2α is more important in RCC [48]. These findings suggest that phosphorylation of Ser 473 in AKT is a key molecular step in the progression of RCC and could be a target for treating these tumors [25]. In agreement with this, our current study showed that the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated and that both were related to tumor metastatic potential, while a high phosphorylated Akt (Ser-473) level in the primary tumor was associated with a poor response to IFN-α ± sorafenib therapy (stable or progressive disease).\nCurrent efforts to achieve the clinical development of mTOR inhibitors are based on the role of mTOR signaling in promoting the proliferation and survival of tumor cells. It has been reported that treatment with mTOR inhibitors can improve the outcome of patients with metastatic RCC [8–10]. On the other hand, the mTOR pathway is also important for IFN-dependent translational responses, and IFN-α is widely used to treat advanced RCC. Although we could not exclude a possible detrimental effect of IFN-α treatment in the patients with metastatic RCC and higher IFNAR2 mRNA levels in their tumors, our findings suggested that there may be different molecular mechanisms of cancer progression in the patients with a good or poor response to IFN-α ± sorafenib. It is possible that IFNAR2 signaling has different biological effects from normal when upregulated in RCC.\nJonasch et al. recently reported that an increase of phosphorylated Akt (Ser-473) was associated with worse survival by microarray analysis of paraffin-embedded specimens [25]. In the present study, the tumors with higher phosphorylated Akt (Ser-473) levels, but not higher phospho-S6 ribosomal protein (Ser-235/236) levels, were resistant to IFN-α ± sorafenib therapy. In addition, the tumor tissue levels of IFNAR2 mRNA and phosphorylated S6 ribosomal protein (Ser-235/236) were positively correlated, and both were related to metastatic potential. Tumor levels of IFNAR2 mRNA also had a weak positive correlation with those of phosphorylated Akt (Ser-473). Moreover, the patients whose tumors had higher levels of phosphorylated Akt (Ser-473), phosphorylated S6 ribosomal protein (Ser-235/236), and IFNAR2 mRNA showed shorter overall survival. Taken together, it is possible that phosphorylation of Ser 473 on Akt is a key molecular step in the progression of RCC and a potential therapeutic target, so that the tumor level of phosphorylated Akt (Ser-473) may be useful for predicting the response to treatment. At present, it remains to be elucidated why upregulation of IFNAR2 expression is linked to the progression of RCC and to a poor response to treatment, and it is unclear how IFNAR2 interacts with mTORC1 and mTORC2, but our findings suggested that the IFNAR2-mTORC1 pathway via phosphorylated S6 ribosomal protein (Ser-235/236) may act locally within tumors to promote proliferation and metastasis by modifying mRNA translation, while the IFNAR2-mTORC2 pathway via phosphorylated Akt (Ser-473) may be associated with tumor resistance. So, these interactions should be elucidated in the future. As Lekmine et al. have indicated, therefore, caution should be exercised when designing clinical trials that combine an mTOR inhibitor and IFN-α due to possible antagonism of antitumor activity [34]. In fact, Huges et al. reported that patients treated with temsirolimus alone had better overall survival than those given IFN-α alone, while patients treated with temsirolimus plus IFN-α did not [49]. In the future, the downstream targets of IFNAR2 should be identified, and the expression or activity of one or two such targets should be studied in cell lines or tissue samples. A better understanding of the IFNAR2 pathway may help to elucidate its role in cancer.\nAlthough none of our patients were treated by sorafenib alone, it would be interesting to assess the expression of not only IFNAR2, phosphorylated Akt (Ser-473), and phosphorylated S6 ribosomal protein (Ser-235/236), but also VEGF receptor and Raf, in tumor cells and the effects of sorafenib, sunitinib, or mTOR inhibitors. Such information could lead to elucidation of the role of the IFNAR2-mTOR pathway in the progression of RCC and the selection of patients who will benefit from treatment with IFN-α, sorafenib, sunitinib, or mTOR inhibitors.", "The serum CRP level is associated with the stage and outcome of RCC [50, 51]. Elevation of CRP is primarily determined by an increase of circulating IL-6 [52], and the IL-6 level is correlated with the serum CRP level as well as with tumor histological grade and tumor metastasis [53]. We previously reported that increased serum levels of CRP and IL-6 were associated with local tumor invasion and metastasis [54]. In the present study, a higher preoperative serum CRP level was associated with local invasion and metastasis of RCC, but not with the response to treatment (data not shown). These findings suggest that the serum CRP level is associated with tumor aggressiveness, so that elevation of CRP might reflect the poorer general condition of the patient rather than the response to therapy. On the other hand, IFN-α has immunomodulatory effects and direct antitumor activity as well as antiangiogenic activity, including inhibition of VEGF [30]. Therefore, it is unclear exactly what serum IFNAR2 reflects, but it is likely to be associated with the overall immune status. Accordingly, we expected to find a relation between the preoperative serum levels of IFNAR2 mRNA and CRP, but no relation was observed.\nAlthough we could not examine how IFN-α binds to IFNAR2 on peripheral blood cells and then exhibits antitumor and antiangiogenic activity, our findings showed that the preoperative serum level of IFNAR2 mRNA was positively correlated with tumor size and was higher in patients with metastatic RCC who showed a good response to IFN-α ± sorafenib therapy than in those with a poor response. Because obtaining blood samples from patients is easier than harvesting tissue samples, chronological analysis of serum IFNAR2 mRNA levels is preferable for evaluation of the role of IFNAR2. While the effect of IFN-α therapy on the serum level of IFNAR2 mRNA is still unclear, our chronological evaluation of IFNAR2 mRNA throughout treatment showed that its serum level remained higher in the IFN-α ± Sor: CR-PR group than in the IFN-α + Sor: SD-PD group. Taken together, these observations suggest that an increased serum level of IFNAR2 mRNA may represent a systemic immunologic and antitumor response to the tumor burden in RCC patients, as well as showing antimicrobial activity if infection occurs.\nRegarding the effect of genetic polymorphism on the response of metastatic RCC to IFN-α, it has been reported that STAT3 polymorphism is a useful diagnostic marker for predicting the response to IFN-α therapy in these patients [55]. An efficient marker of the response to IFN-α is needed to establish individualized optimal treatment strategies, especially when newer therapies are used as first-line treatment for metastatic RCC. Our study showed that patients with higher serum levels of IFNAR2 mRNA may be more likely to respond to IFN-α ± sorafenib therapy and to show a longer overall survival, while patients with higher tumor tissue levels of IFNAR2 mRNA may show poor response and unfavorable overall survival. Although we need a surgical specimen to examine tumor tissue levels of IFNAR2 mRNA and protein, the serum level of IFNAR2 mRNA can be conveniently measured, so it may be more useful for predicting the response to IFN-α ± sorafenib therapy and as a prognostic indicator.", "This study showed that an elevated serum level of IFNAR2 mRNA could predict the response of metastatic RCC to IFN-α ± sorafenib and may be associated with favorable prognosis." ]

[ "introduction", "materials|methods", null, null, null, null, "results", null, null, null, null, null, "discussion", null, null, null, "conclusion" ]

[ "Interferon alpha", "Receptor", "mTOR", "Sorafenib", "Renal cell carcinoma", "Metastasis" ]

[ "Adult", "Disease Progression", "Dose-Response Relationship, Drug", "Female", "Follow-Up Studies", "Humans", "Immunologic Factors", "Interferon-beta", "Magnetic Resonance Imaging", "Male", "Multiple Sclerosis, Relapsing-Remitting", "Product Surveillance, Postmarketing", "Secondary Prevention", "Severity of Illness Index" ]

[ "Background", "Study design and participants", "Outcome measures", "Statistical analysis", "Results", "Participants", "Risk of relapses or worsening in disability after the switch", "Discussion", "Conclusions", "List of abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Three formulation of Interferon beta (IFNB) are nowadays approved as disease modifying agents for subjects with relapsing-remitting (RR) Multiple Sclerosis (MS): IFNB-1b 250 mcg subcutaneous every other day (s.c., e.o.d.) (Betaferon, Bayer Schering, Berlin, Germany); IFNB-1a 30 mcg intramuscular once weekly (i.m., o.w.) (Avonex, Biogen Idec, Cambridge, USA); IFNB-1a 22 or 44 mcg subcutaneous three times per week (s.c., t.p.w.) (Rebif, Merck Serono, Geneva, Switzerland). Large randomised trials demonstrated that IFNB reduces the frequency of MS attacks and the number of new lesions on magnetic resonance imaging (MRI) [1-6]. Also, post-marketing studies suggest a long-term benefit of IFNB treatment on disease activity and disability progression [7-10]. Nevertheless, a considerable number of patients experience a breakthrough disease, i.e. evidence of clinical or imaging disease activity or progression despite the IFNB treatment [11]. Therefore, a significant proportion of subjects requires treatment switching or escalation in order to avoid accumulation of irreversible disability.\nTo date, there is no evidence to guide an alternative therapy in patients with breakthrough disease [11]. However, in daily clinical setting some patients starting with the low-dose (i.m., o.w. 30 mcg, or s.c., t.p.w. 22 mcg IFNB-1a) are routinely switched to the high-dose IFNB (s.c., e.o.d. 250 mcg IFNB-1b, or s.c., t.p.w. 44 mcg IFNB-1a) in case of breakthrough disease, according to suggestions coming from therapeutic recommendation of consensus groups [12,13].\nTwo head-to-head studies comparing i.m. 30 mcg IFNB-1a o.w. with s.c. 250 mcg IFNB-1b e.o.d. and 44 mcg IFNB-1a t.p.w. (INCOMIN and EVIDENCE, respectively) provided evidences that naive patients started with the high-dose IFNB had a better outcome than those receiving a low dose regimen [14,15]. It has also been suggested that increasing the IFNB dose may be useful in reducing mean relapse rate and the accumulation of subclinical lesions as detected on MRI [16,17]. However, these studies have not investigated whether the criteria adopted to switch patients to the high-dose IFNB could have an influence on the clinical outcomes.\nThe 2-year clinical outcomes of patients switched from the low-dose to the high and/or more frequently administered dose of IFNB are reported in the present observational study. Moreover, we attempted to identify which patients with breakthrough disease could benefit from an increase of the IFNB dose.", "Patients affected by RRMS [18] according to Poser [19] or McDonald Criteria [20], and switched to a more frequently administered and/or higher dose of IFNB (Betaferon or Rebif 44) in consequence of breakthrough disease during the low-dose IFNB regimen (Avonex or Rebif 22) were included in this 2-year, independent, observational, post-marketing study at the MS Centre of S. Andrea Hospital in Rome. There was no well-defined protocol for switching patients to the high-dose IFNB (occurrence of one or more relapses and/or MRI activity), but it was a decision taken by the neurologist depending on the growing availability over years of new drugs active against MS (i.e., Natalizumab, experimental treatment, etc).\nPatients were regularly followed-up from the start of the IFNB treatment; clinical and MRI data were prospectively collected and stored into an electronic database (after obtaining an informed consent by each patient). The local Ethical committee board provides exemption of approval for post-marketing prospective studies.\nThe study design with relative time-points are described in Figure 1. We defined as \"switch\" the time-point at which patients interrupted the low-dose and started the higher IFNB dose; therefore, we defined a pre-switch period (i.e., the time frame elapsed between the start of the IFNB treatment and the switch) and a 2-year observational post-switch period.\nStudy design with relative time-points. Note that pre-IFNB and low IFNB dose periods had a different duration for each patient, while after switch all patients were followed for 2 years.\nPatients receiving a low-dose IFNB only for titration, or those discontinuing the IFNB treatment mainly for poor tolerance, or those with previous exposure to immunomodulant or immunosuppressive agents were excluded from this study.\nClinical data, including the occurrence of relapses and the Expanded Disability Status Scale (EDSS) score [21], were evaluated for each subject every 3 months. Unscheduled visits were also performed in suspect of relapse or any other clinically relevant condition.\nAn exacerbation was defined as the appearance or reappearance of one or more symptoms attributable to MS, accompanied by objective deterioration on neurological examination lasting at least 24 hours, in the absence of fever and preceded by neurological stability for at least 30 days [19,20].\nWe also collected yearly brain and spinal cord MRI since the start of treatment with IFNB, focusing on the absence/presence of MRI activity (i.e. the detection of at least one contrast-enhancing lesions); additional MRI scans were performed, if necessary. Clinical examination (including EDSS score) and MRI scan were done in stable patients (i.e. at least 30 days after the last assumption of steroids administered for relapses).", "We considered two main outcomes over the 2-year observation period: (a) the occurrence of at least one relapse, and (b) a sustained progression of 1 or more points on EDSS score (starting after the switch and confirmed in two consecutive neurological visits separated by at least a 6-month interval).", "All values are expressed as a mean ± standard deviation (± SD) or median (range), as appropriate. Differences between groups were tested by using the Chi-square test and the U Mann-Whitney test.\nTwo Cox proportional hazards models were built for identifying the predictors of experiencing a relapse or a sustained progression on EDSS score after the switch. As main time variable for time-to-event analyses we considered the interval (in years) elapsed between the switch and last visit, or high IFNB dose discontinuation, or outcome reach, whichever came first.\nGender, age, MS duration, EDSS score, presence/absence of an active MRI (each considered at the time of switch), pre-IFNB annualised relapse rate, IFNB type (Betaferon or Rebif 44), as well as duration of initial IFNB treatment, number of relapses, and EDSS change occurred during the low-dose IFNB regimen were included as covariates in each model. Variables were added in the models in a forward stepwise fashion, and interactions terms were tested, where appropriate. All models were stratified by IFNB type received by patients before the switch (Avonex or Rebif 22).\nData have been analysed by using the Statistical Package for Social Sciences, version 16.0 (SPSS, Chicago, IL, USA).", "[SUBTITLE] Participants [SUBSECTION] We examined 283 patients starting the low-dose IFNB formulation from October 1997 to March 2008 (Avonex, n = 120; Rebif 22, n = 163). Mean (SD) duration of IFNB treatment was 4.3 (2.3) years, and mean annualised relapse rate was 0.48 (range 0-3). A total of 108 (38.1%) patients had a sustained EDSS increase during IFNB treatment.\nOut of these 283 patients, 121 patients switched to the high-dose IFNB, while 162 continued to receive the low-dose IFNB. Table 1 shows the demographic and clinical characteristics of patients who switched to the high-dose IFNB and those who did not. At the start of treatment with IFNB there were no differences between the two groups.\nDemographic and clinical characteristics of 283 patients starting a low IFNB dose.\nAll value are expressed as mean (standard deviation), unless indicated otherwise.\nAll p-values are > 0.05\nAmong the 121 patients who increased the IFNB dose, therapy switching included: Rebif 22→Rebif 44 (n = 59, 47.9%), Avonex→Rebif 44 (n = 46, 38.4%), Avonex→Betaferon (n = 16, 13.7%). After the switch to the high-dose IFNB, 72 (59.5%) patients experienced a relapse, and 51 (42.1%) patients reached a sustained progression on EDSS score. Overall, 85 (70.3%) patients showed some clinical disease activity (i.e. relapse or disability progression) over the 2-year observation period after the switch.\nWe examined 283 patients starting the low-dose IFNB formulation from October 1997 to March 2008 (Avonex, n = 120; Rebif 22, n = 163). Mean (SD) duration of IFNB treatment was 4.3 (2.3) years, and mean annualised relapse rate was 0.48 (range 0-3). A total of 108 (38.1%) patients had a sustained EDSS increase during IFNB treatment.\nOut of these 283 patients, 121 patients switched to the high-dose IFNB, while 162 continued to receive the low-dose IFNB. Table 1 shows the demographic and clinical characteristics of patients who switched to the high-dose IFNB and those who did not. At the start of treatment with IFNB there were no differences between the two groups.\nDemographic and clinical characteristics of 283 patients starting a low IFNB dose.\nAll value are expressed as mean (standard deviation), unless indicated otherwise.\nAll p-values are > 0.05\nAmong the 121 patients who increased the IFNB dose, therapy switching included: Rebif 22→Rebif 44 (n = 59, 47.9%), Avonex→Rebif 44 (n = 46, 38.4%), Avonex→Betaferon (n = 16, 13.7%). After the switch to the high-dose IFNB, 72 (59.5%) patients experienced a relapse, and 51 (42.1%) patients reached a sustained progression on EDSS score. Overall, 85 (70.3%) patients showed some clinical disease activity (i.e. relapse or disability progression) over the 2-year observation period after the switch.\n[SUBTITLE] Risk of relapses or worsening in disability after the switch [SUBSECTION] One hundred (82.6%) patients switched to the high-dose IFNB due to MS attacks (40 also had an active MRI scan at switch), while 21 (17.4%) patients switched due to evidence of activity on MRI scan, but not relapses, during the low-dose IFNB regimen.\nOnly 5 patients (23.8%) who were switched on the basis of MRI activity experienced a further relapse, while 67 (67.0%) of patients switching because of relapses had a further exacerbation over the 2-year after the increase of IFNB dose (p < 0.001).\nNo significant differences in proportion of patients reaching a sustained progression on EDSS score were observed between patients who increased the IFNB dose because of relapses and those who switched only on the basis of MRI (p = 0.08).\nAccording to the Cox model, when compared with patients switched only on the basis of MRI activity (i.e. no relapses during the low-dose IFNB regimen), those relapsing were more likely of having a further relapse even after the switch (HR: 5.55, 95% C.I. 2.22 - 13.85, p = 0.001). Moreover, this risk of relapsing during the 2-year observational period was related with a younger age at switch (HR: 0.94, 95% C.I. 0.91 - 0.97; p < 0.001), and the number of clinical bouts occurred before the switch: one bout confers a risk ratio of 3.56 (95% C.I. 1.34 - 8.42; p = 0.01), two or more bouts a risk ratio of 7.89 (95% C.I. 3.10 - 19.85; p < 0.001) (see Table 2). The risk of further relapses was not increased in patients who switched to the high-dose IFNB because of a combination of clinical and MRI activity at switch.\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for relapsing after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale\n(*) this value refers to patients switched only on the basis of an active MRI scan\nThe variables predictive for reaching a sustained disability progression after the switch were the EDSS score at baseline (HR: 1.77, 95% C.I. 1.39 - 2.26; p < 0.001), and the combination of clinical and radiological activity at switch (HR: 2.14, 95% C.I. 1.16 - 3.96; p = 0.01) (see Table 3).\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for worsening in disability after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale.\nOne hundred (82.6%) patients switched to the high-dose IFNB due to MS attacks (40 also had an active MRI scan at switch), while 21 (17.4%) patients switched due to evidence of activity on MRI scan, but not relapses, during the low-dose IFNB regimen.\nOnly 5 patients (23.8%) who were switched on the basis of MRI activity experienced a further relapse, while 67 (67.0%) of patients switching because of relapses had a further exacerbation over the 2-year after the increase of IFNB dose (p < 0.001).\nNo significant differences in proportion of patients reaching a sustained progression on EDSS score were observed between patients who increased the IFNB dose because of relapses and those who switched only on the basis of MRI (p = 0.08).\nAccording to the Cox model, when compared with patients switched only on the basis of MRI activity (i.e. no relapses during the low-dose IFNB regimen), those relapsing were more likely of having a further relapse even after the switch (HR: 5.55, 95% C.I. 2.22 - 13.85, p = 0.001). Moreover, this risk of relapsing during the 2-year observational period was related with a younger age at switch (HR: 0.94, 95% C.I. 0.91 - 0.97; p < 0.001), and the number of clinical bouts occurred before the switch: one bout confers a risk ratio of 3.56 (95% C.I. 1.34 - 8.42; p = 0.01), two or more bouts a risk ratio of 7.89 (95% C.I. 3.10 - 19.85; p < 0.001) (see Table 2). The risk of further relapses was not increased in patients who switched to the high-dose IFNB because of a combination of clinical and MRI activity at switch.\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for relapsing after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale\n(*) this value refers to patients switched only on the basis of an active MRI scan\nThe variables predictive for reaching a sustained disability progression after the switch were the EDSS score at baseline (HR: 1.77, 95% C.I. 1.39 - 2.26; p < 0.001), and the combination of clinical and radiological activity at switch (HR: 2.14, 95% C.I. 1.16 - 3.96; p = 0.01) (see Table 3).\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for worsening in disability after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale.", "We examined 283 patients starting the low-dose IFNB formulation from October 1997 to March 2008 (Avonex, n = 120; Rebif 22, n = 163). Mean (SD) duration of IFNB treatment was 4.3 (2.3) years, and mean annualised relapse rate was 0.48 (range 0-3). A total of 108 (38.1%) patients had a sustained EDSS increase during IFNB treatment.\nOut of these 283 patients, 121 patients switched to the high-dose IFNB, while 162 continued to receive the low-dose IFNB. Table 1 shows the demographic and clinical characteristics of patients who switched to the high-dose IFNB and those who did not. At the start of treatment with IFNB there were no differences between the two groups.\nDemographic and clinical characteristics of 283 patients starting a low IFNB dose.\nAll value are expressed as mean (standard deviation), unless indicated otherwise.\nAll p-values are > 0.05\nAmong the 121 patients who increased the IFNB dose, therapy switching included: Rebif 22→Rebif 44 (n = 59, 47.9%), Avonex→Rebif 44 (n = 46, 38.4%), Avonex→Betaferon (n = 16, 13.7%). After the switch to the high-dose IFNB, 72 (59.5%) patients experienced a relapse, and 51 (42.1%) patients reached a sustained progression on EDSS score. Overall, 85 (70.3%) patients showed some clinical disease activity (i.e. relapse or disability progression) over the 2-year observation period after the switch.", "One hundred (82.6%) patients switched to the high-dose IFNB due to MS attacks (40 also had an active MRI scan at switch), while 21 (17.4%) patients switched due to evidence of activity on MRI scan, but not relapses, during the low-dose IFNB regimen.\nOnly 5 patients (23.8%) who were switched on the basis of MRI activity experienced a further relapse, while 67 (67.0%) of patients switching because of relapses had a further exacerbation over the 2-year after the increase of IFNB dose (p < 0.001).\nNo significant differences in proportion of patients reaching a sustained progression on EDSS score were observed between patients who increased the IFNB dose because of relapses and those who switched only on the basis of MRI (p = 0.08).\nAccording to the Cox model, when compared with patients switched only on the basis of MRI activity (i.e. no relapses during the low-dose IFNB regimen), those relapsing were more likely of having a further relapse even after the switch (HR: 5.55, 95% C.I. 2.22 - 13.85, p = 0.001). Moreover, this risk of relapsing during the 2-year observational period was related with a younger age at switch (HR: 0.94, 95% C.I. 0.91 - 0.97; p < 0.001), and the number of clinical bouts occurred before the switch: one bout confers a risk ratio of 3.56 (95% C.I. 1.34 - 8.42; p = 0.01), two or more bouts a risk ratio of 7.89 (95% C.I. 3.10 - 19.85; p < 0.001) (see Table 2). The risk of further relapses was not increased in patients who switched to the high-dose IFNB because of a combination of clinical and MRI activity at switch.\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for relapsing after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale\n(*) this value refers to patients switched only on the basis of an active MRI scan\nThe variables predictive for reaching a sustained disability progression after the switch were the EDSS score at baseline (HR: 1.77, 95% C.I. 1.39 - 2.26; p < 0.001), and the combination of clinical and radiological activity at switch (HR: 2.14, 95% C.I. 1.16 - 3.96; p = 0.01) (see Table 3).\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for worsening in disability after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale.", "Results from the present study suggest that the risk of having a relapse despite the increase of IFNB dose raised according to the number of clinical attacks occurred during the assumption of the low-dose IFNB, and it is reduced when the switching was based on MRI findings rather than on clinical activity. Moreover, the chance of being progression-free after the switch is related to a lower EDSS score and the absence of clinical and MRI activities. At this regard, we may assume that in some patients an incomplete recovery from relapse occurred, with a substantial influence on their EDSS scores.\nOverall, less than 1/3 of our patients remained free from clinical disease activity (i.e. absence of relapse and sustained progression on EDSS score) over the 2-year observational period following the increase of IFNB dose. This could imply that the majority of switchers should be considered poor responder to IFNB therapy regardless the dose and/or frequency of administration, and further supports the hypothesis that a treatment strategy encompassing the increase of IFNB dose should be useful only in some selected cases.\nAfter the increase of IFNB dose, the majority of patients switched only for the evidence of MRI activity had a better clinical outcome than those switched because of the occurrence of relapses. Therefore, we might suggest that monitoring the effect of IFNB treatment with regular MRI scans is recommendable even in absence of clinical relapses.\nIt has been known that conventional MRI represents a powerful tool to monitor latent disease activity, providing a measurable and sensible marker of response to IFNB therapy [22-25]. However, we cannot exclude that patients switched only on the basis of MRI findings might also have had good outcomes without switching to the high-dose IFNB, as the absence of a control group. At this regard, Rio and colleagues showed that patients with only MRI activity and no relapses in the first year of IFNB treatment did not experience an increase of relapses or disability over a 3-year follow-up [26].\nAlthough randomized clinical trials demonstrated a more pronounced effect of high-dose, high-frequency IFNB when compared with both the low-dose, equal-frequency [1,27] and the low-dose, low-frequency regimens [14,15] in naïve patients, data on the effectiveness of increasing the IFNB dose in patients with breakthrough disease are scarce. The open-label extension phase of the EVIDENCE study, involving 223 patients converted from Avonex to Rebif 44, documented a 50% reduction in the annualised relapse rate after the switch [16]. However, we must consider also that in the EVIDENCE study all patients originally randomized to Avonex were offered to receive Rebif 44, independently from the response status during the blind phase of the study. While the authors suggested that increasing the IFNB-1a dose and frequency could rapidly reduce ongoing disease activity, they cannot discharge the hypothesis that the significant reduction in relapse rate might be due, at least in part, to the regression to the mean phenomenon.\nSome open-label studies exploring the usefulness of switching among immunomodulating drugs had different designs and provided conflicting results [28-30]. Two studies reported a decrease in both relapse rate and proportion of relapse-free subjects after the switch [29,30], whilst the QUASIMS study did not provide any support of more favourable outcomes after switching from an IFNB formulation to another [28]. One possible explanation of this discrepancy is that these observational surveys considered different subtype of switch (i.e. from IFNB-1a to IFNB-1b and viceversa, from IFNB to GA, etc.); also, these studies were not specifically aimed to determine the crude effect of an increase of the IFNB dose. Furthermore, in some studies the patients were switched on the basis of tolerability rather than a persistent disease activity.\nIn the present study, patients had variable periods of observation before and after the switch, thus precluding any attempt to estimate the effectiveness of an increase of the IFNB dose in suppressing disease activity and slowing disability progression.\nBeing an observational report, our study suffers from other limits, such as the small sample size, the unavailability of control group, blindness and randomization, as well as the lack of data on neutralising antibodies (NAbs) against IFNB. However, there is evidence that NAbs presence could explain only the 20% of the suboptimal response to IFNB treatment [31].\nDespite these limits, our study might contribute to define a therapeutic algorithm to manage breakthrough disease in patients on treatment with a low-dose IFNB. The identification of patients with subclinical disease activity during the low-dose IFNB treatment, and an early switch to the high-dose IFNB, seem to be effective in achieving a better control of the disease. On the contrary, when relapses occurred during the pre-switch period, especially in combination with MRI activity, patients did not seem benefit from the increase of the IFNB dose in the following years.\nSince efficacy of GA has been demonstrated comparable to IFNB [32-34], switching among immunomodulating treatments may represent an interesting approach in case of treatment failure [35,36]. The scientific rationale for switching to other therapies is strongest for patients on IFNB therapy with persistent high-titre of NAbs [37]. However, a more aggressive approach (Natalizumab or Mitoxantrone) is warranted for patients at high risk of accumulation of fixed disability or with shorter intervals between attacks [13].", "We suggest neurologists to consider the presence of sub-clinical activity, as detected on MRI, in the decision to switch a patient from the low-dose to the high-dose IFNB, even in absence of clinical relapses. Further efforts are warranted to clearly define whether MRI findings might be considered a key element in the choice of increase the IFNB dose when the response to the low-dose IFNB is suboptimal.", "RRMS: relapsing-remitting multiple sclerosis; EDSS: Expanded Disability Status Scale; MRI: magnetic resonance imaging; CELs: contrast enhancing lesions; IFNB: Interferon beta.", "The authors declare that they have no competing interests.", "LP and CP had full access to the data, and take the final responsibility for the content of the present manuscript. LP and CP are responsible for concept and design of the study, contributed to data analysis and manuscript drafting; GB and LL contributed to data analysis and interpretation and manuscript revision; LDG, LL and VB collected the data. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2377/11/26/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study design and participants", "Outcome measures", "Statistical analysis", "Results", "Participants", "Risk of relapses or worsening in disability after the switch", "Discussion", "Conclusions", "List of abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Three formulation of Interferon beta (IFNB) are nowadays approved as disease modifying agents for subjects with relapsing-remitting (RR) Multiple Sclerosis (MS): IFNB-1b 250 mcg subcutaneous every other day (s.c., e.o.d.) (Betaferon, Bayer Schering, Berlin, Germany); IFNB-1a 30 mcg intramuscular once weekly (i.m., o.w.) (Avonex, Biogen Idec, Cambridge, USA); IFNB-1a 22 or 44 mcg subcutaneous three times per week (s.c., t.p.w.) (Rebif, Merck Serono, Geneva, Switzerland). Large randomised trials demonstrated that IFNB reduces the frequency of MS attacks and the number of new lesions on magnetic resonance imaging (MRI) [1-6]. Also, post-marketing studies suggest a long-term benefit of IFNB treatment on disease activity and disability progression [7-10]. Nevertheless, a considerable number of patients experience a breakthrough disease, i.e. evidence of clinical or imaging disease activity or progression despite the IFNB treatment [11]. Therefore, a significant proportion of subjects requires treatment switching or escalation in order to avoid accumulation of irreversible disability.\nTo date, there is no evidence to guide an alternative therapy in patients with breakthrough disease [11]. However, in daily clinical setting some patients starting with the low-dose (i.m., o.w. 30 mcg, or s.c., t.p.w. 22 mcg IFNB-1a) are routinely switched to the high-dose IFNB (s.c., e.o.d. 250 mcg IFNB-1b, or s.c., t.p.w. 44 mcg IFNB-1a) in case of breakthrough disease, according to suggestions coming from therapeutic recommendation of consensus groups [12,13].\nTwo head-to-head studies comparing i.m. 30 mcg IFNB-1a o.w. with s.c. 250 mcg IFNB-1b e.o.d. and 44 mcg IFNB-1a t.p.w. (INCOMIN and EVIDENCE, respectively) provided evidences that naive patients started with the high-dose IFNB had a better outcome than those receiving a low dose regimen [14,15]. It has also been suggested that increasing the IFNB dose may be useful in reducing mean relapse rate and the accumulation of subclinical lesions as detected on MRI [16,17]. However, these studies have not investigated whether the criteria adopted to switch patients to the high-dose IFNB could have an influence on the clinical outcomes.\nThe 2-year clinical outcomes of patients switched from the low-dose to the high and/or more frequently administered dose of IFNB are reported in the present observational study. Moreover, we attempted to identify which patients with breakthrough disease could benefit from an increase of the IFNB dose.", "[SUBTITLE] Study design and participants [SUBSECTION] Patients affected by RRMS [18] according to Poser [19] or McDonald Criteria [20], and switched to a more frequently administered and/or higher dose of IFNB (Betaferon or Rebif 44) in consequence of breakthrough disease during the low-dose IFNB regimen (Avonex or Rebif 22) were included in this 2-year, independent, observational, post-marketing study at the MS Centre of S. Andrea Hospital in Rome. There was no well-defined protocol for switching patients to the high-dose IFNB (occurrence of one or more relapses and/or MRI activity), but it was a decision taken by the neurologist depending on the growing availability over years of new drugs active against MS (i.e., Natalizumab, experimental treatment, etc).\nPatients were regularly followed-up from the start of the IFNB treatment; clinical and MRI data were prospectively collected and stored into an electronic database (after obtaining an informed consent by each patient). The local Ethical committee board provides exemption of approval for post-marketing prospective studies.\nThe study design with relative time-points are described in Figure 1. We defined as \"switch\" the time-point at which patients interrupted the low-dose and started the higher IFNB dose; therefore, we defined a pre-switch period (i.e., the time frame elapsed between the start of the IFNB treatment and the switch) and a 2-year observational post-switch period.\nStudy design with relative time-points. Note that pre-IFNB and low IFNB dose periods had a different duration for each patient, while after switch all patients were followed for 2 years.\nPatients receiving a low-dose IFNB only for titration, or those discontinuing the IFNB treatment mainly for poor tolerance, or those with previous exposure to immunomodulant or immunosuppressive agents were excluded from this study.\nClinical data, including the occurrence of relapses and the Expanded Disability Status Scale (EDSS) score [21], were evaluated for each subject every 3 months. Unscheduled visits were also performed in suspect of relapse or any other clinically relevant condition.\nAn exacerbation was defined as the appearance or reappearance of one or more symptoms attributable to MS, accompanied by objective deterioration on neurological examination lasting at least 24 hours, in the absence of fever and preceded by neurological stability for at least 30 days [19,20].\nWe also collected yearly brain and spinal cord MRI since the start of treatment with IFNB, focusing on the absence/presence of MRI activity (i.e. the detection of at least one contrast-enhancing lesions); additional MRI scans were performed, if necessary. Clinical examination (including EDSS score) and MRI scan were done in stable patients (i.e. at least 30 days after the last assumption of steroids administered for relapses).\nPatients affected by RRMS [18] according to Poser [19] or McDonald Criteria [20], and switched to a more frequently administered and/or higher dose of IFNB (Betaferon or Rebif 44) in consequence of breakthrough disease during the low-dose IFNB regimen (Avonex or Rebif 22) were included in this 2-year, independent, observational, post-marketing study at the MS Centre of S. Andrea Hospital in Rome. There was no well-defined protocol for switching patients to the high-dose IFNB (occurrence of one or more relapses and/or MRI activity), but it was a decision taken by the neurologist depending on the growing availability over years of new drugs active against MS (i.e., Natalizumab, experimental treatment, etc).\nPatients were regularly followed-up from the start of the IFNB treatment; clinical and MRI data were prospectively collected and stored into an electronic database (after obtaining an informed consent by each patient). The local Ethical committee board provides exemption of approval for post-marketing prospective studies.\nThe study design with relative time-points are described in Figure 1. We defined as \"switch\" the time-point at which patients interrupted the low-dose and started the higher IFNB dose; therefore, we defined a pre-switch period (i.e., the time frame elapsed between the start of the IFNB treatment and the switch) and a 2-year observational post-switch period.\nStudy design with relative time-points. Note that pre-IFNB and low IFNB dose periods had a different duration for each patient, while after switch all patients were followed for 2 years.\nPatients receiving a low-dose IFNB only for titration, or those discontinuing the IFNB treatment mainly for poor tolerance, or those with previous exposure to immunomodulant or immunosuppressive agents were excluded from this study.\nClinical data, including the occurrence of relapses and the Expanded Disability Status Scale (EDSS) score [21], were evaluated for each subject every 3 months. Unscheduled visits were also performed in suspect of relapse or any other clinically relevant condition.\nAn exacerbation was defined as the appearance or reappearance of one or more symptoms attributable to MS, accompanied by objective deterioration on neurological examination lasting at least 24 hours, in the absence of fever and preceded by neurological stability for at least 30 days [19,20].\nWe also collected yearly brain and spinal cord MRI since the start of treatment with IFNB, focusing on the absence/presence of MRI activity (i.e. the detection of at least one contrast-enhancing lesions); additional MRI scans were performed, if necessary. Clinical examination (including EDSS score) and MRI scan were done in stable patients (i.e. at least 30 days after the last assumption of steroids administered for relapses).\n[SUBTITLE] Outcome measures [SUBSECTION] We considered two main outcomes over the 2-year observation period: (a) the occurrence of at least one relapse, and (b) a sustained progression of 1 or more points on EDSS score (starting after the switch and confirmed in two consecutive neurological visits separated by at least a 6-month interval).\nWe considered two main outcomes over the 2-year observation period: (a) the occurrence of at least one relapse, and (b) a sustained progression of 1 or more points on EDSS score (starting after the switch and confirmed in two consecutive neurological visits separated by at least a 6-month interval).\n[SUBTITLE] Statistical analysis [SUBSECTION] All values are expressed as a mean ± standard deviation (± SD) or median (range), as appropriate. Differences between groups were tested by using the Chi-square test and the U Mann-Whitney test.\nTwo Cox proportional hazards models were built for identifying the predictors of experiencing a relapse or a sustained progression on EDSS score after the switch. As main time variable for time-to-event analyses we considered the interval (in years) elapsed between the switch and last visit, or high IFNB dose discontinuation, or outcome reach, whichever came first.\nGender, age, MS duration, EDSS score, presence/absence of an active MRI (each considered at the time of switch), pre-IFNB annualised relapse rate, IFNB type (Betaferon or Rebif 44), as well as duration of initial IFNB treatment, number of relapses, and EDSS change occurred during the low-dose IFNB regimen were included as covariates in each model. Variables were added in the models in a forward stepwise fashion, and interactions terms were tested, where appropriate. All models were stratified by IFNB type received by patients before the switch (Avonex or Rebif 22).\nData have been analysed by using the Statistical Package for Social Sciences, version 16.0 (SPSS, Chicago, IL, USA).\nAll values are expressed as a mean ± standard deviation (± SD) or median (range), as appropriate. Differences between groups were tested by using the Chi-square test and the U Mann-Whitney test.\nTwo Cox proportional hazards models were built for identifying the predictors of experiencing a relapse or a sustained progression on EDSS score after the switch. As main time variable for time-to-event analyses we considered the interval (in years) elapsed between the switch and last visit, or high IFNB dose discontinuation, or outcome reach, whichever came first.\nGender, age, MS duration, EDSS score, presence/absence of an active MRI (each considered at the time of switch), pre-IFNB annualised relapse rate, IFNB type (Betaferon or Rebif 44), as well as duration of initial IFNB treatment, number of relapses, and EDSS change occurred during the low-dose IFNB regimen were included as covariates in each model. Variables were added in the models in a forward stepwise fashion, and interactions terms were tested, where appropriate. All models were stratified by IFNB type received by patients before the switch (Avonex or Rebif 22).\nData have been analysed by using the Statistical Package for Social Sciences, version 16.0 (SPSS, Chicago, IL, USA).", "Patients affected by RRMS [18] according to Poser [19] or McDonald Criteria [20], and switched to a more frequently administered and/or higher dose of IFNB (Betaferon or Rebif 44) in consequence of breakthrough disease during the low-dose IFNB regimen (Avonex or Rebif 22) were included in this 2-year, independent, observational, post-marketing study at the MS Centre of S. Andrea Hospital in Rome. There was no well-defined protocol for switching patients to the high-dose IFNB (occurrence of one or more relapses and/or MRI activity), but it was a decision taken by the neurologist depending on the growing availability over years of new drugs active against MS (i.e., Natalizumab, experimental treatment, etc).\nPatients were regularly followed-up from the start of the IFNB treatment; clinical and MRI data were prospectively collected and stored into an electronic database (after obtaining an informed consent by each patient). The local Ethical committee board provides exemption of approval for post-marketing prospective studies.\nThe study design with relative time-points are described in Figure 1. We defined as \"switch\" the time-point at which patients interrupted the low-dose and started the higher IFNB dose; therefore, we defined a pre-switch period (i.e., the time frame elapsed between the start of the IFNB treatment and the switch) and a 2-year observational post-switch period.\nStudy design with relative time-points. Note that pre-IFNB and low IFNB dose periods had a different duration for each patient, while after switch all patients were followed for 2 years.\nPatients receiving a low-dose IFNB only for titration, or those discontinuing the IFNB treatment mainly for poor tolerance, or those with previous exposure to immunomodulant or immunosuppressive agents were excluded from this study.\nClinical data, including the occurrence of relapses and the Expanded Disability Status Scale (EDSS) score [21], were evaluated for each subject every 3 months. Unscheduled visits were also performed in suspect of relapse or any other clinically relevant condition.\nAn exacerbation was defined as the appearance or reappearance of one or more symptoms attributable to MS, accompanied by objective deterioration on neurological examination lasting at least 24 hours, in the absence of fever and preceded by neurological stability for at least 30 days [19,20].\nWe also collected yearly brain and spinal cord MRI since the start of treatment with IFNB, focusing on the absence/presence of MRI activity (i.e. the detection of at least one contrast-enhancing lesions); additional MRI scans were performed, if necessary. Clinical examination (including EDSS score) and MRI scan were done in stable patients (i.e. at least 30 days after the last assumption of steroids administered for relapses).", "We considered two main outcomes over the 2-year observation period: (a) the occurrence of at least one relapse, and (b) a sustained progression of 1 or more points on EDSS score (starting after the switch and confirmed in two consecutive neurological visits separated by at least a 6-month interval).", "All values are expressed as a mean ± standard deviation (± SD) or median (range), as appropriate. Differences between groups were tested by using the Chi-square test and the U Mann-Whitney test.\nTwo Cox proportional hazards models were built for identifying the predictors of experiencing a relapse or a sustained progression on EDSS score after the switch. As main time variable for time-to-event analyses we considered the interval (in years) elapsed between the switch and last visit, or high IFNB dose discontinuation, or outcome reach, whichever came first.\nGender, age, MS duration, EDSS score, presence/absence of an active MRI (each considered at the time of switch), pre-IFNB annualised relapse rate, IFNB type (Betaferon or Rebif 44), as well as duration of initial IFNB treatment, number of relapses, and EDSS change occurred during the low-dose IFNB regimen were included as covariates in each model. Variables were added in the models in a forward stepwise fashion, and interactions terms were tested, where appropriate. All models were stratified by IFNB type received by patients before the switch (Avonex or Rebif 22).\nData have been analysed by using the Statistical Package for Social Sciences, version 16.0 (SPSS, Chicago, IL, USA).", "[SUBTITLE] Participants [SUBSECTION] We examined 283 patients starting the low-dose IFNB formulation from October 1997 to March 2008 (Avonex, n = 120; Rebif 22, n = 163). Mean (SD) duration of IFNB treatment was 4.3 (2.3) years, and mean annualised relapse rate was 0.48 (range 0-3). A total of 108 (38.1%) patients had a sustained EDSS increase during IFNB treatment.\nOut of these 283 patients, 121 patients switched to the high-dose IFNB, while 162 continued to receive the low-dose IFNB. Table 1 shows the demographic and clinical characteristics of patients who switched to the high-dose IFNB and those who did not. At the start of treatment with IFNB there were no differences between the two groups.\nDemographic and clinical characteristics of 283 patients starting a low IFNB dose.\nAll value are expressed as mean (standard deviation), unless indicated otherwise.\nAll p-values are > 0.05\nAmong the 121 patients who increased the IFNB dose, therapy switching included: Rebif 22→Rebif 44 (n = 59, 47.9%), Avonex→Rebif 44 (n = 46, 38.4%), Avonex→Betaferon (n = 16, 13.7%). After the switch to the high-dose IFNB, 72 (59.5%) patients experienced a relapse, and 51 (42.1%) patients reached a sustained progression on EDSS score. Overall, 85 (70.3%) patients showed some clinical disease activity (i.e. relapse or disability progression) over the 2-year observation period after the switch.\nWe examined 283 patients starting the low-dose IFNB formulation from October 1997 to March 2008 (Avonex, n = 120; Rebif 22, n = 163). Mean (SD) duration of IFNB treatment was 4.3 (2.3) years, and mean annualised relapse rate was 0.48 (range 0-3). A total of 108 (38.1%) patients had a sustained EDSS increase during IFNB treatment.\nOut of these 283 patients, 121 patients switched to the high-dose IFNB, while 162 continued to receive the low-dose IFNB. Table 1 shows the demographic and clinical characteristics of patients who switched to the high-dose IFNB and those who did not. At the start of treatment with IFNB there were no differences between the two groups.\nDemographic and clinical characteristics of 283 patients starting a low IFNB dose.\nAll value are expressed as mean (standard deviation), unless indicated otherwise.\nAll p-values are > 0.05\nAmong the 121 patients who increased the IFNB dose, therapy switching included: Rebif 22→Rebif 44 (n = 59, 47.9%), Avonex→Rebif 44 (n = 46, 38.4%), Avonex→Betaferon (n = 16, 13.7%). After the switch to the high-dose IFNB, 72 (59.5%) patients experienced a relapse, and 51 (42.1%) patients reached a sustained progression on EDSS score. Overall, 85 (70.3%) patients showed some clinical disease activity (i.e. relapse or disability progression) over the 2-year observation period after the switch.\n[SUBTITLE] Risk of relapses or worsening in disability after the switch [SUBSECTION] One hundred (82.6%) patients switched to the high-dose IFNB due to MS attacks (40 also had an active MRI scan at switch), while 21 (17.4%) patients switched due to evidence of activity on MRI scan, but not relapses, during the low-dose IFNB regimen.\nOnly 5 patients (23.8%) who were switched on the basis of MRI activity experienced a further relapse, while 67 (67.0%) of patients switching because of relapses had a further exacerbation over the 2-year after the increase of IFNB dose (p < 0.001).\nNo significant differences in proportion of patients reaching a sustained progression on EDSS score were observed between patients who increased the IFNB dose because of relapses and those who switched only on the basis of MRI (p = 0.08).\nAccording to the Cox model, when compared with patients switched only on the basis of MRI activity (i.e. no relapses during the low-dose IFNB regimen), those relapsing were more likely of having a further relapse even after the switch (HR: 5.55, 95% C.I. 2.22 - 13.85, p = 0.001). Moreover, this risk of relapsing during the 2-year observational period was related with a younger age at switch (HR: 0.94, 95% C.I. 0.91 - 0.97; p < 0.001), and the number of clinical bouts occurred before the switch: one bout confers a risk ratio of 3.56 (95% C.I. 1.34 - 8.42; p = 0.01), two or more bouts a risk ratio of 7.89 (95% C.I. 3.10 - 19.85; p < 0.001) (see Table 2). The risk of further relapses was not increased in patients who switched to the high-dose IFNB because of a combination of clinical and MRI activity at switch.\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for relapsing after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale\n(*) this value refers to patients switched only on the basis of an active MRI scan\nThe variables predictive for reaching a sustained disability progression after the switch were the EDSS score at baseline (HR: 1.77, 95% C.I. 1.39 - 2.26; p < 0.001), and the combination of clinical and radiological activity at switch (HR: 2.14, 95% C.I. 1.16 - 3.96; p = 0.01) (see Table 3).\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for worsening in disability after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale.\nOne hundred (82.6%) patients switched to the high-dose IFNB due to MS attacks (40 also had an active MRI scan at switch), while 21 (17.4%) patients switched due to evidence of activity on MRI scan, but not relapses, during the low-dose IFNB regimen.\nOnly 5 patients (23.8%) who were switched on the basis of MRI activity experienced a further relapse, while 67 (67.0%) of patients switching because of relapses had a further exacerbation over the 2-year after the increase of IFNB dose (p < 0.001).\nNo significant differences in proportion of patients reaching a sustained progression on EDSS score were observed between patients who increased the IFNB dose because of relapses and those who switched only on the basis of MRI (p = 0.08).\nAccording to the Cox model, when compared with patients switched only on the basis of MRI activity (i.e. no relapses during the low-dose IFNB regimen), those relapsing were more likely of having a further relapse even after the switch (HR: 5.55, 95% C.I. 2.22 - 13.85, p = 0.001). Moreover, this risk of relapsing during the 2-year observational period was related with a younger age at switch (HR: 0.94, 95% C.I. 0.91 - 0.97; p < 0.001), and the number of clinical bouts occurred before the switch: one bout confers a risk ratio of 3.56 (95% C.I. 1.34 - 8.42; p = 0.01), two or more bouts a risk ratio of 7.89 (95% C.I. 3.10 - 19.85; p < 0.001) (see Table 2). The risk of further relapses was not increased in patients who switched to the high-dose IFNB because of a combination of clinical and MRI activity at switch.\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for relapsing after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale\n(*) this value refers to patients switched only on the basis of an active MRI scan\nThe variables predictive for reaching a sustained disability progression after the switch were the EDSS score at baseline (HR: 1.77, 95% C.I. 1.39 - 2.26; p < 0.001), and the combination of clinical and radiological activity at switch (HR: 2.14, 95% C.I. 1.16 - 3.96; p = 0.01) (see Table 3).\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for worsening in disability after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale.", "We examined 283 patients starting the low-dose IFNB formulation from October 1997 to March 2008 (Avonex, n = 120; Rebif 22, n = 163). Mean (SD) duration of IFNB treatment was 4.3 (2.3) years, and mean annualised relapse rate was 0.48 (range 0-3). A total of 108 (38.1%) patients had a sustained EDSS increase during IFNB treatment.\nOut of these 283 patients, 121 patients switched to the high-dose IFNB, while 162 continued to receive the low-dose IFNB. Table 1 shows the demographic and clinical characteristics of patients who switched to the high-dose IFNB and those who did not. At the start of treatment with IFNB there were no differences between the two groups.\nDemographic and clinical characteristics of 283 patients starting a low IFNB dose.\nAll value are expressed as mean (standard deviation), unless indicated otherwise.\nAll p-values are > 0.05\nAmong the 121 patients who increased the IFNB dose, therapy switching included: Rebif 22→Rebif 44 (n = 59, 47.9%), Avonex→Rebif 44 (n = 46, 38.4%), Avonex→Betaferon (n = 16, 13.7%). After the switch to the high-dose IFNB, 72 (59.5%) patients experienced a relapse, and 51 (42.1%) patients reached a sustained progression on EDSS score. Overall, 85 (70.3%) patients showed some clinical disease activity (i.e. relapse or disability progression) over the 2-year observation period after the switch.", "One hundred (82.6%) patients switched to the high-dose IFNB due to MS attacks (40 also had an active MRI scan at switch), while 21 (17.4%) patients switched due to evidence of activity on MRI scan, but not relapses, during the low-dose IFNB regimen.\nOnly 5 patients (23.8%) who were switched on the basis of MRI activity experienced a further relapse, while 67 (67.0%) of patients switching because of relapses had a further exacerbation over the 2-year after the increase of IFNB dose (p < 0.001).\nNo significant differences in proportion of patients reaching a sustained progression on EDSS score were observed between patients who increased the IFNB dose because of relapses and those who switched only on the basis of MRI (p = 0.08).\nAccording to the Cox model, when compared with patients switched only on the basis of MRI activity (i.e. no relapses during the low-dose IFNB regimen), those relapsing were more likely of having a further relapse even after the switch (HR: 5.55, 95% C.I. 2.22 - 13.85, p = 0.001). Moreover, this risk of relapsing during the 2-year observational period was related with a younger age at switch (HR: 0.94, 95% C.I. 0.91 - 0.97; p < 0.001), and the number of clinical bouts occurred before the switch: one bout confers a risk ratio of 3.56 (95% C.I. 1.34 - 8.42; p = 0.01), two or more bouts a risk ratio of 7.89 (95% C.I. 3.10 - 19.85; p < 0.001) (see Table 2). The risk of further relapses was not increased in patients who switched to the high-dose IFNB because of a combination of clinical and MRI activity at switch.\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for relapsing after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale\n(*) this value refers to patients switched only on the basis of an active MRI scan\nThe variables predictive for reaching a sustained disability progression after the switch were the EDSS score at baseline (HR: 1.77, 95% C.I. 1.39 - 2.26; p < 0.001), and the combination of clinical and radiological activity at switch (HR: 2.14, 95% C.I. 1.16 - 3.96; p = 0.01) (see Table 3).\nFinal model for the stepwise Cox hazard model analysis showing Hazard Ratio (with relative confidence intervals and p-values) for worsening in disability after the increase of IFNB dose.\nIFNB: Interferon Beta; EDSS: Expanded Disability Status Scale.", "Results from the present study suggest that the risk of having a relapse despite the increase of IFNB dose raised according to the number of clinical attacks occurred during the assumption of the low-dose IFNB, and it is reduced when the switching was based on MRI findings rather than on clinical activity. Moreover, the chance of being progression-free after the switch is related to a lower EDSS score and the absence of clinical and MRI activities. At this regard, we may assume that in some patients an incomplete recovery from relapse occurred, with a substantial influence on their EDSS scores.\nOverall, less than 1/3 of our patients remained free from clinical disease activity (i.e. absence of relapse and sustained progression on EDSS score) over the 2-year observational period following the increase of IFNB dose. This could imply that the majority of switchers should be considered poor responder to IFNB therapy regardless the dose and/or frequency of administration, and further supports the hypothesis that a treatment strategy encompassing the increase of IFNB dose should be useful only in some selected cases.\nAfter the increase of IFNB dose, the majority of patients switched only for the evidence of MRI activity had a better clinical outcome than those switched because of the occurrence of relapses. Therefore, we might suggest that monitoring the effect of IFNB treatment with regular MRI scans is recommendable even in absence of clinical relapses.\nIt has been known that conventional MRI represents a powerful tool to monitor latent disease activity, providing a measurable and sensible marker of response to IFNB therapy [22-25]. However, we cannot exclude that patients switched only on the basis of MRI findings might also have had good outcomes without switching to the high-dose IFNB, as the absence of a control group. At this regard, Rio and colleagues showed that patients with only MRI activity and no relapses in the first year of IFNB treatment did not experience an increase of relapses or disability over a 3-year follow-up [26].\nAlthough randomized clinical trials demonstrated a more pronounced effect of high-dose, high-frequency IFNB when compared with both the low-dose, equal-frequency [1,27] and the low-dose, low-frequency regimens [14,15] in naïve patients, data on the effectiveness of increasing the IFNB dose in patients with breakthrough disease are scarce. The open-label extension phase of the EVIDENCE study, involving 223 patients converted from Avonex to Rebif 44, documented a 50% reduction in the annualised relapse rate after the switch [16]. However, we must consider also that in the EVIDENCE study all patients originally randomized to Avonex were offered to receive Rebif 44, independently from the response status during the blind phase of the study. While the authors suggested that increasing the IFNB-1a dose and frequency could rapidly reduce ongoing disease activity, they cannot discharge the hypothesis that the significant reduction in relapse rate might be due, at least in part, to the regression to the mean phenomenon.\nSome open-label studies exploring the usefulness of switching among immunomodulating drugs had different designs and provided conflicting results [28-30]. Two studies reported a decrease in both relapse rate and proportion of relapse-free subjects after the switch [29,30], whilst the QUASIMS study did not provide any support of more favourable outcomes after switching from an IFNB formulation to another [28]. One possible explanation of this discrepancy is that these observational surveys considered different subtype of switch (i.e. from IFNB-1a to IFNB-1b and viceversa, from IFNB to GA, etc.); also, these studies were not specifically aimed to determine the crude effect of an increase of the IFNB dose. Furthermore, in some studies the patients were switched on the basis of tolerability rather than a persistent disease activity.\nIn the present study, patients had variable periods of observation before and after the switch, thus precluding any attempt to estimate the effectiveness of an increase of the IFNB dose in suppressing disease activity and slowing disability progression.\nBeing an observational report, our study suffers from other limits, such as the small sample size, the unavailability of control group, blindness and randomization, as well as the lack of data on neutralising antibodies (NAbs) against IFNB. However, there is evidence that NAbs presence could explain only the 20% of the suboptimal response to IFNB treatment [31].\nDespite these limits, our study might contribute to define a therapeutic algorithm to manage breakthrough disease in patients on treatment with a low-dose IFNB. The identification of patients with subclinical disease activity during the low-dose IFNB treatment, and an early switch to the high-dose IFNB, seem to be effective in achieving a better control of the disease. On the contrary, when relapses occurred during the pre-switch period, especially in combination with MRI activity, patients did not seem benefit from the increase of the IFNB dose in the following years.\nSince efficacy of GA has been demonstrated comparable to IFNB [32-34], switching among immunomodulating treatments may represent an interesting approach in case of treatment failure [35,36]. The scientific rationale for switching to other therapies is strongest for patients on IFNB therapy with persistent high-titre of NAbs [37]. However, a more aggressive approach (Natalizumab or Mitoxantrone) is warranted for patients at high risk of accumulation of fixed disability or with shorter intervals between attacks [13].", "We suggest neurologists to consider the presence of sub-clinical activity, as detected on MRI, in the decision to switch a patient from the low-dose to the high-dose IFNB, even in absence of clinical relapses. Further efforts are warranted to clearly define whether MRI findings might be considered a key element in the choice of increase the IFNB dose when the response to the low-dose IFNB is suboptimal.", "RRMS: relapsing-remitting multiple sclerosis; EDSS: Expanded Disability Status Scale; MRI: magnetic resonance imaging; CELs: contrast enhancing lesions; IFNB: Interferon beta.", "The authors declare that they have no competing interests.", "LP and CP had full access to the data, and take the final responsibility for the content of the present manuscript. LP and CP are responsible for concept and design of the study, contributed to data analysis and manuscript drafting; GB and LL contributed to data analysis and interpretation and manuscript revision; LDG, LL and VB collected the data. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2377/11/26/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Animals", "Carcinoma, Squamous Cell", "Cell Line, Tumor", "Cell Transformation, Neoplastic", "Esophageal Neoplasms", "Gene Expression Profiling", "Gene Expression Regulation, Neoplastic", "Humans", "Mice", "Mice, Nude", "Microarray Analysis", "Receptors, G-Protein-Coupled", "Tissue Array Analysis", "Transplantation, Heterologous", "Up-Regulation" ]

[ "Background", "ESCC cell lines and specimens", "Semiquantitative RT-PCR", "Tissue Microarrays (TMA) and Immunohistochemistry (IHC)", "Tumorigenic function of GPR39", "Flow cytometry assay", "Tumor formation in nude mice", "Migration and invasion assays", "F-actin staining", "RNA interference", "Western blot analysis", "Statistical analysis", "Results", "GPR39 is frequently overexpressed in ESCC", "Clinical significance of GPR39 overexpression in ESCC", "Tumorigenic function of GPR39", "GPR39 promotes G1/S transition", "GPR39 enhances cell motility and invasiveness of ESCCs", "GPR39 induces partial epithelial-mesenchymal transition (EMT)", "Overexpression of GPR39 induced lamellipodia formation", "Silencing GPR39 expression by RNA interference (RNAi)", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Esophageal squamous cell carcinoma (ESCC), the major histological form of esophageal cancer, is one of the most aggressive malignancies with poor prognosis in the world, especially in the Northern part of China [1]. Like other types of solid tumors, the development of ESCC is also the accumulation of the abnormal expression of oncogenes and tumor suppressor genes (TSGs). Several genetic alterations have been associated with the development of ESCC including mutations of p53 and p16, amplification of cyclin D, c-myc, and EGFR, as well as allelic loss on chromosomes 3p, 5q, 8p, 9p, 9q, 13q, 17p, 18q, and 21q [2-5]. Our previous studies have characterized the common deletion regions at 3p and candidate TSGs within frequently deleted regions including PLCD1 and PCAF [6,7]. However, many genes associated with the development and progression of ESCC have not been characterized. To better understand the molecular mechanisms that underlie the ESCC development and progression, cDNA microarray was used to compare the gene expression profiles between 10 primary ESCC tumors and their paired non-tumorous tissues.\nAmong the 185 up-regulated genes, one gene named GPR39 drew our attention. GPR39 belongs to the G protein-coupled receptors (GPCRs) superfamily, which is the largest family of cell-surface molecules involved in signal transmission. It has been reported that GPR39 plays an important role in the regulation of gastrointestinal and metabolic function [8]. GPR39 receptor is now thought to be activated by Zn2+ signals and may have other, as yet unidentified, cognitive ligands [9]. Moreover, GPR39 receptor also displays a strong ligand-independent signaling activity through Gα12/13 as well as Gαq [10,11]. A recent study suggests that overexpression of GPR39 may inhibit cell death induced by oxidative stress, endoplasmic reticulum (ER) stress, and activation of the caspase by Bax overexpression [12]. Emerging evidence indicates that G protein-coupled receptors are crucial players in cancer progression and metastasis [13,14], however, the role of GPR39 in cancer development remains unclear. In this study, we studied GPR39 expression pattern in ESCC. The tumorigenic function of GPR39 was demonstrated by both in vitro and in vivo assays. The tumorigenic mechanism of GPR39 was also addressed. In addition, the clinical significance of GPR39 overexpression in ESCC was investigated.", "Chinese ESCC cell line HKESC1 was kindly provided by Professor Srivastava (Department of Pathology, The University of Hong Kong, Hong Kong, China), and two Chinese ESCC cell lines (EC18 and EC109) were kindly provided by Professor Tsao (Department of Anatomy, The University of Hong Kong). Six Japanese ESCC cell lines (KYSE30, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520) [15] were obtained from DSMZ (Braunschweig, Germany), the German Resource Centre for Biological Material. Fifty pairs of primary ESCCs and their surrounding non-tumorous esophageal tissues were collected immediately after surgical resection at Linzhou Cancer Hospital (Henan, China). Samples used in this study were approved by the Committees for Ethical Review of Research involving Human Subjects at Zhengzhou University and Sun Yat-Sen University.", "Total RNA was extracted from cell lines and frozen ESCC tissues using the Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacture's instruction. Reverse transcripation of total RNA (2 μg) was done using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA), and cDNA was subjected to PCR for a 30-cycle amplification with primers for GPR39Fw: 5'-GCCACCGGGGTCTCACTTGC-3' and GPR39Rv: 5'-GGCCGCAGCCATGATCCTCC-3'. GAPDH (Fw: 5'-CATGAGAAGTATGACAACAGCCT; Rv: 5'-AGTCCTTCCACGATACCAAAGT) was used as an internal control.", "A total of 300 formalin-fixed and paraffin-embedded ESCC tumor specimens were kindly provided by Linzhou Cancer Hospital (Henan, China). TMAs containing 300 pairs of primary ESCC tumor samples and their corresponding nontumourous tissues were constructed as described previously [16]. Standard streptavidin-biotin-peroxidase complex method was used for IHC staining [16]. Briefly, TMA section was deparaffinized, blocked with 10% normal rabbit serum for 10 min, and incubated with rabbit anti-human GPR39 polyclonal antibody (Abcam, 1:100 dilution) overnight at 4°C. The TMA section was then incubated with biotinylated goat anti-rabbit immunoglobulin at a concentration of 1:100 at 37°C for 30 min. All of the IHC staining results were reviewed independently by two pathologists. Positive expression of GPR39 was defined as the brown staining in the cytoplasm. The staining results for GPR39 were scored semiquantitatively. Intensity was estimated in comparison to the control and scored as follows: 0, negative staining; 1, weak staining; 2, moderate staining; and 3, strong staining. Scores representing the percentage of tumor cells stained positive were as follows: 0, <1% positive tumor cells; 1, 1-10%; 2, 10-50%; 3, 50-75%; and 4, >75%. A final score was calculated by adding the scores for percentage and intensity, resulting in scores of 0 and 2-7. A score of 0 was considered negative; 2-3 was considered weak; 4-5 was considered moderate; and 6-7 was considered strong. For statistical analysis, 0-3 were counted as low expression of GPR39, while 4-7 were counted as overexpression of GPR39.", "To test the tumorigenic function of GPR39, full-length GPR39 was PCR amplified, subcoloned into pcDNA3.1(+) vector (Invitrogen, Carlsbad, CA) and stably transfected into ESCC cell line KYSE30. Stable GPR39-expressing clones (GPR39-c1 and GPR39-c4) were selected for further study. Empty-vector transfected KYSE30 cells (Vec-30) were used as control.\nFor foci formation assay, 1 × 103 GPR39-expressing cells or Vec-30 cells were seeded into 6-well plate. After 7 days culture, surviving colonies (>50 cells/colony) were counted with 1% crystal violet staining. Triplicate independent experiments were performed. Colony formation in soft agar was performed by growing 1 × 104 cells in 0.4% Seaplague agar on a base of 0.6% agar in a 6-well plate. After 3 weeks, colonies consisted of more than 80 cells were counted and expressed as the means ± SD of triplicate within the same experiment. To perform cell growth assay, GPR39-expressing cells and control Vec-30 cells were seeded in 96-well plate at a density of 800 cells per well. The cell growth rate was measured using cell counting kit-8 kit (Dojindo, Japan) according to the manufacturer's instruction. Triplicate independent experiments were done.", "GPR39-c4 or Vec-30 cells were cultured in DMEM medium containing 10% FBS. Serum was withdraw from the culture medium when cells were 70% confluent. After 72 hrs, 10% FBS was added in the medium for an additional 8 hrs, Cells were fixed in 70% ethanol, stained with propidium iodide, and DNA content was analyzed by Cytomics FC (Beckman Coulter, Fullerton, CA).", "For in vivo experiment, stable GPR39-expressing KYSE30 cells or control Vec-30 cells (1 × 106) in 200 μL serum-free DMEM (Life Technologies) were injected s.c. into the right and left flank of 4 week-old nude mice (5 mice for GPR39-c1 cells and 5 for GPR39-c4 cells), respectively. The tumor volume was calculated by the formula V = 0.5 × L × W2 [17]. All experiments were done in accordance with institutional standard guidelines of Sun Yat-Sen University for animal experiments.", "For cell migration assay, GPR39-c4 cells or Vec-30 cells were grown to confluence and then mechanically scratched with a sterile pipette tip. Cells were rinsed with PBS and grown in culture medium for additional 24 hrs. The cell motility in terms of wound closure was measured by photographing at three random fields at time points 0 and 24 hr. For invasion assay, GPR39-c4 cells or Vec-30 cells were starved with serum free medium for 24 hrs before the assay. Cells (5 × 104) were suspended in 0.5 ml serum-free medium and loaded on the upper compartment of invasion chamber coated with Matrigel (BD Biosciences). The lower compartment was filled with complete medium as chemoattractant. After 24 hrs, invasive cells were fixed, stained, and counted under a microscope. Triplicate independent experiments were done.", "Cells grown on coverslips were washed three times in PBS, fixed in 4% paraformaldehyde for 20 min, and permeabilized with 0.1% Triton X-100 for 10 min. Cells were then stained with rhodamin-labeled phalloidin (Molecule Probes) in PBS containing 1% bovine serum albumin at room temperature for 30 min. After additional PBS washes, cells were counterstained with DAPI and photographed with a Leica DMRA fluorescence microscope (Rueil-Malmaison, France).", "Small interfering RNA (siRNA) (20 μM) against GPR39 (s6073; Ambion) was transfected into KYSE180 cells in 6-well plates using Lipofectamine 2000 Reagent (Invitrogen) according to the manufacturer's instructions. At 48 hrs after transfection, the effects of gene silencing were measured via RT-PCR.", "Western blot analysis was performed with the standard method with antibodies to GPR39, N-cadherin and GAPDH (Abcam, Cambridge Science Park, Cambridge, UK), cyclin D1, p21, CDK4 and CDK6 (Cell Signalling Technology, Frankfurt, Germany), and E-cadherin (Santa Cruz Biotechnology, Santa Cruz, CA).", "Statistical analysis was performed with the SPSS standard version 16.0 (SPSS Inc., Chicago, IL). The relationship between the expression of GPR39 protein and clinicopathologic characteristics was assessed by χ2 test. Results expressed as mean ± SD were analyzed using the Student t test. Differences were considered significant when P < 0.05.", "[SUBTITLE] GPR39 is frequently overexpressed in ESCC [SUBSECTION] Semi-quantitative RT-PCR was used to study the expression status of GPR39 in 50 primary ESCCs and 9 ESCC cell lines. Compared with their paired non-tumorous tissues, overexpression of GPR39 was detected in 27/50 (54%) of primary ESCCs (Figure 1A). Overexpression of GPR39 was also frequently detected in ESCC cell lines (HKESC1, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520; Figure 1B). GPR39 expression in protein level was further studied in 300 primary ESCCs by IHC using a tissue microarray. Informative IHC results were obtained from 207 pairs of ESCCs. Non-informative samples included lost samples, unrepresentative samples, samples with too few tumor cells, and samples with inappropriate staining; such were not used in data complication. The expression of GPR39 in normal epithelial cells was always negative or weak whereas strong positive staining of GPR39 was observed in 121/207 (58.5%) of informative ESCCs (Figure 1C).\nOverexpression of GPR39 in ESCC. GPR39 was frequently overexpressed in primary ESCCs (A) and ESCC cell lines (B) detected by RT-PCR. For primary ESCCs, expression of GPR39 in tumor tissues (T) was compared with their paired non-tumorous tissues (N). Normal esophageal tissue was used as a normal control. 18S rRNA was used as an internal control. (C) Representative of GPR39 expression in a pair of ESCC (right) and adjacent normal tissue (left) detected by immunostaining with anti-GPR39 antibody (brown). The slide was counterstained with hematoxylin (original magnification × 200).\nSemi-quantitative RT-PCR was used to study the expression status of GPR39 in 50 primary ESCCs and 9 ESCC cell lines. Compared with their paired non-tumorous tissues, overexpression of GPR39 was detected in 27/50 (54%) of primary ESCCs (Figure 1A). Overexpression of GPR39 was also frequently detected in ESCC cell lines (HKESC1, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520; Figure 1B). GPR39 expression in protein level was further studied in 300 primary ESCCs by IHC using a tissue microarray. Informative IHC results were obtained from 207 pairs of ESCCs. Non-informative samples included lost samples, unrepresentative samples, samples with too few tumor cells, and samples with inappropriate staining; such were not used in data complication. The expression of GPR39 in normal epithelial cells was always negative or weak whereas strong positive staining of GPR39 was observed in 121/207 (58.5%) of informative ESCCs (Figure 1C).\nOverexpression of GPR39 in ESCC. GPR39 was frequently overexpressed in primary ESCCs (A) and ESCC cell lines (B) detected by RT-PCR. For primary ESCCs, expression of GPR39 in tumor tissues (T) was compared with their paired non-tumorous tissues (N). Normal esophageal tissue was used as a normal control. 18S rRNA was used as an internal control. (C) Representative of GPR39 expression in a pair of ESCC (right) and adjacent normal tissue (left) detected by immunostaining with anti-GPR39 antibody (brown). The slide was counterstained with hematoxylin (original magnification × 200).\n[SUBTITLE] Clinical significance of GPR39 overexpression in ESCC [SUBSECTION] The correlation between the GPR39 overexpression and clinicopathologic features of ESCC including age (≤60 versus >60), gender (male versus female), tumor invasion (T stage: tumor depth; T3, T4 versus T1, T2), lymph nodes metastasis (N stage; N0 versus N1), TNM stage (I, IIa versus IIb, III-IV), was studied (Table 1). The results showed that overexpression of GPR39 was significantly associated with lymph node metastasis (P = 0.008) and advanced clinical stage (P = 0.004). No correlation was observed between GPR39 overexpression and age (P = 0.735), gender (P = 0.887), tumor differentiation (P = 0.846) and tumor invasion (P = 0.085).\nAssociation between GPR39 expression and clinical characteristics of ESCC patients (n = 207)\n* Statistically significant (P < 0.05)\nThe correlation between the GPR39 overexpression and clinicopathologic features of ESCC including age (≤60 versus >60), gender (male versus female), tumor invasion (T stage: tumor depth; T3, T4 versus T1, T2), lymph nodes metastasis (N stage; N0 versus N1), TNM stage (I, IIa versus IIb, III-IV), was studied (Table 1). The results showed that overexpression of GPR39 was significantly associated with lymph node metastasis (P = 0.008) and advanced clinical stage (P = 0.004). No correlation was observed between GPR39 overexpression and age (P = 0.735), gender (P = 0.887), tumor differentiation (P = 0.846) and tumor invasion (P = 0.085).\nAssociation between GPR39 expression and clinical characteristics of ESCC patients (n = 207)\n* Statistically significant (P < 0.05)\n[SUBTITLE] Tumorigenic function of GPR39 [SUBSECTION] To investigate the tumorigenic potential of GPR39, GPR39-expression vector was stably transfected into KYSE30 cells with silenced GPR39. GPR39 mRNA and protein expression in these clones were confirmed by RT-PCR and Western blot analysis (Figure 2A). The tumorigenic function of GPR39 was assessed by both in vitro and in vivo assays including foci formation, colony formation in soft agar, cell growth rate assays and tumor xenograft experiment. Foci formation assay showed that the frequency of foci formation was significantly increased (P < 0.01) in GPR39-transfectants compared with control cells (Figure 2B). A similar result was shown in soft agar assay (P < 0.01, Figure 2C). Cell growth assay also revealed that the cell growth rates in GPR39-c1 and GPR39-c4 cells were significantly enhanced by GPR39 compared with Vec-30 cells (P < 0.01, Figure 2D). To further explore the in vivo tumorigenic ability of GPR39, tumor formation in nude mice was tested by injection of GPR39-c1 cells (n = 5) or GPR39-c4 cells (n = 5), whereas Vec-30 cells were used as controls. Tumor formation was observed in all tested animals. The results showed that the tumor growth curve of GPR39-overexpressing cells was significantly increased compared to Vec-30 cells (P < 0.01, Figure 2E).\nTumorigenic function of GPR39 in ESCC cells. (A) Expression of GPR39 in GPR39-transfected KYSE30 cells was confirmed by RT-PCR (left) and Western blot analysis (right). c1 and c4 are two independent GPR39-expressing clones. Vec-30 represents empty vector-transfected KYSE30 cells. (B) Representative of foci formation in monolayer culture. Quantitative analyses of foci numbers were shown in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01; independent Student's t-test. (C) Representative of colony formation in soft agar. Percentage of colonies formed was summarized in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01. (D) Growth curves of GPR39-expressing cells were compared with Vec-30 cells by cell growth assay. The results were expressed as mean ± SD of at least three independent experiments. **P < 0.01. (E) Tumor growth curves of GPR39-expressing cells in nude mice were compared with Vec-30 cells by tumor xenograft experiment. The average tumor volume of GPR39-expressing cells vs Vec-30 cells was expressed as mean ± SD in 10 inoculated sites for each group of cells. **P < 0.01. (F) Representative examples of tumors formed in nude mice following injection of GPR39-expressing KYSE30 cells (right) and Vec-30 cells (left).\nTo investigate the tumorigenic potential of GPR39, GPR39-expression vector was stably transfected into KYSE30 cells with silenced GPR39. GPR39 mRNA and protein expression in these clones were confirmed by RT-PCR and Western blot analysis (Figure 2A). The tumorigenic function of GPR39 was assessed by both in vitro and in vivo assays including foci formation, colony formation in soft agar, cell growth rate assays and tumor xenograft experiment. Foci formation assay showed that the frequency of foci formation was significantly increased (P < 0.01) in GPR39-transfectants compared with control cells (Figure 2B). A similar result was shown in soft agar assay (P < 0.01, Figure 2C). Cell growth assay also revealed that the cell growth rates in GPR39-c1 and GPR39-c4 cells were significantly enhanced by GPR39 compared with Vec-30 cells (P < 0.01, Figure 2D). To further explore the in vivo tumorigenic ability of GPR39, tumor formation in nude mice was tested by injection of GPR39-c1 cells (n = 5) or GPR39-c4 cells (n = 5), whereas Vec-30 cells were used as controls. Tumor formation was observed in all tested animals. The results showed that the tumor growth curve of GPR39-overexpressing cells was significantly increased compared to Vec-30 cells (P < 0.01, Figure 2E).\nTumorigenic function of GPR39 in ESCC cells. (A) Expression of GPR39 in GPR39-transfected KYSE30 cells was confirmed by RT-PCR (left) and Western blot analysis (right). c1 and c4 are two independent GPR39-expressing clones. Vec-30 represents empty vector-transfected KYSE30 cells. (B) Representative of foci formation in monolayer culture. Quantitative analyses of foci numbers were shown in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01; independent Student's t-test. (C) Representative of colony formation in soft agar. Percentage of colonies formed was summarized in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01. (D) Growth curves of GPR39-expressing cells were compared with Vec-30 cells by cell growth assay. The results were expressed as mean ± SD of at least three independent experiments. **P < 0.01. (E) Tumor growth curves of GPR39-expressing cells in nude mice were compared with Vec-30 cells by tumor xenograft experiment. The average tumor volume of GPR39-expressing cells vs Vec-30 cells was expressed as mean ± SD in 10 inoculated sites for each group of cells. **P < 0.01. (F) Representative examples of tumors formed in nude mice following injection of GPR39-expressing KYSE30 cells (right) and Vec-30 cells (left).\n[SUBTITLE] GPR39 promotes G1/S transition [SUBSECTION] To explore the mechanism underlying growth promotion by GPR39, the cell cycle distributions of GPR39-c4 and Vec-30 cells were determined by flow cytometry. Before treatment, the percentage of GPR39-c4 cells in G1 phase was obviously reduced in comparison with Vec-30 cells (38.37 ± 1.02% versus 45.87 ± 0.47%, P < 0.05; Figure 3A). After 3 days' serum starvation followed by addition of 10% serum for 8 hrs, the percentage of cells in S phase was significantly increased in GPR39-c4 cells compared to Vec-30 cells (26.43 ± 0.71% versus 8.97 ± 0.31%, P < 0.05; Figure 3A), suggesting that GPR39 was able to promote G1/S transition. To reveal the potential molecular mechanism of GPR39 in cell cycle promotion, expressions of several key cell cycle regulators including p21, cyclin D1, CDK4 and CDK6 were compared between GPR39-c4 and Vec-30 cells. Increased expression of cyclin D1 and CDK6, but not p21 and CDK4, were detected in GPR39-c4 (Figure 3B).\nGPR39 promotes G1/S transition and enhances cell motility. (A) DNA content between GPR39-expressing cells and control Vec-30 cell were compared by Flow-cytometry. Untreated, cells were cultured in DMEM medium with 10% FBS; Withdraw serum, cells were cultured in DMEM medium without serum for 3 days; Add serum, cells were cultured again in DMEM medium with 10% FBS for 8 hr. (B) Expression of p21, cyclin D1, CDK4, and CDK6 were compared between GPR39-expressing cells (c4) and control Vec-30 cells by Western blot analyses. GAPDH was used as loading control. (C) The effect of GPR39 on cell migration was determined by wound-healing assay. During a period of 16 hr, the spreading speed of GPR39-expressing cells along the wound edge was faster than that in control Vec-30 cells. (D) Representative images showed the GPR39-expressing cells and Vec-30 cells that invaded through the matrigel. Number of invaded tumor cells was quantified in the right panel. Columns, mean of triplicate experiments; **P < 0.01.\nTo explore the mechanism underlying growth promotion by GPR39, the cell cycle distributions of GPR39-c4 and Vec-30 cells were determined by flow cytometry. Before treatment, the percentage of GPR39-c4 cells in G1 phase was obviously reduced in comparison with Vec-30 cells (38.37 ± 1.02% versus 45.87 ± 0.47%, P < 0.05; Figure 3A). After 3 days' serum starvation followed by addition of 10% serum for 8 hrs, the percentage of cells in S phase was significantly increased in GPR39-c4 cells compared to Vec-30 cells (26.43 ± 0.71% versus 8.97 ± 0.31%, P < 0.05; Figure 3A), suggesting that GPR39 was able to promote G1/S transition. To reveal the potential molecular mechanism of GPR39 in cell cycle promotion, expressions of several key cell cycle regulators including p21, cyclin D1, CDK4 and CDK6 were compared between GPR39-c4 and Vec-30 cells. Increased expression of cyclin D1 and CDK6, but not p21 and CDK4, were detected in GPR39-c4 (Figure 3B).\nGPR39 promotes G1/S transition and enhances cell motility. (A) DNA content between GPR39-expressing cells and control Vec-30 cell were compared by Flow-cytometry. Untreated, cells were cultured in DMEM medium with 10% FBS; Withdraw serum, cells were cultured in DMEM medium without serum for 3 days; Add serum, cells were cultured again in DMEM medium with 10% FBS for 8 hr. (B) Expression of p21, cyclin D1, CDK4, and CDK6 were compared between GPR39-expressing cells (c4) and control Vec-30 cells by Western blot analyses. GAPDH was used as loading control. (C) The effect of GPR39 on cell migration was determined by wound-healing assay. During a period of 16 hr, the spreading speed of GPR39-expressing cells along the wound edge was faster than that in control Vec-30 cells. (D) Representative images showed the GPR39-expressing cells and Vec-30 cells that invaded through the matrigel. Number of invaded tumor cells was quantified in the right panel. Columns, mean of triplicate experiments; **P < 0.01.\n[SUBTITLE] GPR39 enhances cell motility and invasiveness of ESCCs [SUBSECTION] As the TMA result showed that overexpression of GPR39 was closely associated with ESCC metastasis, the effects of GPR39 on cell migration and invasion were studied by wound-healing and cell invasion assays. Wound-healing assay showed that that the ectopic expression of GPR39 could significantly increase cell migration ability in GPR39-transfected cells compared with empty-vector control (P < 0.05, Figure 3C). Matrigel invasion assay also found that the ectopic expression of GPR39 could significantly enhanced the invasiveness of ESCC cells, as demonstrated by a significant increase in the number of invaded cells (P < 0.01, Figure 3D), in GPR39-transfected cells compared with empty-vector control.\nAs the TMA result showed that overexpression of GPR39 was closely associated with ESCC metastasis, the effects of GPR39 on cell migration and invasion were studied by wound-healing and cell invasion assays. Wound-healing assay showed that that the ectopic expression of GPR39 could significantly increase cell migration ability in GPR39-transfected cells compared with empty-vector control (P < 0.05, Figure 3C). Matrigel invasion assay also found that the ectopic expression of GPR39 could significantly enhanced the invasiveness of ESCC cells, as demonstrated by a significant increase in the number of invaded cells (P < 0.01, Figure 3D), in GPR39-transfected cells compared with empty-vector control.\n[SUBTITLE] GPR39 induces partial epithelial-mesenchymal transition (EMT) [SUBSECTION] In this study, we found that the cell morphology changed obviously after the transfection of GPR39. GPR39-transfected cells showed spindle shape and fibroblastic changes in monolayer culture, whereas empty vector-transfected cells, like KYSE30 parental cells, kept their cobblestone-like phenotype (Figure 4A). To determine whether the effect of GPR39 on cell motility was associated with EMT, expressions of several epithelial markers (E-cadherin, N-cadherin) and mesenchymal markers (vimentin, and fibronectin) were compared between GPR39-c4 and Vec-30 cells by RT-PCR and Western blot analysis. The results showed that E-cadherin was obviously down-regulated in GPR39-c4 cells; however, no obvious difference was observed in the expression of N-cadherin, vimentin and fibronectin between GPR39-c4 and Vec-30 cells (Figure 4B). These findings indicated that GPR39 increased cell motility was partially through the EMT.\nGPR39 promotes cell mobility and invasion by inducing partial EMT and remodeling cytoskeleton. (A) Representatives of cell morphology of GPR39-expressing cells and Vec-30 cells (original magnification × 200). (B) Expressions of epithelial markers E-cadherin and mesenchymal markers fibronectin, N-cadherin, and vimentin, were compared by RT-PCR or Western blotting analysis between GPR39-expressing cells and Vec-30 cells. GAPDH was used as loading control. (C) Representative images of F-actin staining. Formation of lamellipodia (indicated by arrows) was stimulated by GPR39 compared to control cells (magnification × 400).\nIn this study, we found that the cell morphology changed obviously after the transfection of GPR39. GPR39-transfected cells showed spindle shape and fibroblastic changes in monolayer culture, whereas empty vector-transfected cells, like KYSE30 parental cells, kept their cobblestone-like phenotype (Figure 4A). To determine whether the effect of GPR39 on cell motility was associated with EMT, expressions of several epithelial markers (E-cadherin, N-cadherin) and mesenchymal markers (vimentin, and fibronectin) were compared between GPR39-c4 and Vec-30 cells by RT-PCR and Western blot analysis. The results showed that E-cadherin was obviously down-regulated in GPR39-c4 cells; however, no obvious difference was observed in the expression of N-cadherin, vimentin and fibronectin between GPR39-c4 and Vec-30 cells (Figure 4B). These findings indicated that GPR39 increased cell motility was partially through the EMT.\nGPR39 promotes cell mobility and invasion by inducing partial EMT and remodeling cytoskeleton. (A) Representatives of cell morphology of GPR39-expressing cells and Vec-30 cells (original magnification × 200). (B) Expressions of epithelial markers E-cadherin and mesenchymal markers fibronectin, N-cadherin, and vimentin, were compared by RT-PCR or Western blotting analysis between GPR39-expressing cells and Vec-30 cells. GAPDH was used as loading control. (C) Representative images of F-actin staining. Formation of lamellipodia (indicated by arrows) was stimulated by GPR39 compared to control cells (magnification × 400).\n[SUBTITLE] Overexpression of GPR39 induced lamellipodia formation [SUBSECTION] To further explore the molecular mechanism of GPR39 in regulating cancer invasion and metastasis, the role of GPR39 in the polymerized actin was investigated by phalloidin staining. The results showed that GPR39-expressing cells exhibited enhanced lamellipodia formation compared with control cells (Figure 4C), indicating that GPR39 could induce cytoskeleton remodeling to facilitate esophageal cancer cell migration and invasion.\nTo further explore the molecular mechanism of GPR39 in regulating cancer invasion and metastasis, the role of GPR39 in the polymerized actin was investigated by phalloidin staining. The results showed that GPR39-expressing cells exhibited enhanced lamellipodia formation compared with control cells (Figure 4C), indicating that GPR39 could induce cytoskeleton remodeling to facilitate esophageal cancer cell migration and invasion.\n[SUBTITLE] Silencing GPR39 expression by RNA interference (RNAi) [SUBSECTION] ESCC cell line KYSE180, which expresses a high level of endogenous GPR39, was used in the siRNA experiment. Two siRNAs targeting GPR39 (GPR39-si1 and GPR39-si2) were tested and the efficiency of GPR39 gene silencing was detected by RT-PCR. The result showed that the GPR39-si1 had a better silencing effect (Figure 5A). Silencing of GPR39 resulted in a significant inhibition of the cell growth rate (P < 0.01, Figure 5B) and migration (Figure 5C). DNA content analysis by flow cytometry showed that GPR39-si1 was able to inhibit the cell cycle at the G1/S checkpoint (Figure 5D). The percentage of cells in the S phase was significantly reduced in GPR39-si1-treated cells (27.23 ± 1.26%) compared with that in control-si-treated cells (35.13 ± 1.12%; P < 0.05). These findings further supported that the tumorigenic function of GPR39 was through its role in promoting cell proliferation and motility.\nSilencing of GPR39 expression suppresses tumorigenic ability of GPR39. (A) GPR39 expression was efficiently decreased by the treatment of siGPR39 by RT-PCR. Relative expression level was measured by densitometer and summarized in the right panel. **P < 0.01. (B) Growth curve of KYSE180 cells treated with GPR39 siRNA was compared with control siRNA treated cells by cell growth assay. **P < 0.01. (C) Cell migration assay was used to compare the frequency of migratory cells between KYSE180 cells treated with control siRNA and GPR39 siRNA. (D) DNA content between control siRNA and GPR39 siRNA treated cells were compared by Flow-cytometry. *P < 0.05.\nESCC cell line KYSE180, which expresses a high level of endogenous GPR39, was used in the siRNA experiment. Two siRNAs targeting GPR39 (GPR39-si1 and GPR39-si2) were tested and the efficiency of GPR39 gene silencing was detected by RT-PCR. The result showed that the GPR39-si1 had a better silencing effect (Figure 5A). Silencing of GPR39 resulted in a significant inhibition of the cell growth rate (P < 0.01, Figure 5B) and migration (Figure 5C). DNA content analysis by flow cytometry showed that GPR39-si1 was able to inhibit the cell cycle at the G1/S checkpoint (Figure 5D). The percentage of cells in the S phase was significantly reduced in GPR39-si1-treated cells (27.23 ± 1.26%) compared with that in control-si-treated cells (35.13 ± 1.12%; P < 0.05). These findings further supported that the tumorigenic function of GPR39 was through its role in promoting cell proliferation and motility.\nSilencing of GPR39 expression suppresses tumorigenic ability of GPR39. (A) GPR39 expression was efficiently decreased by the treatment of siGPR39 by RT-PCR. Relative expression level was measured by densitometer and summarized in the right panel. **P < 0.01. (B) Growth curve of KYSE180 cells treated with GPR39 siRNA was compared with control siRNA treated cells by cell growth assay. **P < 0.01. (C) Cell migration assay was used to compare the frequency of migratory cells between KYSE180 cells treated with control siRNA and GPR39 siRNA. (D) DNA content between control siRNA and GPR39 siRNA treated cells were compared by Flow-cytometry. *P < 0.05.", "Semi-quantitative RT-PCR was used to study the expression status of GPR39 in 50 primary ESCCs and 9 ESCC cell lines. Compared with their paired non-tumorous tissues, overexpression of GPR39 was detected in 27/50 (54%) of primary ESCCs (Figure 1A). Overexpression of GPR39 was also frequently detected in ESCC cell lines (HKESC1, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520; Figure 1B). GPR39 expression in protein level was further studied in 300 primary ESCCs by IHC using a tissue microarray. Informative IHC results were obtained from 207 pairs of ESCCs. Non-informative samples included lost samples, unrepresentative samples, samples with too few tumor cells, and samples with inappropriate staining; such were not used in data complication. The expression of GPR39 in normal epithelial cells was always negative or weak whereas strong positive staining of GPR39 was observed in 121/207 (58.5%) of informative ESCCs (Figure 1C).\nOverexpression of GPR39 in ESCC. GPR39 was frequently overexpressed in primary ESCCs (A) and ESCC cell lines (B) detected by RT-PCR. For primary ESCCs, expression of GPR39 in tumor tissues (T) was compared with their paired non-tumorous tissues (N). Normal esophageal tissue was used as a normal control. 18S rRNA was used as an internal control. (C) Representative of GPR39 expression in a pair of ESCC (right) and adjacent normal tissue (left) detected by immunostaining with anti-GPR39 antibody (brown). The slide was counterstained with hematoxylin (original magnification × 200).", "The correlation between the GPR39 overexpression and clinicopathologic features of ESCC including age (≤60 versus >60), gender (male versus female), tumor invasion (T stage: tumor depth; T3, T4 versus T1, T2), lymph nodes metastasis (N stage; N0 versus N1), TNM stage (I, IIa versus IIb, III-IV), was studied (Table 1). The results showed that overexpression of GPR39 was significantly associated with lymph node metastasis (P = 0.008) and advanced clinical stage (P = 0.004). No correlation was observed between GPR39 overexpression and age (P = 0.735), gender (P = 0.887), tumor differentiation (P = 0.846) and tumor invasion (P = 0.085).\nAssociation between GPR39 expression and clinical characteristics of ESCC patients (n = 207)\n* Statistically significant (P < 0.05)", "To investigate the tumorigenic potential of GPR39, GPR39-expression vector was stably transfected into KYSE30 cells with silenced GPR39. GPR39 mRNA and protein expression in these clones were confirmed by RT-PCR and Western blot analysis (Figure 2A). The tumorigenic function of GPR39 was assessed by both in vitro and in vivo assays including foci formation, colony formation in soft agar, cell growth rate assays and tumor xenograft experiment. Foci formation assay showed that the frequency of foci formation was significantly increased (P < 0.01) in GPR39-transfectants compared with control cells (Figure 2B). A similar result was shown in soft agar assay (P < 0.01, Figure 2C). Cell growth assay also revealed that the cell growth rates in GPR39-c1 and GPR39-c4 cells were significantly enhanced by GPR39 compared with Vec-30 cells (P < 0.01, Figure 2D). To further explore the in vivo tumorigenic ability of GPR39, tumor formation in nude mice was tested by injection of GPR39-c1 cells (n = 5) or GPR39-c4 cells (n = 5), whereas Vec-30 cells were used as controls. Tumor formation was observed in all tested animals. The results showed that the tumor growth curve of GPR39-overexpressing cells was significantly increased compared to Vec-30 cells (P < 0.01, Figure 2E).\nTumorigenic function of GPR39 in ESCC cells. (A) Expression of GPR39 in GPR39-transfected KYSE30 cells was confirmed by RT-PCR (left) and Western blot analysis (right). c1 and c4 are two independent GPR39-expressing clones. Vec-30 represents empty vector-transfected KYSE30 cells. (B) Representative of foci formation in monolayer culture. Quantitative analyses of foci numbers were shown in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01; independent Student's t-test. (C) Representative of colony formation in soft agar. Percentage of colonies formed was summarized in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01. (D) Growth curves of GPR39-expressing cells were compared with Vec-30 cells by cell growth assay. The results were expressed as mean ± SD of at least three independent experiments. **P < 0.01. (E) Tumor growth curves of GPR39-expressing cells in nude mice were compared with Vec-30 cells by tumor xenograft experiment. The average tumor volume of GPR39-expressing cells vs Vec-30 cells was expressed as mean ± SD in 10 inoculated sites for each group of cells. **P < 0.01. (F) Representative examples of tumors formed in nude mice following injection of GPR39-expressing KYSE30 cells (right) and Vec-30 cells (left).", "To explore the mechanism underlying growth promotion by GPR39, the cell cycle distributions of GPR39-c4 and Vec-30 cells were determined by flow cytometry. Before treatment, the percentage of GPR39-c4 cells in G1 phase was obviously reduced in comparison with Vec-30 cells (38.37 ± 1.02% versus 45.87 ± 0.47%, P < 0.05; Figure 3A). After 3 days' serum starvation followed by addition of 10% serum for 8 hrs, the percentage of cells in S phase was significantly increased in GPR39-c4 cells compared to Vec-30 cells (26.43 ± 0.71% versus 8.97 ± 0.31%, P < 0.05; Figure 3A), suggesting that GPR39 was able to promote G1/S transition. To reveal the potential molecular mechanism of GPR39 in cell cycle promotion, expressions of several key cell cycle regulators including p21, cyclin D1, CDK4 and CDK6 were compared between GPR39-c4 and Vec-30 cells. Increased expression of cyclin D1 and CDK6, but not p21 and CDK4, were detected in GPR39-c4 (Figure 3B).\nGPR39 promotes G1/S transition and enhances cell motility. (A) DNA content between GPR39-expressing cells and control Vec-30 cell were compared by Flow-cytometry. Untreated, cells were cultured in DMEM medium with 10% FBS; Withdraw serum, cells were cultured in DMEM medium without serum for 3 days; Add serum, cells were cultured again in DMEM medium with 10% FBS for 8 hr. (B) Expression of p21, cyclin D1, CDK4, and CDK6 were compared between GPR39-expressing cells (c4) and control Vec-30 cells by Western blot analyses. GAPDH was used as loading control. (C) The effect of GPR39 on cell migration was determined by wound-healing assay. During a period of 16 hr, the spreading speed of GPR39-expressing cells along the wound edge was faster than that in control Vec-30 cells. (D) Representative images showed the GPR39-expressing cells and Vec-30 cells that invaded through the matrigel. Number of invaded tumor cells was quantified in the right panel. Columns, mean of triplicate experiments; **P < 0.01.", "As the TMA result showed that overexpression of GPR39 was closely associated with ESCC metastasis, the effects of GPR39 on cell migration and invasion were studied by wound-healing and cell invasion assays. Wound-healing assay showed that that the ectopic expression of GPR39 could significantly increase cell migration ability in GPR39-transfected cells compared with empty-vector control (P < 0.05, Figure 3C). Matrigel invasion assay also found that the ectopic expression of GPR39 could significantly enhanced the invasiveness of ESCC cells, as demonstrated by a significant increase in the number of invaded cells (P < 0.01, Figure 3D), in GPR39-transfected cells compared with empty-vector control.", "In this study, we found that the cell morphology changed obviously after the transfection of GPR39. GPR39-transfected cells showed spindle shape and fibroblastic changes in monolayer culture, whereas empty vector-transfected cells, like KYSE30 parental cells, kept their cobblestone-like phenotype (Figure 4A). To determine whether the effect of GPR39 on cell motility was associated with EMT, expressions of several epithelial markers (E-cadherin, N-cadherin) and mesenchymal markers (vimentin, and fibronectin) were compared between GPR39-c4 and Vec-30 cells by RT-PCR and Western blot analysis. The results showed that E-cadherin was obviously down-regulated in GPR39-c4 cells; however, no obvious difference was observed in the expression of N-cadherin, vimentin and fibronectin between GPR39-c4 and Vec-30 cells (Figure 4B). These findings indicated that GPR39 increased cell motility was partially through the EMT.\nGPR39 promotes cell mobility and invasion by inducing partial EMT and remodeling cytoskeleton. (A) Representatives of cell morphology of GPR39-expressing cells and Vec-30 cells (original magnification × 200). (B) Expressions of epithelial markers E-cadherin and mesenchymal markers fibronectin, N-cadherin, and vimentin, were compared by RT-PCR or Western blotting analysis between GPR39-expressing cells and Vec-30 cells. GAPDH was used as loading control. (C) Representative images of F-actin staining. Formation of lamellipodia (indicated by arrows) was stimulated by GPR39 compared to control cells (magnification × 400).", "To further explore the molecular mechanism of GPR39 in regulating cancer invasion and metastasis, the role of GPR39 in the polymerized actin was investigated by phalloidin staining. The results showed that GPR39-expressing cells exhibited enhanced lamellipodia formation compared with control cells (Figure 4C), indicating that GPR39 could induce cytoskeleton remodeling to facilitate esophageal cancer cell migration and invasion.", "ESCC cell line KYSE180, which expresses a high level of endogenous GPR39, was used in the siRNA experiment. Two siRNAs targeting GPR39 (GPR39-si1 and GPR39-si2) were tested and the efficiency of GPR39 gene silencing was detected by RT-PCR. The result showed that the GPR39-si1 had a better silencing effect (Figure 5A). Silencing of GPR39 resulted in a significant inhibition of the cell growth rate (P < 0.01, Figure 5B) and migration (Figure 5C). DNA content analysis by flow cytometry showed that GPR39-si1 was able to inhibit the cell cycle at the G1/S checkpoint (Figure 5D). The percentage of cells in the S phase was significantly reduced in GPR39-si1-treated cells (27.23 ± 1.26%) compared with that in control-si-treated cells (35.13 ± 1.12%; P < 0.05). These findings further supported that the tumorigenic function of GPR39 was through its role in promoting cell proliferation and motility.\nSilencing of GPR39 expression suppresses tumorigenic ability of GPR39. (A) GPR39 expression was efficiently decreased by the treatment of siGPR39 by RT-PCR. Relative expression level was measured by densitometer and summarized in the right panel. **P < 0.01. (B) Growth curve of KYSE180 cells treated with GPR39 siRNA was compared with control siRNA treated cells by cell growth assay. **P < 0.01. (C) Cell migration assay was used to compare the frequency of migratory cells between KYSE180 cells treated with control siRNA and GPR39 siRNA. (D) DNA content between control siRNA and GPR39 siRNA treated cells were compared by Flow-cytometry. *P < 0.05.", "Many G protein-coupled receptors (GPCRs) have been found to play critical roles in the development and progression of cancer, including malignant transformation [18,19], tumor growth and survival [20,21], as well as invasion and metastasis [22,23]. Herein, we report that one of the G protein-coupled receptors, GPR39, is frequently overexpressed in human ESCC. To our knowledge, this is the first illustration that GPR39 contributes to the development and progression of ESCC. In the present study, the tumorigenic function of GPR39 was demonstrated by both in vitro and in vivo assays. Functional studies showed that GPR39 could effectively promote ESCC cancer cell growth, increase foci formation and colony formation and enhance tumor formation in nude mice. A recent study suggested that zinc could be a ligand capable of activating the GPR39 receptor [11]. Interestingly, zinc deficiency along with its associated increased cell proliferation can be tumorigenic in the rat esophagus [24,25]. Our study also provided evidence that ectopic expression of GPR39 increased ESCC cancer cell growth, indicating involvement of the GPR39 receptor in the tumorigenesis of esophageal cancer. However, whether GPR39 signaling is activated by zinc in esophageal carcinogenesis needs to be further investigated. Further study revealed that overexpression of GPR39 in esophageal cancer cells KYSE30 promoted G1/S phase transition. We showed for the first time that GPR39 controls cell cycle progression through the activation of CDK6 and its activating protein, cyclin D1. G1/S phase transition is a major checkpoint for cell cycle progression and cyclin D1-CDK6 complex is one of the critical positive regulators during this transition [26,27]. On the other hand, we found that silencing of GPR39 expression could inhibit tumorigenicity in KYSE180 cells through the cell cycle arrest at G1/S checkpoint.\nAnother interesting finding of this study is the promoting effect of GPR39 on tumor metastasis in ESCC. Our data showed that overexpression of GPR39 could promote cell motility and invasiveness of ESCC cells in vitro. This mirrored the findings of GPR39 overexpression in human ESCC samples and its association with advanced clinical stage and lymph node metastasis of ESCC. Conversely, when we knocked down the endogenous GPR39 by RNAi in ESCC cells, the mobility of ESCC cells was significantly reduced, suggesting that GPR39 is closely involved in ESCC invasion and metastasis. Moreover, the observation of overexpression of GPR39 resulting in cell morphological alteration promoted us to further investigate its effect on EMT. We found that GPR39 has some impact on the EMT as shown by decreasing the epithelial molecule E-cadherin, an event critical in tumour invasion and a 'master' regulator of EMT. E-cadherin provides a physical link among adjacent cells and is crucial for the establishment and maintenance of polarity and the structural integrity of epithelia. Indeed, due to the physical and functional link between E-cadherin based complexes and cytoskeletal components, a change in the E-cadherin mediated adhesiveness leads to rearrangement of the cytoskeleton [28]. In view of this, we further explored the role of GPR39 in reorganization of the actin cytoskeleton. As expected, our result showed that GPR39 led to significant alterations on cytoskeleton by inducing the lamellipodia formation in GPR39-transfected ESCC cells. This finding was consistent to previous studies that some G protein-coupled receptors (GPCRs) were able to promote actin reorganization and result in cell shape changes and enhanced cell migration [13,29], indicating that GPR39 might directly alter the cytoskeleton to favor the tumor cell invasion and metastasis in ESCC.\nIn this study, we have also provided evidence that targeting of GPR39 with specific RNAi will reduce the oncogenic characteristics of ESCC tumor cells. To date, some G protein-coupled receptors (GPCRs) provide important practical options for preclinical research, clinical trials, and cancer treatment [30]. Therefore, consideration should be given to the development of novel therapeutics targeting GPR39 for use in GPR39-expressing ESCC tumors.", "In summary, our findings demonstrate that GPR39 plays an important role in ESCC development and progression via promoting cell proliferation, enhancing cell motility and invasiveness, regulating cytoskeleton and inducing EMT. A better understanding of the molecular mechanism of GPR39 in ESCC development and progression would provide novel therapeutic strategies to ESCC cancer patients.", "EMT: epithelial mesenchymal transition; ESCC: esophageal squamous cell carcinoma; GPCR: G protein-coupled receptor; siRNA: small interfering RNA; TMA: tissue microarray; TSG: tumor suppressor gene; L: length; V: volume; W: width.", "The authors declare that they have no competing interests.", "FX and HL performed the experimental procedures with support from YZ, YQ, YD, TZ, LC, CN, TH and YL. FX, LF and XYG were responsible for experimental design, interpretation of the results and writing the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2407/11/86/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "ESCC cell lines and specimens", "Semiquantitative RT-PCR", "Tissue Microarrays (TMA) and Immunohistochemistry (IHC)", "Tumorigenic function of GPR39", "Flow cytometry assay", "Tumor formation in nude mice", "Migration and invasion assays", "F-actin staining", "RNA interference", "Western blot analysis", "Statistical analysis", "Results", "GPR39 is frequently overexpressed in ESCC", "Clinical significance of GPR39 overexpression in ESCC", "Tumorigenic function of GPR39", "GPR39 promotes G1/S transition", "GPR39 enhances cell motility and invasiveness of ESCCs", "GPR39 induces partial epithelial-mesenchymal transition (EMT)", "Overexpression of GPR39 induced lamellipodia formation", "Silencing GPR39 expression by RNA interference (RNAi)", "Discussion", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Esophageal squamous cell carcinoma (ESCC), the major histological form of esophageal cancer, is one of the most aggressive malignancies with poor prognosis in the world, especially in the Northern part of China [1]. Like other types of solid tumors, the development of ESCC is also the accumulation of the abnormal expression of oncogenes and tumor suppressor genes (TSGs). Several genetic alterations have been associated with the development of ESCC including mutations of p53 and p16, amplification of cyclin D, c-myc, and EGFR, as well as allelic loss on chromosomes 3p, 5q, 8p, 9p, 9q, 13q, 17p, 18q, and 21q [2-5]. Our previous studies have characterized the common deletion regions at 3p and candidate TSGs within frequently deleted regions including PLCD1 and PCAF [6,7]. However, many genes associated with the development and progression of ESCC have not been characterized. To better understand the molecular mechanisms that underlie the ESCC development and progression, cDNA microarray was used to compare the gene expression profiles between 10 primary ESCC tumors and their paired non-tumorous tissues.\nAmong the 185 up-regulated genes, one gene named GPR39 drew our attention. GPR39 belongs to the G protein-coupled receptors (GPCRs) superfamily, which is the largest family of cell-surface molecules involved in signal transmission. It has been reported that GPR39 plays an important role in the regulation of gastrointestinal and metabolic function [8]. GPR39 receptor is now thought to be activated by Zn2+ signals and may have other, as yet unidentified, cognitive ligands [9]. Moreover, GPR39 receptor also displays a strong ligand-independent signaling activity through Gα12/13 as well as Gαq [10,11]. A recent study suggests that overexpression of GPR39 may inhibit cell death induced by oxidative stress, endoplasmic reticulum (ER) stress, and activation of the caspase by Bax overexpression [12]. Emerging evidence indicates that G protein-coupled receptors are crucial players in cancer progression and metastasis [13,14], however, the role of GPR39 in cancer development remains unclear. In this study, we studied GPR39 expression pattern in ESCC. The tumorigenic function of GPR39 was demonstrated by both in vitro and in vivo assays. The tumorigenic mechanism of GPR39 was also addressed. In addition, the clinical significance of GPR39 overexpression in ESCC was investigated.", "[SUBTITLE] ESCC cell lines and specimens [SUBSECTION] Chinese ESCC cell line HKESC1 was kindly provided by Professor Srivastava (Department of Pathology, The University of Hong Kong, Hong Kong, China), and two Chinese ESCC cell lines (EC18 and EC109) were kindly provided by Professor Tsao (Department of Anatomy, The University of Hong Kong). Six Japanese ESCC cell lines (KYSE30, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520) [15] were obtained from DSMZ (Braunschweig, Germany), the German Resource Centre for Biological Material. Fifty pairs of primary ESCCs and their surrounding non-tumorous esophageal tissues were collected immediately after surgical resection at Linzhou Cancer Hospital (Henan, China). Samples used in this study were approved by the Committees for Ethical Review of Research involving Human Subjects at Zhengzhou University and Sun Yat-Sen University.\nChinese ESCC cell line HKESC1 was kindly provided by Professor Srivastava (Department of Pathology, The University of Hong Kong, Hong Kong, China), and two Chinese ESCC cell lines (EC18 and EC109) were kindly provided by Professor Tsao (Department of Anatomy, The University of Hong Kong). Six Japanese ESCC cell lines (KYSE30, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520) [15] were obtained from DSMZ (Braunschweig, Germany), the German Resource Centre for Biological Material. Fifty pairs of primary ESCCs and their surrounding non-tumorous esophageal tissues were collected immediately after surgical resection at Linzhou Cancer Hospital (Henan, China). Samples used in this study were approved by the Committees for Ethical Review of Research involving Human Subjects at Zhengzhou University and Sun Yat-Sen University.\n[SUBTITLE] Semiquantitative RT-PCR [SUBSECTION] Total RNA was extracted from cell lines and frozen ESCC tissues using the Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacture's instruction. Reverse transcripation of total RNA (2 μg) was done using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA), and cDNA was subjected to PCR for a 30-cycle amplification with primers for GPR39Fw: 5'-GCCACCGGGGTCTCACTTGC-3' and GPR39Rv: 5'-GGCCGCAGCCATGATCCTCC-3'. GAPDH (Fw: 5'-CATGAGAAGTATGACAACAGCCT; Rv: 5'-AGTCCTTCCACGATACCAAAGT) was used as an internal control.\nTotal RNA was extracted from cell lines and frozen ESCC tissues using the Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacture's instruction. Reverse transcripation of total RNA (2 μg) was done using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA), and cDNA was subjected to PCR for a 30-cycle amplification with primers for GPR39Fw: 5'-GCCACCGGGGTCTCACTTGC-3' and GPR39Rv: 5'-GGCCGCAGCCATGATCCTCC-3'. GAPDH (Fw: 5'-CATGAGAAGTATGACAACAGCCT; Rv: 5'-AGTCCTTCCACGATACCAAAGT) was used as an internal control.\n[SUBTITLE] Tissue Microarrays (TMA) and Immunohistochemistry (IHC) [SUBSECTION] A total of 300 formalin-fixed and paraffin-embedded ESCC tumor specimens were kindly provided by Linzhou Cancer Hospital (Henan, China). TMAs containing 300 pairs of primary ESCC tumor samples and their corresponding nontumourous tissues were constructed as described previously [16]. Standard streptavidin-biotin-peroxidase complex method was used for IHC staining [16]. Briefly, TMA section was deparaffinized, blocked with 10% normal rabbit serum for 10 min, and incubated with rabbit anti-human GPR39 polyclonal antibody (Abcam, 1:100 dilution) overnight at 4°C. The TMA section was then incubated with biotinylated goat anti-rabbit immunoglobulin at a concentration of 1:100 at 37°C for 30 min. All of the IHC staining results were reviewed independently by two pathologists. Positive expression of GPR39 was defined as the brown staining in the cytoplasm. The staining results for GPR39 were scored semiquantitatively. Intensity was estimated in comparison to the control and scored as follows: 0, negative staining; 1, weak staining; 2, moderate staining; and 3, strong staining. Scores representing the percentage of tumor cells stained positive were as follows: 0, <1% positive tumor cells; 1, 1-10%; 2, 10-50%; 3, 50-75%; and 4, >75%. A final score was calculated by adding the scores for percentage and intensity, resulting in scores of 0 and 2-7. A score of 0 was considered negative; 2-3 was considered weak; 4-5 was considered moderate; and 6-7 was considered strong. For statistical analysis, 0-3 were counted as low expression of GPR39, while 4-7 were counted as overexpression of GPR39.\nA total of 300 formalin-fixed and paraffin-embedded ESCC tumor specimens were kindly provided by Linzhou Cancer Hospital (Henan, China). TMAs containing 300 pairs of primary ESCC tumor samples and their corresponding nontumourous tissues were constructed as described previously [16]. Standard streptavidin-biotin-peroxidase complex method was used for IHC staining [16]. Briefly, TMA section was deparaffinized, blocked with 10% normal rabbit serum for 10 min, and incubated with rabbit anti-human GPR39 polyclonal antibody (Abcam, 1:100 dilution) overnight at 4°C. The TMA section was then incubated with biotinylated goat anti-rabbit immunoglobulin at a concentration of 1:100 at 37°C for 30 min. All of the IHC staining results were reviewed independently by two pathologists. Positive expression of GPR39 was defined as the brown staining in the cytoplasm. The staining results for GPR39 were scored semiquantitatively. Intensity was estimated in comparison to the control and scored as follows: 0, negative staining; 1, weak staining; 2, moderate staining; and 3, strong staining. Scores representing the percentage of tumor cells stained positive were as follows: 0, <1% positive tumor cells; 1, 1-10%; 2, 10-50%; 3, 50-75%; and 4, >75%. A final score was calculated by adding the scores for percentage and intensity, resulting in scores of 0 and 2-7. A score of 0 was considered negative; 2-3 was considered weak; 4-5 was considered moderate; and 6-7 was considered strong. For statistical analysis, 0-3 were counted as low expression of GPR39, while 4-7 were counted as overexpression of GPR39.\n[SUBTITLE] Tumorigenic function of GPR39 [SUBSECTION] To test the tumorigenic function of GPR39, full-length GPR39 was PCR amplified, subcoloned into pcDNA3.1(+) vector (Invitrogen, Carlsbad, CA) and stably transfected into ESCC cell line KYSE30. Stable GPR39-expressing clones (GPR39-c1 and GPR39-c4) were selected for further study. Empty-vector transfected KYSE30 cells (Vec-30) were used as control.\nFor foci formation assay, 1 × 103 GPR39-expressing cells or Vec-30 cells were seeded into 6-well plate. After 7 days culture, surviving colonies (>50 cells/colony) were counted with 1% crystal violet staining. Triplicate independent experiments were performed. Colony formation in soft agar was performed by growing 1 × 104 cells in 0.4% Seaplague agar on a base of 0.6% agar in a 6-well plate. After 3 weeks, colonies consisted of more than 80 cells were counted and expressed as the means ± SD of triplicate within the same experiment. To perform cell growth assay, GPR39-expressing cells and control Vec-30 cells were seeded in 96-well plate at a density of 800 cells per well. The cell growth rate was measured using cell counting kit-8 kit (Dojindo, Japan) according to the manufacturer's instruction. Triplicate independent experiments were done.\nTo test the tumorigenic function of GPR39, full-length GPR39 was PCR amplified, subcoloned into pcDNA3.1(+) vector (Invitrogen, Carlsbad, CA) and stably transfected into ESCC cell line KYSE30. Stable GPR39-expressing clones (GPR39-c1 and GPR39-c4) were selected for further study. Empty-vector transfected KYSE30 cells (Vec-30) were used as control.\nFor foci formation assay, 1 × 103 GPR39-expressing cells or Vec-30 cells were seeded into 6-well plate. After 7 days culture, surviving colonies (>50 cells/colony) were counted with 1% crystal violet staining. Triplicate independent experiments were performed. Colony formation in soft agar was performed by growing 1 × 104 cells in 0.4% Seaplague agar on a base of 0.6% agar in a 6-well plate. After 3 weeks, colonies consisted of more than 80 cells were counted and expressed as the means ± SD of triplicate within the same experiment. To perform cell growth assay, GPR39-expressing cells and control Vec-30 cells were seeded in 96-well plate at a density of 800 cells per well. The cell growth rate was measured using cell counting kit-8 kit (Dojindo, Japan) according to the manufacturer's instruction. Triplicate independent experiments were done.\n[SUBTITLE] Flow cytometry assay [SUBSECTION] GPR39-c4 or Vec-30 cells were cultured in DMEM medium containing 10% FBS. Serum was withdraw from the culture medium when cells were 70% confluent. After 72 hrs, 10% FBS was added in the medium for an additional 8 hrs, Cells were fixed in 70% ethanol, stained with propidium iodide, and DNA content was analyzed by Cytomics FC (Beckman Coulter, Fullerton, CA).\nGPR39-c4 or Vec-30 cells were cultured in DMEM medium containing 10% FBS. Serum was withdraw from the culture medium when cells were 70% confluent. After 72 hrs, 10% FBS was added in the medium for an additional 8 hrs, Cells were fixed in 70% ethanol, stained with propidium iodide, and DNA content was analyzed by Cytomics FC (Beckman Coulter, Fullerton, CA).\n[SUBTITLE] Tumor formation in nude mice [SUBSECTION] For in vivo experiment, stable GPR39-expressing KYSE30 cells or control Vec-30 cells (1 × 106) in 200 μL serum-free DMEM (Life Technologies) were injected s.c. into the right and left flank of 4 week-old nude mice (5 mice for GPR39-c1 cells and 5 for GPR39-c4 cells), respectively. The tumor volume was calculated by the formula V = 0.5 × L × W2 [17]. All experiments were done in accordance with institutional standard guidelines of Sun Yat-Sen University for animal experiments.\nFor in vivo experiment, stable GPR39-expressing KYSE30 cells or control Vec-30 cells (1 × 106) in 200 μL serum-free DMEM (Life Technologies) were injected s.c. into the right and left flank of 4 week-old nude mice (5 mice for GPR39-c1 cells and 5 for GPR39-c4 cells), respectively. The tumor volume was calculated by the formula V = 0.5 × L × W2 [17]. All experiments were done in accordance with institutional standard guidelines of Sun Yat-Sen University for animal experiments.\n[SUBTITLE] Migration and invasion assays [SUBSECTION] For cell migration assay, GPR39-c4 cells or Vec-30 cells were grown to confluence and then mechanically scratched with a sterile pipette tip. Cells were rinsed with PBS and grown in culture medium for additional 24 hrs. The cell motility in terms of wound closure was measured by photographing at three random fields at time points 0 and 24 hr. For invasion assay, GPR39-c4 cells or Vec-30 cells were starved with serum free medium for 24 hrs before the assay. Cells (5 × 104) were suspended in 0.5 ml serum-free medium and loaded on the upper compartment of invasion chamber coated with Matrigel (BD Biosciences). The lower compartment was filled with complete medium as chemoattractant. After 24 hrs, invasive cells were fixed, stained, and counted under a microscope. Triplicate independent experiments were done.\nFor cell migration assay, GPR39-c4 cells or Vec-30 cells were grown to confluence and then mechanically scratched with a sterile pipette tip. Cells were rinsed with PBS and grown in culture medium for additional 24 hrs. The cell motility in terms of wound closure was measured by photographing at three random fields at time points 0 and 24 hr. For invasion assay, GPR39-c4 cells or Vec-30 cells were starved with serum free medium for 24 hrs before the assay. Cells (5 × 104) were suspended in 0.5 ml serum-free medium and loaded on the upper compartment of invasion chamber coated with Matrigel (BD Biosciences). The lower compartment was filled with complete medium as chemoattractant. After 24 hrs, invasive cells were fixed, stained, and counted under a microscope. Triplicate independent experiments were done.\n[SUBTITLE] F-actin staining [SUBSECTION] Cells grown on coverslips were washed three times in PBS, fixed in 4% paraformaldehyde for 20 min, and permeabilized with 0.1% Triton X-100 for 10 min. Cells were then stained with rhodamin-labeled phalloidin (Molecule Probes) in PBS containing 1% bovine serum albumin at room temperature for 30 min. After additional PBS washes, cells were counterstained with DAPI and photographed with a Leica DMRA fluorescence microscope (Rueil-Malmaison, France).\nCells grown on coverslips were washed three times in PBS, fixed in 4% paraformaldehyde for 20 min, and permeabilized with 0.1% Triton X-100 for 10 min. Cells were then stained with rhodamin-labeled phalloidin (Molecule Probes) in PBS containing 1% bovine serum albumin at room temperature for 30 min. After additional PBS washes, cells were counterstained with DAPI and photographed with a Leica DMRA fluorescence microscope (Rueil-Malmaison, France).\n[SUBTITLE] RNA interference [SUBSECTION] Small interfering RNA (siRNA) (20 μM) against GPR39 (s6073; Ambion) was transfected into KYSE180 cells in 6-well plates using Lipofectamine 2000 Reagent (Invitrogen) according to the manufacturer's instructions. At 48 hrs after transfection, the effects of gene silencing were measured via RT-PCR.\nSmall interfering RNA (siRNA) (20 μM) against GPR39 (s6073; Ambion) was transfected into KYSE180 cells in 6-well plates using Lipofectamine 2000 Reagent (Invitrogen) according to the manufacturer's instructions. At 48 hrs after transfection, the effects of gene silencing were measured via RT-PCR.\n[SUBTITLE] Western blot analysis [SUBSECTION] Western blot analysis was performed with the standard method with antibodies to GPR39, N-cadherin and GAPDH (Abcam, Cambridge Science Park, Cambridge, UK), cyclin D1, p21, CDK4 and CDK6 (Cell Signalling Technology, Frankfurt, Germany), and E-cadherin (Santa Cruz Biotechnology, Santa Cruz, CA).\nWestern blot analysis was performed with the standard method with antibodies to GPR39, N-cadherin and GAPDH (Abcam, Cambridge Science Park, Cambridge, UK), cyclin D1, p21, CDK4 and CDK6 (Cell Signalling Technology, Frankfurt, Germany), and E-cadherin (Santa Cruz Biotechnology, Santa Cruz, CA).\n[SUBTITLE] Statistical analysis [SUBSECTION] Statistical analysis was performed with the SPSS standard version 16.0 (SPSS Inc., Chicago, IL). The relationship between the expression of GPR39 protein and clinicopathologic characteristics was assessed by χ2 test. Results expressed as mean ± SD were analyzed using the Student t test. Differences were considered significant when P < 0.05.\nStatistical analysis was performed with the SPSS standard version 16.0 (SPSS Inc., Chicago, IL). The relationship between the expression of GPR39 protein and clinicopathologic characteristics was assessed by χ2 test. Results expressed as mean ± SD were analyzed using the Student t test. Differences were considered significant when P < 0.05.", "Chinese ESCC cell line HKESC1 was kindly provided by Professor Srivastava (Department of Pathology, The University of Hong Kong, Hong Kong, China), and two Chinese ESCC cell lines (EC18 and EC109) were kindly provided by Professor Tsao (Department of Anatomy, The University of Hong Kong). Six Japanese ESCC cell lines (KYSE30, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520) [15] were obtained from DSMZ (Braunschweig, Germany), the German Resource Centre for Biological Material. Fifty pairs of primary ESCCs and their surrounding non-tumorous esophageal tissues were collected immediately after surgical resection at Linzhou Cancer Hospital (Henan, China). Samples used in this study were approved by the Committees for Ethical Review of Research involving Human Subjects at Zhengzhou University and Sun Yat-Sen University.", "Total RNA was extracted from cell lines and frozen ESCC tissues using the Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacture's instruction. Reverse transcripation of total RNA (2 μg) was done using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA), and cDNA was subjected to PCR for a 30-cycle amplification with primers for GPR39Fw: 5'-GCCACCGGGGTCTCACTTGC-3' and GPR39Rv: 5'-GGCCGCAGCCATGATCCTCC-3'. GAPDH (Fw: 5'-CATGAGAAGTATGACAACAGCCT; Rv: 5'-AGTCCTTCCACGATACCAAAGT) was used as an internal control.", "A total of 300 formalin-fixed and paraffin-embedded ESCC tumor specimens were kindly provided by Linzhou Cancer Hospital (Henan, China). TMAs containing 300 pairs of primary ESCC tumor samples and their corresponding nontumourous tissues were constructed as described previously [16]. Standard streptavidin-biotin-peroxidase complex method was used for IHC staining [16]. Briefly, TMA section was deparaffinized, blocked with 10% normal rabbit serum for 10 min, and incubated with rabbit anti-human GPR39 polyclonal antibody (Abcam, 1:100 dilution) overnight at 4°C. The TMA section was then incubated with biotinylated goat anti-rabbit immunoglobulin at a concentration of 1:100 at 37°C for 30 min. All of the IHC staining results were reviewed independently by two pathologists. Positive expression of GPR39 was defined as the brown staining in the cytoplasm. The staining results for GPR39 were scored semiquantitatively. Intensity was estimated in comparison to the control and scored as follows: 0, negative staining; 1, weak staining; 2, moderate staining; and 3, strong staining. Scores representing the percentage of tumor cells stained positive were as follows: 0, <1% positive tumor cells; 1, 1-10%; 2, 10-50%; 3, 50-75%; and 4, >75%. A final score was calculated by adding the scores for percentage and intensity, resulting in scores of 0 and 2-7. A score of 0 was considered negative; 2-3 was considered weak; 4-5 was considered moderate; and 6-7 was considered strong. For statistical analysis, 0-3 were counted as low expression of GPR39, while 4-7 were counted as overexpression of GPR39.", "To test the tumorigenic function of GPR39, full-length GPR39 was PCR amplified, subcoloned into pcDNA3.1(+) vector (Invitrogen, Carlsbad, CA) and stably transfected into ESCC cell line KYSE30. Stable GPR39-expressing clones (GPR39-c1 and GPR39-c4) were selected for further study. Empty-vector transfected KYSE30 cells (Vec-30) were used as control.\nFor foci formation assay, 1 × 103 GPR39-expressing cells or Vec-30 cells were seeded into 6-well plate. After 7 days culture, surviving colonies (>50 cells/colony) were counted with 1% crystal violet staining. Triplicate independent experiments were performed. Colony formation in soft agar was performed by growing 1 × 104 cells in 0.4% Seaplague agar on a base of 0.6% agar in a 6-well plate. After 3 weeks, colonies consisted of more than 80 cells were counted and expressed as the means ± SD of triplicate within the same experiment. To perform cell growth assay, GPR39-expressing cells and control Vec-30 cells were seeded in 96-well plate at a density of 800 cells per well. The cell growth rate was measured using cell counting kit-8 kit (Dojindo, Japan) according to the manufacturer's instruction. Triplicate independent experiments were done.", "GPR39-c4 or Vec-30 cells were cultured in DMEM medium containing 10% FBS. Serum was withdraw from the culture medium when cells were 70% confluent. After 72 hrs, 10% FBS was added in the medium for an additional 8 hrs, Cells were fixed in 70% ethanol, stained with propidium iodide, and DNA content was analyzed by Cytomics FC (Beckman Coulter, Fullerton, CA).", "For in vivo experiment, stable GPR39-expressing KYSE30 cells or control Vec-30 cells (1 × 106) in 200 μL serum-free DMEM (Life Technologies) were injected s.c. into the right and left flank of 4 week-old nude mice (5 mice for GPR39-c1 cells and 5 for GPR39-c4 cells), respectively. The tumor volume was calculated by the formula V = 0.5 × L × W2 [17]. All experiments were done in accordance with institutional standard guidelines of Sun Yat-Sen University for animal experiments.", "For cell migration assay, GPR39-c4 cells or Vec-30 cells were grown to confluence and then mechanically scratched with a sterile pipette tip. Cells were rinsed with PBS and grown in culture medium for additional 24 hrs. The cell motility in terms of wound closure was measured by photographing at three random fields at time points 0 and 24 hr. For invasion assay, GPR39-c4 cells or Vec-30 cells were starved with serum free medium for 24 hrs before the assay. Cells (5 × 104) were suspended in 0.5 ml serum-free medium and loaded on the upper compartment of invasion chamber coated with Matrigel (BD Biosciences). The lower compartment was filled with complete medium as chemoattractant. After 24 hrs, invasive cells were fixed, stained, and counted under a microscope. Triplicate independent experiments were done.", "Cells grown on coverslips were washed three times in PBS, fixed in 4% paraformaldehyde for 20 min, and permeabilized with 0.1% Triton X-100 for 10 min. Cells were then stained with rhodamin-labeled phalloidin (Molecule Probes) in PBS containing 1% bovine serum albumin at room temperature for 30 min. After additional PBS washes, cells were counterstained with DAPI and photographed with a Leica DMRA fluorescence microscope (Rueil-Malmaison, France).", "Small interfering RNA (siRNA) (20 μM) against GPR39 (s6073; Ambion) was transfected into KYSE180 cells in 6-well plates using Lipofectamine 2000 Reagent (Invitrogen) according to the manufacturer's instructions. At 48 hrs after transfection, the effects of gene silencing were measured via RT-PCR.", "Western blot analysis was performed with the standard method with antibodies to GPR39, N-cadherin and GAPDH (Abcam, Cambridge Science Park, Cambridge, UK), cyclin D1, p21, CDK4 and CDK6 (Cell Signalling Technology, Frankfurt, Germany), and E-cadherin (Santa Cruz Biotechnology, Santa Cruz, CA).", "Statistical analysis was performed with the SPSS standard version 16.0 (SPSS Inc., Chicago, IL). The relationship between the expression of GPR39 protein and clinicopathologic characteristics was assessed by χ2 test. Results expressed as mean ± SD were analyzed using the Student t test. Differences were considered significant when P < 0.05.", "[SUBTITLE] GPR39 is frequently overexpressed in ESCC [SUBSECTION] Semi-quantitative RT-PCR was used to study the expression status of GPR39 in 50 primary ESCCs and 9 ESCC cell lines. Compared with their paired non-tumorous tissues, overexpression of GPR39 was detected in 27/50 (54%) of primary ESCCs (Figure 1A). Overexpression of GPR39 was also frequently detected in ESCC cell lines (HKESC1, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520; Figure 1B). GPR39 expression in protein level was further studied in 300 primary ESCCs by IHC using a tissue microarray. Informative IHC results were obtained from 207 pairs of ESCCs. Non-informative samples included lost samples, unrepresentative samples, samples with too few tumor cells, and samples with inappropriate staining; such were not used in data complication. The expression of GPR39 in normal epithelial cells was always negative or weak whereas strong positive staining of GPR39 was observed in 121/207 (58.5%) of informative ESCCs (Figure 1C).\nOverexpression of GPR39 in ESCC. GPR39 was frequently overexpressed in primary ESCCs (A) and ESCC cell lines (B) detected by RT-PCR. For primary ESCCs, expression of GPR39 in tumor tissues (T) was compared with their paired non-tumorous tissues (N). Normal esophageal tissue was used as a normal control. 18S rRNA was used as an internal control. (C) Representative of GPR39 expression in a pair of ESCC (right) and adjacent normal tissue (left) detected by immunostaining with anti-GPR39 antibody (brown). The slide was counterstained with hematoxylin (original magnification × 200).\nSemi-quantitative RT-PCR was used to study the expression status of GPR39 in 50 primary ESCCs and 9 ESCC cell lines. Compared with their paired non-tumorous tissues, overexpression of GPR39 was detected in 27/50 (54%) of primary ESCCs (Figure 1A). Overexpression of GPR39 was also frequently detected in ESCC cell lines (HKESC1, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520; Figure 1B). GPR39 expression in protein level was further studied in 300 primary ESCCs by IHC using a tissue microarray. Informative IHC results were obtained from 207 pairs of ESCCs. Non-informative samples included lost samples, unrepresentative samples, samples with too few tumor cells, and samples with inappropriate staining; such were not used in data complication. The expression of GPR39 in normal epithelial cells was always negative or weak whereas strong positive staining of GPR39 was observed in 121/207 (58.5%) of informative ESCCs (Figure 1C).\nOverexpression of GPR39 in ESCC. GPR39 was frequently overexpressed in primary ESCCs (A) and ESCC cell lines (B) detected by RT-PCR. For primary ESCCs, expression of GPR39 in tumor tissues (T) was compared with their paired non-tumorous tissues (N). Normal esophageal tissue was used as a normal control. 18S rRNA was used as an internal control. (C) Representative of GPR39 expression in a pair of ESCC (right) and adjacent normal tissue (left) detected by immunostaining with anti-GPR39 antibody (brown). The slide was counterstained with hematoxylin (original magnification × 200).\n[SUBTITLE] Clinical significance of GPR39 overexpression in ESCC [SUBSECTION] The correlation between the GPR39 overexpression and clinicopathologic features of ESCC including age (≤60 versus >60), gender (male versus female), tumor invasion (T stage: tumor depth; T3, T4 versus T1, T2), lymph nodes metastasis (N stage; N0 versus N1), TNM stage (I, IIa versus IIb, III-IV), was studied (Table 1). The results showed that overexpression of GPR39 was significantly associated with lymph node metastasis (P = 0.008) and advanced clinical stage (P = 0.004). No correlation was observed between GPR39 overexpression and age (P = 0.735), gender (P = 0.887), tumor differentiation (P = 0.846) and tumor invasion (P = 0.085).\nAssociation between GPR39 expression and clinical characteristics of ESCC patients (n = 207)\n* Statistically significant (P < 0.05)\nThe correlation between the GPR39 overexpression and clinicopathologic features of ESCC including age (≤60 versus >60), gender (male versus female), tumor invasion (T stage: tumor depth; T3, T4 versus T1, T2), lymph nodes metastasis (N stage; N0 versus N1), TNM stage (I, IIa versus IIb, III-IV), was studied (Table 1). The results showed that overexpression of GPR39 was significantly associated with lymph node metastasis (P = 0.008) and advanced clinical stage (P = 0.004). No correlation was observed between GPR39 overexpression and age (P = 0.735), gender (P = 0.887), tumor differentiation (P = 0.846) and tumor invasion (P = 0.085).\nAssociation between GPR39 expression and clinical characteristics of ESCC patients (n = 207)\n* Statistically significant (P < 0.05)\n[SUBTITLE] Tumorigenic function of GPR39 [SUBSECTION] To investigate the tumorigenic potential of GPR39, GPR39-expression vector was stably transfected into KYSE30 cells with silenced GPR39. GPR39 mRNA and protein expression in these clones were confirmed by RT-PCR and Western blot analysis (Figure 2A). The tumorigenic function of GPR39 was assessed by both in vitro and in vivo assays including foci formation, colony formation in soft agar, cell growth rate assays and tumor xenograft experiment. Foci formation assay showed that the frequency of foci formation was significantly increased (P < 0.01) in GPR39-transfectants compared with control cells (Figure 2B). A similar result was shown in soft agar assay (P < 0.01, Figure 2C). Cell growth assay also revealed that the cell growth rates in GPR39-c1 and GPR39-c4 cells were significantly enhanced by GPR39 compared with Vec-30 cells (P < 0.01, Figure 2D). To further explore the in vivo tumorigenic ability of GPR39, tumor formation in nude mice was tested by injection of GPR39-c1 cells (n = 5) or GPR39-c4 cells (n = 5), whereas Vec-30 cells were used as controls. Tumor formation was observed in all tested animals. The results showed that the tumor growth curve of GPR39-overexpressing cells was significantly increased compared to Vec-30 cells (P < 0.01, Figure 2E).\nTumorigenic function of GPR39 in ESCC cells. (A) Expression of GPR39 in GPR39-transfected KYSE30 cells was confirmed by RT-PCR (left) and Western blot analysis (right). c1 and c4 are two independent GPR39-expressing clones. Vec-30 represents empty vector-transfected KYSE30 cells. (B) Representative of foci formation in monolayer culture. Quantitative analyses of foci numbers were shown in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01; independent Student's t-test. (C) Representative of colony formation in soft agar. Percentage of colonies formed was summarized in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01. (D) Growth curves of GPR39-expressing cells were compared with Vec-30 cells by cell growth assay. The results were expressed as mean ± SD of at least three independent experiments. **P < 0.01. (E) Tumor growth curves of GPR39-expressing cells in nude mice were compared with Vec-30 cells by tumor xenograft experiment. The average tumor volume of GPR39-expressing cells vs Vec-30 cells was expressed as mean ± SD in 10 inoculated sites for each group of cells. **P < 0.01. (F) Representative examples of tumors formed in nude mice following injection of GPR39-expressing KYSE30 cells (right) and Vec-30 cells (left).\nTo investigate the tumorigenic potential of GPR39, GPR39-expression vector was stably transfected into KYSE30 cells with silenced GPR39. GPR39 mRNA and protein expression in these clones were confirmed by RT-PCR and Western blot analysis (Figure 2A). The tumorigenic function of GPR39 was assessed by both in vitro and in vivo assays including foci formation, colony formation in soft agar, cell growth rate assays and tumor xenograft experiment. Foci formation assay showed that the frequency of foci formation was significantly increased (P < 0.01) in GPR39-transfectants compared with control cells (Figure 2B). A similar result was shown in soft agar assay (P < 0.01, Figure 2C). Cell growth assay also revealed that the cell growth rates in GPR39-c1 and GPR39-c4 cells were significantly enhanced by GPR39 compared with Vec-30 cells (P < 0.01, Figure 2D). To further explore the in vivo tumorigenic ability of GPR39, tumor formation in nude mice was tested by injection of GPR39-c1 cells (n = 5) or GPR39-c4 cells (n = 5), whereas Vec-30 cells were used as controls. Tumor formation was observed in all tested animals. The results showed that the tumor growth curve of GPR39-overexpressing cells was significantly increased compared to Vec-30 cells (P < 0.01, Figure 2E).\nTumorigenic function of GPR39 in ESCC cells. (A) Expression of GPR39 in GPR39-transfected KYSE30 cells was confirmed by RT-PCR (left) and Western blot analysis (right). c1 and c4 are two independent GPR39-expressing clones. Vec-30 represents empty vector-transfected KYSE30 cells. (B) Representative of foci formation in monolayer culture. Quantitative analyses of foci numbers were shown in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01; independent Student's t-test. (C) Representative of colony formation in soft agar. Percentage of colonies formed was summarized in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01. (D) Growth curves of GPR39-expressing cells were compared with Vec-30 cells by cell growth assay. The results were expressed as mean ± SD of at least three independent experiments. **P < 0.01. (E) Tumor growth curves of GPR39-expressing cells in nude mice were compared with Vec-30 cells by tumor xenograft experiment. The average tumor volume of GPR39-expressing cells vs Vec-30 cells was expressed as mean ± SD in 10 inoculated sites for each group of cells. **P < 0.01. (F) Representative examples of tumors formed in nude mice following injection of GPR39-expressing KYSE30 cells (right) and Vec-30 cells (left).\n[SUBTITLE] GPR39 promotes G1/S transition [SUBSECTION] To explore the mechanism underlying growth promotion by GPR39, the cell cycle distributions of GPR39-c4 and Vec-30 cells were determined by flow cytometry. Before treatment, the percentage of GPR39-c4 cells in G1 phase was obviously reduced in comparison with Vec-30 cells (38.37 ± 1.02% versus 45.87 ± 0.47%, P < 0.05; Figure 3A). After 3 days' serum starvation followed by addition of 10% serum for 8 hrs, the percentage of cells in S phase was significantly increased in GPR39-c4 cells compared to Vec-30 cells (26.43 ± 0.71% versus 8.97 ± 0.31%, P < 0.05; Figure 3A), suggesting that GPR39 was able to promote G1/S transition. To reveal the potential molecular mechanism of GPR39 in cell cycle promotion, expressions of several key cell cycle regulators including p21, cyclin D1, CDK4 and CDK6 were compared between GPR39-c4 and Vec-30 cells. Increased expression of cyclin D1 and CDK6, but not p21 and CDK4, were detected in GPR39-c4 (Figure 3B).\nGPR39 promotes G1/S transition and enhances cell motility. (A) DNA content between GPR39-expressing cells and control Vec-30 cell were compared by Flow-cytometry. Untreated, cells were cultured in DMEM medium with 10% FBS; Withdraw serum, cells were cultured in DMEM medium without serum for 3 days; Add serum, cells were cultured again in DMEM medium with 10% FBS for 8 hr. (B) Expression of p21, cyclin D1, CDK4, and CDK6 were compared between GPR39-expressing cells (c4) and control Vec-30 cells by Western blot analyses. GAPDH was used as loading control. (C) The effect of GPR39 on cell migration was determined by wound-healing assay. During a period of 16 hr, the spreading speed of GPR39-expressing cells along the wound edge was faster than that in control Vec-30 cells. (D) Representative images showed the GPR39-expressing cells and Vec-30 cells that invaded through the matrigel. Number of invaded tumor cells was quantified in the right panel. Columns, mean of triplicate experiments; **P < 0.01.\nTo explore the mechanism underlying growth promotion by GPR39, the cell cycle distributions of GPR39-c4 and Vec-30 cells were determined by flow cytometry. Before treatment, the percentage of GPR39-c4 cells in G1 phase was obviously reduced in comparison with Vec-30 cells (38.37 ± 1.02% versus 45.87 ± 0.47%, P < 0.05; Figure 3A). After 3 days' serum starvation followed by addition of 10% serum for 8 hrs, the percentage of cells in S phase was significantly increased in GPR39-c4 cells compared to Vec-30 cells (26.43 ± 0.71% versus 8.97 ± 0.31%, P < 0.05; Figure 3A), suggesting that GPR39 was able to promote G1/S transition. To reveal the potential molecular mechanism of GPR39 in cell cycle promotion, expressions of several key cell cycle regulators including p21, cyclin D1, CDK4 and CDK6 were compared between GPR39-c4 and Vec-30 cells. Increased expression of cyclin D1 and CDK6, but not p21 and CDK4, were detected in GPR39-c4 (Figure 3B).\nGPR39 promotes G1/S transition and enhances cell motility. (A) DNA content between GPR39-expressing cells and control Vec-30 cell were compared by Flow-cytometry. Untreated, cells were cultured in DMEM medium with 10% FBS; Withdraw serum, cells were cultured in DMEM medium without serum for 3 days; Add serum, cells were cultured again in DMEM medium with 10% FBS for 8 hr. (B) Expression of p21, cyclin D1, CDK4, and CDK6 were compared between GPR39-expressing cells (c4) and control Vec-30 cells by Western blot analyses. GAPDH was used as loading control. (C) The effect of GPR39 on cell migration was determined by wound-healing assay. During a period of 16 hr, the spreading speed of GPR39-expressing cells along the wound edge was faster than that in control Vec-30 cells. (D) Representative images showed the GPR39-expressing cells and Vec-30 cells that invaded through the matrigel. Number of invaded tumor cells was quantified in the right panel. Columns, mean of triplicate experiments; **P < 0.01.\n[SUBTITLE] GPR39 enhances cell motility and invasiveness of ESCCs [SUBSECTION] As the TMA result showed that overexpression of GPR39 was closely associated with ESCC metastasis, the effects of GPR39 on cell migration and invasion were studied by wound-healing and cell invasion assays. Wound-healing assay showed that that the ectopic expression of GPR39 could significantly increase cell migration ability in GPR39-transfected cells compared with empty-vector control (P < 0.05, Figure 3C). Matrigel invasion assay also found that the ectopic expression of GPR39 could significantly enhanced the invasiveness of ESCC cells, as demonstrated by a significant increase in the number of invaded cells (P < 0.01, Figure 3D), in GPR39-transfected cells compared with empty-vector control.\nAs the TMA result showed that overexpression of GPR39 was closely associated with ESCC metastasis, the effects of GPR39 on cell migration and invasion were studied by wound-healing and cell invasion assays. Wound-healing assay showed that that the ectopic expression of GPR39 could significantly increase cell migration ability in GPR39-transfected cells compared with empty-vector control (P < 0.05, Figure 3C). Matrigel invasion assay also found that the ectopic expression of GPR39 could significantly enhanced the invasiveness of ESCC cells, as demonstrated by a significant increase in the number of invaded cells (P < 0.01, Figure 3D), in GPR39-transfected cells compared with empty-vector control.\n[SUBTITLE] GPR39 induces partial epithelial-mesenchymal transition (EMT) [SUBSECTION] In this study, we found that the cell morphology changed obviously after the transfection of GPR39. GPR39-transfected cells showed spindle shape and fibroblastic changes in monolayer culture, whereas empty vector-transfected cells, like KYSE30 parental cells, kept their cobblestone-like phenotype (Figure 4A). To determine whether the effect of GPR39 on cell motility was associated with EMT, expressions of several epithelial markers (E-cadherin, N-cadherin) and mesenchymal markers (vimentin, and fibronectin) were compared between GPR39-c4 and Vec-30 cells by RT-PCR and Western blot analysis. The results showed that E-cadherin was obviously down-regulated in GPR39-c4 cells; however, no obvious difference was observed in the expression of N-cadherin, vimentin and fibronectin between GPR39-c4 and Vec-30 cells (Figure 4B). These findings indicated that GPR39 increased cell motility was partially through the EMT.\nGPR39 promotes cell mobility and invasion by inducing partial EMT and remodeling cytoskeleton. (A) Representatives of cell morphology of GPR39-expressing cells and Vec-30 cells (original magnification × 200). (B) Expressions of epithelial markers E-cadherin and mesenchymal markers fibronectin, N-cadherin, and vimentin, were compared by RT-PCR or Western blotting analysis between GPR39-expressing cells and Vec-30 cells. GAPDH was used as loading control. (C) Representative images of F-actin staining. Formation of lamellipodia (indicated by arrows) was stimulated by GPR39 compared to control cells (magnification × 400).\nIn this study, we found that the cell morphology changed obviously after the transfection of GPR39. GPR39-transfected cells showed spindle shape and fibroblastic changes in monolayer culture, whereas empty vector-transfected cells, like KYSE30 parental cells, kept their cobblestone-like phenotype (Figure 4A). To determine whether the effect of GPR39 on cell motility was associated with EMT, expressions of several epithelial markers (E-cadherin, N-cadherin) and mesenchymal markers (vimentin, and fibronectin) were compared between GPR39-c4 and Vec-30 cells by RT-PCR and Western blot analysis. The results showed that E-cadherin was obviously down-regulated in GPR39-c4 cells; however, no obvious difference was observed in the expression of N-cadherin, vimentin and fibronectin between GPR39-c4 and Vec-30 cells (Figure 4B). These findings indicated that GPR39 increased cell motility was partially through the EMT.\nGPR39 promotes cell mobility and invasion by inducing partial EMT and remodeling cytoskeleton. (A) Representatives of cell morphology of GPR39-expressing cells and Vec-30 cells (original magnification × 200). (B) Expressions of epithelial markers E-cadherin and mesenchymal markers fibronectin, N-cadherin, and vimentin, were compared by RT-PCR or Western blotting analysis between GPR39-expressing cells and Vec-30 cells. GAPDH was used as loading control. (C) Representative images of F-actin staining. Formation of lamellipodia (indicated by arrows) was stimulated by GPR39 compared to control cells (magnification × 400).\n[SUBTITLE] Overexpression of GPR39 induced lamellipodia formation [SUBSECTION] To further explore the molecular mechanism of GPR39 in regulating cancer invasion and metastasis, the role of GPR39 in the polymerized actin was investigated by phalloidin staining. The results showed that GPR39-expressing cells exhibited enhanced lamellipodia formation compared with control cells (Figure 4C), indicating that GPR39 could induce cytoskeleton remodeling to facilitate esophageal cancer cell migration and invasion.\nTo further explore the molecular mechanism of GPR39 in regulating cancer invasion and metastasis, the role of GPR39 in the polymerized actin was investigated by phalloidin staining. The results showed that GPR39-expressing cells exhibited enhanced lamellipodia formation compared with control cells (Figure 4C), indicating that GPR39 could induce cytoskeleton remodeling to facilitate esophageal cancer cell migration and invasion.\n[SUBTITLE] Silencing GPR39 expression by RNA interference (RNAi) [SUBSECTION] ESCC cell line KYSE180, which expresses a high level of endogenous GPR39, was used in the siRNA experiment. Two siRNAs targeting GPR39 (GPR39-si1 and GPR39-si2) were tested and the efficiency of GPR39 gene silencing was detected by RT-PCR. The result showed that the GPR39-si1 had a better silencing effect (Figure 5A). Silencing of GPR39 resulted in a significant inhibition of the cell growth rate (P < 0.01, Figure 5B) and migration (Figure 5C). DNA content analysis by flow cytometry showed that GPR39-si1 was able to inhibit the cell cycle at the G1/S checkpoint (Figure 5D). The percentage of cells in the S phase was significantly reduced in GPR39-si1-treated cells (27.23 ± 1.26%) compared with that in control-si-treated cells (35.13 ± 1.12%; P < 0.05). These findings further supported that the tumorigenic function of GPR39 was through its role in promoting cell proliferation and motility.\nSilencing of GPR39 expression suppresses tumorigenic ability of GPR39. (A) GPR39 expression was efficiently decreased by the treatment of siGPR39 by RT-PCR. Relative expression level was measured by densitometer and summarized in the right panel. **P < 0.01. (B) Growth curve of KYSE180 cells treated with GPR39 siRNA was compared with control siRNA treated cells by cell growth assay. **P < 0.01. (C) Cell migration assay was used to compare the frequency of migratory cells between KYSE180 cells treated with control siRNA and GPR39 siRNA. (D) DNA content between control siRNA and GPR39 siRNA treated cells were compared by Flow-cytometry. *P < 0.05.\nESCC cell line KYSE180, which expresses a high level of endogenous GPR39, was used in the siRNA experiment. Two siRNAs targeting GPR39 (GPR39-si1 and GPR39-si2) were tested and the efficiency of GPR39 gene silencing was detected by RT-PCR. The result showed that the GPR39-si1 had a better silencing effect (Figure 5A). Silencing of GPR39 resulted in a significant inhibition of the cell growth rate (P < 0.01, Figure 5B) and migration (Figure 5C). DNA content analysis by flow cytometry showed that GPR39-si1 was able to inhibit the cell cycle at the G1/S checkpoint (Figure 5D). The percentage of cells in the S phase was significantly reduced in GPR39-si1-treated cells (27.23 ± 1.26%) compared with that in control-si-treated cells (35.13 ± 1.12%; P < 0.05). These findings further supported that the tumorigenic function of GPR39 was through its role in promoting cell proliferation and motility.\nSilencing of GPR39 expression suppresses tumorigenic ability of GPR39. (A) GPR39 expression was efficiently decreased by the treatment of siGPR39 by RT-PCR. Relative expression level was measured by densitometer and summarized in the right panel. **P < 0.01. (B) Growth curve of KYSE180 cells treated with GPR39 siRNA was compared with control siRNA treated cells by cell growth assay. **P < 0.01. (C) Cell migration assay was used to compare the frequency of migratory cells between KYSE180 cells treated with control siRNA and GPR39 siRNA. (D) DNA content between control siRNA and GPR39 siRNA treated cells were compared by Flow-cytometry. *P < 0.05.", "Semi-quantitative RT-PCR was used to study the expression status of GPR39 in 50 primary ESCCs and 9 ESCC cell lines. Compared with their paired non-tumorous tissues, overexpression of GPR39 was detected in 27/50 (54%) of primary ESCCs (Figure 1A). Overexpression of GPR39 was also frequently detected in ESCC cell lines (HKESC1, KYSE140, KYSE180, KYSE410, KYSE510 and KYSE520; Figure 1B). GPR39 expression in protein level was further studied in 300 primary ESCCs by IHC using a tissue microarray. Informative IHC results were obtained from 207 pairs of ESCCs. Non-informative samples included lost samples, unrepresentative samples, samples with too few tumor cells, and samples with inappropriate staining; such were not used in data complication. The expression of GPR39 in normal epithelial cells was always negative or weak whereas strong positive staining of GPR39 was observed in 121/207 (58.5%) of informative ESCCs (Figure 1C).\nOverexpression of GPR39 in ESCC. GPR39 was frequently overexpressed in primary ESCCs (A) and ESCC cell lines (B) detected by RT-PCR. For primary ESCCs, expression of GPR39 in tumor tissues (T) was compared with their paired non-tumorous tissues (N). Normal esophageal tissue was used as a normal control. 18S rRNA was used as an internal control. (C) Representative of GPR39 expression in a pair of ESCC (right) and adjacent normal tissue (left) detected by immunostaining with anti-GPR39 antibody (brown). The slide was counterstained with hematoxylin (original magnification × 200).", "The correlation between the GPR39 overexpression and clinicopathologic features of ESCC including age (≤60 versus >60), gender (male versus female), tumor invasion (T stage: tumor depth; T3, T4 versus T1, T2), lymph nodes metastasis (N stage; N0 versus N1), TNM stage (I, IIa versus IIb, III-IV), was studied (Table 1). The results showed that overexpression of GPR39 was significantly associated with lymph node metastasis (P = 0.008) and advanced clinical stage (P = 0.004). No correlation was observed between GPR39 overexpression and age (P = 0.735), gender (P = 0.887), tumor differentiation (P = 0.846) and tumor invasion (P = 0.085).\nAssociation between GPR39 expression and clinical characteristics of ESCC patients (n = 207)\n* Statistically significant (P < 0.05)", "To investigate the tumorigenic potential of GPR39, GPR39-expression vector was stably transfected into KYSE30 cells with silenced GPR39. GPR39 mRNA and protein expression in these clones were confirmed by RT-PCR and Western blot analysis (Figure 2A). The tumorigenic function of GPR39 was assessed by both in vitro and in vivo assays including foci formation, colony formation in soft agar, cell growth rate assays and tumor xenograft experiment. Foci formation assay showed that the frequency of foci formation was significantly increased (P < 0.01) in GPR39-transfectants compared with control cells (Figure 2B). A similar result was shown in soft agar assay (P < 0.01, Figure 2C). Cell growth assay also revealed that the cell growth rates in GPR39-c1 and GPR39-c4 cells were significantly enhanced by GPR39 compared with Vec-30 cells (P < 0.01, Figure 2D). To further explore the in vivo tumorigenic ability of GPR39, tumor formation in nude mice was tested by injection of GPR39-c1 cells (n = 5) or GPR39-c4 cells (n = 5), whereas Vec-30 cells were used as controls. Tumor formation was observed in all tested animals. The results showed that the tumor growth curve of GPR39-overexpressing cells was significantly increased compared to Vec-30 cells (P < 0.01, Figure 2E).\nTumorigenic function of GPR39 in ESCC cells. (A) Expression of GPR39 in GPR39-transfected KYSE30 cells was confirmed by RT-PCR (left) and Western blot analysis (right). c1 and c4 are two independent GPR39-expressing clones. Vec-30 represents empty vector-transfected KYSE30 cells. (B) Representative of foci formation in monolayer culture. Quantitative analyses of foci numbers were shown in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01; independent Student's t-test. (C) Representative of colony formation in soft agar. Percentage of colonies formed was summarized in the right panel. Values were the mean ± SD of at least three independent experiments. **P < 0.01. (D) Growth curves of GPR39-expressing cells were compared with Vec-30 cells by cell growth assay. The results were expressed as mean ± SD of at least three independent experiments. **P < 0.01. (E) Tumor growth curves of GPR39-expressing cells in nude mice were compared with Vec-30 cells by tumor xenograft experiment. The average tumor volume of GPR39-expressing cells vs Vec-30 cells was expressed as mean ± SD in 10 inoculated sites for each group of cells. **P < 0.01. (F) Representative examples of tumors formed in nude mice following injection of GPR39-expressing KYSE30 cells (right) and Vec-30 cells (left).", "To explore the mechanism underlying growth promotion by GPR39, the cell cycle distributions of GPR39-c4 and Vec-30 cells were determined by flow cytometry. Before treatment, the percentage of GPR39-c4 cells in G1 phase was obviously reduced in comparison with Vec-30 cells (38.37 ± 1.02% versus 45.87 ± 0.47%, P < 0.05; Figure 3A). After 3 days' serum starvation followed by addition of 10% serum for 8 hrs, the percentage of cells in S phase was significantly increased in GPR39-c4 cells compared to Vec-30 cells (26.43 ± 0.71% versus 8.97 ± 0.31%, P < 0.05; Figure 3A), suggesting that GPR39 was able to promote G1/S transition. To reveal the potential molecular mechanism of GPR39 in cell cycle promotion, expressions of several key cell cycle regulators including p21, cyclin D1, CDK4 and CDK6 were compared between GPR39-c4 and Vec-30 cells. Increased expression of cyclin D1 and CDK6, but not p21 and CDK4, were detected in GPR39-c4 (Figure 3B).\nGPR39 promotes G1/S transition and enhances cell motility. (A) DNA content between GPR39-expressing cells and control Vec-30 cell were compared by Flow-cytometry. Untreated, cells were cultured in DMEM medium with 10% FBS; Withdraw serum, cells were cultured in DMEM medium without serum for 3 days; Add serum, cells were cultured again in DMEM medium with 10% FBS for 8 hr. (B) Expression of p21, cyclin D1, CDK4, and CDK6 were compared between GPR39-expressing cells (c4) and control Vec-30 cells by Western blot analyses. GAPDH was used as loading control. (C) The effect of GPR39 on cell migration was determined by wound-healing assay. During a period of 16 hr, the spreading speed of GPR39-expressing cells along the wound edge was faster than that in control Vec-30 cells. (D) Representative images showed the GPR39-expressing cells and Vec-30 cells that invaded through the matrigel. Number of invaded tumor cells was quantified in the right panel. Columns, mean of triplicate experiments; **P < 0.01.", "As the TMA result showed that overexpression of GPR39 was closely associated with ESCC metastasis, the effects of GPR39 on cell migration and invasion were studied by wound-healing and cell invasion assays. Wound-healing assay showed that that the ectopic expression of GPR39 could significantly increase cell migration ability in GPR39-transfected cells compared with empty-vector control (P < 0.05, Figure 3C). Matrigel invasion assay also found that the ectopic expression of GPR39 could significantly enhanced the invasiveness of ESCC cells, as demonstrated by a significant increase in the number of invaded cells (P < 0.01, Figure 3D), in GPR39-transfected cells compared with empty-vector control.", "In this study, we found that the cell morphology changed obviously after the transfection of GPR39. GPR39-transfected cells showed spindle shape and fibroblastic changes in monolayer culture, whereas empty vector-transfected cells, like KYSE30 parental cells, kept their cobblestone-like phenotype (Figure 4A). To determine whether the effect of GPR39 on cell motility was associated with EMT, expressions of several epithelial markers (E-cadherin, N-cadherin) and mesenchymal markers (vimentin, and fibronectin) were compared between GPR39-c4 and Vec-30 cells by RT-PCR and Western blot analysis. The results showed that E-cadherin was obviously down-regulated in GPR39-c4 cells; however, no obvious difference was observed in the expression of N-cadherin, vimentin and fibronectin between GPR39-c4 and Vec-30 cells (Figure 4B). These findings indicated that GPR39 increased cell motility was partially through the EMT.\nGPR39 promotes cell mobility and invasion by inducing partial EMT and remodeling cytoskeleton. (A) Representatives of cell morphology of GPR39-expressing cells and Vec-30 cells (original magnification × 200). (B) Expressions of epithelial markers E-cadherin and mesenchymal markers fibronectin, N-cadherin, and vimentin, were compared by RT-PCR or Western blotting analysis between GPR39-expressing cells and Vec-30 cells. GAPDH was used as loading control. (C) Representative images of F-actin staining. Formation of lamellipodia (indicated by arrows) was stimulated by GPR39 compared to control cells (magnification × 400).", "To further explore the molecular mechanism of GPR39 in regulating cancer invasion and metastasis, the role of GPR39 in the polymerized actin was investigated by phalloidin staining. The results showed that GPR39-expressing cells exhibited enhanced lamellipodia formation compared with control cells (Figure 4C), indicating that GPR39 could induce cytoskeleton remodeling to facilitate esophageal cancer cell migration and invasion.", "ESCC cell line KYSE180, which expresses a high level of endogenous GPR39, was used in the siRNA experiment. Two siRNAs targeting GPR39 (GPR39-si1 and GPR39-si2) were tested and the efficiency of GPR39 gene silencing was detected by RT-PCR. The result showed that the GPR39-si1 had a better silencing effect (Figure 5A). Silencing of GPR39 resulted in a significant inhibition of the cell growth rate (P < 0.01, Figure 5B) and migration (Figure 5C). DNA content analysis by flow cytometry showed that GPR39-si1 was able to inhibit the cell cycle at the G1/S checkpoint (Figure 5D). The percentage of cells in the S phase was significantly reduced in GPR39-si1-treated cells (27.23 ± 1.26%) compared with that in control-si-treated cells (35.13 ± 1.12%; P < 0.05). These findings further supported that the tumorigenic function of GPR39 was through its role in promoting cell proliferation and motility.\nSilencing of GPR39 expression suppresses tumorigenic ability of GPR39. (A) GPR39 expression was efficiently decreased by the treatment of siGPR39 by RT-PCR. Relative expression level was measured by densitometer and summarized in the right panel. **P < 0.01. (B) Growth curve of KYSE180 cells treated with GPR39 siRNA was compared with control siRNA treated cells by cell growth assay. **P < 0.01. (C) Cell migration assay was used to compare the frequency of migratory cells between KYSE180 cells treated with control siRNA and GPR39 siRNA. (D) DNA content between control siRNA and GPR39 siRNA treated cells were compared by Flow-cytometry. *P < 0.05.", "Many G protein-coupled receptors (GPCRs) have been found to play critical roles in the development and progression of cancer, including malignant transformation [18,19], tumor growth and survival [20,21], as well as invasion and metastasis [22,23]. Herein, we report that one of the G protein-coupled receptors, GPR39, is frequently overexpressed in human ESCC. To our knowledge, this is the first illustration that GPR39 contributes to the development and progression of ESCC. In the present study, the tumorigenic function of GPR39 was demonstrated by both in vitro and in vivo assays. Functional studies showed that GPR39 could effectively promote ESCC cancer cell growth, increase foci formation and colony formation and enhance tumor formation in nude mice. A recent study suggested that zinc could be a ligand capable of activating the GPR39 receptor [11]. Interestingly, zinc deficiency along with its associated increased cell proliferation can be tumorigenic in the rat esophagus [24,25]. Our study also provided evidence that ectopic expression of GPR39 increased ESCC cancer cell growth, indicating involvement of the GPR39 receptor in the tumorigenesis of esophageal cancer. However, whether GPR39 signaling is activated by zinc in esophageal carcinogenesis needs to be further investigated. Further study revealed that overexpression of GPR39 in esophageal cancer cells KYSE30 promoted G1/S phase transition. We showed for the first time that GPR39 controls cell cycle progression through the activation of CDK6 and its activating protein, cyclin D1. G1/S phase transition is a major checkpoint for cell cycle progression and cyclin D1-CDK6 complex is one of the critical positive regulators during this transition [26,27]. On the other hand, we found that silencing of GPR39 expression could inhibit tumorigenicity in KYSE180 cells through the cell cycle arrest at G1/S checkpoint.\nAnother interesting finding of this study is the promoting effect of GPR39 on tumor metastasis in ESCC. Our data showed that overexpression of GPR39 could promote cell motility and invasiveness of ESCC cells in vitro. This mirrored the findings of GPR39 overexpression in human ESCC samples and its association with advanced clinical stage and lymph node metastasis of ESCC. Conversely, when we knocked down the endogenous GPR39 by RNAi in ESCC cells, the mobility of ESCC cells was significantly reduced, suggesting that GPR39 is closely involved in ESCC invasion and metastasis. Moreover, the observation of overexpression of GPR39 resulting in cell morphological alteration promoted us to further investigate its effect on EMT. We found that GPR39 has some impact on the EMT as shown by decreasing the epithelial molecule E-cadherin, an event critical in tumour invasion and a 'master' regulator of EMT. E-cadherin provides a physical link among adjacent cells and is crucial for the establishment and maintenance of polarity and the structural integrity of epithelia. Indeed, due to the physical and functional link between E-cadherin based complexes and cytoskeletal components, a change in the E-cadherin mediated adhesiveness leads to rearrangement of the cytoskeleton [28]. In view of this, we further explored the role of GPR39 in reorganization of the actin cytoskeleton. As expected, our result showed that GPR39 led to significant alterations on cytoskeleton by inducing the lamellipodia formation in GPR39-transfected ESCC cells. This finding was consistent to previous studies that some G protein-coupled receptors (GPCRs) were able to promote actin reorganization and result in cell shape changes and enhanced cell migration [13,29], indicating that GPR39 might directly alter the cytoskeleton to favor the tumor cell invasion and metastasis in ESCC.\nIn this study, we have also provided evidence that targeting of GPR39 with specific RNAi will reduce the oncogenic characteristics of ESCC tumor cells. To date, some G protein-coupled receptors (GPCRs) provide important practical options for preclinical research, clinical trials, and cancer treatment [30]. Therefore, consideration should be given to the development of novel therapeutics targeting GPR39 for use in GPR39-expressing ESCC tumors.", "In summary, our findings demonstrate that GPR39 plays an important role in ESCC development and progression via promoting cell proliferation, enhancing cell motility and invasiveness, regulating cytoskeleton and inducing EMT. A better understanding of the molecular mechanism of GPR39 in ESCC development and progression would provide novel therapeutic strategies to ESCC cancer patients.", "EMT: epithelial mesenchymal transition; ESCC: esophageal squamous cell carcinoma; GPCR: G protein-coupled receptor; siRNA: small interfering RNA; TMA: tissue microarray; TSG: tumor suppressor gene; L: length; V: volume; W: width.", "The authors declare that they have no competing interests.", "FX and HL performed the experimental procedures with support from YZ, YQ, YD, TZ, LC, CN, TH and YL. FX, LF and XYG were responsible for experimental design, interpretation of the results and writing the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2407/11/86/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adolescent", "Adult", "Female", "Health Surveys", "Humans", "Longitudinal Studies", "Male", "Mexico", "Middle Aged", "Smoking", "Smoking Cessation", "Young Adult" ]

[ "Background", "Study sample", "Measures", "Smoking and quitting behaviour", "Singles use and perceptions", "Sociodemographic variables", "Analysis", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "The World Health Organization Framework Convention on Tobacco Control (WHO-FCTC) recommends that the sale of single cigarettes be banned [1,2]. Mexico has banned singles since 1999, before it ratified the WHO-FCTC in 2004. Nevertheless, the sale of single cigarettes appears prevalent in Mexico [3-5], as in other jurisdictions that have banned them [6-10]. The rationale for banning singles primarily reflects concerns about facilitating youth access [11]. However, how adult smokers respond to the availability of single cigarettes is less well understood and merits clarification, especially as the availability and consumption of singles cigarettes could either lower or increase successful smoking cessation.\nAccess to single cigarettes may promote smoking among adult smokers. For example, because the price of a single cigarette is lower than for a cigarette pack [12-14], smokers who may otherwise quit because of affordability issues may continue to nourish their addiction, albeit while smoking fewer cigarettes. Perhaps reflecting affordability issues, younger smokers and smokers with lower incomes are more likely to smoke singles in Mexico [3]. Furthermore, when a tax increase went into effect in Mexico, consumption went down overall, while the prevalence of singles use increased from 10% to 20% [13]. The visibility of single cigarettes in the surrounding environment may also cue smoking behavior or promote relapse [15]. Indeed, Mexican smokers who experience more frequent cravings to smoke because of seeing singles cigarettes for sale are less likely to intend to quit than Mexican smokers who do not experience such cues or cravings [3]. Similarly, the visibility of singles sales both within and outside of more formal points of sale may support perceptions of the normative nature of smoking. Indeed, the frequency of brand impressions may increase with the visible presence of cigarette packages from which vendors sell singles. Such brand imagery exposures may increase in importance as other advertising channels are banned [16]. Finally, people who consume singles may be less likely to be exposed to health warning labels on cigarette packs compared to people who buy and carry with them the packages that contain these warnings.\nThe availability of singles cigarettes may also contribute to reductions in cigarette consumption among adults who smoke. In disadvantaged urban settings in the US [14], as well as in Mexico [3], adult smokers report buying single cigarettes as a method to keep consumption down and to quit. Single cigarettes cost around twice as much, per stick, as cigarettes bought in a standard package, so smokers impose upon themselves a steeper monetary cost to limit consumption. Furthermore, purchasing singles imposes greater search costs for each cigarette [17], often involving a substantial increase in the amount of time spent going to the place where single cigarettes are sold, when compared to the time spent reaching into a cigarette package that is kept at hand. In cross-sectional analyses, the frequency of using singles to limit smoking behavior was positively associated with quit intentions among Mexican smokers [3]. Longitudinal analyses are needed to clarify whether the availability of single cigarettes on balance promotes or inhibits smoking among adults.\nThis study uses longitudinal data from a cohort of smokers representative of six major Mexican cities in order to determine the relationship between consumption of single cigarettes and downstream quit behavior. First, we examine correlates of single cigarette consumption. Second, we examine whether baseline consumption and perceptions of single cigarettes are associated with quit attempts and quitting after baseline.", "As part of the International Tobacco Control (ITC) Policy Evaluation Survey, data have been collected on an annual basis from adult smokers in Mexico since 2006. The longitudinal analytic sample for the current study consists of data from wave 3 (November to December 2008) and wave 4 (January to February 2010) of the ITC-Mexico survey. Between previous survey administrations, taxes were increased [13], and advertising restrictions and smoke-free policies were implemented [18]; however, no major tobacco control policies were implemented between waves 3 and 4. In six cities (i.e., Mexico City, Monterrey, Guadalajara, Puebla, Tijuana, Mérida) data were collected at both of these waves. A stratified, multi-stage sampling strategy was used within the urban limits designated for each city. Within selected block groups, face-to-face interviews were conducted with randomly selected adult smokers, defined as those who has smoked at least once the previous week and at least 100 lifetime cigarettes (for details, see Thrasher et al., 2009). The wave 3 analytic sample (n = 1649) includes 73% (524/717) of those who were successfully followed up from the wave 2 sample cities (i.e., Mexico City, Guadalajara, Tijuana), of whom 117 were excluded because they had quit by wave 3. The sample also includes a replenishment sample of 203 smokers randomly selected from the same census tracts as those selected for the original sample, as well as new samples of smokers in Mérida, Monterrey and Puebla (n = 813) as well as an augmented sample in Mexico City (n = 135). Household contact and cooperation rates for wave 3 was 79% and 70%, respectively. Sampling weights were developed to account for the likelihood of participant selection. To produce more efficient estimates of association [19], the weights used for model estimation were rescaled to sum to the sample size within each city. The protocol for this study was approved by the IRB at the Mexican National Institute of Public Health. Data are not publically available, but may be requested through http://www.itcproject.org.", "[SUBTITLE] Smoking and quitting behaviour [SUBSECTION] Baseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].\nBaseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].\n[SUBTITLE] Singles use and perceptions [SUBSECTION] At baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.\nAt baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.\n[SUBTITLE] Sociodemographic variables [SUBSECTION] At baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.\nAt baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.", "Baseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].", "At baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.", "At baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.", "Analyses were conducted using STATA, version 11.0. Univariate descriptions of the study sample and attrition analyses of differences between the sample that was and was not followed up were calculated without taking into account the complex survey design and sampling weights; however, all other analyses took into account the design and weights. Logistic regression was used to estimate crude and adjusted odds ratios that expressed associations between study variables and buying single cigarettes at the last purchase, as well as when determinng the odds of subsequent attempts to quit and being quit. Ordinal regression models were used to estimate the bivariate and multivariate adjusted relationships between study variables and frequency of purchasing single cigarettes. For models that regressed singles purchase behavior on study variables, the baseline (i.e., wave 3) analytic sample of smokers was assessed, whether participants were successfully followed up or not. For logistic models that regressed subsequent quit behavior (i.e., wave 4) on baseline variables (i.e., wave 3), only data were analyzed from participants who were successfully surveyed at both waves.", "The authors declare that they have no competing interests.", "JT conceptualized and wrote the majority of the manuscript. VV conducted the statistical analyses and wrote initial drafts of the methods and results. JB, RS and RO provided substantial contributions to the background and implications of the study. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2458/11/134/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study sample", "Measures", "Smoking and quitting behaviour", "Singles use and perceptions", "Sociodemographic variables", "Analysis", "Results", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "The World Health Organization Framework Convention on Tobacco Control (WHO-FCTC) recommends that the sale of single cigarettes be banned [1,2]. Mexico has banned singles since 1999, before it ratified the WHO-FCTC in 2004. Nevertheless, the sale of single cigarettes appears prevalent in Mexico [3-5], as in other jurisdictions that have banned them [6-10]. The rationale for banning singles primarily reflects concerns about facilitating youth access [11]. However, how adult smokers respond to the availability of single cigarettes is less well understood and merits clarification, especially as the availability and consumption of singles cigarettes could either lower or increase successful smoking cessation.\nAccess to single cigarettes may promote smoking among adult smokers. For example, because the price of a single cigarette is lower than for a cigarette pack [12-14], smokers who may otherwise quit because of affordability issues may continue to nourish their addiction, albeit while smoking fewer cigarettes. Perhaps reflecting affordability issues, younger smokers and smokers with lower incomes are more likely to smoke singles in Mexico [3]. Furthermore, when a tax increase went into effect in Mexico, consumption went down overall, while the prevalence of singles use increased from 10% to 20% [13]. The visibility of single cigarettes in the surrounding environment may also cue smoking behavior or promote relapse [15]. Indeed, Mexican smokers who experience more frequent cravings to smoke because of seeing singles cigarettes for sale are less likely to intend to quit than Mexican smokers who do not experience such cues or cravings [3]. Similarly, the visibility of singles sales both within and outside of more formal points of sale may support perceptions of the normative nature of smoking. Indeed, the frequency of brand impressions may increase with the visible presence of cigarette packages from which vendors sell singles. Such brand imagery exposures may increase in importance as other advertising channels are banned [16]. Finally, people who consume singles may be less likely to be exposed to health warning labels on cigarette packs compared to people who buy and carry with them the packages that contain these warnings.\nThe availability of singles cigarettes may also contribute to reductions in cigarette consumption among adults who smoke. In disadvantaged urban settings in the US [14], as well as in Mexico [3], adult smokers report buying single cigarettes as a method to keep consumption down and to quit. Single cigarettes cost around twice as much, per stick, as cigarettes bought in a standard package, so smokers impose upon themselves a steeper monetary cost to limit consumption. Furthermore, purchasing singles imposes greater search costs for each cigarette [17], often involving a substantial increase in the amount of time spent going to the place where single cigarettes are sold, when compared to the time spent reaching into a cigarette package that is kept at hand. In cross-sectional analyses, the frequency of using singles to limit smoking behavior was positively associated with quit intentions among Mexican smokers [3]. Longitudinal analyses are needed to clarify whether the availability of single cigarettes on balance promotes or inhibits smoking among adults.\nThis study uses longitudinal data from a cohort of smokers representative of six major Mexican cities in order to determine the relationship between consumption of single cigarettes and downstream quit behavior. First, we examine correlates of single cigarette consumption. Second, we examine whether baseline consumption and perceptions of single cigarettes are associated with quit attempts and quitting after baseline.", "[SUBTITLE] Study sample [SUBSECTION] As part of the International Tobacco Control (ITC) Policy Evaluation Survey, data have been collected on an annual basis from adult smokers in Mexico since 2006. The longitudinal analytic sample for the current study consists of data from wave 3 (November to December 2008) and wave 4 (January to February 2010) of the ITC-Mexico survey. Between previous survey administrations, taxes were increased [13], and advertising restrictions and smoke-free policies were implemented [18]; however, no major tobacco control policies were implemented between waves 3 and 4. In six cities (i.e., Mexico City, Monterrey, Guadalajara, Puebla, Tijuana, Mérida) data were collected at both of these waves. A stratified, multi-stage sampling strategy was used within the urban limits designated for each city. Within selected block groups, face-to-face interviews were conducted with randomly selected adult smokers, defined as those who has smoked at least once the previous week and at least 100 lifetime cigarettes (for details, see Thrasher et al., 2009). The wave 3 analytic sample (n = 1649) includes 73% (524/717) of those who were successfully followed up from the wave 2 sample cities (i.e., Mexico City, Guadalajara, Tijuana), of whom 117 were excluded because they had quit by wave 3. The sample also includes a replenishment sample of 203 smokers randomly selected from the same census tracts as those selected for the original sample, as well as new samples of smokers in Mérida, Monterrey and Puebla (n = 813) as well as an augmented sample in Mexico City (n = 135). Household contact and cooperation rates for wave 3 was 79% and 70%, respectively. Sampling weights were developed to account for the likelihood of participant selection. To produce more efficient estimates of association [19], the weights used for model estimation were rescaled to sum to the sample size within each city. The protocol for this study was approved by the IRB at the Mexican National Institute of Public Health. Data are not publically available, but may be requested through http://www.itcproject.org.\nAs part of the International Tobacco Control (ITC) Policy Evaluation Survey, data have been collected on an annual basis from adult smokers in Mexico since 2006. The longitudinal analytic sample for the current study consists of data from wave 3 (November to December 2008) and wave 4 (January to February 2010) of the ITC-Mexico survey. Between previous survey administrations, taxes were increased [13], and advertising restrictions and smoke-free policies were implemented [18]; however, no major tobacco control policies were implemented between waves 3 and 4. In six cities (i.e., Mexico City, Monterrey, Guadalajara, Puebla, Tijuana, Mérida) data were collected at both of these waves. A stratified, multi-stage sampling strategy was used within the urban limits designated for each city. Within selected block groups, face-to-face interviews were conducted with randomly selected adult smokers, defined as those who has smoked at least once the previous week and at least 100 lifetime cigarettes (for details, see Thrasher et al., 2009). The wave 3 analytic sample (n = 1649) includes 73% (524/717) of those who were successfully followed up from the wave 2 sample cities (i.e., Mexico City, Guadalajara, Tijuana), of whom 117 were excluded because they had quit by wave 3. The sample also includes a replenishment sample of 203 smokers randomly selected from the same census tracts as those selected for the original sample, as well as new samples of smokers in Mérida, Monterrey and Puebla (n = 813) as well as an augmented sample in Mexico City (n = 135). Household contact and cooperation rates for wave 3 was 79% and 70%, respectively. Sampling weights were developed to account for the likelihood of participant selection. To produce more efficient estimates of association [19], the weights used for model estimation were rescaled to sum to the sample size within each city. The protocol for this study was approved by the IRB at the Mexican National Institute of Public Health. Data are not publically available, but may be requested through http://www.itcproject.org.\n[SUBTITLE] Measures [SUBSECTION] [SUBTITLE] Smoking and quitting behaviour [SUBSECTION] Baseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].\nBaseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].\n[SUBTITLE] Singles use and perceptions [SUBSECTION] At baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.\nAt baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.\n[SUBTITLE] Sociodemographic variables [SUBSECTION] At baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.\nAt baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.\n[SUBTITLE] Smoking and quitting behaviour [SUBSECTION] Baseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].\nBaseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].\n[SUBTITLE] Singles use and perceptions [SUBSECTION] At baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.\nAt baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.\n[SUBTITLE] Sociodemographic variables [SUBSECTION] At baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.\nAt baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.\n[SUBTITLE] Analysis [SUBSECTION] Analyses were conducted using STATA, version 11.0. Univariate descriptions of the study sample and attrition analyses of differences between the sample that was and was not followed up were calculated without taking into account the complex survey design and sampling weights; however, all other analyses took into account the design and weights. Logistic regression was used to estimate crude and adjusted odds ratios that expressed associations between study variables and buying single cigarettes at the last purchase, as well as when determinng the odds of subsequent attempts to quit and being quit. Ordinal regression models were used to estimate the bivariate and multivariate adjusted relationships between study variables and frequency of purchasing single cigarettes. For models that regressed singles purchase behavior on study variables, the baseline (i.e., wave 3) analytic sample of smokers was assessed, whether participants were successfully followed up or not. For logistic models that regressed subsequent quit behavior (i.e., wave 4) on baseline variables (i.e., wave 3), only data were analyzed from participants who were successfully surveyed at both waves.\nAnalyses were conducted using STATA, version 11.0. Univariate descriptions of the study sample and attrition analyses of differences between the sample that was and was not followed up were calculated without taking into account the complex survey design and sampling weights; however, all other analyses took into account the design and weights. Logistic regression was used to estimate crude and adjusted odds ratios that expressed associations between study variables and buying single cigarettes at the last purchase, as well as when determinng the odds of subsequent attempts to quit and being quit. Ordinal regression models were used to estimate the bivariate and multivariate adjusted relationships between study variables and frequency of purchasing single cigarettes. For models that regressed singles purchase behavior on study variables, the baseline (i.e., wave 3) analytic sample of smokers was assessed, whether participants were successfully followed up or not. For logistic models that regressed subsequent quit behavior (i.e., wave 4) on baseline variables (i.e., wave 3), only data were analyzed from participants who were successfully surveyed at both waves.", "As part of the International Tobacco Control (ITC) Policy Evaluation Survey, data have been collected on an annual basis from adult smokers in Mexico since 2006. The longitudinal analytic sample for the current study consists of data from wave 3 (November to December 2008) and wave 4 (January to February 2010) of the ITC-Mexico survey. Between previous survey administrations, taxes were increased [13], and advertising restrictions and smoke-free policies were implemented [18]; however, no major tobacco control policies were implemented between waves 3 and 4. In six cities (i.e., Mexico City, Monterrey, Guadalajara, Puebla, Tijuana, Mérida) data were collected at both of these waves. A stratified, multi-stage sampling strategy was used within the urban limits designated for each city. Within selected block groups, face-to-face interviews were conducted with randomly selected adult smokers, defined as those who has smoked at least once the previous week and at least 100 lifetime cigarettes (for details, see Thrasher et al., 2009). The wave 3 analytic sample (n = 1649) includes 73% (524/717) of those who were successfully followed up from the wave 2 sample cities (i.e., Mexico City, Guadalajara, Tijuana), of whom 117 were excluded because they had quit by wave 3. The sample also includes a replenishment sample of 203 smokers randomly selected from the same census tracts as those selected for the original sample, as well as new samples of smokers in Mérida, Monterrey and Puebla (n = 813) as well as an augmented sample in Mexico City (n = 135). Household contact and cooperation rates for wave 3 was 79% and 70%, respectively. Sampling weights were developed to account for the likelihood of participant selection. To produce more efficient estimates of association [19], the weights used for model estimation were rescaled to sum to the sample size within each city. The protocol for this study was approved by the IRB at the Mexican National Institute of Public Health. Data are not publically available, but may be requested through http://www.itcproject.org.", "[SUBTITLE] Smoking and quitting behaviour [SUBSECTION] Baseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].\nBaseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].\n[SUBTITLE] Singles use and perceptions [SUBSECTION] At baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.\nAt baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.\n[SUBTITLE] Sociodemographic variables [SUBSECTION] At baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.\nAt baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.", "Baseline (i.e., wave 3) frequency of smoking was used to categorize respondents as nondaily, daily < 5 cigarettes a day or daily > = 5 cigarettes a day. These were dummy coded, with nondaily as the reference group. Baseline intention to quit involved classifying participants as intending to quit in the next 6 months (1) or not (0). Baseline quit behavior was assessed through self-report of attempting to quit in the previous year (1) or not (0). At followup, participants self-reported whether they had attempted to quit in the last year, that is, since wave 3. Participants were also asked at follow up if they continued to smoke or not, and if not, how long they had been quit. Those that had been quit for at least 30 days at follow-up were classified as being quit, as recommended for survey research[20]. A similar period of 4 weeks or more abstinence has been recommended for short clinical trials [21].", "At baseline (i.e., wave 3), participants were asked whether their last purchase of cigarettes was a single cigarette, a pack, or a carton of packs, with responses recoded to indicate purchasing either singles (1) or a pack or carton (0). Participants were asked four additional questions about singles, two regarding consumption and two regarding smoking cues. How often did they buy singles and how often did they buy singles to reduce the amount they smoke refer to the former. How often did they see singles being sold and how often did the feel cravings to smoke upon seeing singles being sold refer to the latter. Response options were daily; not daily, but once or more a week; once to three times a month; a few times in the last six months; not in the last 6 months. Due to relatively small sample size in some categories, the first two responses (i.e., daily; not daily, but once or more a week) and the second set of responses (i.e., monthly; a few times in the last six months) were collapsed for analyses where these variables were treated as independent variables. In bi-variate and multivariate analyses, these were dummy coded, with not having exhibited the characteristic in the previous 6 months as the reference group. Interactions among these variables and intentions to quit were assessed by multipliying the ordinal variable by intentions to quit dummy variable.", "At baseline (i.e., wave 3), respondents were asked to report their age, sex, highest educational level completed, and monthly income (reported in Mexican pesos; at the time of data collection, $1 USD ≈ $12.7 to $13.6 pesos). Education and income were reclassified to the four categories that reflected the most uniform distribution possible (i.e., less than middle school, middle school, high school or technical school, and more than high school; $0-3000, $3001 to 5000, $5001 to 8000 and $8001 or more pesos per month), and dummy variables were created with the lowest level as the reference group. For income, respondents with missing data were assigned a dummy variable, in order to avoid losing their information in multivariate analyses.", "Analyses were conducted using STATA, version 11.0. Univariate descriptions of the study sample and attrition analyses of differences between the sample that was and was not followed up were calculated without taking into account the complex survey design and sampling weights; however, all other analyses took into account the design and weights. Logistic regression was used to estimate crude and adjusted odds ratios that expressed associations between study variables and buying single cigarettes at the last purchase, as well as when determinng the odds of subsequent attempts to quit and being quit. Ordinal regression models were used to estimate the bivariate and multivariate adjusted relationships between study variables and frequency of purchasing single cigarettes. For models that regressed singles purchase behavior on study variables, the baseline (i.e., wave 3) analytic sample of smokers was assessed, whether participants were successfully followed up or not. For logistic models that regressed subsequent quit behavior (i.e., wave 4) on baseline variables (i.e., wave 3), only data were analyzed from participants who were successfully surveyed at both waves.", "Of the smokers who participated in the baseline (i.e., wave 3) survey, 72% (n = 1206/1649) were successfully followed up 14 months year later. Table 1 shows the baseline sample, including characteristics of those who were and were not followed up. Compared to those who were followed up, those who were not followed up were more likely to be male (68% vs. 61%), younger, have higher educational attainment, were less likely to have attempted to quit in the year before baseline (30% vs 35%), and were less likely to have noticed the sale of single cigarettes. Otherwise, the samples were comparable.\nSample characteristics of Mexican smokers, including those who were and were followed up over 14 months, November/December 2008 to January/February 2010*\na: p < 0.05; b: p < 0.01; c: p < 0.001 for comparing those who were and were not followed up.\n*Raw estimates, not taking into account complex sample design\nTable 2 shows the results of bi-variate and multivariate adjusted logistic regression models, where having purchased singles at last purchase is regressed on the study variables. In bivariate models, higher age and greater household income were associated with lower odds of purchasing singles. However, these associations became non-significant in multivariate models, with the exception of those with middle household income having lower odds of purchasing singles compared to those with the lowest income. Both heavier and lighter daily smokers were less likely than nondaily smokers to have purchased singles, with this association maintaining significance in multivariate models. People who intended to quit in the next six months were more likely than those who did not to have purchased singles in both bivariate (OR = 2.33, 95%CI 1.56, 3.49) and multivariate models (AOR = 2.17, 95%CI 1.45, 3.27). Those who most felt urges to smoke when seeing singles for sale were also much more likely to have purchased singles than those who did not feel these urges.\nCorrelates of purchasing singles, adult Mexican smokers\na: p < 0.05; b: p < 0.01; c: p < 0.001.\n*logistic regression models estimated to determine purchasing singles at last purchase; all variables in models assessed at baseline only.\n**ordinal regression model estimated to determine frequency of purchasing singles; all variables in models assessed at baseline only.\n***models adjust for all variables shown in the table\nOrdinal regression models regressed frequency of purchasing singles on study variables (see Table 2), with results that were generally consistent with models of buying singles at last purchase. In both bivariate and multivariate models, more frequent purchases of singles were made by younger smokers (i.e., 18 to 24 vs. 40 to 54 and vs. 55 and older), lighter smokers (i.e., nondaily vs. daily 5 or more cigarettes a day) and those who intended to quit compared to those who did not. Also, those who were prompted to smoke upon seeing singles for sale more frequently purchased singles, compared to those who did not experience such urges. Bivariate and multivariate models that regressed the frequency of buying singles to reduce consumption on study variables found the same pattern of results for most sociodemographic variables and all smoking-related variables (see Table 2).\nLogistic models were estimated to determine whether singles consumption and cueing predicted followup self-report of any quit attempt (see Table 3). Of the variables assessed, only one had a statistically significant association with subsequent quit attempts: those who purchased singles at least once a week to control their consumption had a greater likelihood trying to quit in bivariate models (OR = 1.81; 95% CI 1.31, 2.51), but not in the adjusted models (AOR = 1.56; 95% CI 0.89, 2.74). A series of additional multivariate models were estimated to assess multiplicative interactions between singles consumption variables and other key variables, while adjusting for the same study variables (results not shown). The multiplicative interaction between the dichotomous intention to quit variable and the three level variable of frequency of purchasing singles to control consumption was not statistically significant (p = 0.35). Multiplicative interactions between daily consumption and both the three level purchasing singles to control consumption and the frequency of urges to smoke were not statistically significant (p = 0.31 and 0.08, respectively).\nPredictors of trying to quit during 14 months of followup, adult Mexican smokers*\na: p < 0.05; b: p < 0.01; c: p < 0.001.\n*logistic regression models estimated; all independent variables were assessed at baseline.\n**model adjusts for all variables shown in the table, as well as baseline age, sex, education, income, smoking intensity, previous year quit attempts, and intention to quit.\nLogistic models also regressed being quit after 14 months of followup on study variables (see Table 4). In bivariate and adjusted models, frequency of buying singles to reduce consumption predicted being quit, but the association was curvilinear. The increased odds of being quit was only statistically significant when comparing those who had not bought singles to reduce consumption with those who did so at a more irregular basis (AOR = 2.30; 95% CI 1.19, 4.45), whereas those who did so more regularly were no more likely to be quit at followup. In an additional multivariate adjusted model that included an interaction between intention to quit and frequency of purchasing singles to control consumption, the interaction was not statistically significant (p = 0.36). Two additional multivariate adjusted models were run to assess interactions. One model included a multiplicative interaction between daily consumption and the three-level purchasing singles to control consumption variable. The other model included a multiplicative interaction between daily consumption and frequency of urges to smoke. In neither model was the interaction term statistically significant (p = 0.98 and 0.94, respectively).\nPredictors of being quit for 30 days or more after 14 months of followup, adult Mexican smokers*\na: p < 0.05; b: p < 0.01; c: p < 0.001.\n*logistic regression models estimated; all independent variables assessed at baseline\n**model adjusts for all variables shown in the table, as well as baseline age, sex, education, income, smoking intensity, previous year quit attempts, and intention to quit.", "The results from our study indicate that many smokers see and consume single cigarettes in Mexico, in spite of the illegality of their sale. Approximately 30% of smokers saw singles for sale on a daily basis, 18% bought singles at their last purchase, and 31% bought singles in the previous month. These estimates are generally higher than estimates from 2006 [3], although they are not as high as found in a convenience sample of young, disadvantaged adults in the US, where 77% had purchased singles in the previous month [14]. Our results are mostly consistent with the notion that single cigarette use in Mexico is most prevalent among younger smokers and those with lower income and educational achievement. However, associations between singles consumption and these characteristics were somewhat inconsistent across bivariate and multivariate models. More consistent positive associations were found between singles consumption and lighter intensity of cigarette consumption, greater intention to quit, and more frequent urges to smoke upon seeing the sale of singles. This suggests that singles availability may facilitate the early stages of smoking uptake among young people, but they may also maintain low levels of smoking or be used as a method to quit, as has been reported previously [3,14].\nSmokers who reported more frequent urges to smoke upon viewing single cigarettes for sale were more likely to purchase singles. However, these same people were no less likely to quit at followup than those who did not report these urges. This lack of association contrasts with cross-sectional research, which found that smokers who reported more frequent urges to smoke because of seeing singles for sale were less likely to intend to quit (AOR = 0.40) [3]. Although our longitudinal results suggest that cueing due to the availability and visibility of singles may not maintain smoking among Mexicans, our self-report measure may not have adequately captured the cueing phenomenon, which can operate at unconscious levels. Cues to smoke have been studied very little in natural settings or through surveys [22], and future research should assess the reliability and validity of self-report and other measures, in order to better understand how cueing works in naturalistic settings. For example, environmental scans indicating widely varying prevalence of singles availability in Mexico [5], could be linked to other data on smoking among people who inhabit these environments.\nOur longitudinal results regarding the use of singles as a method to quit were inconsistent. Smokers who frequently purchased singles to control their consumption were no more likely to attempt to quit than those who did not. Estimates of factors that predicted being quit for a month or more produced more inconsistent results, with no increased likelihood of being quit among smokers with the highest frequency of purchasing singles to control consumption; however, less frequent singles consumption was associated with a greater likelihood of being quit at followup (AORno urge vs. less frequent urges = 2.77, 95% CI 1.77, 4.53). When examining either quit attempts or quit success, interactions between intention to quit and the use of singles to reduce consumption were not statistically significant. Hence, it appears that when smokers impose the additional economic costs and search costs of consuming singles as a method to quit, the population-level effect as a harm reduction strategy is unclear. Furthermore, non-statistically significant results around the interactions between consumption intensity and singles consumption suggested that stratification of the data by consumption intensity does not clarify these relationships.\nThe current study's population-based, longitudinal nature lends strength to these conclusions. However, there were some limitations, including the need for longer studies to better understand relapse, which could be greater in environments with higher prevalence of cues to smoke due to the availability of singles. Furthermore, attrition may have biased results, particularly as the followup sample was slightly more likely to notice singles than those who were not followed up (31% vs. 27% noticing daily). Nevertheless, this difference was not substantial and there were no differences between the analytic sample and those lost to followup on the primary indicators of singles consumption and perceptions. Participation in the study may have been biased in ways that preclude generalization to the sampling frame, although the direction of this bias is not possible to ascertain because of the lack of data on nonparticipants.\nThe results from this study may not generalize to Mexican populations outside of the sampling frame. However, data were collected in the largest cities in Mexico, and 70% of the Mexican population lives in urban areas [23]. Furthermore, the prevalence of smoking is three times higher in urban areas than in rural areas [24]. Hence, the results likely generalize to the segment of the population that bears a substantial part of the tobacco-related disease burden in Mexico. Similar studies should be conducted with smokers in other settings, including in populations outside of Mexico that have heavier smoking patterns, as the light smoking pattern among Mexicans may restrict these conclusions to Mexico. Finally, a fuller treatment of the implications of singles availability for tobacco control policy development would attempt to address whether smokers who switch from packs to singles would have otherwise quit in the face of interventions, such as tax increases. This type of assessment would likely demand quasiexperimental designs which could compare smoking behavior in countries with different levels of singles availability.", "This study provides the first longitudinal assessment of the relationship between perceptions of single cigarettes, their consumption and quitting behavior. The results suggest that the public health impact of singles consumption among adult smokers in Mexico is unclear. Further studies should more squarely focus on the issue of relapse, switching to singles instead of quitting, and the translatability of this phenomenon to other contexts where single cigarettes are at different levels of prevalence. Without more compelling evidence of their potential to reduce the harms of smoking, countries should be urged to effectively implement and enforce the WHO-FCTC's recommendation to ban singles sales [2].", "The authors declare that they have no competing interests.", "JT conceptualized and wrote the majority of the manuscript. VV conducted the statistical analyses and wrote initial drafts of the methods and results. JB, RS and RO provided substantial contributions to the background and implications of the study. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2458/11/134/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, "results", "discussion", "conclusions", null, null, null ]

[]

[ "Adult", "Aged", "Cross-Sectional Studies", "Disease Progression", "Fatigue", "Female", "Health Surveys", "Hot Temperature", "Humans", "Immunosuppressive Agents", "Male", "Middle Aged", "Multiple Sclerosis", "Self Report", "Severity of Illness Index", "Thermosensing" ]

[ "Background", "Study group", "Data collection", "Statistical analyses", "Ethical considerations", "Results", "Study group", "Heat sensitivity", "Discussion", "Heat sensitivity intensifies MS symptoms", "The mechanism behind heat sensitivity", "The subjective phenomenon of heat sensitivity", "Fatigue as a side effect of immune modulating drugs", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Between 60 and 80% of individuals diagnosed with the neurological disease Multiple Sclerosis (MS) have been reported as being sensitive to environmental heat [1]. In a multinational Internet-based survey of MS patients (n = 2 529), 70% reported that high temperatures worsened their MS [2]. Clinically, increased body temperature can result in increased neurological signs and MS symptoms. Blurred vision, known as Uthoff's phenomenon and first described in 1890, is caused by increased body temperature due to physical exercise or physical restraint [3]. The body temperature is found to influence nerve impulses, which are blocked or slowed down in a damaged nerve [4-6]. After normalization of the temperature, signs and symptoms improve or disappear [1,3].\nHeat sensitivity has been described as a significant correlate of the symptom of fatigue in MS [5,7-9]. Together with divided attention and reduced muscular endurance, both heat sensitivity and fatigue have been reported as predictors of accidental falls [10]. Recently, Marino [6] stressed that heat sensitivity has been disregarded in studies focusing on fatigue.\nIn studies covering several decades, the occurrence rate of fatigue among individuals with MS is described as varying from 60 to more than 90% [5,11-13]. Irrespective of disease course, already in 1984 Freal and co-workers [5] reported fatigue as the very first symptom in about a third of patients. This initial symptom has been found to persist over the disease trajectory [14,15]. None of these studies, however, has focused on heat sensitivity.\nDuring recent decades, individuals diagnosed with MS have had increased opportunities for treatment through the development of immune-modulating medications. These products affect the frequency of relapses and slow the progression of disability. Some immune-modulating products (e.g. beta-interferon) have side effects of influenza-like symptoms and fatigue, and may also increase depressive symptoms [16]. However, in treatment with beta-interferon, the initial side effects of influenza-like symptoms and perceived fatigue often decrease when the treatment spans a longer period. Another product, glatiramer acetate, is clinically considered neutral in this respect [17]. A relatively new immune-modulating product (natalizumab) has been reported to decrease the symptom of fatigue [18].\nThe aim of this study was to investigate the occurrence of heat sensitivity and its relations to disease course, disability, common MS-related symptoms (especially fatigue), and ongoing immunosuppressive treatments among individuals 65 years of age or younger diagnosed with MS and living in an eastern county of Sweden.", "In order to reach individuals diagnosed with MS living in the area, a cross-sectional designed survey was undertaken, addressed to the individuals registered in the Swedish MS register (SwMS-reg). Together with information about the study, a questionnaire and a pre-paid reply envelope were sent to 334 individuals fulfilling the following criteria: i) being diagnosed with MS, ii) having an Expanded Disability Status Score (EDSS) [19] in the interval 0≤EDSS≤6.5, and iii) being of working age, i.e. 20-65 years, as 65 is the official retirement age. The questionnaires were distributed during 2007 to individuals with EDSS 1-6.5, and a complementary distribution was sent to individuals with EDSS = 0 in 2008. To those who had not answered within three weeks, one reminder was sent.", "The individual's age, sex, disease course, i.e. relapsing-remitting (RR), secondary progressive (SP), primary progressive (PP) and function, measured with the EDSS [19], together with date of onset of MS, were obtained from the SwMS register. The diagnostic criteria used in the SwMS register are both Poser criteria [20] (older cases) and McDonald criteria [21] (recent time). Secondary progressive course of MS (SP) is a clinical neurological status whose onset can be hard to pinpoint. In clinical trials and other studies, this status is often not defined in detail. In the SwMS register the following definition is chosen: A phase of the disease where one can secure a slowly increasing pyramidal para- or even tetraparesis, sometimes also a bladder syndrome. In some cases can it be a cerebellar syndrome. A crucial criterion is that you observe increasing pyramidal (or cerebellar) symptoms without any signs of bouts [22].\nThe patients' EDSS and disease-course are continuously upgraded at physicians call at the neurological policlinics. For the majority of the participants in this study, both the EDSS and disease-course had been determined within a year before the data collection started. Common background factors, such as civil status, family, level of education and ongoing medical treatment, were requested in the questionnaire. Occurrence of heat sensitivity was asked about in a single question, \"Are you sensitive to heat?\" (Yes/No), with follow-up questions concerning how this was experienced when in sunshine, in a warm room, or taking a hot bath or shower, and what room-temperature the patient preferred. Occurrence and perceived severity of fatigue were requested as well. Furthermore, the following instruments were included: the Fatigue Severity Scale (FSS) [7,11]; the MS-related symptom checklist [23] and the Perceived Deficit Questionnaire (PDQ) [24].\nIn this study, and based on earlier studies [12,14,25], the EDSS was grouped according to normal neurological condition (EDSS = 0), mild disability (1.0≤EDSS≤3.5), moderate disability (4.0≤EDSS≤5.5) and severe disability (6.0≤EDSS≤6.5).\nThe FSS [7,11] comprises nine items covering perceived severity of fatigue, and each item is graded from least fatigue (1) to severe fatigue (7). As in earlier studies [e.g. [9,14], in this study an FSS mean score ≤4 was regarded as indicating no fatigue, >4 but <5 borderline fatigue, and ≥5 severe fatigue. The FSS has been used in earlier studies in Sweden [14] and is also used clinically. In this study, concurrent validity was assessed through correlations between the FSS with questions about the impact of fatigue on daily life as well as the occurrence of fatigue, rated on the MS-related symptom checklist [23], and resulted in r = 0.79, (P < 0.01) and r = 0.77 (P < 0.01), respectively. Reliability was assessed using Cronbach's alpha, and the alpha coefficient was 0.93.\nThe MS-related symptom checklist [23] was used to report 25 common MS symptoms such as fatigue, weakness in arms and legs, balance problems, pain, numbness, blurred vision, depression and difficulties with urination such as frequency and urgency. The occurrence of each symptom was graded in six steps from never (1) to always (6), which were then dichotomized, into never/sometimes (1-3) to 1 and usually/always (4-6) to 2. The checklist has been found to be closely related to the EDSS [23], and has been used in an earlier study in Sweden [13].\nCognitive dysfunction was assessed using the PDQ [24], measuring perceived problems with memory, attention and concentration. The original English version of the PDQ was translated into Swedish and then back-translated in accordance with Streiner and Norman [26], with good agreement. Twenty items are graded from never (0) to always (4). Summated scores vary from 0 to 80, where higher scores indicate greater cognitive problems. In this study, concurrent validity, assessed through correlations between the PDQ sum score and the MS-related symptoms forgetfulness and concentration difficulties, was r = 0.75 (P < 0.010) and r = 0.73 (P < 0.010), respectively. Reliability was tested using Cronbach's alpha (α = 0.95) and the split-half technique (r = 0.87).", "The data have been treated and analysed in relation to perceived heat sensitivity. The descriptive statistics used are in congruence with measurement scale level, except for the FSS which is treated on interval level as in earlier studies [e.g. [9,11,14]. Differences between groups have been tested using the chi-square test on data on nominal level, the Mann Whitney U-test on data on ordinal level, and Student's t-test on data on interval level. To test associations between variables, Pearson's and Spearman's correlations were calculated. In the logistic regression analysis (enter), MS symptoms that were significantly more frequent among heat-sensitive participants were dichotomized and entered as dependent variables, and EDSS (0-6.5), disease course (relapsing-remitting/progressive forms), heat sensitivity (yes/no), time since onset, and age and sex (female/male) were independent variables. In the linear regression analysis (enter) mean FSS and summarized PDQ were entered as dependent variables and EDSS (0-6.5), disease course (relapsing/progressive forms), time since onset and heat sensitivity age and sex (female/male) were independent variables. In this study, and due to the number of statistical analyses, a P-value < 0.010 has been considered a significant value.", "The study was guided by common ethical principals in research and according to the Declaration of Helsinki [27]. Approval to use the SwMS register was received from the responsible local administrator. The Regional Ethical Review Board at the Faculty of Health Sciences, Linköping University, Sweden (Dnr M13-07) also approved the study. A completed and returned questionnaire was regarded as granted informed consent.", "[SUBTITLE] Study group [SUBSECTION] Two hundred and fifty-six patients, 195 (76%) women and 61 (24%) men, aged 65 years or younger and with an EDSS score between 0 and 6.5, answered the questionnaires. The response rate was 79.3%. Between participants and non-participants there were no statistically significant differences regarding sex (P = 0.052), disease course (P = 0.664) or EDSS score (P = 0.142). The non-participants were about five years younger than the participants, m = 42.7 years, (SD 10.5) (P < 0.001) and also had fewer years since onset, Md = 10 (Q1 = 5.75; Q3 = 15.25) (P = 0.014) (Table 1).\nCharacteristics of the participants (n = 256)\nOf the participants, the women were somewhat older (P = 0.022) than the men, 48.3 (SD 11) and 44.8 (SD 10.1), respectively. At onset the median age was 31 years (Q1 = 25; Q3 = 38). In 5.5% of the participants, onset occurred before 20 years of age, and the youngest had been 12 years old. The main course of the disease was relapsing-remitting (73%). Ten percent had an EDSS score of 0, but most of the patients (61%) had mild disease severity according to the EDSS. Two hundred and two participants (79%) were being treated pharmacologically. One hundred and forty-four participants (56%) were being treated with immune modulating medication and 32 participants (13%) with medication for fatigue (Table 1).\nTwo hundred and fifty-six patients, 195 (76%) women and 61 (24%) men, aged 65 years or younger and with an EDSS score between 0 and 6.5, answered the questionnaires. The response rate was 79.3%. Between participants and non-participants there were no statistically significant differences regarding sex (P = 0.052), disease course (P = 0.664) or EDSS score (P = 0.142). The non-participants were about five years younger than the participants, m = 42.7 years, (SD 10.5) (P < 0.001) and also had fewer years since onset, Md = 10 (Q1 = 5.75; Q3 = 15.25) (P = 0.014) (Table 1).\nCharacteristics of the participants (n = 256)\nOf the participants, the women were somewhat older (P = 0.022) than the men, 48.3 (SD 11) and 44.8 (SD 10.1), respectively. At onset the median age was 31 years (Q1 = 25; Q3 = 38). In 5.5% of the participants, onset occurred before 20 years of age, and the youngest had been 12 years old. The main course of the disease was relapsing-remitting (73%). Ten percent had an EDSS score of 0, but most of the patients (61%) had mild disease severity according to the EDSS. Two hundred and two participants (79%) were being treated pharmacologically. One hundred and forty-four participants (56%) were being treated with immune modulating medication and 32 participants (13%) with medication for fatigue (Table 1).\n[SUBTITLE] Heat sensitivity [SUBSECTION] One hundred and forty-nine participants (58%) reported heat sensitivity, with no difference between women and men (P = 0.102) or in relation to disease course (P = 0.226). In relation to EDSS (Table 1), eight of the participants (31%) with normal neurological condition, 98 (62%) with mild disability, 16 (53%) with moderate disability and 27 (63%) with a severe disability reported heat sensitivity. Of those who were sensitive to heat, 70% preferred a room temperature of less than 20 C°, while 73% of those who were not heat sensitive preferred a room temperature of more than 20°C (P < 0.001). Eleven of the heat-sensitive participants (n = 149) annotated in the questionnaire that they also were 'cold'.\nOf common MS-related symptoms, the heat-sensitive participants reported a significantly higher occurrence of several symptoms than did those who were not heat sensitive, for example fatigue, weakness in legs, concentration difficulties, pain, and urination urgency (Table 2). Weakness in arms and legs as well as balance problems correlated significantly with the EDSS, r = 0.17 (p < 0.010), r = 0.49 (p < 0.010), r = 0.51 (p < 0.010), respectively.\nSignificant dichotomized MS symptoms (never to sometimes and usually to always) in relation to reported heat sensitivity (n = 256)\n*) Missing values between two and six\nIn the logistic regression analysis, heat sensitivity significantly explained several MS symptoms, such as fatigue, concentration difficulties and pain. Two of the symptoms, spasms and balance problems, were explained by the EDSS alone (Table 3) while some were explained by more than one variable. Together with heat sensitivity and the EDSS, increasing age also explained the occurrence of weakness in arms (OR = 1.05, 95%CI = 1.02-1.08, P = 0.001) and urination frequency (OR = 1.04, 95%CI = 1.01-1.07, P = 0.007). Other interesting findings with regard to heat sensitivity, EDSS, and increasing age were weakness in legs (OR = 1.03, 95%CI = 1.0-1.06, P = 0.042) and pain (OR = 1.04, 95%CI = 1.0-1.07, P = 0.025). Urination urgency was also explained by gender, female (Or = 0.23, 95%CI = 0.07-0.7, P = 0.01).\nLogistic regression analyses of dichotomized common MS symptoms (1 = never to sometimes, 2 = usually to always) as dependent variables, and disability (EDSS 0-6.5) and heat sensitivity (1 = yes/0 = no) as independent variables\nThe difference in the occurrence of fatigue was confirmed by the FSS. Of those who were heat sensitive, 63% reported severe fatigue, FSS≥5, while among those who were not heat sensitive 45% reported mild or no fatigue, FSS≤4 (P < 0.001) (Table 4). One hundred and ten (57%) of the women (n = 192) and 21 (35%) of the men (n = 60) scored an FSS mean value ≥5, classified as fatigued (P = 0.010). About a third of the participants who were heat sensitive had concentration difficulties significantly more often (P < 0.001). This difference was confirmed by the PDQ, with Md = 29.5 (Q1 = 19; Q3 = 38) among the heat sensitive participants and Md = 21 (Q1 = 12; Q3 = 33) in the non-heat-sensitive participants (P < 0.001). In the linear regression analyses, heat sensitivity predicted both fatigue and concentration difficulties (Table 5).\nFatigue assessed in the Fatigue Severity Scale (FSS)* in relation to reported heat sensitivity (n = 252)**\n*FSS≤4 = Non-fatigue; 4 > FSS <5 = Borderline fatigue; FSS≥5 = Severe fatigue;\n**Missing value, n = 4\nLinear regression analyses of fatigue (FSS) and perceived deficit questionnaire (PDQ) as dependent variables and EDSS (0-6.5) and heat sensitivity (yes/no) as independent variables\na) R 0.421; R2 0.177; Adjusted R2 0.161\nb) R 0.239; R 0.057; Adjusted R2 0.038\nThere was no difference between heat-sensitive and non-heat sensitive participants in the use of immune-modulating drugs or treatment for fatigue. Of those being treated with the immune-modulating beta-interferon, fewer reported severe fatigue, FSS≥5.0, compared to those being treated with glatiramer acetate or natalizumab (P = 0.021). Of those who were being treated for fatigue (n = 32) a higher proportion reported severe fatigue, FSS≥5.0 (P = 0.016), compared to those not being treated for fatigue.\nOne hundred and forty-nine participants (58%) reported heat sensitivity, with no difference between women and men (P = 0.102) or in relation to disease course (P = 0.226). In relation to EDSS (Table 1), eight of the participants (31%) with normal neurological condition, 98 (62%) with mild disability, 16 (53%) with moderate disability and 27 (63%) with a severe disability reported heat sensitivity. Of those who were sensitive to heat, 70% preferred a room temperature of less than 20 C°, while 73% of those who were not heat sensitive preferred a room temperature of more than 20°C (P < 0.001). Eleven of the heat-sensitive participants (n = 149) annotated in the questionnaire that they also were 'cold'.\nOf common MS-related symptoms, the heat-sensitive participants reported a significantly higher occurrence of several symptoms than did those who were not heat sensitive, for example fatigue, weakness in legs, concentration difficulties, pain, and urination urgency (Table 2). Weakness in arms and legs as well as balance problems correlated significantly with the EDSS, r = 0.17 (p < 0.010), r = 0.49 (p < 0.010), r = 0.51 (p < 0.010), respectively.\nSignificant dichotomized MS symptoms (never to sometimes and usually to always) in relation to reported heat sensitivity (n = 256)\n*) Missing values between two and six\nIn the logistic regression analysis, heat sensitivity significantly explained several MS symptoms, such as fatigue, concentration difficulties and pain. Two of the symptoms, spasms and balance problems, were explained by the EDSS alone (Table 3) while some were explained by more than one variable. Together with heat sensitivity and the EDSS, increasing age also explained the occurrence of weakness in arms (OR = 1.05, 95%CI = 1.02-1.08, P = 0.001) and urination frequency (OR = 1.04, 95%CI = 1.01-1.07, P = 0.007). Other interesting findings with regard to heat sensitivity, EDSS, and increasing age were weakness in legs (OR = 1.03, 95%CI = 1.0-1.06, P = 0.042) and pain (OR = 1.04, 95%CI = 1.0-1.07, P = 0.025). Urination urgency was also explained by gender, female (Or = 0.23, 95%CI = 0.07-0.7, P = 0.01).\nLogistic regression analyses of dichotomized common MS symptoms (1 = never to sometimes, 2 = usually to always) as dependent variables, and disability (EDSS 0-6.5) and heat sensitivity (1 = yes/0 = no) as independent variables\nThe difference in the occurrence of fatigue was confirmed by the FSS. Of those who were heat sensitive, 63% reported severe fatigue, FSS≥5, while among those who were not heat sensitive 45% reported mild or no fatigue, FSS≤4 (P < 0.001) (Table 4). One hundred and ten (57%) of the women (n = 192) and 21 (35%) of the men (n = 60) scored an FSS mean value ≥5, classified as fatigued (P = 0.010). About a third of the participants who were heat sensitive had concentration difficulties significantly more often (P < 0.001). This difference was confirmed by the PDQ, with Md = 29.5 (Q1 = 19; Q3 = 38) among the heat sensitive participants and Md = 21 (Q1 = 12; Q3 = 33) in the non-heat-sensitive participants (P < 0.001). In the linear regression analyses, heat sensitivity predicted both fatigue and concentration difficulties (Table 5).\nFatigue assessed in the Fatigue Severity Scale (FSS)* in relation to reported heat sensitivity (n = 252)**\n*FSS≤4 = Non-fatigue; 4 > FSS <5 = Borderline fatigue; FSS≥5 = Severe fatigue;\n**Missing value, n = 4\nLinear regression analyses of fatigue (FSS) and perceived deficit questionnaire (PDQ) as dependent variables and EDSS (0-6.5) and heat sensitivity (yes/no) as independent variables\na) R 0.421; R2 0.177; Adjusted R2 0.161\nb) R 0.239; R 0.057; Adjusted R2 0.038\nThere was no difference between heat-sensitive and non-heat sensitive participants in the use of immune-modulating drugs or treatment for fatigue. Of those being treated with the immune-modulating beta-interferon, fewer reported severe fatigue, FSS≥5.0, compared to those being treated with glatiramer acetate or natalizumab (P = 0.021). Of those who were being treated for fatigue (n = 32) a higher proportion reported severe fatigue, FSS≥5.0 (P = 0.016), compared to those not being treated for fatigue.", "Two hundred and fifty-six patients, 195 (76%) women and 61 (24%) men, aged 65 years or younger and with an EDSS score between 0 and 6.5, answered the questionnaires. The response rate was 79.3%. Between participants and non-participants there were no statistically significant differences regarding sex (P = 0.052), disease course (P = 0.664) or EDSS score (P = 0.142). The non-participants were about five years younger than the participants, m = 42.7 years, (SD 10.5) (P < 0.001) and also had fewer years since onset, Md = 10 (Q1 = 5.75; Q3 = 15.25) (P = 0.014) (Table 1).\nCharacteristics of the participants (n = 256)\nOf the participants, the women were somewhat older (P = 0.022) than the men, 48.3 (SD 11) and 44.8 (SD 10.1), respectively. At onset the median age was 31 years (Q1 = 25; Q3 = 38). In 5.5% of the participants, onset occurred before 20 years of age, and the youngest had been 12 years old. The main course of the disease was relapsing-remitting (73%). Ten percent had an EDSS score of 0, but most of the patients (61%) had mild disease severity according to the EDSS. Two hundred and two participants (79%) were being treated pharmacologically. One hundred and forty-four participants (56%) were being treated with immune modulating medication and 32 participants (13%) with medication for fatigue (Table 1).", "One hundred and forty-nine participants (58%) reported heat sensitivity, with no difference between women and men (P = 0.102) or in relation to disease course (P = 0.226). In relation to EDSS (Table 1), eight of the participants (31%) with normal neurological condition, 98 (62%) with mild disability, 16 (53%) with moderate disability and 27 (63%) with a severe disability reported heat sensitivity. Of those who were sensitive to heat, 70% preferred a room temperature of less than 20 C°, while 73% of those who were not heat sensitive preferred a room temperature of more than 20°C (P < 0.001). Eleven of the heat-sensitive participants (n = 149) annotated in the questionnaire that they also were 'cold'.\nOf common MS-related symptoms, the heat-sensitive participants reported a significantly higher occurrence of several symptoms than did those who were not heat sensitive, for example fatigue, weakness in legs, concentration difficulties, pain, and urination urgency (Table 2). Weakness in arms and legs as well as balance problems correlated significantly with the EDSS, r = 0.17 (p < 0.010), r = 0.49 (p < 0.010), r = 0.51 (p < 0.010), respectively.\nSignificant dichotomized MS symptoms (never to sometimes and usually to always) in relation to reported heat sensitivity (n = 256)\n*) Missing values between two and six\nIn the logistic regression analysis, heat sensitivity significantly explained several MS symptoms, such as fatigue, concentration difficulties and pain. Two of the symptoms, spasms and balance problems, were explained by the EDSS alone (Table 3) while some were explained by more than one variable. Together with heat sensitivity and the EDSS, increasing age also explained the occurrence of weakness in arms (OR = 1.05, 95%CI = 1.02-1.08, P = 0.001) and urination frequency (OR = 1.04, 95%CI = 1.01-1.07, P = 0.007). Other interesting findings with regard to heat sensitivity, EDSS, and increasing age were weakness in legs (OR = 1.03, 95%CI = 1.0-1.06, P = 0.042) and pain (OR = 1.04, 95%CI = 1.0-1.07, P = 0.025). Urination urgency was also explained by gender, female (Or = 0.23, 95%CI = 0.07-0.7, P = 0.01).\nLogistic regression analyses of dichotomized common MS symptoms (1 = never to sometimes, 2 = usually to always) as dependent variables, and disability (EDSS 0-6.5) and heat sensitivity (1 = yes/0 = no) as independent variables\nThe difference in the occurrence of fatigue was confirmed by the FSS. Of those who were heat sensitive, 63% reported severe fatigue, FSS≥5, while among those who were not heat sensitive 45% reported mild or no fatigue, FSS≤4 (P < 0.001) (Table 4). One hundred and ten (57%) of the women (n = 192) and 21 (35%) of the men (n = 60) scored an FSS mean value ≥5, classified as fatigued (P = 0.010). About a third of the participants who were heat sensitive had concentration difficulties significantly more often (P < 0.001). This difference was confirmed by the PDQ, with Md = 29.5 (Q1 = 19; Q3 = 38) among the heat sensitive participants and Md = 21 (Q1 = 12; Q3 = 33) in the non-heat-sensitive participants (P < 0.001). In the linear regression analyses, heat sensitivity predicted both fatigue and concentration difficulties (Table 5).\nFatigue assessed in the Fatigue Severity Scale (FSS)* in relation to reported heat sensitivity (n = 252)**\n*FSS≤4 = Non-fatigue; 4 > FSS <5 = Borderline fatigue; FSS≥5 = Severe fatigue;\n**Missing value, n = 4\nLinear regression analyses of fatigue (FSS) and perceived deficit questionnaire (PDQ) as dependent variables and EDSS (0-6.5) and heat sensitivity (yes/no) as independent variables\na) R 0.421; R2 0.177; Adjusted R2 0.161\nb) R 0.239; R 0.057; Adjusted R2 0.038\nThere was no difference between heat-sensitive and non-heat sensitive participants in the use of immune-modulating drugs or treatment for fatigue. Of those being treated with the immune-modulating beta-interferon, fewer reported severe fatigue, FSS≥5.0, compared to those being treated with glatiramer acetate or natalizumab (P = 0.021). Of those who were being treated for fatigue (n = 32) a higher proportion reported severe fatigue, FSS≥5.0 (P = 0.016), compared to those not being treated for fatigue.", "This is an explorative study aimed at investigating the occurrence of heat sensitivity in a group of individuals diagnosed with MS. Further, it aimed to investigate relations between heat sensitivity and common symptoms in MS, disease course and disability.\nThe participants were recruited from the Swedish MS register in a local county in eastern Sweden. In the chosen EDSS interval, a greater proportion of the patients had an RR course of the disease and were receiving immune-modulating treatment, with regular follow-up visits at the neurological policlinics once a year. Furthermore, some patients with secondary progressive MS and no treatment are seen by general practitioners. Thus, in this study, there is a selection of patients towards an RR course of the disease and who are treated patients. The EDSS scores used were assessed within a year before the data collection and, thus, there might be some random variation in the group. However, if any changes occurred this should be noted in the registry.\nOur results indicate that heat sensitivity is highly correlated with several symptoms commonly reported by MS patients. That heat sensitivity is associated with fatigue is well known and has been reported by several authors [4,6-8]. However, our data also reveal that heat sensitivity is a key factor associated with the occurrence of a wide variety of other MS symptoms, for example pain, concentration problems and urination urgency.\n[SUBTITLE] Heat sensitivity intensifies MS symptoms [SUBSECTION] The most striking result in this study is that heat sensitivity significantly correlated with - and in the logistic regression analyses, appeared as an explaining factor for - the most incapacitating symptoms of MS, viz. fatigue, concentration problems and pain. This result discloses heat sensitivity as a key clinical factor. Our findings are consistent with a Norwegian study by Nortvedt and co-workers [28] stating that bodily pain and low vitality are important problems of MS with a significant impact on quality of life. Low vitality [28] can be interpreted as strongly associated to the symptom of fatigue.\nWe found that many other MS symptoms are also correlated with heat sensitivity. Interestingly, similar observations were actually reported in the early works of Uthoff, although subsequent citations narrowed the interpretation of Uthoff's phenomenon to blurring of the vision [6]. In the present study, however, no relationship between the symptom of blurred vision and heat sensitivity was confirmed.\nWhat might be the cause of the observed co-occurrence of heat sensitivity and pain, fatigue and cognitive problems? Can this indicate something about the pathogenic mechanism? Based on studies of post-stroke patients, Craig [29] suggests that central pain might be caused by a thermoregulatory dysfunction. A lesion in the thalamus releases or disinhibits the feeling of burning pain by removing the normal inhibition of pain by cold. The role of the thalamus is as ascending relay of spinothalamic activity to the cortex [29]. This is similar to Österberg's [30] proposal that central pain in MS seems to be generated through lesions affecting the spinothalamo-cortical pathways [30]. A third of the MS patients with central pain had visible lesions in the thalamus [31]. Thus, one can speculate that the thalamus might be involved.\nIt is also interesting to think about the effects of decreased sweating in terms of electrolytic imbalance and secondary neuronal effects. Recent results by Saari and co-workers [32] disclose such an impairment to thermoregulatory sweating in MS. One idea that may explain thermoregulatory dysfunction is that lesions can affect important cerebral areas such as the hypothalamus. The study by Saari [32] also demonstrated a correlation between increasing sweating impairment and increasing disability (EDSS), which is congruent with our results showing a correlation between heat sensitivity and increasing EDSS.\nFatigue in MS has been eagerly studied by many researchers during the past decade. Fatigue and heat sensitivity are related in many aspects, and a subsequent question is whether they have common pathogenic features. Recently, Marino [6] stated that MS fatigue is likely to be a central rather than a peripheral phenomenon. Heat sensitivity in MS, however, is not understood as clearly.\nThe most striking result in this study is that heat sensitivity significantly correlated with - and in the logistic regression analyses, appeared as an explaining factor for - the most incapacitating symptoms of MS, viz. fatigue, concentration problems and pain. This result discloses heat sensitivity as a key clinical factor. Our findings are consistent with a Norwegian study by Nortvedt and co-workers [28] stating that bodily pain and low vitality are important problems of MS with a significant impact on quality of life. Low vitality [28] can be interpreted as strongly associated to the symptom of fatigue.\nWe found that many other MS symptoms are also correlated with heat sensitivity. Interestingly, similar observations were actually reported in the early works of Uthoff, although subsequent citations narrowed the interpretation of Uthoff's phenomenon to blurring of the vision [6]. In the present study, however, no relationship between the symptom of blurred vision and heat sensitivity was confirmed.\nWhat might be the cause of the observed co-occurrence of heat sensitivity and pain, fatigue and cognitive problems? Can this indicate something about the pathogenic mechanism? Based on studies of post-stroke patients, Craig [29] suggests that central pain might be caused by a thermoregulatory dysfunction. A lesion in the thalamus releases or disinhibits the feeling of burning pain by removing the normal inhibition of pain by cold. The role of the thalamus is as ascending relay of spinothalamic activity to the cortex [29]. This is similar to Österberg's [30] proposal that central pain in MS seems to be generated through lesions affecting the spinothalamo-cortical pathways [30]. A third of the MS patients with central pain had visible lesions in the thalamus [31]. Thus, one can speculate that the thalamus might be involved.\nIt is also interesting to think about the effects of decreased sweating in terms of electrolytic imbalance and secondary neuronal effects. Recent results by Saari and co-workers [32] disclose such an impairment to thermoregulatory sweating in MS. One idea that may explain thermoregulatory dysfunction is that lesions can affect important cerebral areas such as the hypothalamus. The study by Saari [32] also demonstrated a correlation between increasing sweating impairment and increasing disability (EDSS), which is congruent with our results showing a correlation between heat sensitivity and increasing EDSS.\nFatigue in MS has been eagerly studied by many researchers during the past decade. Fatigue and heat sensitivity are related in many aspects, and a subsequent question is whether they have common pathogenic features. Recently, Marino [6] stated that MS fatigue is likely to be a central rather than a peripheral phenomenon. Heat sensitivity in MS, however, is not understood as clearly.\n[SUBTITLE] The mechanism behind heat sensitivity [SUBSECTION] The mechanism of heat sensitivity in MS is reviewed in a recent article by Marino [6]. The heat reaction blocks the action potential of the demyelinated neuron (frequency-dependent conduction block, - FDCB) [3]. The demyelinization results in a slower nerve conduction velocity. A conduction block can occur because the damaged axons transmit only single or low frequency impulses, instead of high frequency impulse trains like in healthy nerve tissue [33]. Marino [6] states that this observation is very important, especially when an increase in temperature blocks nerve impulses in demyelinated fibres [34]. Interestingly, very small increases in temperature can also block action potentials [6]. The heat sensitivity in MS is described by Baker as secondary to both environmental heat and environmental humidity, as well as to exercise [35]. Both passive and active body temperature increases give heat reactions. Clinical reports from individual patients in our MS clinic (i.e. outside this study) reveal that their experience of temperature aberrations can vary greatly, indicating that the mechanisms may be multiple.\nThe mechanism of heat sensitivity in MS is reviewed in a recent article by Marino [6]. The heat reaction blocks the action potential of the demyelinated neuron (frequency-dependent conduction block, - FDCB) [3]. The demyelinization results in a slower nerve conduction velocity. A conduction block can occur because the damaged axons transmit only single or low frequency impulses, instead of high frequency impulse trains like in healthy nerve tissue [33]. Marino [6] states that this observation is very important, especially when an increase in temperature blocks nerve impulses in demyelinated fibres [34]. Interestingly, very small increases in temperature can also block action potentials [6]. The heat sensitivity in MS is described by Baker as secondary to both environmental heat and environmental humidity, as well as to exercise [35]. Both passive and active body temperature increases give heat reactions. Clinical reports from individual patients in our MS clinic (i.e. outside this study) reveal that their experience of temperature aberrations can vary greatly, indicating that the mechanisms may be multiple.\n[SUBTITLE] The subjective phenomenon of heat sensitivity [SUBSECTION] Many patients with MS are aware of their heat sensitivity, and have experienced increased clinical symptoms when becoming warm from fever, hot environmental temperature or physical exercise. In this study, patients who did or did not report heat sensitivity were dichotomized and analysed according to this categorization. It is worth noting that some patients reported both heat sensitivity and a subjective feeling of \"cold\". This suggests that temperature aberrations in MS can be complex and individual. In our clinical practice we also encounter MS patients with complex subjective temperature aberrations, e.g. patients who deny heat sensitivity when asked, yet prefer to sleep in cold bedrooms because it alleviates their MS symptoms. Such patients may not yet be aware of a mild heat sensitivity. This factor also suggests that milder forms of heat sensitivity may be underreported in a study like ours. We have also met one patient who likes to be in the sun all day without any side effects, but after jogging has to take repeated showers to prevent incapacitating fatigue. Finally, it is well known that patients with severe MS can sometimes develop hypothermia, often without any major subjective symptoms. We suggest that qualitative studies should be initiated to describe patients' experiences of subjective phenomena like heat sensitivity, and feeling cold as well as combinations of the two. One cannot exclude that central and peripheral mechanisms interact in the \"temperature syndrome\" of MS, which includes more features than classical heat sensitivity.\nMany patients with MS are aware of their heat sensitivity, and have experienced increased clinical symptoms when becoming warm from fever, hot environmental temperature or physical exercise. In this study, patients who did or did not report heat sensitivity were dichotomized and analysed according to this categorization. It is worth noting that some patients reported both heat sensitivity and a subjective feeling of \"cold\". This suggests that temperature aberrations in MS can be complex and individual. In our clinical practice we also encounter MS patients with complex subjective temperature aberrations, e.g. patients who deny heat sensitivity when asked, yet prefer to sleep in cold bedrooms because it alleviates their MS symptoms. Such patients may not yet be aware of a mild heat sensitivity. This factor also suggests that milder forms of heat sensitivity may be underreported in a study like ours. We have also met one patient who likes to be in the sun all day without any side effects, but after jogging has to take repeated showers to prevent incapacitating fatigue. Finally, it is well known that patients with severe MS can sometimes develop hypothermia, often without any major subjective symptoms. We suggest that qualitative studies should be initiated to describe patients' experiences of subjective phenomena like heat sensitivity, and feeling cold as well as combinations of the two. One cannot exclude that central and peripheral mechanisms interact in the \"temperature syndrome\" of MS, which includes more features than classical heat sensitivity.\n[SUBTITLE] Fatigue as a side effect of immune modulating drugs [SUBSECTION] It has recently been suggested that beta-interferon can increase fatigue through influenza-like side effects including hyperthermia, but that glatiramer-acetate, on the other hand, is a neutral agent in this respect, and that natalizumab at least in some cases decreases fatigue. Some scientific reports support these observations [16-18]. In our study, however, none of these relationships could be demonstrated. However, as this study has a descriptive design, no conclusion regarding what impact immune modulating treatment has on fatigue could be drawn.\nIt has recently been suggested that beta-interferon can increase fatigue through influenza-like side effects including hyperthermia, but that glatiramer-acetate, on the other hand, is a neutral agent in this respect, and that natalizumab at least in some cases decreases fatigue. Some scientific reports support these observations [16-18]. In our study, however, none of these relationships could be demonstrated. However, as this study has a descriptive design, no conclusion regarding what impact immune modulating treatment has on fatigue could be drawn.", "The most striking result in this study is that heat sensitivity significantly correlated with - and in the logistic regression analyses, appeared as an explaining factor for - the most incapacitating symptoms of MS, viz. fatigue, concentration problems and pain. This result discloses heat sensitivity as a key clinical factor. Our findings are consistent with a Norwegian study by Nortvedt and co-workers [28] stating that bodily pain and low vitality are important problems of MS with a significant impact on quality of life. Low vitality [28] can be interpreted as strongly associated to the symptom of fatigue.\nWe found that many other MS symptoms are also correlated with heat sensitivity. Interestingly, similar observations were actually reported in the early works of Uthoff, although subsequent citations narrowed the interpretation of Uthoff's phenomenon to blurring of the vision [6]. In the present study, however, no relationship between the symptom of blurred vision and heat sensitivity was confirmed.\nWhat might be the cause of the observed co-occurrence of heat sensitivity and pain, fatigue and cognitive problems? Can this indicate something about the pathogenic mechanism? Based on studies of post-stroke patients, Craig [29] suggests that central pain might be caused by a thermoregulatory dysfunction. A lesion in the thalamus releases or disinhibits the feeling of burning pain by removing the normal inhibition of pain by cold. The role of the thalamus is as ascending relay of spinothalamic activity to the cortex [29]. This is similar to Österberg's [30] proposal that central pain in MS seems to be generated through lesions affecting the spinothalamo-cortical pathways [30]. A third of the MS patients with central pain had visible lesions in the thalamus [31]. Thus, one can speculate that the thalamus might be involved.\nIt is also interesting to think about the effects of decreased sweating in terms of electrolytic imbalance and secondary neuronal effects. Recent results by Saari and co-workers [32] disclose such an impairment to thermoregulatory sweating in MS. One idea that may explain thermoregulatory dysfunction is that lesions can affect important cerebral areas such as the hypothalamus. The study by Saari [32] also demonstrated a correlation between increasing sweating impairment and increasing disability (EDSS), which is congruent with our results showing a correlation between heat sensitivity and increasing EDSS.\nFatigue in MS has been eagerly studied by many researchers during the past decade. Fatigue and heat sensitivity are related in many aspects, and a subsequent question is whether they have common pathogenic features. Recently, Marino [6] stated that MS fatigue is likely to be a central rather than a peripheral phenomenon. Heat sensitivity in MS, however, is not understood as clearly.", "The mechanism of heat sensitivity in MS is reviewed in a recent article by Marino [6]. The heat reaction blocks the action potential of the demyelinated neuron (frequency-dependent conduction block, - FDCB) [3]. The demyelinization results in a slower nerve conduction velocity. A conduction block can occur because the damaged axons transmit only single or low frequency impulses, instead of high frequency impulse trains like in healthy nerve tissue [33]. Marino [6] states that this observation is very important, especially when an increase in temperature blocks nerve impulses in demyelinated fibres [34]. Interestingly, very small increases in temperature can also block action potentials [6]. The heat sensitivity in MS is described by Baker as secondary to both environmental heat and environmental humidity, as well as to exercise [35]. Both passive and active body temperature increases give heat reactions. Clinical reports from individual patients in our MS clinic (i.e. outside this study) reveal that their experience of temperature aberrations can vary greatly, indicating that the mechanisms may be multiple.", "Many patients with MS are aware of their heat sensitivity, and have experienced increased clinical symptoms when becoming warm from fever, hot environmental temperature or physical exercise. In this study, patients who did or did not report heat sensitivity were dichotomized and analysed according to this categorization. It is worth noting that some patients reported both heat sensitivity and a subjective feeling of \"cold\". This suggests that temperature aberrations in MS can be complex and individual. In our clinical practice we also encounter MS patients with complex subjective temperature aberrations, e.g. patients who deny heat sensitivity when asked, yet prefer to sleep in cold bedrooms because it alleviates their MS symptoms. Such patients may not yet be aware of a mild heat sensitivity. This factor also suggests that milder forms of heat sensitivity may be underreported in a study like ours. We have also met one patient who likes to be in the sun all day without any side effects, but after jogging has to take repeated showers to prevent incapacitating fatigue. Finally, it is well known that patients with severe MS can sometimes develop hypothermia, often without any major subjective symptoms. We suggest that qualitative studies should be initiated to describe patients' experiences of subjective phenomena like heat sensitivity, and feeling cold as well as combinations of the two. One cannot exclude that central and peripheral mechanisms interact in the \"temperature syndrome\" of MS, which includes more features than classical heat sensitivity.", "It has recently been suggested that beta-interferon can increase fatigue through influenza-like side effects including hyperthermia, but that glatiramer-acetate, on the other hand, is a neutral agent in this respect, and that natalizumab at least in some cases decreases fatigue. Some scientific reports support these observations [16-18]. In our study, however, none of these relationships could be demonstrated. However, as this study has a descriptive design, no conclusion regarding what impact immune modulating treatment has on fatigue could be drawn.", "In conclusion, although heat sensitivity in MS was described as early as the late 19th century and is a well-known phenomenon today, it has to date been disregarded in studies of fatigue in MS [6]. The findings in this study underline the importance of heat sensitivity in MS patients as a key symptom that is highly correlated with disabling symptoms such as fatigue, concentration difficulty and urination urgency. A majority of the participants rated the symptom of fatigue as their most impairing symptom. Furthermore, a significantly higher proportion among the heat-sensitive participants rated higher levels of fatigue compared to the participants who were not heat sensitive.\nThe results of our study put heat sensitivity in the position of a key clinical symptom. Our findings emphasize the need to further investigate the mechanism of heat sensitivity: What is the role of the sweating impairment? Is there thalamic involvement? What is the role of the immune system? One should also analyse what it means for patients and in the care of MS patients. Finally, a challenging topic to investigate is how heat sensitivity can be treated clinically, for example using thermotherapy.", "The authors declare that they have no competing interests.", "All authors contributed to the planning of the study. GF collected all data and, together with A-CE, conducted the statistical analyses. In collaboration, GF, A-CE and A-ML wrote the drafts, which were repeatedly read, discussed and revised by all the authors. Finally, all authors read and approved the final manuscript before its submission.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2377/11/27/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Study group", "Data collection", "Statistical analyses", "Ethical considerations", "Results", "Study group", "Heat sensitivity", "Discussion", "Heat sensitivity intensifies MS symptoms", "The mechanism behind heat sensitivity", "The subjective phenomenon of heat sensitivity", "Fatigue as a side effect of immune modulating drugs", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Between 60 and 80% of individuals diagnosed with the neurological disease Multiple Sclerosis (MS) have been reported as being sensitive to environmental heat [1]. In a multinational Internet-based survey of MS patients (n = 2 529), 70% reported that high temperatures worsened their MS [2]. Clinically, increased body temperature can result in increased neurological signs and MS symptoms. Blurred vision, known as Uthoff's phenomenon and first described in 1890, is caused by increased body temperature due to physical exercise or physical restraint [3]. The body temperature is found to influence nerve impulses, which are blocked or slowed down in a damaged nerve [4-6]. After normalization of the temperature, signs and symptoms improve or disappear [1,3].\nHeat sensitivity has been described as a significant correlate of the symptom of fatigue in MS [5,7-9]. Together with divided attention and reduced muscular endurance, both heat sensitivity and fatigue have been reported as predictors of accidental falls [10]. Recently, Marino [6] stressed that heat sensitivity has been disregarded in studies focusing on fatigue.\nIn studies covering several decades, the occurrence rate of fatigue among individuals with MS is described as varying from 60 to more than 90% [5,11-13]. Irrespective of disease course, already in 1984 Freal and co-workers [5] reported fatigue as the very first symptom in about a third of patients. This initial symptom has been found to persist over the disease trajectory [14,15]. None of these studies, however, has focused on heat sensitivity.\nDuring recent decades, individuals diagnosed with MS have had increased opportunities for treatment through the development of immune-modulating medications. These products affect the frequency of relapses and slow the progression of disability. Some immune-modulating products (e.g. beta-interferon) have side effects of influenza-like symptoms and fatigue, and may also increase depressive symptoms [16]. However, in treatment with beta-interferon, the initial side effects of influenza-like symptoms and perceived fatigue often decrease when the treatment spans a longer period. Another product, glatiramer acetate, is clinically considered neutral in this respect [17]. A relatively new immune-modulating product (natalizumab) has been reported to decrease the symptom of fatigue [18].\nThe aim of this study was to investigate the occurrence of heat sensitivity and its relations to disease course, disability, common MS-related symptoms (especially fatigue), and ongoing immunosuppressive treatments among individuals 65 years of age or younger diagnosed with MS and living in an eastern county of Sweden.", "[SUBTITLE] Study group [SUBSECTION] In order to reach individuals diagnosed with MS living in the area, a cross-sectional designed survey was undertaken, addressed to the individuals registered in the Swedish MS register (SwMS-reg). Together with information about the study, a questionnaire and a pre-paid reply envelope were sent to 334 individuals fulfilling the following criteria: i) being diagnosed with MS, ii) having an Expanded Disability Status Score (EDSS) [19] in the interval 0≤EDSS≤6.5, and iii) being of working age, i.e. 20-65 years, as 65 is the official retirement age. The questionnaires were distributed during 2007 to individuals with EDSS 1-6.5, and a complementary distribution was sent to individuals with EDSS = 0 in 2008. To those who had not answered within three weeks, one reminder was sent.\nIn order to reach individuals diagnosed with MS living in the area, a cross-sectional designed survey was undertaken, addressed to the individuals registered in the Swedish MS register (SwMS-reg). Together with information about the study, a questionnaire and a pre-paid reply envelope were sent to 334 individuals fulfilling the following criteria: i) being diagnosed with MS, ii) having an Expanded Disability Status Score (EDSS) [19] in the interval 0≤EDSS≤6.5, and iii) being of working age, i.e. 20-65 years, as 65 is the official retirement age. The questionnaires were distributed during 2007 to individuals with EDSS 1-6.5, and a complementary distribution was sent to individuals with EDSS = 0 in 2008. To those who had not answered within three weeks, one reminder was sent.\n[SUBTITLE] Data collection [SUBSECTION] The individual's age, sex, disease course, i.e. relapsing-remitting (RR), secondary progressive (SP), primary progressive (PP) and function, measured with the EDSS [19], together with date of onset of MS, were obtained from the SwMS register. The diagnostic criteria used in the SwMS register are both Poser criteria [20] (older cases) and McDonald criteria [21] (recent time). Secondary progressive course of MS (SP) is a clinical neurological status whose onset can be hard to pinpoint. In clinical trials and other studies, this status is often not defined in detail. In the SwMS register the following definition is chosen: A phase of the disease where one can secure a slowly increasing pyramidal para- or even tetraparesis, sometimes also a bladder syndrome. In some cases can it be a cerebellar syndrome. A crucial criterion is that you observe increasing pyramidal (or cerebellar) symptoms without any signs of bouts [22].\nThe patients' EDSS and disease-course are continuously upgraded at physicians call at the neurological policlinics. For the majority of the participants in this study, both the EDSS and disease-course had been determined within a year before the data collection started. Common background factors, such as civil status, family, level of education and ongoing medical treatment, were requested in the questionnaire. Occurrence of heat sensitivity was asked about in a single question, \"Are you sensitive to heat?\" (Yes/No), with follow-up questions concerning how this was experienced when in sunshine, in a warm room, or taking a hot bath or shower, and what room-temperature the patient preferred. Occurrence and perceived severity of fatigue were requested as well. Furthermore, the following instruments were included: the Fatigue Severity Scale (FSS) [7,11]; the MS-related symptom checklist [23] and the Perceived Deficit Questionnaire (PDQ) [24].\nIn this study, and based on earlier studies [12,14,25], the EDSS was grouped according to normal neurological condition (EDSS = 0), mild disability (1.0≤EDSS≤3.5), moderate disability (4.0≤EDSS≤5.5) and severe disability (6.0≤EDSS≤6.5).\nThe FSS [7,11] comprises nine items covering perceived severity of fatigue, and each item is graded from least fatigue (1) to severe fatigue (7). As in earlier studies [e.g. [9,14], in this study an FSS mean score ≤4 was regarded as indicating no fatigue, >4 but <5 borderline fatigue, and ≥5 severe fatigue. The FSS has been used in earlier studies in Sweden [14] and is also used clinically. In this study, concurrent validity was assessed through correlations between the FSS with questions about the impact of fatigue on daily life as well as the occurrence of fatigue, rated on the MS-related symptom checklist [23], and resulted in r = 0.79, (P < 0.01) and r = 0.77 (P < 0.01), respectively. Reliability was assessed using Cronbach's alpha, and the alpha coefficient was 0.93.\nThe MS-related symptom checklist [23] was used to report 25 common MS symptoms such as fatigue, weakness in arms and legs, balance problems, pain, numbness, blurred vision, depression and difficulties with urination such as frequency and urgency. The occurrence of each symptom was graded in six steps from never (1) to always (6), which were then dichotomized, into never/sometimes (1-3) to 1 and usually/always (4-6) to 2. The checklist has been found to be closely related to the EDSS [23], and has been used in an earlier study in Sweden [13].\nCognitive dysfunction was assessed using the PDQ [24], measuring perceived problems with memory, attention and concentration. The original English version of the PDQ was translated into Swedish and then back-translated in accordance with Streiner and Norman [26], with good agreement. Twenty items are graded from never (0) to always (4). Summated scores vary from 0 to 80, where higher scores indicate greater cognitive problems. In this study, concurrent validity, assessed through correlations between the PDQ sum score and the MS-related symptoms forgetfulness and concentration difficulties, was r = 0.75 (P < 0.010) and r = 0.73 (P < 0.010), respectively. Reliability was tested using Cronbach's alpha (α = 0.95) and the split-half technique (r = 0.87).\nThe individual's age, sex, disease course, i.e. relapsing-remitting (RR), secondary progressive (SP), primary progressive (PP) and function, measured with the EDSS [19], together with date of onset of MS, were obtained from the SwMS register. The diagnostic criteria used in the SwMS register are both Poser criteria [20] (older cases) and McDonald criteria [21] (recent time). Secondary progressive course of MS (SP) is a clinical neurological status whose onset can be hard to pinpoint. In clinical trials and other studies, this status is often not defined in detail. In the SwMS register the following definition is chosen: A phase of the disease where one can secure a slowly increasing pyramidal para- or even tetraparesis, sometimes also a bladder syndrome. In some cases can it be a cerebellar syndrome. A crucial criterion is that you observe increasing pyramidal (or cerebellar) symptoms without any signs of bouts [22].\nThe patients' EDSS and disease-course are continuously upgraded at physicians call at the neurological policlinics. For the majority of the participants in this study, both the EDSS and disease-course had been determined within a year before the data collection started. Common background factors, such as civil status, family, level of education and ongoing medical treatment, were requested in the questionnaire. Occurrence of heat sensitivity was asked about in a single question, \"Are you sensitive to heat?\" (Yes/No), with follow-up questions concerning how this was experienced when in sunshine, in a warm room, or taking a hot bath or shower, and what room-temperature the patient preferred. Occurrence and perceived severity of fatigue were requested as well. Furthermore, the following instruments were included: the Fatigue Severity Scale (FSS) [7,11]; the MS-related symptom checklist [23] and the Perceived Deficit Questionnaire (PDQ) [24].\nIn this study, and based on earlier studies [12,14,25], the EDSS was grouped according to normal neurological condition (EDSS = 0), mild disability (1.0≤EDSS≤3.5), moderate disability (4.0≤EDSS≤5.5) and severe disability (6.0≤EDSS≤6.5).\nThe FSS [7,11] comprises nine items covering perceived severity of fatigue, and each item is graded from least fatigue (1) to severe fatigue (7). As in earlier studies [e.g. [9,14], in this study an FSS mean score ≤4 was regarded as indicating no fatigue, >4 but <5 borderline fatigue, and ≥5 severe fatigue. The FSS has been used in earlier studies in Sweden [14] and is also used clinically. In this study, concurrent validity was assessed through correlations between the FSS with questions about the impact of fatigue on daily life as well as the occurrence of fatigue, rated on the MS-related symptom checklist [23], and resulted in r = 0.79, (P < 0.01) and r = 0.77 (P < 0.01), respectively. Reliability was assessed using Cronbach's alpha, and the alpha coefficient was 0.93.\nThe MS-related symptom checklist [23] was used to report 25 common MS symptoms such as fatigue, weakness in arms and legs, balance problems, pain, numbness, blurred vision, depression and difficulties with urination such as frequency and urgency. The occurrence of each symptom was graded in six steps from never (1) to always (6), which were then dichotomized, into never/sometimes (1-3) to 1 and usually/always (4-6) to 2. The checklist has been found to be closely related to the EDSS [23], and has been used in an earlier study in Sweden [13].\nCognitive dysfunction was assessed using the PDQ [24], measuring perceived problems with memory, attention and concentration. The original English version of the PDQ was translated into Swedish and then back-translated in accordance with Streiner and Norman [26], with good agreement. Twenty items are graded from never (0) to always (4). Summated scores vary from 0 to 80, where higher scores indicate greater cognitive problems. In this study, concurrent validity, assessed through correlations between the PDQ sum score and the MS-related symptoms forgetfulness and concentration difficulties, was r = 0.75 (P < 0.010) and r = 0.73 (P < 0.010), respectively. Reliability was tested using Cronbach's alpha (α = 0.95) and the split-half technique (r = 0.87).\n[SUBTITLE] Statistical analyses [SUBSECTION] The data have been treated and analysed in relation to perceived heat sensitivity. The descriptive statistics used are in congruence with measurement scale level, except for the FSS which is treated on interval level as in earlier studies [e.g. [9,11,14]. Differences between groups have been tested using the chi-square test on data on nominal level, the Mann Whitney U-test on data on ordinal level, and Student's t-test on data on interval level. To test associations between variables, Pearson's and Spearman's correlations were calculated. In the logistic regression analysis (enter), MS symptoms that were significantly more frequent among heat-sensitive participants were dichotomized and entered as dependent variables, and EDSS (0-6.5), disease course (relapsing-remitting/progressive forms), heat sensitivity (yes/no), time since onset, and age and sex (female/male) were independent variables. In the linear regression analysis (enter) mean FSS and summarized PDQ were entered as dependent variables and EDSS (0-6.5), disease course (relapsing/progressive forms), time since onset and heat sensitivity age and sex (female/male) were independent variables. In this study, and due to the number of statistical analyses, a P-value < 0.010 has been considered a significant value.\nThe data have been treated and analysed in relation to perceived heat sensitivity. The descriptive statistics used are in congruence with measurement scale level, except for the FSS which is treated on interval level as in earlier studies [e.g. [9,11,14]. Differences between groups have been tested using the chi-square test on data on nominal level, the Mann Whitney U-test on data on ordinal level, and Student's t-test on data on interval level. To test associations between variables, Pearson's and Spearman's correlations were calculated. In the logistic regression analysis (enter), MS symptoms that were significantly more frequent among heat-sensitive participants were dichotomized and entered as dependent variables, and EDSS (0-6.5), disease course (relapsing-remitting/progressive forms), heat sensitivity (yes/no), time since onset, and age and sex (female/male) were independent variables. In the linear regression analysis (enter) mean FSS and summarized PDQ were entered as dependent variables and EDSS (0-6.5), disease course (relapsing/progressive forms), time since onset and heat sensitivity age and sex (female/male) were independent variables. In this study, and due to the number of statistical analyses, a P-value < 0.010 has been considered a significant value.\n[SUBTITLE] Ethical considerations [SUBSECTION] The study was guided by common ethical principals in research and according to the Declaration of Helsinki [27]. Approval to use the SwMS register was received from the responsible local administrator. The Regional Ethical Review Board at the Faculty of Health Sciences, Linköping University, Sweden (Dnr M13-07) also approved the study. A completed and returned questionnaire was regarded as granted informed consent.\nThe study was guided by common ethical principals in research and according to the Declaration of Helsinki [27]. Approval to use the SwMS register was received from the responsible local administrator. The Regional Ethical Review Board at the Faculty of Health Sciences, Linköping University, Sweden (Dnr M13-07) also approved the study. A completed and returned questionnaire was regarded as granted informed consent.", "In order to reach individuals diagnosed with MS living in the area, a cross-sectional designed survey was undertaken, addressed to the individuals registered in the Swedish MS register (SwMS-reg). Together with information about the study, a questionnaire and a pre-paid reply envelope were sent to 334 individuals fulfilling the following criteria: i) being diagnosed with MS, ii) having an Expanded Disability Status Score (EDSS) [19] in the interval 0≤EDSS≤6.5, and iii) being of working age, i.e. 20-65 years, as 65 is the official retirement age. The questionnaires were distributed during 2007 to individuals with EDSS 1-6.5, and a complementary distribution was sent to individuals with EDSS = 0 in 2008. To those who had not answered within three weeks, one reminder was sent.", "The individual's age, sex, disease course, i.e. relapsing-remitting (RR), secondary progressive (SP), primary progressive (PP) and function, measured with the EDSS [19], together with date of onset of MS, were obtained from the SwMS register. The diagnostic criteria used in the SwMS register are both Poser criteria [20] (older cases) and McDonald criteria [21] (recent time). Secondary progressive course of MS (SP) is a clinical neurological status whose onset can be hard to pinpoint. In clinical trials and other studies, this status is often not defined in detail. In the SwMS register the following definition is chosen: A phase of the disease where one can secure a slowly increasing pyramidal para- or even tetraparesis, sometimes also a bladder syndrome. In some cases can it be a cerebellar syndrome. A crucial criterion is that you observe increasing pyramidal (or cerebellar) symptoms without any signs of bouts [22].\nThe patients' EDSS and disease-course are continuously upgraded at physicians call at the neurological policlinics. For the majority of the participants in this study, both the EDSS and disease-course had been determined within a year before the data collection started. Common background factors, such as civil status, family, level of education and ongoing medical treatment, were requested in the questionnaire. Occurrence of heat sensitivity was asked about in a single question, \"Are you sensitive to heat?\" (Yes/No), with follow-up questions concerning how this was experienced when in sunshine, in a warm room, or taking a hot bath or shower, and what room-temperature the patient preferred. Occurrence and perceived severity of fatigue were requested as well. Furthermore, the following instruments were included: the Fatigue Severity Scale (FSS) [7,11]; the MS-related symptom checklist [23] and the Perceived Deficit Questionnaire (PDQ) [24].\nIn this study, and based on earlier studies [12,14,25], the EDSS was grouped according to normal neurological condition (EDSS = 0), mild disability (1.0≤EDSS≤3.5), moderate disability (4.0≤EDSS≤5.5) and severe disability (6.0≤EDSS≤6.5).\nThe FSS [7,11] comprises nine items covering perceived severity of fatigue, and each item is graded from least fatigue (1) to severe fatigue (7). As in earlier studies [e.g. [9,14], in this study an FSS mean score ≤4 was regarded as indicating no fatigue, >4 but <5 borderline fatigue, and ≥5 severe fatigue. The FSS has been used in earlier studies in Sweden [14] and is also used clinically. In this study, concurrent validity was assessed through correlations between the FSS with questions about the impact of fatigue on daily life as well as the occurrence of fatigue, rated on the MS-related symptom checklist [23], and resulted in r = 0.79, (P < 0.01) and r = 0.77 (P < 0.01), respectively. Reliability was assessed using Cronbach's alpha, and the alpha coefficient was 0.93.\nThe MS-related symptom checklist [23] was used to report 25 common MS symptoms such as fatigue, weakness in arms and legs, balance problems, pain, numbness, blurred vision, depression and difficulties with urination such as frequency and urgency. The occurrence of each symptom was graded in six steps from never (1) to always (6), which were then dichotomized, into never/sometimes (1-3) to 1 and usually/always (4-6) to 2. The checklist has been found to be closely related to the EDSS [23], and has been used in an earlier study in Sweden [13].\nCognitive dysfunction was assessed using the PDQ [24], measuring perceived problems with memory, attention and concentration. The original English version of the PDQ was translated into Swedish and then back-translated in accordance with Streiner and Norman [26], with good agreement. Twenty items are graded from never (0) to always (4). Summated scores vary from 0 to 80, where higher scores indicate greater cognitive problems. In this study, concurrent validity, assessed through correlations between the PDQ sum score and the MS-related symptoms forgetfulness and concentration difficulties, was r = 0.75 (P < 0.010) and r = 0.73 (P < 0.010), respectively. Reliability was tested using Cronbach's alpha (α = 0.95) and the split-half technique (r = 0.87).", "The data have been treated and analysed in relation to perceived heat sensitivity. The descriptive statistics used are in congruence with measurement scale level, except for the FSS which is treated on interval level as in earlier studies [e.g. [9,11,14]. Differences between groups have been tested using the chi-square test on data on nominal level, the Mann Whitney U-test on data on ordinal level, and Student's t-test on data on interval level. To test associations between variables, Pearson's and Spearman's correlations were calculated. In the logistic regression analysis (enter), MS symptoms that were significantly more frequent among heat-sensitive participants were dichotomized and entered as dependent variables, and EDSS (0-6.5), disease course (relapsing-remitting/progressive forms), heat sensitivity (yes/no), time since onset, and age and sex (female/male) were independent variables. In the linear regression analysis (enter) mean FSS and summarized PDQ were entered as dependent variables and EDSS (0-6.5), disease course (relapsing/progressive forms), time since onset and heat sensitivity age and sex (female/male) were independent variables. In this study, and due to the number of statistical analyses, a P-value < 0.010 has been considered a significant value.", "The study was guided by common ethical principals in research and according to the Declaration of Helsinki [27]. Approval to use the SwMS register was received from the responsible local administrator. The Regional Ethical Review Board at the Faculty of Health Sciences, Linköping University, Sweden (Dnr M13-07) also approved the study. A completed and returned questionnaire was regarded as granted informed consent.", "[SUBTITLE] Study group [SUBSECTION] Two hundred and fifty-six patients, 195 (76%) women and 61 (24%) men, aged 65 years or younger and with an EDSS score between 0 and 6.5, answered the questionnaires. The response rate was 79.3%. Between participants and non-participants there were no statistically significant differences regarding sex (P = 0.052), disease course (P = 0.664) or EDSS score (P = 0.142). The non-participants were about five years younger than the participants, m = 42.7 years, (SD 10.5) (P < 0.001) and also had fewer years since onset, Md = 10 (Q1 = 5.75; Q3 = 15.25) (P = 0.014) (Table 1).\nCharacteristics of the participants (n = 256)\nOf the participants, the women were somewhat older (P = 0.022) than the men, 48.3 (SD 11) and 44.8 (SD 10.1), respectively. At onset the median age was 31 years (Q1 = 25; Q3 = 38). In 5.5% of the participants, onset occurred before 20 years of age, and the youngest had been 12 years old. The main course of the disease was relapsing-remitting (73%). Ten percent had an EDSS score of 0, but most of the patients (61%) had mild disease severity according to the EDSS. Two hundred and two participants (79%) were being treated pharmacologically. One hundred and forty-four participants (56%) were being treated with immune modulating medication and 32 participants (13%) with medication for fatigue (Table 1).\nTwo hundred and fifty-six patients, 195 (76%) women and 61 (24%) men, aged 65 years or younger and with an EDSS score between 0 and 6.5, answered the questionnaires. The response rate was 79.3%. Between participants and non-participants there were no statistically significant differences regarding sex (P = 0.052), disease course (P = 0.664) or EDSS score (P = 0.142). The non-participants were about five years younger than the participants, m = 42.7 years, (SD 10.5) (P < 0.001) and also had fewer years since onset, Md = 10 (Q1 = 5.75; Q3 = 15.25) (P = 0.014) (Table 1).\nCharacteristics of the participants (n = 256)\nOf the participants, the women were somewhat older (P = 0.022) than the men, 48.3 (SD 11) and 44.8 (SD 10.1), respectively. At onset the median age was 31 years (Q1 = 25; Q3 = 38). In 5.5% of the participants, onset occurred before 20 years of age, and the youngest had been 12 years old. The main course of the disease was relapsing-remitting (73%). Ten percent had an EDSS score of 0, but most of the patients (61%) had mild disease severity according to the EDSS. Two hundred and two participants (79%) were being treated pharmacologically. One hundred and forty-four participants (56%) were being treated with immune modulating medication and 32 participants (13%) with medication for fatigue (Table 1).\n[SUBTITLE] Heat sensitivity [SUBSECTION] One hundred and forty-nine participants (58%) reported heat sensitivity, with no difference between women and men (P = 0.102) or in relation to disease course (P = 0.226). In relation to EDSS (Table 1), eight of the participants (31%) with normal neurological condition, 98 (62%) with mild disability, 16 (53%) with moderate disability and 27 (63%) with a severe disability reported heat sensitivity. Of those who were sensitive to heat, 70% preferred a room temperature of less than 20 C°, while 73% of those who were not heat sensitive preferred a room temperature of more than 20°C (P < 0.001). Eleven of the heat-sensitive participants (n = 149) annotated in the questionnaire that they also were 'cold'.\nOf common MS-related symptoms, the heat-sensitive participants reported a significantly higher occurrence of several symptoms than did those who were not heat sensitive, for example fatigue, weakness in legs, concentration difficulties, pain, and urination urgency (Table 2). Weakness in arms and legs as well as balance problems correlated significantly with the EDSS, r = 0.17 (p < 0.010), r = 0.49 (p < 0.010), r = 0.51 (p < 0.010), respectively.\nSignificant dichotomized MS symptoms (never to sometimes and usually to always) in relation to reported heat sensitivity (n = 256)\n*) Missing values between two and six\nIn the logistic regression analysis, heat sensitivity significantly explained several MS symptoms, such as fatigue, concentration difficulties and pain. Two of the symptoms, spasms and balance problems, were explained by the EDSS alone (Table 3) while some were explained by more than one variable. Together with heat sensitivity and the EDSS, increasing age also explained the occurrence of weakness in arms (OR = 1.05, 95%CI = 1.02-1.08, P = 0.001) and urination frequency (OR = 1.04, 95%CI = 1.01-1.07, P = 0.007). Other interesting findings with regard to heat sensitivity, EDSS, and increasing age were weakness in legs (OR = 1.03, 95%CI = 1.0-1.06, P = 0.042) and pain (OR = 1.04, 95%CI = 1.0-1.07, P = 0.025). Urination urgency was also explained by gender, female (Or = 0.23, 95%CI = 0.07-0.7, P = 0.01).\nLogistic regression analyses of dichotomized common MS symptoms (1 = never to sometimes, 2 = usually to always) as dependent variables, and disability (EDSS 0-6.5) and heat sensitivity (1 = yes/0 = no) as independent variables\nThe difference in the occurrence of fatigue was confirmed by the FSS. Of those who were heat sensitive, 63% reported severe fatigue, FSS≥5, while among those who were not heat sensitive 45% reported mild or no fatigue, FSS≤4 (P < 0.001) (Table 4). One hundred and ten (57%) of the women (n = 192) and 21 (35%) of the men (n = 60) scored an FSS mean value ≥5, classified as fatigued (P = 0.010). About a third of the participants who were heat sensitive had concentration difficulties significantly more often (P < 0.001). This difference was confirmed by the PDQ, with Md = 29.5 (Q1 = 19; Q3 = 38) among the heat sensitive participants and Md = 21 (Q1 = 12; Q3 = 33) in the non-heat-sensitive participants (P < 0.001). In the linear regression analyses, heat sensitivity predicted both fatigue and concentration difficulties (Table 5).\nFatigue assessed in the Fatigue Severity Scale (FSS)* in relation to reported heat sensitivity (n = 252)**\n*FSS≤4 = Non-fatigue; 4 > FSS <5 = Borderline fatigue; FSS≥5 = Severe fatigue;\n**Missing value, n = 4\nLinear regression analyses of fatigue (FSS) and perceived deficit questionnaire (PDQ) as dependent variables and EDSS (0-6.5) and heat sensitivity (yes/no) as independent variables\na) R 0.421; R2 0.177; Adjusted R2 0.161\nb) R 0.239; R 0.057; Adjusted R2 0.038\nThere was no difference between heat-sensitive and non-heat sensitive participants in the use of immune-modulating drugs or treatment for fatigue. Of those being treated with the immune-modulating beta-interferon, fewer reported severe fatigue, FSS≥5.0, compared to those being treated with glatiramer acetate or natalizumab (P = 0.021). Of those who were being treated for fatigue (n = 32) a higher proportion reported severe fatigue, FSS≥5.0 (P = 0.016), compared to those not being treated for fatigue.\nOne hundred and forty-nine participants (58%) reported heat sensitivity, with no difference between women and men (P = 0.102) or in relation to disease course (P = 0.226). In relation to EDSS (Table 1), eight of the participants (31%) with normal neurological condition, 98 (62%) with mild disability, 16 (53%) with moderate disability and 27 (63%) with a severe disability reported heat sensitivity. Of those who were sensitive to heat, 70% preferred a room temperature of less than 20 C°, while 73% of those who were not heat sensitive preferred a room temperature of more than 20°C (P < 0.001). Eleven of the heat-sensitive participants (n = 149) annotated in the questionnaire that they also were 'cold'.\nOf common MS-related symptoms, the heat-sensitive participants reported a significantly higher occurrence of several symptoms than did those who were not heat sensitive, for example fatigue, weakness in legs, concentration difficulties, pain, and urination urgency (Table 2). Weakness in arms and legs as well as balance problems correlated significantly with the EDSS, r = 0.17 (p < 0.010), r = 0.49 (p < 0.010), r = 0.51 (p < 0.010), respectively.\nSignificant dichotomized MS symptoms (never to sometimes and usually to always) in relation to reported heat sensitivity (n = 256)\n*) Missing values between two and six\nIn the logistic regression analysis, heat sensitivity significantly explained several MS symptoms, such as fatigue, concentration difficulties and pain. Two of the symptoms, spasms and balance problems, were explained by the EDSS alone (Table 3) while some were explained by more than one variable. Together with heat sensitivity and the EDSS, increasing age also explained the occurrence of weakness in arms (OR = 1.05, 95%CI = 1.02-1.08, P = 0.001) and urination frequency (OR = 1.04, 95%CI = 1.01-1.07, P = 0.007). Other interesting findings with regard to heat sensitivity, EDSS, and increasing age were weakness in legs (OR = 1.03, 95%CI = 1.0-1.06, P = 0.042) and pain (OR = 1.04, 95%CI = 1.0-1.07, P = 0.025). Urination urgency was also explained by gender, female (Or = 0.23, 95%CI = 0.07-0.7, P = 0.01).\nLogistic regression analyses of dichotomized common MS symptoms (1 = never to sometimes, 2 = usually to always) as dependent variables, and disability (EDSS 0-6.5) and heat sensitivity (1 = yes/0 = no) as independent variables\nThe difference in the occurrence of fatigue was confirmed by the FSS. Of those who were heat sensitive, 63% reported severe fatigue, FSS≥5, while among those who were not heat sensitive 45% reported mild or no fatigue, FSS≤4 (P < 0.001) (Table 4). One hundred and ten (57%) of the women (n = 192) and 21 (35%) of the men (n = 60) scored an FSS mean value ≥5, classified as fatigued (P = 0.010). About a third of the participants who were heat sensitive had concentration difficulties significantly more often (P < 0.001). This difference was confirmed by the PDQ, with Md = 29.5 (Q1 = 19; Q3 = 38) among the heat sensitive participants and Md = 21 (Q1 = 12; Q3 = 33) in the non-heat-sensitive participants (P < 0.001). In the linear regression analyses, heat sensitivity predicted both fatigue and concentration difficulties (Table 5).\nFatigue assessed in the Fatigue Severity Scale (FSS)* in relation to reported heat sensitivity (n = 252)**\n*FSS≤4 = Non-fatigue; 4 > FSS <5 = Borderline fatigue; FSS≥5 = Severe fatigue;\n**Missing value, n = 4\nLinear regression analyses of fatigue (FSS) and perceived deficit questionnaire (PDQ) as dependent variables and EDSS (0-6.5) and heat sensitivity (yes/no) as independent variables\na) R 0.421; R2 0.177; Adjusted R2 0.161\nb) R 0.239; R 0.057; Adjusted R2 0.038\nThere was no difference between heat-sensitive and non-heat sensitive participants in the use of immune-modulating drugs or treatment for fatigue. Of those being treated with the immune-modulating beta-interferon, fewer reported severe fatigue, FSS≥5.0, compared to those being treated with glatiramer acetate or natalizumab (P = 0.021). Of those who were being treated for fatigue (n = 32) a higher proportion reported severe fatigue, FSS≥5.0 (P = 0.016), compared to those not being treated for fatigue.", "Two hundred and fifty-six patients, 195 (76%) women and 61 (24%) men, aged 65 years or younger and with an EDSS score between 0 and 6.5, answered the questionnaires. The response rate was 79.3%. Between participants and non-participants there were no statistically significant differences regarding sex (P = 0.052), disease course (P = 0.664) or EDSS score (P = 0.142). The non-participants were about five years younger than the participants, m = 42.7 years, (SD 10.5) (P < 0.001) and also had fewer years since onset, Md = 10 (Q1 = 5.75; Q3 = 15.25) (P = 0.014) (Table 1).\nCharacteristics of the participants (n = 256)\nOf the participants, the women were somewhat older (P = 0.022) than the men, 48.3 (SD 11) and 44.8 (SD 10.1), respectively. At onset the median age was 31 years (Q1 = 25; Q3 = 38). In 5.5% of the participants, onset occurred before 20 years of age, and the youngest had been 12 years old. The main course of the disease was relapsing-remitting (73%). Ten percent had an EDSS score of 0, but most of the patients (61%) had mild disease severity according to the EDSS. Two hundred and two participants (79%) were being treated pharmacologically. One hundred and forty-four participants (56%) were being treated with immune modulating medication and 32 participants (13%) with medication for fatigue (Table 1).", "One hundred and forty-nine participants (58%) reported heat sensitivity, with no difference between women and men (P = 0.102) or in relation to disease course (P = 0.226). In relation to EDSS (Table 1), eight of the participants (31%) with normal neurological condition, 98 (62%) with mild disability, 16 (53%) with moderate disability and 27 (63%) with a severe disability reported heat sensitivity. Of those who were sensitive to heat, 70% preferred a room temperature of less than 20 C°, while 73% of those who were not heat sensitive preferred a room temperature of more than 20°C (P < 0.001). Eleven of the heat-sensitive participants (n = 149) annotated in the questionnaire that they also were 'cold'.\nOf common MS-related symptoms, the heat-sensitive participants reported a significantly higher occurrence of several symptoms than did those who were not heat sensitive, for example fatigue, weakness in legs, concentration difficulties, pain, and urination urgency (Table 2). Weakness in arms and legs as well as balance problems correlated significantly with the EDSS, r = 0.17 (p < 0.010), r = 0.49 (p < 0.010), r = 0.51 (p < 0.010), respectively.\nSignificant dichotomized MS symptoms (never to sometimes and usually to always) in relation to reported heat sensitivity (n = 256)\n*) Missing values between two and six\nIn the logistic regression analysis, heat sensitivity significantly explained several MS symptoms, such as fatigue, concentration difficulties and pain. Two of the symptoms, spasms and balance problems, were explained by the EDSS alone (Table 3) while some were explained by more than one variable. Together with heat sensitivity and the EDSS, increasing age also explained the occurrence of weakness in arms (OR = 1.05, 95%CI = 1.02-1.08, P = 0.001) and urination frequency (OR = 1.04, 95%CI = 1.01-1.07, P = 0.007). Other interesting findings with regard to heat sensitivity, EDSS, and increasing age were weakness in legs (OR = 1.03, 95%CI = 1.0-1.06, P = 0.042) and pain (OR = 1.04, 95%CI = 1.0-1.07, P = 0.025). Urination urgency was also explained by gender, female (Or = 0.23, 95%CI = 0.07-0.7, P = 0.01).\nLogistic regression analyses of dichotomized common MS symptoms (1 = never to sometimes, 2 = usually to always) as dependent variables, and disability (EDSS 0-6.5) and heat sensitivity (1 = yes/0 = no) as independent variables\nThe difference in the occurrence of fatigue was confirmed by the FSS. Of those who were heat sensitive, 63% reported severe fatigue, FSS≥5, while among those who were not heat sensitive 45% reported mild or no fatigue, FSS≤4 (P < 0.001) (Table 4). One hundred and ten (57%) of the women (n = 192) and 21 (35%) of the men (n = 60) scored an FSS mean value ≥5, classified as fatigued (P = 0.010). About a third of the participants who were heat sensitive had concentration difficulties significantly more often (P < 0.001). This difference was confirmed by the PDQ, with Md = 29.5 (Q1 = 19; Q3 = 38) among the heat sensitive participants and Md = 21 (Q1 = 12; Q3 = 33) in the non-heat-sensitive participants (P < 0.001). In the linear regression analyses, heat sensitivity predicted both fatigue and concentration difficulties (Table 5).\nFatigue assessed in the Fatigue Severity Scale (FSS)* in relation to reported heat sensitivity (n = 252)**\n*FSS≤4 = Non-fatigue; 4 > FSS <5 = Borderline fatigue; FSS≥5 = Severe fatigue;\n**Missing value, n = 4\nLinear regression analyses of fatigue (FSS) and perceived deficit questionnaire (PDQ) as dependent variables and EDSS (0-6.5) and heat sensitivity (yes/no) as independent variables\na) R 0.421; R2 0.177; Adjusted R2 0.161\nb) R 0.239; R 0.057; Adjusted R2 0.038\nThere was no difference between heat-sensitive and non-heat sensitive participants in the use of immune-modulating drugs or treatment for fatigue. Of those being treated with the immune-modulating beta-interferon, fewer reported severe fatigue, FSS≥5.0, compared to those being treated with glatiramer acetate or natalizumab (P = 0.021). Of those who were being treated for fatigue (n = 32) a higher proportion reported severe fatigue, FSS≥5.0 (P = 0.016), compared to those not being treated for fatigue.", "This is an explorative study aimed at investigating the occurrence of heat sensitivity in a group of individuals diagnosed with MS. Further, it aimed to investigate relations between heat sensitivity and common symptoms in MS, disease course and disability.\nThe participants were recruited from the Swedish MS register in a local county in eastern Sweden. In the chosen EDSS interval, a greater proportion of the patients had an RR course of the disease and were receiving immune-modulating treatment, with regular follow-up visits at the neurological policlinics once a year. Furthermore, some patients with secondary progressive MS and no treatment are seen by general practitioners. Thus, in this study, there is a selection of patients towards an RR course of the disease and who are treated patients. The EDSS scores used were assessed within a year before the data collection and, thus, there might be some random variation in the group. However, if any changes occurred this should be noted in the registry.\nOur results indicate that heat sensitivity is highly correlated with several symptoms commonly reported by MS patients. That heat sensitivity is associated with fatigue is well known and has been reported by several authors [4,6-8]. However, our data also reveal that heat sensitivity is a key factor associated with the occurrence of a wide variety of other MS symptoms, for example pain, concentration problems and urination urgency.\n[SUBTITLE] Heat sensitivity intensifies MS symptoms [SUBSECTION] The most striking result in this study is that heat sensitivity significantly correlated with - and in the logistic regression analyses, appeared as an explaining factor for - the most incapacitating symptoms of MS, viz. fatigue, concentration problems and pain. This result discloses heat sensitivity as a key clinical factor. Our findings are consistent with a Norwegian study by Nortvedt and co-workers [28] stating that bodily pain and low vitality are important problems of MS with a significant impact on quality of life. Low vitality [28] can be interpreted as strongly associated to the symptom of fatigue.\nWe found that many other MS symptoms are also correlated with heat sensitivity. Interestingly, similar observations were actually reported in the early works of Uthoff, although subsequent citations narrowed the interpretation of Uthoff's phenomenon to blurring of the vision [6]. In the present study, however, no relationship between the symptom of blurred vision and heat sensitivity was confirmed.\nWhat might be the cause of the observed co-occurrence of heat sensitivity and pain, fatigue and cognitive problems? Can this indicate something about the pathogenic mechanism? Based on studies of post-stroke patients, Craig [29] suggests that central pain might be caused by a thermoregulatory dysfunction. A lesion in the thalamus releases or disinhibits the feeling of burning pain by removing the normal inhibition of pain by cold. The role of the thalamus is as ascending relay of spinothalamic activity to the cortex [29]. This is similar to Österberg's [30] proposal that central pain in MS seems to be generated through lesions affecting the spinothalamo-cortical pathways [30]. A third of the MS patients with central pain had visible lesions in the thalamus [31]. Thus, one can speculate that the thalamus might be involved.\nIt is also interesting to think about the effects of decreased sweating in terms of electrolytic imbalance and secondary neuronal effects. Recent results by Saari and co-workers [32] disclose such an impairment to thermoregulatory sweating in MS. One idea that may explain thermoregulatory dysfunction is that lesions can affect important cerebral areas such as the hypothalamus. The study by Saari [32] also demonstrated a correlation between increasing sweating impairment and increasing disability (EDSS), which is congruent with our results showing a correlation between heat sensitivity and increasing EDSS.\nFatigue in MS has been eagerly studied by many researchers during the past decade. Fatigue and heat sensitivity are related in many aspects, and a subsequent question is whether they have common pathogenic features. Recently, Marino [6] stated that MS fatigue is likely to be a central rather than a peripheral phenomenon. Heat sensitivity in MS, however, is not understood as clearly.\nThe most striking result in this study is that heat sensitivity significantly correlated with - and in the logistic regression analyses, appeared as an explaining factor for - the most incapacitating symptoms of MS, viz. fatigue, concentration problems and pain. This result discloses heat sensitivity as a key clinical factor. Our findings are consistent with a Norwegian study by Nortvedt and co-workers [28] stating that bodily pain and low vitality are important problems of MS with a significant impact on quality of life. Low vitality [28] can be interpreted as strongly associated to the symptom of fatigue.\nWe found that many other MS symptoms are also correlated with heat sensitivity. Interestingly, similar observations were actually reported in the early works of Uthoff, although subsequent citations narrowed the interpretation of Uthoff's phenomenon to blurring of the vision [6]. In the present study, however, no relationship between the symptom of blurred vision and heat sensitivity was confirmed.\nWhat might be the cause of the observed co-occurrence of heat sensitivity and pain, fatigue and cognitive problems? Can this indicate something about the pathogenic mechanism? Based on studies of post-stroke patients, Craig [29] suggests that central pain might be caused by a thermoregulatory dysfunction. A lesion in the thalamus releases or disinhibits the feeling of burning pain by removing the normal inhibition of pain by cold. The role of the thalamus is as ascending relay of spinothalamic activity to the cortex [29]. This is similar to Österberg's [30] proposal that central pain in MS seems to be generated through lesions affecting the spinothalamo-cortical pathways [30]. A third of the MS patients with central pain had visible lesions in the thalamus [31]. Thus, one can speculate that the thalamus might be involved.\nIt is also interesting to think about the effects of decreased sweating in terms of electrolytic imbalance and secondary neuronal effects. Recent results by Saari and co-workers [32] disclose such an impairment to thermoregulatory sweating in MS. One idea that may explain thermoregulatory dysfunction is that lesions can affect important cerebral areas such as the hypothalamus. The study by Saari [32] also demonstrated a correlation between increasing sweating impairment and increasing disability (EDSS), which is congruent with our results showing a correlation between heat sensitivity and increasing EDSS.\nFatigue in MS has been eagerly studied by many researchers during the past decade. Fatigue and heat sensitivity are related in many aspects, and a subsequent question is whether they have common pathogenic features. Recently, Marino [6] stated that MS fatigue is likely to be a central rather than a peripheral phenomenon. Heat sensitivity in MS, however, is not understood as clearly.\n[SUBTITLE] The mechanism behind heat sensitivity [SUBSECTION] The mechanism of heat sensitivity in MS is reviewed in a recent article by Marino [6]. The heat reaction blocks the action potential of the demyelinated neuron (frequency-dependent conduction block, - FDCB) [3]. The demyelinization results in a slower nerve conduction velocity. A conduction block can occur because the damaged axons transmit only single or low frequency impulses, instead of high frequency impulse trains like in healthy nerve tissue [33]. Marino [6] states that this observation is very important, especially when an increase in temperature blocks nerve impulses in demyelinated fibres [34]. Interestingly, very small increases in temperature can also block action potentials [6]. The heat sensitivity in MS is described by Baker as secondary to both environmental heat and environmental humidity, as well as to exercise [35]. Both passive and active body temperature increases give heat reactions. Clinical reports from individual patients in our MS clinic (i.e. outside this study) reveal that their experience of temperature aberrations can vary greatly, indicating that the mechanisms may be multiple.\nThe mechanism of heat sensitivity in MS is reviewed in a recent article by Marino [6]. The heat reaction blocks the action potential of the demyelinated neuron (frequency-dependent conduction block, - FDCB) [3]. The demyelinization results in a slower nerve conduction velocity. A conduction block can occur because the damaged axons transmit only single or low frequency impulses, instead of high frequency impulse trains like in healthy nerve tissue [33]. Marino [6] states that this observation is very important, especially when an increase in temperature blocks nerve impulses in demyelinated fibres [34]. Interestingly, very small increases in temperature can also block action potentials [6]. The heat sensitivity in MS is described by Baker as secondary to both environmental heat and environmental humidity, as well as to exercise [35]. Both passive and active body temperature increases give heat reactions. Clinical reports from individual patients in our MS clinic (i.e. outside this study) reveal that their experience of temperature aberrations can vary greatly, indicating that the mechanisms may be multiple.\n[SUBTITLE] The subjective phenomenon of heat sensitivity [SUBSECTION] Many patients with MS are aware of their heat sensitivity, and have experienced increased clinical symptoms when becoming warm from fever, hot environmental temperature or physical exercise. In this study, patients who did or did not report heat sensitivity were dichotomized and analysed according to this categorization. It is worth noting that some patients reported both heat sensitivity and a subjective feeling of \"cold\". This suggests that temperature aberrations in MS can be complex and individual. In our clinical practice we also encounter MS patients with complex subjective temperature aberrations, e.g. patients who deny heat sensitivity when asked, yet prefer to sleep in cold bedrooms because it alleviates their MS symptoms. Such patients may not yet be aware of a mild heat sensitivity. This factor also suggests that milder forms of heat sensitivity may be underreported in a study like ours. We have also met one patient who likes to be in the sun all day without any side effects, but after jogging has to take repeated showers to prevent incapacitating fatigue. Finally, it is well known that patients with severe MS can sometimes develop hypothermia, often without any major subjective symptoms. We suggest that qualitative studies should be initiated to describe patients' experiences of subjective phenomena like heat sensitivity, and feeling cold as well as combinations of the two. One cannot exclude that central and peripheral mechanisms interact in the \"temperature syndrome\" of MS, which includes more features than classical heat sensitivity.\nMany patients with MS are aware of their heat sensitivity, and have experienced increased clinical symptoms when becoming warm from fever, hot environmental temperature or physical exercise. In this study, patients who did or did not report heat sensitivity were dichotomized and analysed according to this categorization. It is worth noting that some patients reported both heat sensitivity and a subjective feeling of \"cold\". This suggests that temperature aberrations in MS can be complex and individual. In our clinical practice we also encounter MS patients with complex subjective temperature aberrations, e.g. patients who deny heat sensitivity when asked, yet prefer to sleep in cold bedrooms because it alleviates their MS symptoms. Such patients may not yet be aware of a mild heat sensitivity. This factor also suggests that milder forms of heat sensitivity may be underreported in a study like ours. We have also met one patient who likes to be in the sun all day without any side effects, but after jogging has to take repeated showers to prevent incapacitating fatigue. Finally, it is well known that patients with severe MS can sometimes develop hypothermia, often without any major subjective symptoms. We suggest that qualitative studies should be initiated to describe patients' experiences of subjective phenomena like heat sensitivity, and feeling cold as well as combinations of the two. One cannot exclude that central and peripheral mechanisms interact in the \"temperature syndrome\" of MS, which includes more features than classical heat sensitivity.\n[SUBTITLE] Fatigue as a side effect of immune modulating drugs [SUBSECTION] It has recently been suggested that beta-interferon can increase fatigue through influenza-like side effects including hyperthermia, but that glatiramer-acetate, on the other hand, is a neutral agent in this respect, and that natalizumab at least in some cases decreases fatigue. Some scientific reports support these observations [16-18]. In our study, however, none of these relationships could be demonstrated. However, as this study has a descriptive design, no conclusion regarding what impact immune modulating treatment has on fatigue could be drawn.\nIt has recently been suggested that beta-interferon can increase fatigue through influenza-like side effects including hyperthermia, but that glatiramer-acetate, on the other hand, is a neutral agent in this respect, and that natalizumab at least in some cases decreases fatigue. Some scientific reports support these observations [16-18]. In our study, however, none of these relationships could be demonstrated. However, as this study has a descriptive design, no conclusion regarding what impact immune modulating treatment has on fatigue could be drawn.", "The most striking result in this study is that heat sensitivity significantly correlated with - and in the logistic regression analyses, appeared as an explaining factor for - the most incapacitating symptoms of MS, viz. fatigue, concentration problems and pain. This result discloses heat sensitivity as a key clinical factor. Our findings are consistent with a Norwegian study by Nortvedt and co-workers [28] stating that bodily pain and low vitality are important problems of MS with a significant impact on quality of life. Low vitality [28] can be interpreted as strongly associated to the symptom of fatigue.\nWe found that many other MS symptoms are also correlated with heat sensitivity. Interestingly, similar observations were actually reported in the early works of Uthoff, although subsequent citations narrowed the interpretation of Uthoff's phenomenon to blurring of the vision [6]. In the present study, however, no relationship between the symptom of blurred vision and heat sensitivity was confirmed.\nWhat might be the cause of the observed co-occurrence of heat sensitivity and pain, fatigue and cognitive problems? Can this indicate something about the pathogenic mechanism? Based on studies of post-stroke patients, Craig [29] suggests that central pain might be caused by a thermoregulatory dysfunction. A lesion in the thalamus releases or disinhibits the feeling of burning pain by removing the normal inhibition of pain by cold. The role of the thalamus is as ascending relay of spinothalamic activity to the cortex [29]. This is similar to Österberg's [30] proposal that central pain in MS seems to be generated through lesions affecting the spinothalamo-cortical pathways [30]. A third of the MS patients with central pain had visible lesions in the thalamus [31]. Thus, one can speculate that the thalamus might be involved.\nIt is also interesting to think about the effects of decreased sweating in terms of electrolytic imbalance and secondary neuronal effects. Recent results by Saari and co-workers [32] disclose such an impairment to thermoregulatory sweating in MS. One idea that may explain thermoregulatory dysfunction is that lesions can affect important cerebral areas such as the hypothalamus. The study by Saari [32] also demonstrated a correlation between increasing sweating impairment and increasing disability (EDSS), which is congruent with our results showing a correlation between heat sensitivity and increasing EDSS.\nFatigue in MS has been eagerly studied by many researchers during the past decade. Fatigue and heat sensitivity are related in many aspects, and a subsequent question is whether they have common pathogenic features. Recently, Marino [6] stated that MS fatigue is likely to be a central rather than a peripheral phenomenon. Heat sensitivity in MS, however, is not understood as clearly.", "The mechanism of heat sensitivity in MS is reviewed in a recent article by Marino [6]. The heat reaction blocks the action potential of the demyelinated neuron (frequency-dependent conduction block, - FDCB) [3]. The demyelinization results in a slower nerve conduction velocity. A conduction block can occur because the damaged axons transmit only single or low frequency impulses, instead of high frequency impulse trains like in healthy nerve tissue [33]. Marino [6] states that this observation is very important, especially when an increase in temperature blocks nerve impulses in demyelinated fibres [34]. Interestingly, very small increases in temperature can also block action potentials [6]. The heat sensitivity in MS is described by Baker as secondary to both environmental heat and environmental humidity, as well as to exercise [35]. Both passive and active body temperature increases give heat reactions. Clinical reports from individual patients in our MS clinic (i.e. outside this study) reveal that their experience of temperature aberrations can vary greatly, indicating that the mechanisms may be multiple.", "Many patients with MS are aware of their heat sensitivity, and have experienced increased clinical symptoms when becoming warm from fever, hot environmental temperature or physical exercise. In this study, patients who did or did not report heat sensitivity were dichotomized and analysed according to this categorization. It is worth noting that some patients reported both heat sensitivity and a subjective feeling of \"cold\". This suggests that temperature aberrations in MS can be complex and individual. In our clinical practice we also encounter MS patients with complex subjective temperature aberrations, e.g. patients who deny heat sensitivity when asked, yet prefer to sleep in cold bedrooms because it alleviates their MS symptoms. Such patients may not yet be aware of a mild heat sensitivity. This factor also suggests that milder forms of heat sensitivity may be underreported in a study like ours. We have also met one patient who likes to be in the sun all day without any side effects, but after jogging has to take repeated showers to prevent incapacitating fatigue. Finally, it is well known that patients with severe MS can sometimes develop hypothermia, often without any major subjective symptoms. We suggest that qualitative studies should be initiated to describe patients' experiences of subjective phenomena like heat sensitivity, and feeling cold as well as combinations of the two. One cannot exclude that central and peripheral mechanisms interact in the \"temperature syndrome\" of MS, which includes more features than classical heat sensitivity.", "It has recently been suggested that beta-interferon can increase fatigue through influenza-like side effects including hyperthermia, but that glatiramer-acetate, on the other hand, is a neutral agent in this respect, and that natalizumab at least in some cases decreases fatigue. Some scientific reports support these observations [16-18]. In our study, however, none of these relationships could be demonstrated. However, as this study has a descriptive design, no conclusion regarding what impact immune modulating treatment has on fatigue could be drawn.", "In conclusion, although heat sensitivity in MS was described as early as the late 19th century and is a well-known phenomenon today, it has to date been disregarded in studies of fatigue in MS [6]. The findings in this study underline the importance of heat sensitivity in MS patients as a key symptom that is highly correlated with disabling symptoms such as fatigue, concentration difficulty and urination urgency. A majority of the participants rated the symptom of fatigue as their most impairing symptom. Furthermore, a significantly higher proportion among the heat-sensitive participants rated higher levels of fatigue compared to the participants who were not heat sensitive.\nThe results of our study put heat sensitivity in the position of a key clinical symptom. Our findings emphasize the need to further investigate the mechanism of heat sensitivity: What is the role of the sweating impairment? Is there thalamic involvement? What is the role of the immune system? One should also analyse what it means for patients and in the care of MS patients. Finally, a challenging topic to investigate is how heat sensitivity can be treated clinically, for example using thermotherapy.", "The authors declare that they have no competing interests.", "All authors contributed to the planning of the study. GF collected all data and, together with A-CE, conducted the statistical analyses. In collaboration, GF, A-CE and A-ML wrote the drafts, which were repeatedly read, discussed and revised by all the authors. Finally, all authors read and approved the final manuscript before its submission.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2377/11/27/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adult", "Female", "Forecasting", "Humans", "Male", "Middle Aged", "Smoking", "Social Class", "Texas", "Time Factors" ]

[ "Background", "Procedures", "Participants", "Measures", "Sociodemographic and Smoking Characteristics", "Subjective Social Status", "Smoking Abstinence", "Data Analysis", "Results", "Sample Characteristics", "Preliminary Analyses", "Primary Analyses", "Moderator Analyses", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Subjective Social Status (SSS) refers to the self-perception of one's status in the social hierarchy [1]. In general, those with greater financial resources typically endorse higher SSS. However, determinants of SSS extend beyond objective socioeconomic status (SES) indicators (such as income, education, and occupational status) to include satisfaction with financial resources, social trust, beliefs about upcoming opportunities, acculturation, and the anticipation of future security [2]. Although highly correlated with SES, several studies have indicated that SSS contributes unique variance in the prediction of self-rated health [3,4], health indicators [5], depression [4], negative affect [6], and health behaviors (e.g., fruit and vegetable consumption [7]). Most recently, research has shown that after adjusting for the effects of SES, higher SSS predicted greater rates of short-term smoking abstinence (i.e., through 2 weeks post-quit) during a smoking quit attempt [8]. Whether SSS predicts longer-term abstinence, however, is yet unknown. As smoking is becoming increasingly concentrated among individuals of low SES, incremental predictors of smoking cessation are useful to identify higher risk subgroups of smokers who may need more intensive interventions to successfully quit smoking.\nSSS may represent an incremental predictor of various health-related outcomes over traditional indicators of SES because it captures the nuances of SES that might affect social standing (e.g., quality of education, prestige associated with a particular workplace, personal control in a particular job), taps into rarely assessed SES components (i.e., wealth), and captures the experience of societal inequalities and inequities [1,2,9,10]. These unique components of SSS may affect health-related outcomes through their associations with psychological distress (e.g., stress, depression) and physiologic dysfunction (in the case of low SSS) or psychological and physiological wellness (in the case of high SSS) [11]. They may also be particularly relevant to racial/ethnic minority groups or other segments of the population (e.g., women) that experience discrimination. Because these unique SSS components and their effects on the mechanisms underlying health-related outcomes may vary based on race/ethnicity and/or sex, the strength of the associations between SSS and health-related outcomes might also vary by race/ethnicity and/or sex. Indeed, racial/ethnic and gender differences have been cited in previous studies investigating associations between SSS and health outcomes [3,4,9]. However, although race/ethnicity and/or sex may moderate some SSS-health relationships, whether race/ethnicity and/or sex affect the relationship between SSS and smoking abstinence during a quit attempt is unknown. For a number of decades, smoking has remained the leading cause of preventable death and disability in the United States (U.S.) [12], and a clearer understanding of the factors predicting smoking cessation among historically understudied and underserved groups (e.g., racial/ethnic minorities, women) may facilitate the targeting of interventions to increase cessation rates, and help to identify groups in need of more intense interventions to quit smoking.\nThe purpose of this study was to examine the independent effect of SSS on long-term smoking abstinence during a specific quit attempt, while controlling for race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked. A complementary aim was to explore whether the relationship between SSS and long-term abstinence differed by race/ethnicity and/or sex. This study represents an extension of our earlier work linking SSS with short-term smoking abstinence (through 2 weeks post-quit) [8] to examine the effects of SSS on long-term smoking abstinence (through 26 weeks post-quit).", "Data for the current study were collected in Houston, Texas, as part of a longitudinal, cohort study designed to examine social disparities in smoking cessation [13],which was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center. Houston is currently the 4th largest city in the U.S. and the most populous of Texas' metropolitan statistical areas. The enrollment period for this study was from April 2005 - April 2007. Recruitment was via local print and radio advertisements. Inclusion criteria included: ≥21 years of age, daily smoker (≥5 cigarettes per day for the last year), motivated to quit in ≤30 days, and ≥6th grade English proficiency. Potential participants were excluded if the nicotine patch was contraindicated, they reported use of tobacco products other than cigarettes, or they reported participation in a smoking cessation program within the past 90 days. Data for the current study were collected one week before quitting (baseline) and post-quit weeks 1, 2, 4, and 26.", "The parent study enrolled 424 racially/ethnically diverse participants. Sample size for the parent study was based on 80% power to detect a .25 standard deviation difference on questionnaire measures between any two racial/ethnic groups at any single point in time. All participants received standard smoking cessation treatment including six weeks of nicotine patch therapy, six brief smoking cessation counseling sessions based on the Treating Tobacco Use and Dependence Clinical Practice Guideline, [14] and self-help materials. Three participants in the parent study failed to answer the SSS item and were excluded from this study.", "All measures in this study were administered and completed via computer.", "Sociodemographic and smoking characteristics were measured at baseline and included race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate (the number of cigarettes smoked per day), and the number of years smoked. Race/ethnicity was self-reported as: non-Latino White, non-Latino Black, or Latino. SES was represented by three variables: annual household income, educational level, and employment status. Categories of income were: <$20,000 a year or ≥$20,000. Categories of education were: < high school education, high school education (or GED), some college (no degree), or college degree. Categories of employment were: employed or not employed. Partner status categories were: married or living with partner versus other.", "Subjective social status was measured at baseline with the SES version of the MacArthur Scale of Subjective Social Status, developed by the John D. and Catherine T. MacArthur Research Network on Socioeconomic Status and Health [15]. This measure presents a 10-rung ladder to the participant with instructions to imagine that it represents where people stand in society, with higher rungs representing higher status (i.e., more money, more education, and better jobs) [1]. Participants are asked to select the rung that best represents where they think they stand relative to others in the U.S., resulting in a continuous variable ranging from 1 to 10.", "Continuous abstinence from smoking was defined as a self-report of abstinence from smoking since the quit date (not even a puff), which was verified by expired carbon monoxide levels of ≤10 ppm. Smoking status was assessed at weeks 1, 2, 4, and 26 post-quit. Because the focus was on continuous abstinence, relapse at any post-quit week resulted in classification as relapsed from that point forward. The percentage of participants with missing smoking status was 18.7% at week 1, 16.1% at week 2, 14.3% at week 4, and 14.5% at week 26. An intention-to-treat procedure was followed, whereby participants with missing smoking status were considered not abstinent (i.e., relapsed).", "All analyses were performed using Statistical Analysis Software (version 9.1). Preliminary analyses explored differences in sociodemographic and smoking characteristics by race/ethnicity and sex using Chi-Square tests for categorical variables and Analyses of Variance for continuous variables. Primary analyses examined the relationship between SSS and long-term abstinence. Because continuous abstinence was the outcome, continuation ratio (CR) logit models (SAS PROC GENMOD; [16-19]) were used to examine the influence of SSS on abstinence across weeks 1, 2, 4, and 26 post-quit. CR logit models are appropriate when ordered categories (e.g., relapsed at week 1, abstinent at week 1 but relapsed at week 2, abstinent at week 2 but relapsed at week 4, abstinent at week 4 but relapsed at week 26, and abstinent through week 26) represent a progression through stages [16-18]. The CR logit models operate by modeling the conditional probability of being abstinent at the current assessment point given that a participant has been abstinent through the most recent assessment point. Analyses adjusted for stage only were followed by analyses adjusted for stage and the following covariates: race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked. These covariates were selected based on their established relationships with smoking relapse (for example, see [20-22]), and included in order to isolate the effect of subjective social status on long-term, continuous abstinence over and above the effects of these variables. Next, additional adjusted CR models were used to examine whether race/ethnicity or sex (respectively) were potential moderators of the effect of SSS on long-term smoking abstinence.", "[SUBTITLE] Sample Characteristics [SUBSECTION] Participants (N = 421, 46% male) were 33% non-Latino White (n = 137), 34% non-Latino Black (n = 144), and 33% Latino (n = 140). They were 41.2 years of age on average (SD = 11.2, median = 42, range 21-73), and 34% reported being married or living with a partner. With regard to SES variables, 38% of respondents reported less than $20,000 in annual household income, 14% lacked a high school diploma or equivalency, and 42% reported current unemployment. At baseline, participants smoked an average of 21.1 (SD = 10.3) cigarettes per day for an average of 21.6 years (SD = 11.1).\nParticipants (N = 421, 46% male) were 33% non-Latino White (n = 137), 34% non-Latino Black (n = 144), and 33% Latino (n = 140). They were 41.2 years of age on average (SD = 11.2, median = 42, range 21-73), and 34% reported being married or living with a partner. With regard to SES variables, 38% of respondents reported less than $20,000 in annual household income, 14% lacked a high school diploma or equivalency, and 42% reported current unemployment. At baseline, participants smoked an average of 21.1 (SD = 10.3) cigarettes per day for an average of 21.6 years (SD = 11.1).\n[SUBTITLE] Preliminary Analyses [SUBSECTION] Racial/ethnic groups were examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 1). Results indicated that White participants smoked more cigarettes per day than both Black and Latino participants, and both Black and White participants smoked for more years than Latinos. Latinos were younger than the Black and White groups in this sample and more likely than the other racial/ethnic groups to be male, employed, and earning $20,000 or more in annual household income. Of all racial/ethnic groups, Black participants had the highest unemployment rates and the highest proportion of individuals earning less than $20,000 annually. There were no significant differences between racial/ethnic groups on SSS.\nParticipants' Characteristics at Baseline by Race/Ethnicity\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Diploma.\nSex groups were also examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 2). Results indicated that women smoked fewer cigarettes per day than men prior to quitting. Notably, women were significantly more likely to endorse an annual household income of less than $20,000 and to be unemployed. Women also endorsed lower SSS than men.\nParticipants' Characteristics at Baseline by Sex\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree.\nRacial/ethnic groups were examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 1). Results indicated that White participants smoked more cigarettes per day than both Black and Latino participants, and both Black and White participants smoked for more years than Latinos. Latinos were younger than the Black and White groups in this sample and more likely than the other racial/ethnic groups to be male, employed, and earning $20,000 or more in annual household income. Of all racial/ethnic groups, Black participants had the highest unemployment rates and the highest proportion of individuals earning less than $20,000 annually. There were no significant differences between racial/ethnic groups on SSS.\nParticipants' Characteristics at Baseline by Race/Ethnicity\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Diploma.\nSex groups were also examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 2). Results indicated that women smoked fewer cigarettes per day than men prior to quitting. Notably, women were significantly more likely to endorse an annual household income of less than $20,000 and to be unemployed. Women also endorsed lower SSS than men.\nParticipants' Characteristics at Baseline by Sex\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree.\n[SUBTITLE] Primary Analyses [SUBSECTION] In unadjusted analyses, SSS predicted long-term smoking abstinence [β = .20, SE = .05; χ2 (1) = 14.83; OR = 1.23 (95% CI: 1.11-1.36); p = .0001]. In analyses controlling for race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked, SSS remained a significant predictor of long-term smoking abstinence [β = .13, SE = .06; χ2 (1) = 4.07; OR = 1.14, 95% CI: 1.00 - 1.28; p = .044; Table 3]. The SSS by stage interaction was not significant, indicating that the effect of SSS on abstinence was consistent across post-quit weeks 1, 2, 4, and 26 (p = .145).\nResults of Unadjusted and Adjusted Models for Subjective Social Status Predicting Smoking Abstinence\nNote: OR = Odds Ratio. CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree. Both unadjusted and adjusted models controlled for stage.\nIn unadjusted analyses, SSS predicted long-term smoking abstinence [β = .20, SE = .05; χ2 (1) = 14.83; OR = 1.23 (95% CI: 1.11-1.36); p = .0001]. In analyses controlling for race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked, SSS remained a significant predictor of long-term smoking abstinence [β = .13, SE = .06; χ2 (1) = 4.07; OR = 1.14, 95% CI: 1.00 - 1.28; p = .044; Table 3]. The SSS by stage interaction was not significant, indicating that the effect of SSS on abstinence was consistent across post-quit weeks 1, 2, 4, and 26 (p = .145).\nResults of Unadjusted and Adjusted Models for Subjective Social Status Predicting Smoking Abstinence\nNote: OR = Odds Ratio. CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree. Both unadjusted and adjusted models controlled for stage.\n[SUBTITLE] Moderator Analyses [SUBSECTION] The relationship between SSS and long-term abstinence was not moderated by race/ethnicity or sex in unadjusted [race/ethnicity: χ2 (2) = 1.94, p = .379; sex: χ2 (1) = 0.28, p = .597] or adjusted [race/ethnicity: χ2 (2) = 0.56, p = .755; sex: χ2 (2) = 0.01, p = .924] analyses.\nThe relationship between SSS and long-term abstinence was not moderated by race/ethnicity or sex in unadjusted [race/ethnicity: χ2 (2) = 1.94, p = .379; sex: χ2 (1) = 0.28, p = .597] or adjusted [race/ethnicity: χ2 (2) = 0.56, p = .755; sex: χ2 (2) = 0.01, p = .924] analyses.", "Participants (N = 421, 46% male) were 33% non-Latino White (n = 137), 34% non-Latino Black (n = 144), and 33% Latino (n = 140). They were 41.2 years of age on average (SD = 11.2, median = 42, range 21-73), and 34% reported being married or living with a partner. With regard to SES variables, 38% of respondents reported less than $20,000 in annual household income, 14% lacked a high school diploma or equivalency, and 42% reported current unemployment. At baseline, participants smoked an average of 21.1 (SD = 10.3) cigarettes per day for an average of 21.6 years (SD = 11.1).", "Racial/ethnic groups were examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 1). Results indicated that White participants smoked more cigarettes per day than both Black and Latino participants, and both Black and White participants smoked for more years than Latinos. Latinos were younger than the Black and White groups in this sample and more likely than the other racial/ethnic groups to be male, employed, and earning $20,000 or more in annual household income. Of all racial/ethnic groups, Black participants had the highest unemployment rates and the highest proportion of individuals earning less than $20,000 annually. There were no significant differences between racial/ethnic groups on SSS.\nParticipants' Characteristics at Baseline by Race/Ethnicity\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Diploma.\nSex groups were also examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 2). Results indicated that women smoked fewer cigarettes per day than men prior to quitting. Notably, women were significantly more likely to endorse an annual household income of less than $20,000 and to be unemployed. Women also endorsed lower SSS than men.\nParticipants' Characteristics at Baseline by Sex\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree.", "In unadjusted analyses, SSS predicted long-term smoking abstinence [β = .20, SE = .05; χ2 (1) = 14.83; OR = 1.23 (95% CI: 1.11-1.36); p = .0001]. In analyses controlling for race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked, SSS remained a significant predictor of long-term smoking abstinence [β = .13, SE = .06; χ2 (1) = 4.07; OR = 1.14, 95% CI: 1.00 - 1.28; p = .044; Table 3]. The SSS by stage interaction was not significant, indicating that the effect of SSS on abstinence was consistent across post-quit weeks 1, 2, 4, and 26 (p = .145).\nResults of Unadjusted and Adjusted Models for Subjective Social Status Predicting Smoking Abstinence\nNote: OR = Odds Ratio. CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree. Both unadjusted and adjusted models controlled for stage.", "The relationship between SSS and long-term abstinence was not moderated by race/ethnicity or sex in unadjusted [race/ethnicity: χ2 (2) = 1.94, p = .379; sex: χ2 (1) = 0.28, p = .597] or adjusted [race/ethnicity: χ2 (2) = 0.56, p = .755; sex: χ2 (2) = 0.01, p = .924] analyses.", "The purpose of this study was to examine whether SSS predicted long-term smoking abstinence (through 26 weeks post-quit) among a racially/ethnically diverse sample of smokers undergoing a specific quit attempt. The results demonstrated that SSS conveys unique predictive information with respect to long-term smoking abstinence, with higher SSS being associated with a greater likelihood of maintaining abstinence over time. While there have been a number of cross-sectional studies linking SSS with self-rated health and health indicators, few studies have used a longitudinal design and even fewer have done so in the examination of the effect of SSS on health behaviors. The current research builds on our previous work demonstrating that SSS predicted short-term smoking abstinence in this racially/ethnically diverse sample [8], and suggests that quitting smokers endorsing lower SSS may face significant hurdles in maintaining both short- and long-term abstinence. These results add to a growing body of literature documenting that SSS contributes to the prediction of self-rated health [3,4], health indicators [4-6], and health behaviors [7] over and above the influence of objective SES indicators.\nA secondary aim of this study was to assess whether race/ethnicity and/or sex moderated the relationship between SSS and long-term smoking abstinence. Although previous studies have supported that relationships between SSS and health-related outcomes are moderated by race/ethnicity and sex [3,4,9], this study failed to support the moderating effects of these variables on the relationship between SSS and long-term smoking abstinence. One possible explanation for this finding is that the unique factors that SSS captures over SES are common to all population groups examined, and therefore affected smoking abstinence equally among these groups. However, this may be unlikely given prior research indicating that determinants of SSS differ by race/ethnicity and sex [3,4,9] and, in some cases, may be relevant only to certain racial/ethnic groups (e.g., acculturation [23]). Therefore, another explanation might be that the unique components of SSS, although varying in significance and weighted differently between race/ethnicities and the sexes, have a similar effect on the mechanisms underlying smoking cessation. Many studies have confirmed the role of depression, negative affect, and stress in smoking relapse (e.g., [24-27]), and it may be that these are the mechanisms through which SSS uniquely predicts smoking abstinence among all racial/ethnic and sex groups. This supposition is partially supported by previous research finding that depression mediated the relationship between SSS and short-term smoking abstinence [8]. However, a post-hoc analysis indicated that further adjusting our models for baseline depressive symptoms as measured by the Center for Epidemiologic Studies Depression Scale [28] did not alter the pattern of results. A better understanding of the mechanisms underlying the relationship between SSS and smoking cessation, and how these may vary over the course of the quit attempt, is needed.\nStrengths of this study include the longitudinal design and the use of a racially/ethnically diverse sample of participants. Because smoking is the leading cause of morbidity and mortality in the U.S. and is a major contributor to health disparities [12], enhancing understanding of the predictors of smoking abstinence is vital to the development of interventions and the targeting of treatment to affected groups. Results of the current study clearly demonstrate that smokers endorsing lower SSS are at increased risk of relapse during a smoking quit attempt as compared with smokers endorsing higher SSS scores. Results also suggest that individuals endorsing lower SSS might benefit from targeted interventions to help facilitate smoking cessation. Whether such targeted interventions might include a more intense (e.g., greater coverage of issues) or higher dosage (e.g., increased number of sessions) intervention for smoking cessation than standard treatment is fodder for future research. However, the standard smoking cessation intervention provided to all participants in this study (i.e., brief counseling, self-help materials, and the \"patch\") was not sufficient to enable cessation among lower SSS smokers in this study. As the SSS scale is a relatively quick and easy tool to administer [15], smoking cessation programs could use it as a screener for smokers who may require additional attention to facilitate abstinence, although more information is needed on what might constitute a meaningful SSS cut-off for additional service provision. Future research should explore how to better treat lower SSS smokers, which would benefit from a greater understanding of the mechanisms driving the relationship between SSS and smoking abstinence.\nAnother potential application of this finding could entail the development of interventions designed to facilitate smoking cessation while also affecting the perceived social status of lower SSS individuals. This is a completely unexplored area but might entail building social trust [2,10] by encouraging involvement in community or church activities, enhancing social standing through the adoption of community leadership roles (e.g., homeowners association board, neighborhood watch), or facilitating hope for the future by assisting in the identification and utilization of local resources (e.g., job placement agencies, food banks, reduced-rate health care options). It is unknown whether such interventions might affect SSS and facilitate smoking cessation, but these possibilities could be examined in future research.\nLimitations of the current study include the potential for limited power to detect interaction effects, as the parent study was not specifically designed to explore moderators of SSS effects on long-term cessation. Future studies should replicate these analyses with a more purposeful design. Future studies in this area might also adjust for other SES variables not collected (e.g., wealth) in the present study in order to further isolate the unique effects of SSS on health indicators and outcomes. Also, participants in this study were self-selected, treatment-seeking smokers who may differ from smokers who attempt to quit without treatment in important ways, and the influence of SSS on cessation among the latter group remains unknown. It is also worthy of note that this study relied on self-report of racial/ethnic group membership, which may not perfectly align with biologically-based classifications of race/ethnicity. Also, the parent study was designed to recruit Whites, Blacks and Latino smokers in equal proportions, which may not be representative of the local population in terms of racial/ethnic distribution or other factors. Thus, results may not generalize to the larger population of smokers in Houston. Importantly, our interpretation of the data assumes that the unique predictive ability of SSS over SES indicators is not based on measurement error. Although these assumptions have been supported in previous literature (e.g., [1,29]), we cannot rule out that the presence of unknown and unmeasured confounders might have influenced these results.", "This study examined whether SSS predicted long-term abstinence from smoking during a specific quit attempt, and assessed potential moderation by racial/ethnic group or sex. Our results indicated that smokers endorsing higher SSS were more likely to maintain long-term smoking abstinence during a specific quit attempt than smokers with lower social status ratings, regardless of their race/ethnicity or sex. Although a number of studies have supported inverse relationships between SES indicators (e.g., income, education) and smoking cessation (cf. [13]), the current study indicates that SSS is an even stronger predictor. SSS, the self-reported perception of relative standing in the social hierarchy, captures SES but also taps into the experience of societal inequities and a person's feelings of hope about future status change [1,10]. It may be these, or other, unique components of SSS that are especially important influences on health and health behaviors. More research is needed to understand the mechanisms underlying the relationship between SSS and long-term smoking abstinence, and to appropriately tailor treatment to facilitate abstinence among smokers endorsing lower SSS.", "The authors do not have any competing interests directly pertaining to this work; however, we would like to report that PM Cinciripini has served on the scientific advisory board of Pfizer Pharmaceuticals and has conducted educational talks for physicians on smoking cessation that were sponsored by Pfizer within the past five years.", "LRR, MSB, DEK, YCastro, and DWW conceptualized the research question and wrote the manuscript. YL and YCao conducted the data analysis, interpreted results, and reviewed the data analysis and results sections. DWW is the senior author of the paper, and the principal investigator on the supporting grant (National Institute of Drug Abuse; R01DA014818). CAM, LC-W, and PMC helped with the conceptualization of the overall project and methodology, and reviewed and edited manuscript drafts. PMC is a co-investigator on the supporting grant. All authors have read, contributed to, and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2458/11/135/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Procedures", "Participants", "Measures", "Sociodemographic and Smoking Characteristics", "Subjective Social Status", "Smoking Abstinence", "Data Analysis", "Results", "Sample Characteristics", "Preliminary Analyses", "Primary Analyses", "Moderator Analyses", "Discussion", "Conclusions", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Subjective Social Status (SSS) refers to the self-perception of one's status in the social hierarchy [1]. In general, those with greater financial resources typically endorse higher SSS. However, determinants of SSS extend beyond objective socioeconomic status (SES) indicators (such as income, education, and occupational status) to include satisfaction with financial resources, social trust, beliefs about upcoming opportunities, acculturation, and the anticipation of future security [2]. Although highly correlated with SES, several studies have indicated that SSS contributes unique variance in the prediction of self-rated health [3,4], health indicators [5], depression [4], negative affect [6], and health behaviors (e.g., fruit and vegetable consumption [7]). Most recently, research has shown that after adjusting for the effects of SES, higher SSS predicted greater rates of short-term smoking abstinence (i.e., through 2 weeks post-quit) during a smoking quit attempt [8]. Whether SSS predicts longer-term abstinence, however, is yet unknown. As smoking is becoming increasingly concentrated among individuals of low SES, incremental predictors of smoking cessation are useful to identify higher risk subgroups of smokers who may need more intensive interventions to successfully quit smoking.\nSSS may represent an incremental predictor of various health-related outcomes over traditional indicators of SES because it captures the nuances of SES that might affect social standing (e.g., quality of education, prestige associated with a particular workplace, personal control in a particular job), taps into rarely assessed SES components (i.e., wealth), and captures the experience of societal inequalities and inequities [1,2,9,10]. These unique components of SSS may affect health-related outcomes through their associations with psychological distress (e.g., stress, depression) and physiologic dysfunction (in the case of low SSS) or psychological and physiological wellness (in the case of high SSS) [11]. They may also be particularly relevant to racial/ethnic minority groups or other segments of the population (e.g., women) that experience discrimination. Because these unique SSS components and their effects on the mechanisms underlying health-related outcomes may vary based on race/ethnicity and/or sex, the strength of the associations between SSS and health-related outcomes might also vary by race/ethnicity and/or sex. Indeed, racial/ethnic and gender differences have been cited in previous studies investigating associations between SSS and health outcomes [3,4,9]. However, although race/ethnicity and/or sex may moderate some SSS-health relationships, whether race/ethnicity and/or sex affect the relationship between SSS and smoking abstinence during a quit attempt is unknown. For a number of decades, smoking has remained the leading cause of preventable death and disability in the United States (U.S.) [12], and a clearer understanding of the factors predicting smoking cessation among historically understudied and underserved groups (e.g., racial/ethnic minorities, women) may facilitate the targeting of interventions to increase cessation rates, and help to identify groups in need of more intense interventions to quit smoking.\nThe purpose of this study was to examine the independent effect of SSS on long-term smoking abstinence during a specific quit attempt, while controlling for race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked. A complementary aim was to explore whether the relationship between SSS and long-term abstinence differed by race/ethnicity and/or sex. This study represents an extension of our earlier work linking SSS with short-term smoking abstinence (through 2 weeks post-quit) [8] to examine the effects of SSS on long-term smoking abstinence (through 26 weeks post-quit).", "[SUBTITLE] Procedures [SUBSECTION] Data for the current study were collected in Houston, Texas, as part of a longitudinal, cohort study designed to examine social disparities in smoking cessation [13],which was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center. Houston is currently the 4th largest city in the U.S. and the most populous of Texas' metropolitan statistical areas. The enrollment period for this study was from April 2005 - April 2007. Recruitment was via local print and radio advertisements. Inclusion criteria included: ≥21 years of age, daily smoker (≥5 cigarettes per day for the last year), motivated to quit in ≤30 days, and ≥6th grade English proficiency. Potential participants were excluded if the nicotine patch was contraindicated, they reported use of tobacco products other than cigarettes, or they reported participation in a smoking cessation program within the past 90 days. Data for the current study were collected one week before quitting (baseline) and post-quit weeks 1, 2, 4, and 26.\nData for the current study were collected in Houston, Texas, as part of a longitudinal, cohort study designed to examine social disparities in smoking cessation [13],which was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center. Houston is currently the 4th largest city in the U.S. and the most populous of Texas' metropolitan statistical areas. The enrollment period for this study was from April 2005 - April 2007. Recruitment was via local print and radio advertisements. Inclusion criteria included: ≥21 years of age, daily smoker (≥5 cigarettes per day for the last year), motivated to quit in ≤30 days, and ≥6th grade English proficiency. Potential participants were excluded if the nicotine patch was contraindicated, they reported use of tobacco products other than cigarettes, or they reported participation in a smoking cessation program within the past 90 days. Data for the current study were collected one week before quitting (baseline) and post-quit weeks 1, 2, 4, and 26.\n[SUBTITLE] Participants [SUBSECTION] The parent study enrolled 424 racially/ethnically diverse participants. Sample size for the parent study was based on 80% power to detect a .25 standard deviation difference on questionnaire measures between any two racial/ethnic groups at any single point in time. All participants received standard smoking cessation treatment including six weeks of nicotine patch therapy, six brief smoking cessation counseling sessions based on the Treating Tobacco Use and Dependence Clinical Practice Guideline, [14] and self-help materials. Three participants in the parent study failed to answer the SSS item and were excluded from this study.\nThe parent study enrolled 424 racially/ethnically diverse participants. Sample size for the parent study was based on 80% power to detect a .25 standard deviation difference on questionnaire measures between any two racial/ethnic groups at any single point in time. All participants received standard smoking cessation treatment including six weeks of nicotine patch therapy, six brief smoking cessation counseling sessions based on the Treating Tobacco Use and Dependence Clinical Practice Guideline, [14] and self-help materials. Three participants in the parent study failed to answer the SSS item and were excluded from this study.\n[SUBTITLE] Measures [SUBSECTION] All measures in this study were administered and completed via computer.\nAll measures in this study were administered and completed via computer.\n[SUBTITLE] Sociodemographic and Smoking Characteristics [SUBSECTION] Sociodemographic and smoking characteristics were measured at baseline and included race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate (the number of cigarettes smoked per day), and the number of years smoked. Race/ethnicity was self-reported as: non-Latino White, non-Latino Black, or Latino. SES was represented by three variables: annual household income, educational level, and employment status. Categories of income were: <$20,000 a year or ≥$20,000. Categories of education were: < high school education, high school education (or GED), some college (no degree), or college degree. Categories of employment were: employed or not employed. Partner status categories were: married or living with partner versus other.\nSociodemographic and smoking characteristics were measured at baseline and included race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate (the number of cigarettes smoked per day), and the number of years smoked. Race/ethnicity was self-reported as: non-Latino White, non-Latino Black, or Latino. SES was represented by three variables: annual household income, educational level, and employment status. Categories of income were: <$20,000 a year or ≥$20,000. Categories of education were: < high school education, high school education (or GED), some college (no degree), or college degree. Categories of employment were: employed or not employed. Partner status categories were: married or living with partner versus other.\n[SUBTITLE] Subjective Social Status [SUBSECTION] Subjective social status was measured at baseline with the SES version of the MacArthur Scale of Subjective Social Status, developed by the John D. and Catherine T. MacArthur Research Network on Socioeconomic Status and Health [15]. This measure presents a 10-rung ladder to the participant with instructions to imagine that it represents where people stand in society, with higher rungs representing higher status (i.e., more money, more education, and better jobs) [1]. Participants are asked to select the rung that best represents where they think they stand relative to others in the U.S., resulting in a continuous variable ranging from 1 to 10.\nSubjective social status was measured at baseline with the SES version of the MacArthur Scale of Subjective Social Status, developed by the John D. and Catherine T. MacArthur Research Network on Socioeconomic Status and Health [15]. This measure presents a 10-rung ladder to the participant with instructions to imagine that it represents where people stand in society, with higher rungs representing higher status (i.e., more money, more education, and better jobs) [1]. Participants are asked to select the rung that best represents where they think they stand relative to others in the U.S., resulting in a continuous variable ranging from 1 to 10.\n[SUBTITLE] Smoking Abstinence [SUBSECTION] Continuous abstinence from smoking was defined as a self-report of abstinence from smoking since the quit date (not even a puff), which was verified by expired carbon monoxide levels of ≤10 ppm. Smoking status was assessed at weeks 1, 2, 4, and 26 post-quit. Because the focus was on continuous abstinence, relapse at any post-quit week resulted in classification as relapsed from that point forward. The percentage of participants with missing smoking status was 18.7% at week 1, 16.1% at week 2, 14.3% at week 4, and 14.5% at week 26. An intention-to-treat procedure was followed, whereby participants with missing smoking status were considered not abstinent (i.e., relapsed).\nContinuous abstinence from smoking was defined as a self-report of abstinence from smoking since the quit date (not even a puff), which was verified by expired carbon monoxide levels of ≤10 ppm. Smoking status was assessed at weeks 1, 2, 4, and 26 post-quit. Because the focus was on continuous abstinence, relapse at any post-quit week resulted in classification as relapsed from that point forward. The percentage of participants with missing smoking status was 18.7% at week 1, 16.1% at week 2, 14.3% at week 4, and 14.5% at week 26. An intention-to-treat procedure was followed, whereby participants with missing smoking status were considered not abstinent (i.e., relapsed).\n[SUBTITLE] Data Analysis [SUBSECTION] All analyses were performed using Statistical Analysis Software (version 9.1). Preliminary analyses explored differences in sociodemographic and smoking characteristics by race/ethnicity and sex using Chi-Square tests for categorical variables and Analyses of Variance for continuous variables. Primary analyses examined the relationship between SSS and long-term abstinence. Because continuous abstinence was the outcome, continuation ratio (CR) logit models (SAS PROC GENMOD; [16-19]) were used to examine the influence of SSS on abstinence across weeks 1, 2, 4, and 26 post-quit. CR logit models are appropriate when ordered categories (e.g., relapsed at week 1, abstinent at week 1 but relapsed at week 2, abstinent at week 2 but relapsed at week 4, abstinent at week 4 but relapsed at week 26, and abstinent through week 26) represent a progression through stages [16-18]. The CR logit models operate by modeling the conditional probability of being abstinent at the current assessment point given that a participant has been abstinent through the most recent assessment point. Analyses adjusted for stage only were followed by analyses adjusted for stage and the following covariates: race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked. These covariates were selected based on their established relationships with smoking relapse (for example, see [20-22]), and included in order to isolate the effect of subjective social status on long-term, continuous abstinence over and above the effects of these variables. Next, additional adjusted CR models were used to examine whether race/ethnicity or sex (respectively) were potential moderators of the effect of SSS on long-term smoking abstinence.\nAll analyses were performed using Statistical Analysis Software (version 9.1). Preliminary analyses explored differences in sociodemographic and smoking characteristics by race/ethnicity and sex using Chi-Square tests for categorical variables and Analyses of Variance for continuous variables. Primary analyses examined the relationship between SSS and long-term abstinence. Because continuous abstinence was the outcome, continuation ratio (CR) logit models (SAS PROC GENMOD; [16-19]) were used to examine the influence of SSS on abstinence across weeks 1, 2, 4, and 26 post-quit. CR logit models are appropriate when ordered categories (e.g., relapsed at week 1, abstinent at week 1 but relapsed at week 2, abstinent at week 2 but relapsed at week 4, abstinent at week 4 but relapsed at week 26, and abstinent through week 26) represent a progression through stages [16-18]. The CR logit models operate by modeling the conditional probability of being abstinent at the current assessment point given that a participant has been abstinent through the most recent assessment point. Analyses adjusted for stage only were followed by analyses adjusted for stage and the following covariates: race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked. These covariates were selected based on their established relationships with smoking relapse (for example, see [20-22]), and included in order to isolate the effect of subjective social status on long-term, continuous abstinence over and above the effects of these variables. Next, additional adjusted CR models were used to examine whether race/ethnicity or sex (respectively) were potential moderators of the effect of SSS on long-term smoking abstinence.", "Data for the current study were collected in Houston, Texas, as part of a longitudinal, cohort study designed to examine social disparities in smoking cessation [13],which was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center. Houston is currently the 4th largest city in the U.S. and the most populous of Texas' metropolitan statistical areas. The enrollment period for this study was from April 2005 - April 2007. Recruitment was via local print and radio advertisements. Inclusion criteria included: ≥21 years of age, daily smoker (≥5 cigarettes per day for the last year), motivated to quit in ≤30 days, and ≥6th grade English proficiency. Potential participants were excluded if the nicotine patch was contraindicated, they reported use of tobacco products other than cigarettes, or they reported participation in a smoking cessation program within the past 90 days. Data for the current study were collected one week before quitting (baseline) and post-quit weeks 1, 2, 4, and 26.", "The parent study enrolled 424 racially/ethnically diverse participants. Sample size for the parent study was based on 80% power to detect a .25 standard deviation difference on questionnaire measures between any two racial/ethnic groups at any single point in time. All participants received standard smoking cessation treatment including six weeks of nicotine patch therapy, six brief smoking cessation counseling sessions based on the Treating Tobacco Use and Dependence Clinical Practice Guideline, [14] and self-help materials. Three participants in the parent study failed to answer the SSS item and were excluded from this study.", "All measures in this study were administered and completed via computer.", "Sociodemographic and smoking characteristics were measured at baseline and included race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate (the number of cigarettes smoked per day), and the number of years smoked. Race/ethnicity was self-reported as: non-Latino White, non-Latino Black, or Latino. SES was represented by three variables: annual household income, educational level, and employment status. Categories of income were: <$20,000 a year or ≥$20,000. Categories of education were: < high school education, high school education (or GED), some college (no degree), or college degree. Categories of employment were: employed or not employed. Partner status categories were: married or living with partner versus other.", "Subjective social status was measured at baseline with the SES version of the MacArthur Scale of Subjective Social Status, developed by the John D. and Catherine T. MacArthur Research Network on Socioeconomic Status and Health [15]. This measure presents a 10-rung ladder to the participant with instructions to imagine that it represents where people stand in society, with higher rungs representing higher status (i.e., more money, more education, and better jobs) [1]. Participants are asked to select the rung that best represents where they think they stand relative to others in the U.S., resulting in a continuous variable ranging from 1 to 10.", "Continuous abstinence from smoking was defined as a self-report of abstinence from smoking since the quit date (not even a puff), which was verified by expired carbon monoxide levels of ≤10 ppm. Smoking status was assessed at weeks 1, 2, 4, and 26 post-quit. Because the focus was on continuous abstinence, relapse at any post-quit week resulted in classification as relapsed from that point forward. The percentage of participants with missing smoking status was 18.7% at week 1, 16.1% at week 2, 14.3% at week 4, and 14.5% at week 26. An intention-to-treat procedure was followed, whereby participants with missing smoking status were considered not abstinent (i.e., relapsed).", "All analyses were performed using Statistical Analysis Software (version 9.1). Preliminary analyses explored differences in sociodemographic and smoking characteristics by race/ethnicity and sex using Chi-Square tests for categorical variables and Analyses of Variance for continuous variables. Primary analyses examined the relationship between SSS and long-term abstinence. Because continuous abstinence was the outcome, continuation ratio (CR) logit models (SAS PROC GENMOD; [16-19]) were used to examine the influence of SSS on abstinence across weeks 1, 2, 4, and 26 post-quit. CR logit models are appropriate when ordered categories (e.g., relapsed at week 1, abstinent at week 1 but relapsed at week 2, abstinent at week 2 but relapsed at week 4, abstinent at week 4 but relapsed at week 26, and abstinent through week 26) represent a progression through stages [16-18]. The CR logit models operate by modeling the conditional probability of being abstinent at the current assessment point given that a participant has been abstinent through the most recent assessment point. Analyses adjusted for stage only were followed by analyses adjusted for stage and the following covariates: race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked. These covariates were selected based on their established relationships with smoking relapse (for example, see [20-22]), and included in order to isolate the effect of subjective social status on long-term, continuous abstinence over and above the effects of these variables. Next, additional adjusted CR models were used to examine whether race/ethnicity or sex (respectively) were potential moderators of the effect of SSS on long-term smoking abstinence.", "[SUBTITLE] Sample Characteristics [SUBSECTION] Participants (N = 421, 46% male) were 33% non-Latino White (n = 137), 34% non-Latino Black (n = 144), and 33% Latino (n = 140). They were 41.2 years of age on average (SD = 11.2, median = 42, range 21-73), and 34% reported being married or living with a partner. With regard to SES variables, 38% of respondents reported less than $20,000 in annual household income, 14% lacked a high school diploma or equivalency, and 42% reported current unemployment. At baseline, participants smoked an average of 21.1 (SD = 10.3) cigarettes per day for an average of 21.6 years (SD = 11.1).\nParticipants (N = 421, 46% male) were 33% non-Latino White (n = 137), 34% non-Latino Black (n = 144), and 33% Latino (n = 140). They were 41.2 years of age on average (SD = 11.2, median = 42, range 21-73), and 34% reported being married or living with a partner. With regard to SES variables, 38% of respondents reported less than $20,000 in annual household income, 14% lacked a high school diploma or equivalency, and 42% reported current unemployment. At baseline, participants smoked an average of 21.1 (SD = 10.3) cigarettes per day for an average of 21.6 years (SD = 11.1).\n[SUBTITLE] Preliminary Analyses [SUBSECTION] Racial/ethnic groups were examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 1). Results indicated that White participants smoked more cigarettes per day than both Black and Latino participants, and both Black and White participants smoked for more years than Latinos. Latinos were younger than the Black and White groups in this sample and more likely than the other racial/ethnic groups to be male, employed, and earning $20,000 or more in annual household income. Of all racial/ethnic groups, Black participants had the highest unemployment rates and the highest proportion of individuals earning less than $20,000 annually. There were no significant differences between racial/ethnic groups on SSS.\nParticipants' Characteristics at Baseline by Race/Ethnicity\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Diploma.\nSex groups were also examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 2). Results indicated that women smoked fewer cigarettes per day than men prior to quitting. Notably, women were significantly more likely to endorse an annual household income of less than $20,000 and to be unemployed. Women also endorsed lower SSS than men.\nParticipants' Characteristics at Baseline by Sex\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree.\nRacial/ethnic groups were examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 1). Results indicated that White participants smoked more cigarettes per day than both Black and Latino participants, and both Black and White participants smoked for more years than Latinos. Latinos were younger than the Black and White groups in this sample and more likely than the other racial/ethnic groups to be male, employed, and earning $20,000 or more in annual household income. Of all racial/ethnic groups, Black participants had the highest unemployment rates and the highest proportion of individuals earning less than $20,000 annually. There were no significant differences between racial/ethnic groups on SSS.\nParticipants' Characteristics at Baseline by Race/Ethnicity\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Diploma.\nSex groups were also examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 2). Results indicated that women smoked fewer cigarettes per day than men prior to quitting. Notably, women were significantly more likely to endorse an annual household income of less than $20,000 and to be unemployed. Women also endorsed lower SSS than men.\nParticipants' Characteristics at Baseline by Sex\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree.\n[SUBTITLE] Primary Analyses [SUBSECTION] In unadjusted analyses, SSS predicted long-term smoking abstinence [β = .20, SE = .05; χ2 (1) = 14.83; OR = 1.23 (95% CI: 1.11-1.36); p = .0001]. In analyses controlling for race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked, SSS remained a significant predictor of long-term smoking abstinence [β = .13, SE = .06; χ2 (1) = 4.07; OR = 1.14, 95% CI: 1.00 - 1.28; p = .044; Table 3]. The SSS by stage interaction was not significant, indicating that the effect of SSS on abstinence was consistent across post-quit weeks 1, 2, 4, and 26 (p = .145).\nResults of Unadjusted and Adjusted Models for Subjective Social Status Predicting Smoking Abstinence\nNote: OR = Odds Ratio. CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree. Both unadjusted and adjusted models controlled for stage.\nIn unadjusted analyses, SSS predicted long-term smoking abstinence [β = .20, SE = .05; χ2 (1) = 14.83; OR = 1.23 (95% CI: 1.11-1.36); p = .0001]. In analyses controlling for race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked, SSS remained a significant predictor of long-term smoking abstinence [β = .13, SE = .06; χ2 (1) = 4.07; OR = 1.14, 95% CI: 1.00 - 1.28; p = .044; Table 3]. The SSS by stage interaction was not significant, indicating that the effect of SSS on abstinence was consistent across post-quit weeks 1, 2, 4, and 26 (p = .145).\nResults of Unadjusted and Adjusted Models for Subjective Social Status Predicting Smoking Abstinence\nNote: OR = Odds Ratio. CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree. Both unadjusted and adjusted models controlled for stage.\n[SUBTITLE] Moderator Analyses [SUBSECTION] The relationship between SSS and long-term abstinence was not moderated by race/ethnicity or sex in unadjusted [race/ethnicity: χ2 (2) = 1.94, p = .379; sex: χ2 (1) = 0.28, p = .597] or adjusted [race/ethnicity: χ2 (2) = 0.56, p = .755; sex: χ2 (2) = 0.01, p = .924] analyses.\nThe relationship between SSS and long-term abstinence was not moderated by race/ethnicity or sex in unadjusted [race/ethnicity: χ2 (2) = 1.94, p = .379; sex: χ2 (1) = 0.28, p = .597] or adjusted [race/ethnicity: χ2 (2) = 0.56, p = .755; sex: χ2 (2) = 0.01, p = .924] analyses.", "Participants (N = 421, 46% male) were 33% non-Latino White (n = 137), 34% non-Latino Black (n = 144), and 33% Latino (n = 140). They were 41.2 years of age on average (SD = 11.2, median = 42, range 21-73), and 34% reported being married or living with a partner. With regard to SES variables, 38% of respondents reported less than $20,000 in annual household income, 14% lacked a high school diploma or equivalency, and 42% reported current unemployment. At baseline, participants smoked an average of 21.1 (SD = 10.3) cigarettes per day for an average of 21.6 years (SD = 11.1).", "Racial/ethnic groups were examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 1). Results indicated that White participants smoked more cigarettes per day than both Black and Latino participants, and both Black and White participants smoked for more years than Latinos. Latinos were younger than the Black and White groups in this sample and more likely than the other racial/ethnic groups to be male, employed, and earning $20,000 or more in annual household income. Of all racial/ethnic groups, Black participants had the highest unemployment rates and the highest proportion of individuals earning less than $20,000 annually. There were no significant differences between racial/ethnic groups on SSS.\nParticipants' Characteristics at Baseline by Race/Ethnicity\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Diploma.\nSex groups were also examined for significant differences on sociodemographics and smoking characteristics at baseline (Table 2). Results indicated that women smoked fewer cigarettes per day than men prior to quitting. Notably, women were significantly more likely to endorse an annual household income of less than $20,000 and to be unemployed. Women also endorsed lower SSS than men.\nParticipants' Characteristics at Baseline by Sex\nNote: CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree.", "In unadjusted analyses, SSS predicted long-term smoking abstinence [β = .20, SE = .05; χ2 (1) = 14.83; OR = 1.23 (95% CI: 1.11-1.36); p = .0001]. In analyses controlling for race/ethnicity, sex, SES, age, partner status, pre-quit smoking rate, and the number of years smoked, SSS remained a significant predictor of long-term smoking abstinence [β = .13, SE = .06; χ2 (1) = 4.07; OR = 1.14, 95% CI: 1.00 - 1.28; p = .044; Table 3]. The SSS by stage interaction was not significant, indicating that the effect of SSS on abstinence was consistent across post-quit weeks 1, 2, 4, and 26 (p = .145).\nResults of Unadjusted and Adjusted Models for Subjective Social Status Predicting Smoking Abstinence\nNote: OR = Odds Ratio. CPD = Self-reported number of cigarettes smoked per day at baseline. GED = General Equivalency Degree. Both unadjusted and adjusted models controlled for stage.", "The relationship between SSS and long-term abstinence was not moderated by race/ethnicity or sex in unadjusted [race/ethnicity: χ2 (2) = 1.94, p = .379; sex: χ2 (1) = 0.28, p = .597] or adjusted [race/ethnicity: χ2 (2) = 0.56, p = .755; sex: χ2 (2) = 0.01, p = .924] analyses.", "The purpose of this study was to examine whether SSS predicted long-term smoking abstinence (through 26 weeks post-quit) among a racially/ethnically diverse sample of smokers undergoing a specific quit attempt. The results demonstrated that SSS conveys unique predictive information with respect to long-term smoking abstinence, with higher SSS being associated with a greater likelihood of maintaining abstinence over time. While there have been a number of cross-sectional studies linking SSS with self-rated health and health indicators, few studies have used a longitudinal design and even fewer have done so in the examination of the effect of SSS on health behaviors. The current research builds on our previous work demonstrating that SSS predicted short-term smoking abstinence in this racially/ethnically diverse sample [8], and suggests that quitting smokers endorsing lower SSS may face significant hurdles in maintaining both short- and long-term abstinence. These results add to a growing body of literature documenting that SSS contributes to the prediction of self-rated health [3,4], health indicators [4-6], and health behaviors [7] over and above the influence of objective SES indicators.\nA secondary aim of this study was to assess whether race/ethnicity and/or sex moderated the relationship between SSS and long-term smoking abstinence. Although previous studies have supported that relationships between SSS and health-related outcomes are moderated by race/ethnicity and sex [3,4,9], this study failed to support the moderating effects of these variables on the relationship between SSS and long-term smoking abstinence. One possible explanation for this finding is that the unique factors that SSS captures over SES are common to all population groups examined, and therefore affected smoking abstinence equally among these groups. However, this may be unlikely given prior research indicating that determinants of SSS differ by race/ethnicity and sex [3,4,9] and, in some cases, may be relevant only to certain racial/ethnic groups (e.g., acculturation [23]). Therefore, another explanation might be that the unique components of SSS, although varying in significance and weighted differently between race/ethnicities and the sexes, have a similar effect on the mechanisms underlying smoking cessation. Many studies have confirmed the role of depression, negative affect, and stress in smoking relapse (e.g., [24-27]), and it may be that these are the mechanisms through which SSS uniquely predicts smoking abstinence among all racial/ethnic and sex groups. This supposition is partially supported by previous research finding that depression mediated the relationship between SSS and short-term smoking abstinence [8]. However, a post-hoc analysis indicated that further adjusting our models for baseline depressive symptoms as measured by the Center for Epidemiologic Studies Depression Scale [28] did not alter the pattern of results. A better understanding of the mechanisms underlying the relationship between SSS and smoking cessation, and how these may vary over the course of the quit attempt, is needed.\nStrengths of this study include the longitudinal design and the use of a racially/ethnically diverse sample of participants. Because smoking is the leading cause of morbidity and mortality in the U.S. and is a major contributor to health disparities [12], enhancing understanding of the predictors of smoking abstinence is vital to the development of interventions and the targeting of treatment to affected groups. Results of the current study clearly demonstrate that smokers endorsing lower SSS are at increased risk of relapse during a smoking quit attempt as compared with smokers endorsing higher SSS scores. Results also suggest that individuals endorsing lower SSS might benefit from targeted interventions to help facilitate smoking cessation. Whether such targeted interventions might include a more intense (e.g., greater coverage of issues) or higher dosage (e.g., increased number of sessions) intervention for smoking cessation than standard treatment is fodder for future research. However, the standard smoking cessation intervention provided to all participants in this study (i.e., brief counseling, self-help materials, and the \"patch\") was not sufficient to enable cessation among lower SSS smokers in this study. As the SSS scale is a relatively quick and easy tool to administer [15], smoking cessation programs could use it as a screener for smokers who may require additional attention to facilitate abstinence, although more information is needed on what might constitute a meaningful SSS cut-off for additional service provision. Future research should explore how to better treat lower SSS smokers, which would benefit from a greater understanding of the mechanisms driving the relationship between SSS and smoking abstinence.\nAnother potential application of this finding could entail the development of interventions designed to facilitate smoking cessation while also affecting the perceived social status of lower SSS individuals. This is a completely unexplored area but might entail building social trust [2,10] by encouraging involvement in community or church activities, enhancing social standing through the adoption of community leadership roles (e.g., homeowners association board, neighborhood watch), or facilitating hope for the future by assisting in the identification and utilization of local resources (e.g., job placement agencies, food banks, reduced-rate health care options). It is unknown whether such interventions might affect SSS and facilitate smoking cessation, but these possibilities could be examined in future research.\nLimitations of the current study include the potential for limited power to detect interaction effects, as the parent study was not specifically designed to explore moderators of SSS effects on long-term cessation. Future studies should replicate these analyses with a more purposeful design. Future studies in this area might also adjust for other SES variables not collected (e.g., wealth) in the present study in order to further isolate the unique effects of SSS on health indicators and outcomes. Also, participants in this study were self-selected, treatment-seeking smokers who may differ from smokers who attempt to quit without treatment in important ways, and the influence of SSS on cessation among the latter group remains unknown. It is also worthy of note that this study relied on self-report of racial/ethnic group membership, which may not perfectly align with biologically-based classifications of race/ethnicity. Also, the parent study was designed to recruit Whites, Blacks and Latino smokers in equal proportions, which may not be representative of the local population in terms of racial/ethnic distribution or other factors. Thus, results may not generalize to the larger population of smokers in Houston. Importantly, our interpretation of the data assumes that the unique predictive ability of SSS over SES indicators is not based on measurement error. Although these assumptions have been supported in previous literature (e.g., [1,29]), we cannot rule out that the presence of unknown and unmeasured confounders might have influenced these results.", "This study examined whether SSS predicted long-term abstinence from smoking during a specific quit attempt, and assessed potential moderation by racial/ethnic group or sex. Our results indicated that smokers endorsing higher SSS were more likely to maintain long-term smoking abstinence during a specific quit attempt than smokers with lower social status ratings, regardless of their race/ethnicity or sex. Although a number of studies have supported inverse relationships between SES indicators (e.g., income, education) and smoking cessation (cf. [13]), the current study indicates that SSS is an even stronger predictor. SSS, the self-reported perception of relative standing in the social hierarchy, captures SES but also taps into the experience of societal inequities and a person's feelings of hope about future status change [1,10]. It may be these, or other, unique components of SSS that are especially important influences on health and health behaviors. More research is needed to understand the mechanisms underlying the relationship between SSS and long-term smoking abstinence, and to appropriately tailor treatment to facilitate abstinence among smokers endorsing lower SSS.", "The authors do not have any competing interests directly pertaining to this work; however, we would like to report that PM Cinciripini has served on the scientific advisory board of Pfizer Pharmaceuticals and has conducted educational talks for physicians on smoking cessation that were sponsored by Pfizer within the past five years.", "LRR, MSB, DEK, YCastro, and DWW conceptualized the research question and wrote the manuscript. YL and YCao conducted the data analysis, interpreted results, and reviewed the data analysis and results sections. DWW is the senior author of the paper, and the principal investigator on the supporting grant (National Institute of Drug Abuse; R01DA014818). CAM, LC-W, and PMC helped with the conceptualization of the overall project and methodology, and reviewed and edited manuscript drafts. PMC is a co-investigator on the supporting grant. All authors have read, contributed to, and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2458/11/135/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Adolescent", "Anti-Mullerian Hormone", "Biomarkers", "Female", "Follicle Stimulating Hormone", "Gonadal Disorders", "Hematopoietic Stem Cell Transplantation", "Humans", "Inhibins", "Luteinizing Hormone", "Male", "Ovary", "Retrospective Studies", "Testis", "Transplantation Conditioning" ]

[ "Background", "Patients", "Protocol", "Methods", "Results", "1. Boys", "2. Girls", "Discussion", "1. Boys", "2. Girls", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Conditioning for hematopoietic cell transplantation (HCT) may alter the production of gonadal hormones (testosterone in boys, estradiol and progesterone in girls) and the viability of the germ cells. Gonadal failure may result in incomplete sexual development and growth at puberty, and sterility in adulthood. Thus, gonadal hormones are required for the development of secondary sexual characteristics and the growth spurt which normally occurs at puberty.\nThere are few reports of patients given HCT during childhood being fertile (2 boys and 3 girls in Salooja [1], 2 boys and 9 girls in Sanders [2]). Sanders et al [2] showed that 15 (13%) of 114 prepubertal boys developed normal testicular function and the partners of 2 of them became pregnant; the great majority of those who recovered testicular function had been given cyclophosphamide without irradiation. In parallel, 23 (28%) of 82 prepubertal girls developed normal ovarian function, 9 of whom became pregnant; the pregnancies of all 5 given total body irradiation (TBI) ended in spontaneous abortion.\nIt is difficult to predict the reproductive capacity of a child before pubertal age because the plasma concentrations of follicle-stimulating hormone (FSH) and luteinizing hormone (LH) are not informative and no spermogram can be done. The plasma concentrations of inhibin B and anti-Müllerian hormone (AMH) might be helpful at this age. In boys, inhibin B is produced by the Sertoli cells. Its plasma concentration is the best plasma marker of spermatogenesis [3-5]. A study of 218 subfertile men found that their inhibin B concentration accurately (95%) differentiated between competent and impaired spermatogenesis, while the FSH concentration was less accurate (80%) [5]. In girls, inhibin B is produced only by the granulosa cells of small antral follicles, while AMH is produced by the granulosa cells of pre-antral follicles, i.e. the ovarian reserve. Its plasma concentration decreases as the number of follicles decreases with age, with a strong correlation between age at menopause and AMH measured between the ages of 20 and 36 years [6].\nWe evaluated the gonadal function of 72 patients given HCT during childhood after they had reached pubertal age. The objectives were: 1) to compare the evaluation of gonadal function by pubertal stage, testicular volume in boys and menstrual pattern in girls, plasma FSH and LH concentrations to the plasma concentrations of inhibin B and AMH; 2) to evaluate the influence of the age at HCT, and the interval between HCT and evaluation, conditioning and, for chemotherapy, the use of cyclophosphamide, busulfan and melphalan on gonadal function.", "This retrospective single center study included 72 patients (38 boys, 34 girls) given HCT between 1976 and 2006 (median 1991, Table 1) and followed by one of us (R. Brauner) in a university pediatric hospital. The boys were younger than 15 years (8.2 ± 0.6 yr) at HCT and the girls younger than 13 years (7.0 ± 0.6 yr). Puberty began in 3 boys and 2 girls at HCT. They had reached pubertal age (over 13 years for boys and 11 years for girls) when their gonadal function was evaluated. The interval between HCT and evaluation was 8.3 ± 0.6 yr in boys and 6.9 ± 0.6 yr in girls.\nPatient characteristics\nThe initial diseases were malignant [acute lymphoblastic leukemia (n = 31), acute myeloid leukemia (n = 16), chronic myelogenous leukemia (n = 3), lymphoma (n = 5), neuroblastoma (n = 5), nephroblastoma (n = 1)], or non-malignant [severe aplastic anemia (n = 6), congenital immunodeficiency (n = 4), myelodysplasia (n = 1)]. Of the patients with malignant disease, 34 were given HCT in first remission and 27 in second or third remission. The patients were allografted (70%) or autografted (30%). The conditioning protocol for HCT included chemotherapy in all, TBI 12 Grays (Gy) as six fractions of 2 Gy over 3 consecutive days, or a single dose 10 Gy TBI, 5 or 6 Gy total lymphoid irradiation (TLI) as single 4-h exposures or chemotherapy alone (Table 1). The total chemotherapy doses were cyclophosphamide (120, 150 or 200 mg/kg according to the disease), melphalan (140 mg/m2) and/or busulfan (600 mg/m2). The other drugs were cytarabine in 12 (18 or 24 g/m2), etoposide in 5 (400 mg/m2), methotrexate in 4 and vincristine in 5.\nFifteen other boys and 11 girls were excluded because no samples were available for measuring inhibin B. Their characteristics were similar to those who were included. Another 75 patients seen during the same period and fulfilling these criteria were excluded because they had factors other than conditioning for HCT that might have interfered with their gonadal function: the initial disease [Fanconi's anemia (n = 19), Blackfan-Diamond anemia (n = 3), thalassemia (n = 3), drepanocytosis (n = 2), Seckel disease (n = 1)], central nervous system involvement or additional irradiation (n = 47).", "Informed consent for evaluation and treatment was obtained from the patients and their parents. The Ethical Review Committee (Comité de Protection des Personnes Ile de France III) stated that ''This research was found to conform to generally accepted scientific principles and research ethical standards and to be in conformity with the laws and regulations of the country in which the research experiment was performed\".\nWe recorded testicular dimensions [7] and pubic hair development in the boys [8]. Because the testicular dimensions may be altered by the tubular failure caused by the conditioning protocol, the pubertal stage was defined by the pubic hair development and the plasma testosterone concentration: P1 - below 0.5 ng/mL, P2 - 0.5-2 ng/mL, P3 - 2-3 ng/mL and P4-5 over 3 ng/mL [adapted from 9]. We recorded the age at breast development and the occurrence and progress of their menstruations in the girls [10]. Plasma samples were collected before the substitutive treatment except in 5 girls in whom the estradiol treatment was interrupted for at least 2 months (see below), and a long time after graft versus host disease. We measured the basal plasma concentrations of FSH, LH and testosterone in boys and estradiol in girls. Aliquots of plasma were frozen at -20°C and used to measure concomitant plasma inhibin B (in all) and AMH (31 boys and 25 girls) concentrations. We used the last sample taken from patients who had undergone more than one laboratory evaluation.\nNormal gonadal function was defined by the occurrence of spontaneous puberty in both sexes, plus regular menstruations in girls, and normal basal plasma FSH (< 9 IU/L) and LH (< 5 IU/L) concentrations. In boys, tubular failure was defined by an increased plasma FSH concentration, and Leydig cell failure by an increased plasma LH concentration with a normal (partial failure) or low (complete failure) plasma testosterone concentration. The normal basal testosterone concentration in adult boys is 3.5-8.5 ng/mL. In girls, ovarian failure was defined by increased plasma FSH and/or LH concentrations, which is partial when pubertal development is spontaneous and plasma estradiol is normal, and complete when pubertal development is partial or absent and plasma estradiol low.\nTwo boys with partial and the one with complete Leydig cell failure were given testosterone heptylate (25 mg i.m. every 14 days) at the age of around 13 years. Seven girls with partial and the 16 with complete ovarian failure were given oral ethinyl estradiol (2 μg/day) at the age of around 12 years. The doses were increased to adult levels, and associated with cyclical progestin in girls when they had finished growing.\nThe growth hormone (GH) secretion of the patients given TBI who had a decreased growth rate was evaluated by a stimulation test. The test was repeated in those with a low peak to decide on GH treatment, and in those whose growth rate remained low despite a normal GH peak [11]. Twenty-four patients were given GH. Plasma cortisol and prolactin concentrations were normal in all patients. The 32 patients with high plasma thyroid stimulating hormone concentrations after TBI or TLI were given thyroxin (50 μg/m2/day).", "When the assay method for a given hormone was changed during the study period, it was cross-correlated with the earlier method. Thus, the results are comparable throughout the whole period.\nThe plasma concentrations of inhibin B and AMH were measured in serum by enzyme immunometric assays (Oxford Bio-Innovation reagents, Serotec, Oxford, UK for inhibin B and Immunotech reagents, Beckman Coulter Company, Marseille, France for AMH). The lower limit of detection was 10 pg/mL for inhibin B and 1 pmol/L for AMH. Their concentrations were compared to the normal values [9,12]\nData are expressed as means ± se. The differences between groups were analyzed by a Kruskall Wallis test, followed by Mann-Whitney tests. Correlations were made with the Spearman rank test. We also analysed the factors associated with abnormal gonadal function using standard statistical tools and Weka Data Mining software [13].", "Gonadal function and the plasma inhibin B concentrations did not differ with the type of HCT (allograft or autograft), or of TBI (single 10 or fractionated 12 Gy). We therefore analysed all the data together.\n[SUBTITLE] 1. Boys [SUBSECTION] The plasma FSH and LH concentrations indicated that 10 (26%) boys had normal testicular function and 28 (74%) had abnormal function (Table 2). Of the latter, 16 (42%) had isolated tubular failure and 12 (32%) also had Leydig cell failure (11 partial and one complete). The mean plasma inhibin B and AMH concentrations were significantly higher in the boys with normal testicular function than in those with tubular failure, and among these latter higher in those with isolated tubular failure than in those with tubular and Leydig cell failures. Among the 3 boys given HCT after the onset of puberty, the two given TBI and cyclophosphamide had normal testicular function and the one given TBI, melphalan and vincristin had tubular failure.\nFeatures of testis function after HCT\nmean±se\nP*<0.0001 compared to normal \nP**=0.01 compared to tubular and Leydig cell failures\nP***<0.005 compared to normal testis function \nP****<0.02 compared to tubular and Leydig cell failure \nThe 10 boys with normal FSH and LH concentrations included 8 who were given TBI for acute lymphoblastic leukemia (n = 2), acute myeloid leukemia (n = 5), and chronic myelogenous leukemia (n = 1), one who was given TLI for severe aplastic anemia, and one who was given chemotherapy alone (aracytine and busulfan) at one year for acute lymphoblastic leukemia. All, except the 3 treated for acute lymphoblastic leukemia, were given cyclophosphamide alone.\nAll 16 boys who were given melphalan had increased plasma FSH, and 14/16 had decreased inhibin B concentrations (Figure 1). The melphalan was combined with TLI in one case and with TBI in the others (single 10 Gy in four and fractionated 12 Gy in eleven cases).\nDistributions of plasma inhibin B and antimüllerian hormone (AMH) in boys given hematopoietic cell transplantation according to the conditioning protocol. Broken lines to the 5th and 95th percentiles9,12; * indicates those given melphalan; a indicates the 2 boys with increased FSH (10 and 11 IU/L) and normal inhibin B.\nThe plasma AMH concentrations were normal in 25 patients and decreased in 6, all of whom had increased FSH and low inhibin B concentrations (Figure 1). The plasma inhibin B concentrations were normal in 10 and decreased in 28 boys. There were dissociations between the plasma FSH and inhibin B concentrations in 3 patients: one boy with normal FSH had low inhibin B (55 pg/mL); conversely two boys with increased FSH (10 and 11 IU/L) had normal inhibin B (79 and 95 pg/mL). These 3 patients included 2 with testicular volumes greater than 10 mL.\nThe plasma inhibin B concentrations were positively correlated with those of AMH and negatively correlated with those of FSH (Rho = 0.71, P < 0.0001 for both), but not with the age at HCT, interval between HCT and evaluation, testicular volume, which was negatively correlated with the plasma FSH concentrations (Rho = -0.46, P < 0.006).\nData Mining analysis showed that the factors associated with increased plasma FSH concentrations were melphalan treatment, no cyclophosphamide treatment, and a low inhibin B concentration: all 22 boys with an inhibin B concentration below 55 pg/mL had increased plasma FSH concentrations, while all 7 of the boys whose inhibin B concentration was above 100 pg/mL had normal plasma FSH and LH concentrations.\nThe plasma FSH and LH concentrations indicated that 10 (26%) boys had normal testicular function and 28 (74%) had abnormal function (Table 2). Of the latter, 16 (42%) had isolated tubular failure and 12 (32%) also had Leydig cell failure (11 partial and one complete). The mean plasma inhibin B and AMH concentrations were significantly higher in the boys with normal testicular function than in those with tubular failure, and among these latter higher in those with isolated tubular failure than in those with tubular and Leydig cell failures. Among the 3 boys given HCT after the onset of puberty, the two given TBI and cyclophosphamide had normal testicular function and the one given TBI, melphalan and vincristin had tubular failure.\nFeatures of testis function after HCT\nmean±se\nP*<0.0001 compared to normal \nP**=0.01 compared to tubular and Leydig cell failures\nP***<0.005 compared to normal testis function \nP****<0.02 compared to tubular and Leydig cell failure \nThe 10 boys with normal FSH and LH concentrations included 8 who were given TBI for acute lymphoblastic leukemia (n = 2), acute myeloid leukemia (n = 5), and chronic myelogenous leukemia (n = 1), one who was given TLI for severe aplastic anemia, and one who was given chemotherapy alone (aracytine and busulfan) at one year for acute lymphoblastic leukemia. All, except the 3 treated for acute lymphoblastic leukemia, were given cyclophosphamide alone.\nAll 16 boys who were given melphalan had increased plasma FSH, and 14/16 had decreased inhibin B concentrations (Figure 1). The melphalan was combined with TLI in one case and with TBI in the others (single 10 Gy in four and fractionated 12 Gy in eleven cases).\nDistributions of plasma inhibin B and antimüllerian hormone (AMH) in boys given hematopoietic cell transplantation according to the conditioning protocol. Broken lines to the 5th and 95th percentiles9,12; * indicates those given melphalan; a indicates the 2 boys with increased FSH (10 and 11 IU/L) and normal inhibin B.\nThe plasma AMH concentrations were normal in 25 patients and decreased in 6, all of whom had increased FSH and low inhibin B concentrations (Figure 1). The plasma inhibin B concentrations were normal in 10 and decreased in 28 boys. There were dissociations between the plasma FSH and inhibin B concentrations in 3 patients: one boy with normal FSH had low inhibin B (55 pg/mL); conversely two boys with increased FSH (10 and 11 IU/L) had normal inhibin B (79 and 95 pg/mL). These 3 patients included 2 with testicular volumes greater than 10 mL.\nThe plasma inhibin B concentrations were positively correlated with those of AMH and negatively correlated with those of FSH (Rho = 0.71, P < 0.0001 for both), but not with the age at HCT, interval between HCT and evaluation, testicular volume, which was negatively correlated with the plasma FSH concentrations (Rho = -0.46, P < 0.006).\nData Mining analysis showed that the factors associated with increased plasma FSH concentrations were melphalan treatment, no cyclophosphamide treatment, and a low inhibin B concentration: all 22 boys with an inhibin B concentration below 55 pg/mL had increased plasma FSH concentrations, while all 7 of the boys whose inhibin B concentration was above 100 pg/mL had normal plasma FSH and LH concentrations.\n[SUBTITLE] 2. Girls [SUBSECTION] Menarche had occurred spontaneously in 11 girls, 6 of whom had normal FSH and LH concentrations and 5 had partial ovarian failure.\nThe plasma FSH and LH concentrations indicated that 7 (21%) girls had normal ovarian function and 27 (79%) had ovarian failure (11 partial and 16 complete, Table 3). The mean plasma inhibin B concentrations were significantly higher in the girls with normal ovarian function than in those with ovarian failure, and among these latter higher in those with partial ovarian failure than in those with complete failure. The two girls given HCT after the onset of puberty suffered from ovarian failure.\nFeatures of ovarian function after HCT\nmean ± se\na: NS\nP* = 0.0003 compared to normal ovarian function\nThe 7 girls with normal FSH and LH concentrations included 3 who were given TBI for acute lymphoblastic leukemia (n = 2) or acute myeloid leukemia (n = 1), one who was given TLI for severe aplastic anemia and 3 who were given chemotherapy alone for congenital immunodeficiency (= 2) or nephroblastoma.\nAmong the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH and LH concentrations and 7/8 had low inhibin B with dissociation in one case.\nThe plasma AMH concentrations were very low, between 0 and 2 pmol/L, including in the 7 with normal ovarian function. The plasma inhibin B concentrations were normal in 6 and decreased in 28 girls (Figure 2). There were dissociations between the plasma FSH/LH and inhibin B concentrations in 2 patients: one girl given cyclophosphamide and busulfan without irradiation had normal FSH and LH and low inhibin B (22 pg/mL) at 12.2 years; conversely, one girl with increased FSH/LH had normal inhibin B (124 pg/mL) with few spontaneous menstruations.\nDistributions of plasma inhibin B in girls given hematopoietic cell transplantation according to the ovarian function and conditioning protocol. * indicates those given busulfan; a indicates the 2 girls with dissociations between gonadotropin and inhibin B concentrations.\nThe plasma inhibin B concentrations were negatively correlated with those of FSH (Rho = -0.6, P = 0.0006), but not with the age at HCT or the interval between HCT and evaluation.\nData Mining analysis showed that the factors associated with increased plasma FSH/LH concentrations were busulfan and a low inhibin B concentration: the inhibin B concentration was below 40 pg/mL in 26 of the 27 girls with increased LH/FSH concentrations, while it was above 40 pg/mL in 6 of the 7 girls (above 100 pg/mL in 5 girls) with normal plasma FSH and LH concentrations.\nMenarche had occurred spontaneously in 11 girls, 6 of whom had normal FSH and LH concentrations and 5 had partial ovarian failure.\nThe plasma FSH and LH concentrations indicated that 7 (21%) girls had normal ovarian function and 27 (79%) had ovarian failure (11 partial and 16 complete, Table 3). The mean plasma inhibin B concentrations were significantly higher in the girls with normal ovarian function than in those with ovarian failure, and among these latter higher in those with partial ovarian failure than in those with complete failure. The two girls given HCT after the onset of puberty suffered from ovarian failure.\nFeatures of ovarian function after HCT\nmean ± se\na: NS\nP* = 0.0003 compared to normal ovarian function\nThe 7 girls with normal FSH and LH concentrations included 3 who were given TBI for acute lymphoblastic leukemia (n = 2) or acute myeloid leukemia (n = 1), one who was given TLI for severe aplastic anemia and 3 who were given chemotherapy alone for congenital immunodeficiency (= 2) or nephroblastoma.\nAmong the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH and LH concentrations and 7/8 had low inhibin B with dissociation in one case.\nThe plasma AMH concentrations were very low, between 0 and 2 pmol/L, including in the 7 with normal ovarian function. The plasma inhibin B concentrations were normal in 6 and decreased in 28 girls (Figure 2). There were dissociations between the plasma FSH/LH and inhibin B concentrations in 2 patients: one girl given cyclophosphamide and busulfan without irradiation had normal FSH and LH and low inhibin B (22 pg/mL) at 12.2 years; conversely, one girl with increased FSH/LH had normal inhibin B (124 pg/mL) with few spontaneous menstruations.\nDistributions of plasma inhibin B in girls given hematopoietic cell transplantation according to the ovarian function and conditioning protocol. * indicates those given busulfan; a indicates the 2 girls with dissociations between gonadotropin and inhibin B concentrations.\nThe plasma inhibin B concentrations were negatively correlated with those of FSH (Rho = -0.6, P = 0.0006), but not with the age at HCT or the interval between HCT and evaluation.\nData Mining analysis showed that the factors associated with increased plasma FSH/LH concentrations were busulfan and a low inhibin B concentration: the inhibin B concentration was below 40 pg/mL in 26 of the 27 girls with increased LH/FSH concentrations, while it was above 40 pg/mL in 6 of the 7 girls (above 100 pg/mL in 5 girls) with normal plasma FSH and LH concentrations.", "The plasma FSH and LH concentrations indicated that 10 (26%) boys had normal testicular function and 28 (74%) had abnormal function (Table 2). Of the latter, 16 (42%) had isolated tubular failure and 12 (32%) also had Leydig cell failure (11 partial and one complete). The mean plasma inhibin B and AMH concentrations were significantly higher in the boys with normal testicular function than in those with tubular failure, and among these latter higher in those with isolated tubular failure than in those with tubular and Leydig cell failures. Among the 3 boys given HCT after the onset of puberty, the two given TBI and cyclophosphamide had normal testicular function and the one given TBI, melphalan and vincristin had tubular failure.\nFeatures of testis function after HCT\nmean±se\nP*<0.0001 compared to normal \nP**=0.01 compared to tubular and Leydig cell failures\nP***<0.005 compared to normal testis function \nP****<0.02 compared to tubular and Leydig cell failure \nThe 10 boys with normal FSH and LH concentrations included 8 who were given TBI for acute lymphoblastic leukemia (n = 2), acute myeloid leukemia (n = 5), and chronic myelogenous leukemia (n = 1), one who was given TLI for severe aplastic anemia, and one who was given chemotherapy alone (aracytine and busulfan) at one year for acute lymphoblastic leukemia. All, except the 3 treated for acute lymphoblastic leukemia, were given cyclophosphamide alone.\nAll 16 boys who were given melphalan had increased plasma FSH, and 14/16 had decreased inhibin B concentrations (Figure 1). The melphalan was combined with TLI in one case and with TBI in the others (single 10 Gy in four and fractionated 12 Gy in eleven cases).\nDistributions of plasma inhibin B and antimüllerian hormone (AMH) in boys given hematopoietic cell transplantation according to the conditioning protocol. Broken lines to the 5th and 95th percentiles9,12; * indicates those given melphalan; a indicates the 2 boys with increased FSH (10 and 11 IU/L) and normal inhibin B.\nThe plasma AMH concentrations were normal in 25 patients and decreased in 6, all of whom had increased FSH and low inhibin B concentrations (Figure 1). The plasma inhibin B concentrations were normal in 10 and decreased in 28 boys. There were dissociations between the plasma FSH and inhibin B concentrations in 3 patients: one boy with normal FSH had low inhibin B (55 pg/mL); conversely two boys with increased FSH (10 and 11 IU/L) had normal inhibin B (79 and 95 pg/mL). These 3 patients included 2 with testicular volumes greater than 10 mL.\nThe plasma inhibin B concentrations were positively correlated with those of AMH and negatively correlated with those of FSH (Rho = 0.71, P < 0.0001 for both), but not with the age at HCT, interval between HCT and evaluation, testicular volume, which was negatively correlated with the plasma FSH concentrations (Rho = -0.46, P < 0.006).\nData Mining analysis showed that the factors associated with increased plasma FSH concentrations were melphalan treatment, no cyclophosphamide treatment, and a low inhibin B concentration: all 22 boys with an inhibin B concentration below 55 pg/mL had increased plasma FSH concentrations, while all 7 of the boys whose inhibin B concentration was above 100 pg/mL had normal plasma FSH and LH concentrations.", "Menarche had occurred spontaneously in 11 girls, 6 of whom had normal FSH and LH concentrations and 5 had partial ovarian failure.\nThe plasma FSH and LH concentrations indicated that 7 (21%) girls had normal ovarian function and 27 (79%) had ovarian failure (11 partial and 16 complete, Table 3). The mean plasma inhibin B concentrations were significantly higher in the girls with normal ovarian function than in those with ovarian failure, and among these latter higher in those with partial ovarian failure than in those with complete failure. The two girls given HCT after the onset of puberty suffered from ovarian failure.\nFeatures of ovarian function after HCT\nmean ± se\na: NS\nP* = 0.0003 compared to normal ovarian function\nThe 7 girls with normal FSH and LH concentrations included 3 who were given TBI for acute lymphoblastic leukemia (n = 2) or acute myeloid leukemia (n = 1), one who was given TLI for severe aplastic anemia and 3 who were given chemotherapy alone for congenital immunodeficiency (= 2) or nephroblastoma.\nAmong the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH and LH concentrations and 7/8 had low inhibin B with dissociation in one case.\nThe plasma AMH concentrations were very low, between 0 and 2 pmol/L, including in the 7 with normal ovarian function. The plasma inhibin B concentrations were normal in 6 and decreased in 28 girls (Figure 2). There were dissociations between the plasma FSH/LH and inhibin B concentrations in 2 patients: one girl given cyclophosphamide and busulfan without irradiation had normal FSH and LH and low inhibin B (22 pg/mL) at 12.2 years; conversely, one girl with increased FSH/LH had normal inhibin B (124 pg/mL) with few spontaneous menstruations.\nDistributions of plasma inhibin B in girls given hematopoietic cell transplantation according to the ovarian function and conditioning protocol. * indicates those given busulfan; a indicates the 2 girls with dissociations between gonadotropin and inhibin B concentrations.\nThe plasma inhibin B concentrations were negatively correlated with those of FSH (Rho = -0.6, P = 0.0006), but not with the age at HCT or the interval between HCT and evaluation.\nData Mining analysis showed that the factors associated with increased plasma FSH/LH concentrations were busulfan and a low inhibin B concentration: the inhibin B concentration was below 40 pg/mL in 26 of the 27 girls with increased LH/FSH concentrations, while it was above 40 pg/mL in 6 of the 7 girls (above 100 pg/mL in 5 girls) with normal plasma FSH and LH concentrations.", "We have evaluated the specific effect of conditioning for HCT on the gonads. Any patient having other factors that might have interfered with their gonadal function was excluded, except the 27 given HCT in the second or third remission who were previously given chemotherapy without irradiation.\n[SUBTITLE] 1. Boys [SUBSECTION] Their plasma FSH and LH concentrations indicated that 26% of our boys had normal testicular function. These figures are higher than those reported by Sanders et al [2] (13%) and Van Casteren et al [14] who reported that the 10 boys given TBI 7.5 or 12 Gy before HCT had almost undetectable (< 26 pg/mL, n = 8) or 26-62 pg/mL (n = 2) inhibin B concentrations; the severity of gonadal dysfunction in this study may be partly due to the additional testicular radiotherapy given to 5/10 boys. The differences might also be explained by differences in the age at HCT and/or the interval between HCT and evaluation. However, these two factors did not influence testicular function in our study. The wide ranges of the age at HCT and the interval between HCT and evaluation, and the limited study population might partly explain the absence of any significant impact of these factors in our study.\nMelphalan seemed to be more gonadotoxic, giving rise to increased FSH (in all 16) and decreased inhibin B (in 14 of them) concentrations. This effect of melphalan was also seen by Crofton et al [15], who showed that inhibin B was normal before, during and after treatment in 16 boys given chemotherapy for solid tumors or acute lymphoblastic leukemia, except in one who had undetectable inhibin B after cyclophosphamide plus cisplatin plus melphalan. However conclusions cannot easily be drawn as the effect of the drug cannot be separated from that of irradiation.\nOur study confirms the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B that was reported by Cicognani et al [16]. They showed that all boys who have increased FSH (> 6.1 IU/L) or LH (> 5.8 IU/L) concentrations also had inhibin B concentrations below 112 pg/mL (-2 standard deviations). Unlike Lähteenmäki et al [17], we found no correlation between testicular volume and inhibin B (P = 0.06). They reported that patients with small testes (< 10 mL) after treatment for childhood malignancy had inhibin B concentrations below 42 pg/mL, and 6 out of 7 had FSH concentrations above 9 IU/L; patients with a testicular volume above 13 mL had inhibin B concentrations above 100 pg/mL. This difference might be explained by variations in the age at which the boys in our study were given HCT and evaluated.\nWe find that the plasma AMH concentration does not seem to be a good marker, as it normally decreases when testosterone increases at puberty.\nWe did not carry out a sperm analysis because our boys were too young. However, the reported relationship between inhibin B and the sperm count in normal men and in boys surviving a childhood malignancy suggests that 10/38 (based on FSH or inhibin B) or 7/38 (based on both) of the boys we studied will be normospermic. Thomson et al [18] showed that inhibin B was barely detectable in azoospermic patients whose germ cells had been destroyed, despite preservation of their Sertoli cells, as confirmed by testicular biopsy. Van Beek et al [19] showed that inhibin B is a better marker of spermatogenesis than is the FSH concentration in men given chemotherapy for Hodgkin's lymphoma as children. Multivariate analysis with inhibin B and FSH concentrations as determinants of sperm concentration showed that inhibin B was the only significant determinant; all the men with a plasma inhibin B concentration above 75 pg/mL were normospermic. Lähteenmäki et al [20] showed that the inhibin B and FSH concentrations and the testicular volume explain 44% of the variance in the sperm count. Van Casteren et al [14] found that their 5 patients who were normospermic had inhibin B concentrations of 125-234 pg/mL, while the corresponding values in the 9 azoospermic patients was 0-79 pg/mL, with a correlation between sperm count and inhibin B concentrations.\nTheir plasma FSH and LH concentrations indicated that 26% of our boys had normal testicular function. These figures are higher than those reported by Sanders et al [2] (13%) and Van Casteren et al [14] who reported that the 10 boys given TBI 7.5 or 12 Gy before HCT had almost undetectable (< 26 pg/mL, n = 8) or 26-62 pg/mL (n = 2) inhibin B concentrations; the severity of gonadal dysfunction in this study may be partly due to the additional testicular radiotherapy given to 5/10 boys. The differences might also be explained by differences in the age at HCT and/or the interval between HCT and evaluation. However, these two factors did not influence testicular function in our study. The wide ranges of the age at HCT and the interval between HCT and evaluation, and the limited study population might partly explain the absence of any significant impact of these factors in our study.\nMelphalan seemed to be more gonadotoxic, giving rise to increased FSH (in all 16) and decreased inhibin B (in 14 of them) concentrations. This effect of melphalan was also seen by Crofton et al [15], who showed that inhibin B was normal before, during and after treatment in 16 boys given chemotherapy for solid tumors or acute lymphoblastic leukemia, except in one who had undetectable inhibin B after cyclophosphamide plus cisplatin plus melphalan. However conclusions cannot easily be drawn as the effect of the drug cannot be separated from that of irradiation.\nOur study confirms the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B that was reported by Cicognani et al [16]. They showed that all boys who have increased FSH (> 6.1 IU/L) or LH (> 5.8 IU/L) concentrations also had inhibin B concentrations below 112 pg/mL (-2 standard deviations). Unlike Lähteenmäki et al [17], we found no correlation between testicular volume and inhibin B (P = 0.06). They reported that patients with small testes (< 10 mL) after treatment for childhood malignancy had inhibin B concentrations below 42 pg/mL, and 6 out of 7 had FSH concentrations above 9 IU/L; patients with a testicular volume above 13 mL had inhibin B concentrations above 100 pg/mL. This difference might be explained by variations in the age at which the boys in our study were given HCT and evaluated.\nWe find that the plasma AMH concentration does not seem to be a good marker, as it normally decreases when testosterone increases at puberty.\nWe did not carry out a sperm analysis because our boys were too young. However, the reported relationship between inhibin B and the sperm count in normal men and in boys surviving a childhood malignancy suggests that 10/38 (based on FSH or inhibin B) or 7/38 (based on both) of the boys we studied will be normospermic. Thomson et al [18] showed that inhibin B was barely detectable in azoospermic patients whose germ cells had been destroyed, despite preservation of their Sertoli cells, as confirmed by testicular biopsy. Van Beek et al [19] showed that inhibin B is a better marker of spermatogenesis than is the FSH concentration in men given chemotherapy for Hodgkin's lymphoma as children. Multivariate analysis with inhibin B and FSH concentrations as determinants of sperm concentration showed that inhibin B was the only significant determinant; all the men with a plasma inhibin B concentration above 75 pg/mL were normospermic. Lähteenmäki et al [20] showed that the inhibin B and FSH concentrations and the testicular volume explain 44% of the variance in the sperm count. Van Casteren et al [14] found that their 5 patients who were normospermic had inhibin B concentrations of 125-234 pg/mL, while the corresponding values in the 9 azoospermic patients was 0-79 pg/mL, with a correlation between sperm count and inhibin B concentrations.\n[SUBTITLE] 2. Girls [SUBSECTION] Their plasma FSH and LH concentrations indicate that 21% of our girls had normal ovarian function. These figures are similar to those reported by others [2,21,22]. However, the girls in our study with spontaneous puberty were no younger at HCT than were those with ovarian failure. Similarly, the FSH, inhibin B and AMH concentrations were not influenced by age or pubertal status at treatment. Fong et al [23] studied women treated for cancer during childhood and found that those with undetectable AMH were significantly older at diagnosis, only 2/17 had regular menstrual cycles, and more of them had been given body or abdomen radiotherapy than those with greater AMH concentrations. The preeminence of the role of the irradiation and of busulfan might partly explain why the age at HCT has no significant impact on ovarian function in our study.\nBusulfan seemed to be more gonadotoxic. Of the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH/LH concentrations and 7/8 had low inhibin B. This effect of busulfan was also seen by Teinturier et al [24], who showed that the 10 girls given busulfan (600 mg/m2) without TBI before autologous HCT all developed severe, persistent ovarian failure. They identified 26 of 29 girls in published reports given busulfan during prepuberty or puberty as having signs of ovarian failure.\nLarsen et al [25] used multiple linear regression analysis of 100 girls who had survived childhood malignancy to predict the total number of antral follicles per ovary. Ovarian irradiation, alkylating chemotherapy, older age at diagnosis and a long period off treatment all reduced the number of follicles. Among the 10 girls conditioned for HCT by TBI, the 3 with spontaneous cycles had smaller ovaries, fewer total follicles and higher FSH than the others.\nWe have therefore confirmed the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B, and agreement with the clinical evolution, while those of AMH were very low. The 7 girls with normal puberty and FSH and LH all had AMH concentrations similar to those with ovarian failure. Three [23,26,27] studies have evaluated the capacity of plasma AMH concentrations to assess the ovarian damage in adults surviving childhood malignancy. Bath et al [26] compared 10 girls treated for cancer during childhood, all with regular menstrual cycles, with 11 controls. The treated girls had smaller ovaries than the controls and significantly higher FSH and lower AMH concentrations, while the other hormone concentrations were unchanged. Thus, the AMH concentration indicates a depletion of ovarian reserve, despite regular menstrual cycles. Van Beek et al [27] showed that girls treated with MOPP (mechlorethamine, oncovin, prednisone, procarbazine) for Hodgkin's lymphoma during childhood developed into women with elevated FSH concentrations and decreased AMH concentrations. Three of the women with normal FSH had decreased AMH.\nTheir plasma FSH and LH concentrations indicate that 21% of our girls had normal ovarian function. These figures are similar to those reported by others [2,21,22]. However, the girls in our study with spontaneous puberty were no younger at HCT than were those with ovarian failure. Similarly, the FSH, inhibin B and AMH concentrations were not influenced by age or pubertal status at treatment. Fong et al [23] studied women treated for cancer during childhood and found that those with undetectable AMH were significantly older at diagnosis, only 2/17 had regular menstrual cycles, and more of them had been given body or abdomen radiotherapy than those with greater AMH concentrations. The preeminence of the role of the irradiation and of busulfan might partly explain why the age at HCT has no significant impact on ovarian function in our study.\nBusulfan seemed to be more gonadotoxic. Of the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH/LH concentrations and 7/8 had low inhibin B. This effect of busulfan was also seen by Teinturier et al [24], who showed that the 10 girls given busulfan (600 mg/m2) without TBI before autologous HCT all developed severe, persistent ovarian failure. They identified 26 of 29 girls in published reports given busulfan during prepuberty or puberty as having signs of ovarian failure.\nLarsen et al [25] used multiple linear regression analysis of 100 girls who had survived childhood malignancy to predict the total number of antral follicles per ovary. Ovarian irradiation, alkylating chemotherapy, older age at diagnosis and a long period off treatment all reduced the number of follicles. Among the 10 girls conditioned for HCT by TBI, the 3 with spontaneous cycles had smaller ovaries, fewer total follicles and higher FSH than the others.\nWe have therefore confirmed the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B, and agreement with the clinical evolution, while those of AMH were very low. The 7 girls with normal puberty and FSH and LH all had AMH concentrations similar to those with ovarian failure. Three [23,26,27] studies have evaluated the capacity of plasma AMH concentrations to assess the ovarian damage in adults surviving childhood malignancy. Bath et al [26] compared 10 girls treated for cancer during childhood, all with regular menstrual cycles, with 11 controls. The treated girls had smaller ovaries than the controls and significantly higher FSH and lower AMH concentrations, while the other hormone concentrations were unchanged. Thus, the AMH concentration indicates a depletion of ovarian reserve, despite regular menstrual cycles. Van Beek et al [27] showed that girls treated with MOPP (mechlorethamine, oncovin, prednisone, procarbazine) for Hodgkin's lymphoma during childhood developed into women with elevated FSH concentrations and decreased AMH concentrations. Three of the women with normal FSH had decreased AMH.", "Their plasma FSH and LH concentrations indicated that 26% of our boys had normal testicular function. These figures are higher than those reported by Sanders et al [2] (13%) and Van Casteren et al [14] who reported that the 10 boys given TBI 7.5 or 12 Gy before HCT had almost undetectable (< 26 pg/mL, n = 8) or 26-62 pg/mL (n = 2) inhibin B concentrations; the severity of gonadal dysfunction in this study may be partly due to the additional testicular radiotherapy given to 5/10 boys. The differences might also be explained by differences in the age at HCT and/or the interval between HCT and evaluation. However, these two factors did not influence testicular function in our study. The wide ranges of the age at HCT and the interval between HCT and evaluation, and the limited study population might partly explain the absence of any significant impact of these factors in our study.\nMelphalan seemed to be more gonadotoxic, giving rise to increased FSH (in all 16) and decreased inhibin B (in 14 of them) concentrations. This effect of melphalan was also seen by Crofton et al [15], who showed that inhibin B was normal before, during and after treatment in 16 boys given chemotherapy for solid tumors or acute lymphoblastic leukemia, except in one who had undetectable inhibin B after cyclophosphamide plus cisplatin plus melphalan. However conclusions cannot easily be drawn as the effect of the drug cannot be separated from that of irradiation.\nOur study confirms the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B that was reported by Cicognani et al [16]. They showed that all boys who have increased FSH (> 6.1 IU/L) or LH (> 5.8 IU/L) concentrations also had inhibin B concentrations below 112 pg/mL (-2 standard deviations). Unlike Lähteenmäki et al [17], we found no correlation between testicular volume and inhibin B (P = 0.06). They reported that patients with small testes (< 10 mL) after treatment for childhood malignancy had inhibin B concentrations below 42 pg/mL, and 6 out of 7 had FSH concentrations above 9 IU/L; patients with a testicular volume above 13 mL had inhibin B concentrations above 100 pg/mL. This difference might be explained by variations in the age at which the boys in our study were given HCT and evaluated.\nWe find that the plasma AMH concentration does not seem to be a good marker, as it normally decreases when testosterone increases at puberty.\nWe did not carry out a sperm analysis because our boys were too young. However, the reported relationship between inhibin B and the sperm count in normal men and in boys surviving a childhood malignancy suggests that 10/38 (based on FSH or inhibin B) or 7/38 (based on both) of the boys we studied will be normospermic. Thomson et al [18] showed that inhibin B was barely detectable in azoospermic patients whose germ cells had been destroyed, despite preservation of their Sertoli cells, as confirmed by testicular biopsy. Van Beek et al [19] showed that inhibin B is a better marker of spermatogenesis than is the FSH concentration in men given chemotherapy for Hodgkin's lymphoma as children. Multivariate analysis with inhibin B and FSH concentrations as determinants of sperm concentration showed that inhibin B was the only significant determinant; all the men with a plasma inhibin B concentration above 75 pg/mL were normospermic. Lähteenmäki et al [20] showed that the inhibin B and FSH concentrations and the testicular volume explain 44% of the variance in the sperm count. Van Casteren et al [14] found that their 5 patients who were normospermic had inhibin B concentrations of 125-234 pg/mL, while the corresponding values in the 9 azoospermic patients was 0-79 pg/mL, with a correlation between sperm count and inhibin B concentrations.", "Their plasma FSH and LH concentrations indicate that 21% of our girls had normal ovarian function. These figures are similar to those reported by others [2,21,22]. However, the girls in our study with spontaneous puberty were no younger at HCT than were those with ovarian failure. Similarly, the FSH, inhibin B and AMH concentrations were not influenced by age or pubertal status at treatment. Fong et al [23] studied women treated for cancer during childhood and found that those with undetectable AMH were significantly older at diagnosis, only 2/17 had regular menstrual cycles, and more of them had been given body or abdomen radiotherapy than those with greater AMH concentrations. The preeminence of the role of the irradiation and of busulfan might partly explain why the age at HCT has no significant impact on ovarian function in our study.\nBusulfan seemed to be more gonadotoxic. Of the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH/LH concentrations and 7/8 had low inhibin B. This effect of busulfan was also seen by Teinturier et al [24], who showed that the 10 girls given busulfan (600 mg/m2) without TBI before autologous HCT all developed severe, persistent ovarian failure. They identified 26 of 29 girls in published reports given busulfan during prepuberty or puberty as having signs of ovarian failure.\nLarsen et al [25] used multiple linear regression analysis of 100 girls who had survived childhood malignancy to predict the total number of antral follicles per ovary. Ovarian irradiation, alkylating chemotherapy, older age at diagnosis and a long period off treatment all reduced the number of follicles. Among the 10 girls conditioned for HCT by TBI, the 3 with spontaneous cycles had smaller ovaries, fewer total follicles and higher FSH than the others.\nWe have therefore confirmed the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B, and agreement with the clinical evolution, while those of AMH were very low. The 7 girls with normal puberty and FSH and LH all had AMH concentrations similar to those with ovarian failure. Three [23,26,27] studies have evaluated the capacity of plasma AMH concentrations to assess the ovarian damage in adults surviving childhood malignancy. Bath et al [26] compared 10 girls treated for cancer during childhood, all with regular menstrual cycles, with 11 controls. The treated girls had smaller ovaries than the controls and significantly higher FSH and lower AMH concentrations, while the other hormone concentrations were unchanged. Thus, the AMH concentration indicates a depletion of ovarian reserve, despite regular menstrual cycles. Van Beek et al [27] showed that girls treated with MOPP (mechlorethamine, oncovin, prednisone, procarbazine) for Hodgkin's lymphoma during childhood developed into women with elevated FSH concentrations and decreased AMH concentrations. Three of the women with normal FSH had decreased AMH.", "In boys, the concordance between plasma FSH and inhibin B concentrations suggests that measuring inhibin B may help in counselling at pubertal age. Plasma AMH concentrations do not seem to be a good marker as they normally decrease when testosterone increases at puberty. All or most of the boys given melphalan, when combined with TBI or TLI, will be azoospermic.\nIn girls, there is also a good concordance between plasma FSH and inhibin B concentrations. Both were normal in 6/34 girls who had normal puberty and regular menstruations, but their very low AMH concentrations suggest that there is a major loss of primordial follicles.\nOne limitation of this study is the range of drugs used. The fact that the drug effect could not be separated from that of irradiation significantly limits our ability to draw any conclusion about the effects of melphalan or busulfan.", "AMH: antimüllerian hormone; FSH: follicle-stimulating hormone, growth hormone - GH; Gy: Grays; HCT: hematopoietic cell transplantation; LH: luteinizing hormone; TLI: total lymphoid irradiation; TBI: total body irradiation", "The authors declare that they have no competing interests.", "SL and AC-CS participated in the design of the study, data acquisition and analysis. ST and SBT carried out the immunoassays. PL and CT performed the statistical analyses. CT prepared the tables and the figures. HE, JM, AB and AF followed the patients and participated in the design of the study. RB directed the work and prepared the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2431/11/20/prepub\n" ]

[ null, null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[ "Background", "Methods", "Patients", "Protocol", "Methods", "Results", "1. Boys", "2. Girls", "Discussion", "1. Boys", "2. Girls", "Conclusions", "Abbreviations", "Competing interests", "Authors' contributions", "Pre-publication history" ]

[ "Conditioning for hematopoietic cell transplantation (HCT) may alter the production of gonadal hormones (testosterone in boys, estradiol and progesterone in girls) and the viability of the germ cells. Gonadal failure may result in incomplete sexual development and growth at puberty, and sterility in adulthood. Thus, gonadal hormones are required for the development of secondary sexual characteristics and the growth spurt which normally occurs at puberty.\nThere are few reports of patients given HCT during childhood being fertile (2 boys and 3 girls in Salooja [1], 2 boys and 9 girls in Sanders [2]). Sanders et al [2] showed that 15 (13%) of 114 prepubertal boys developed normal testicular function and the partners of 2 of them became pregnant; the great majority of those who recovered testicular function had been given cyclophosphamide without irradiation. In parallel, 23 (28%) of 82 prepubertal girls developed normal ovarian function, 9 of whom became pregnant; the pregnancies of all 5 given total body irradiation (TBI) ended in spontaneous abortion.\nIt is difficult to predict the reproductive capacity of a child before pubertal age because the plasma concentrations of follicle-stimulating hormone (FSH) and luteinizing hormone (LH) are not informative and no spermogram can be done. The plasma concentrations of inhibin B and anti-Müllerian hormone (AMH) might be helpful at this age. In boys, inhibin B is produced by the Sertoli cells. Its plasma concentration is the best plasma marker of spermatogenesis [3-5]. A study of 218 subfertile men found that their inhibin B concentration accurately (95%) differentiated between competent and impaired spermatogenesis, while the FSH concentration was less accurate (80%) [5]. In girls, inhibin B is produced only by the granulosa cells of small antral follicles, while AMH is produced by the granulosa cells of pre-antral follicles, i.e. the ovarian reserve. Its plasma concentration decreases as the number of follicles decreases with age, with a strong correlation between age at menopause and AMH measured between the ages of 20 and 36 years [6].\nWe evaluated the gonadal function of 72 patients given HCT during childhood after they had reached pubertal age. The objectives were: 1) to compare the evaluation of gonadal function by pubertal stage, testicular volume in boys and menstrual pattern in girls, plasma FSH and LH concentrations to the plasma concentrations of inhibin B and AMH; 2) to evaluate the influence of the age at HCT, and the interval between HCT and evaluation, conditioning and, for chemotherapy, the use of cyclophosphamide, busulfan and melphalan on gonadal function.", "[SUBTITLE] Patients [SUBSECTION] This retrospective single center study included 72 patients (38 boys, 34 girls) given HCT between 1976 and 2006 (median 1991, Table 1) and followed by one of us (R. Brauner) in a university pediatric hospital. The boys were younger than 15 years (8.2 ± 0.6 yr) at HCT and the girls younger than 13 years (7.0 ± 0.6 yr). Puberty began in 3 boys and 2 girls at HCT. They had reached pubertal age (over 13 years for boys and 11 years for girls) when their gonadal function was evaluated. The interval between HCT and evaluation was 8.3 ± 0.6 yr in boys and 6.9 ± 0.6 yr in girls.\nPatient characteristics\nThe initial diseases were malignant [acute lymphoblastic leukemia (n = 31), acute myeloid leukemia (n = 16), chronic myelogenous leukemia (n = 3), lymphoma (n = 5), neuroblastoma (n = 5), nephroblastoma (n = 1)], or non-malignant [severe aplastic anemia (n = 6), congenital immunodeficiency (n = 4), myelodysplasia (n = 1)]. Of the patients with malignant disease, 34 were given HCT in first remission and 27 in second or third remission. The patients were allografted (70%) or autografted (30%). The conditioning protocol for HCT included chemotherapy in all, TBI 12 Grays (Gy) as six fractions of 2 Gy over 3 consecutive days, or a single dose 10 Gy TBI, 5 or 6 Gy total lymphoid irradiation (TLI) as single 4-h exposures or chemotherapy alone (Table 1). The total chemotherapy doses were cyclophosphamide (120, 150 or 200 mg/kg according to the disease), melphalan (140 mg/m2) and/or busulfan (600 mg/m2). The other drugs were cytarabine in 12 (18 or 24 g/m2), etoposide in 5 (400 mg/m2), methotrexate in 4 and vincristine in 5.\nFifteen other boys and 11 girls were excluded because no samples were available for measuring inhibin B. Their characteristics were similar to those who were included. Another 75 patients seen during the same period and fulfilling these criteria were excluded because they had factors other than conditioning for HCT that might have interfered with their gonadal function: the initial disease [Fanconi's anemia (n = 19), Blackfan-Diamond anemia (n = 3), thalassemia (n = 3), drepanocytosis (n = 2), Seckel disease (n = 1)], central nervous system involvement or additional irradiation (n = 47).\nThis retrospective single center study included 72 patients (38 boys, 34 girls) given HCT between 1976 and 2006 (median 1991, Table 1) and followed by one of us (R. Brauner) in a university pediatric hospital. The boys were younger than 15 years (8.2 ± 0.6 yr) at HCT and the girls younger than 13 years (7.0 ± 0.6 yr). Puberty began in 3 boys and 2 girls at HCT. They had reached pubertal age (over 13 years for boys and 11 years for girls) when their gonadal function was evaluated. The interval between HCT and evaluation was 8.3 ± 0.6 yr in boys and 6.9 ± 0.6 yr in girls.\nPatient characteristics\nThe initial diseases were malignant [acute lymphoblastic leukemia (n = 31), acute myeloid leukemia (n = 16), chronic myelogenous leukemia (n = 3), lymphoma (n = 5), neuroblastoma (n = 5), nephroblastoma (n = 1)], or non-malignant [severe aplastic anemia (n = 6), congenital immunodeficiency (n = 4), myelodysplasia (n = 1)]. Of the patients with malignant disease, 34 were given HCT in first remission and 27 in second or third remission. The patients were allografted (70%) or autografted (30%). The conditioning protocol for HCT included chemotherapy in all, TBI 12 Grays (Gy) as six fractions of 2 Gy over 3 consecutive days, or a single dose 10 Gy TBI, 5 or 6 Gy total lymphoid irradiation (TLI) as single 4-h exposures or chemotherapy alone (Table 1). The total chemotherapy doses were cyclophosphamide (120, 150 or 200 mg/kg according to the disease), melphalan (140 mg/m2) and/or busulfan (600 mg/m2). The other drugs were cytarabine in 12 (18 or 24 g/m2), etoposide in 5 (400 mg/m2), methotrexate in 4 and vincristine in 5.\nFifteen other boys and 11 girls were excluded because no samples were available for measuring inhibin B. Their characteristics were similar to those who were included. Another 75 patients seen during the same period and fulfilling these criteria were excluded because they had factors other than conditioning for HCT that might have interfered with their gonadal function: the initial disease [Fanconi's anemia (n = 19), Blackfan-Diamond anemia (n = 3), thalassemia (n = 3), drepanocytosis (n = 2), Seckel disease (n = 1)], central nervous system involvement or additional irradiation (n = 47).\n[SUBTITLE] Protocol [SUBSECTION] Informed consent for evaluation and treatment was obtained from the patients and their parents. The Ethical Review Committee (Comité de Protection des Personnes Ile de France III) stated that ''This research was found to conform to generally accepted scientific principles and research ethical standards and to be in conformity with the laws and regulations of the country in which the research experiment was performed\".\nWe recorded testicular dimensions [7] and pubic hair development in the boys [8]. Because the testicular dimensions may be altered by the tubular failure caused by the conditioning protocol, the pubertal stage was defined by the pubic hair development and the plasma testosterone concentration: P1 - below 0.5 ng/mL, P2 - 0.5-2 ng/mL, P3 - 2-3 ng/mL and P4-5 over 3 ng/mL [adapted from 9]. We recorded the age at breast development and the occurrence and progress of their menstruations in the girls [10]. Plasma samples were collected before the substitutive treatment except in 5 girls in whom the estradiol treatment was interrupted for at least 2 months (see below), and a long time after graft versus host disease. We measured the basal plasma concentrations of FSH, LH and testosterone in boys and estradiol in girls. Aliquots of plasma were frozen at -20°C and used to measure concomitant plasma inhibin B (in all) and AMH (31 boys and 25 girls) concentrations. We used the last sample taken from patients who had undergone more than one laboratory evaluation.\nNormal gonadal function was defined by the occurrence of spontaneous puberty in both sexes, plus regular menstruations in girls, and normal basal plasma FSH (< 9 IU/L) and LH (< 5 IU/L) concentrations. In boys, tubular failure was defined by an increased plasma FSH concentration, and Leydig cell failure by an increased plasma LH concentration with a normal (partial failure) or low (complete failure) plasma testosterone concentration. The normal basal testosterone concentration in adult boys is 3.5-8.5 ng/mL. In girls, ovarian failure was defined by increased plasma FSH and/or LH concentrations, which is partial when pubertal development is spontaneous and plasma estradiol is normal, and complete when pubertal development is partial or absent and plasma estradiol low.\nTwo boys with partial and the one with complete Leydig cell failure were given testosterone heptylate (25 mg i.m. every 14 days) at the age of around 13 years. Seven girls with partial and the 16 with complete ovarian failure were given oral ethinyl estradiol (2 μg/day) at the age of around 12 years. The doses were increased to adult levels, and associated with cyclical progestin in girls when they had finished growing.\nThe growth hormone (GH) secretion of the patients given TBI who had a decreased growth rate was evaluated by a stimulation test. The test was repeated in those with a low peak to decide on GH treatment, and in those whose growth rate remained low despite a normal GH peak [11]. Twenty-four patients were given GH. Plasma cortisol and prolactin concentrations were normal in all patients. The 32 patients with high plasma thyroid stimulating hormone concentrations after TBI or TLI were given thyroxin (50 μg/m2/day).\nInformed consent for evaluation and treatment was obtained from the patients and their parents. The Ethical Review Committee (Comité de Protection des Personnes Ile de France III) stated that ''This research was found to conform to generally accepted scientific principles and research ethical standards and to be in conformity with the laws and regulations of the country in which the research experiment was performed\".\nWe recorded testicular dimensions [7] and pubic hair development in the boys [8]. Because the testicular dimensions may be altered by the tubular failure caused by the conditioning protocol, the pubertal stage was defined by the pubic hair development and the plasma testosterone concentration: P1 - below 0.5 ng/mL, P2 - 0.5-2 ng/mL, P3 - 2-3 ng/mL and P4-5 over 3 ng/mL [adapted from 9]. We recorded the age at breast development and the occurrence and progress of their menstruations in the girls [10]. Plasma samples were collected before the substitutive treatment except in 5 girls in whom the estradiol treatment was interrupted for at least 2 months (see below), and a long time after graft versus host disease. We measured the basal plasma concentrations of FSH, LH and testosterone in boys and estradiol in girls. Aliquots of plasma were frozen at -20°C and used to measure concomitant plasma inhibin B (in all) and AMH (31 boys and 25 girls) concentrations. We used the last sample taken from patients who had undergone more than one laboratory evaluation.\nNormal gonadal function was defined by the occurrence of spontaneous puberty in both sexes, plus regular menstruations in girls, and normal basal plasma FSH (< 9 IU/L) and LH (< 5 IU/L) concentrations. In boys, tubular failure was defined by an increased plasma FSH concentration, and Leydig cell failure by an increased plasma LH concentration with a normal (partial failure) or low (complete failure) plasma testosterone concentration. The normal basal testosterone concentration in adult boys is 3.5-8.5 ng/mL. In girls, ovarian failure was defined by increased plasma FSH and/or LH concentrations, which is partial when pubertal development is spontaneous and plasma estradiol is normal, and complete when pubertal development is partial or absent and plasma estradiol low.\nTwo boys with partial and the one with complete Leydig cell failure were given testosterone heptylate (25 mg i.m. every 14 days) at the age of around 13 years. Seven girls with partial and the 16 with complete ovarian failure were given oral ethinyl estradiol (2 μg/day) at the age of around 12 years. The doses were increased to adult levels, and associated with cyclical progestin in girls when they had finished growing.\nThe growth hormone (GH) secretion of the patients given TBI who had a decreased growth rate was evaluated by a stimulation test. The test was repeated in those with a low peak to decide on GH treatment, and in those whose growth rate remained low despite a normal GH peak [11]. Twenty-four patients were given GH. Plasma cortisol and prolactin concentrations were normal in all patients. The 32 patients with high plasma thyroid stimulating hormone concentrations after TBI or TLI were given thyroxin (50 μg/m2/day).\n[SUBTITLE] Methods [SUBSECTION] When the assay method for a given hormone was changed during the study period, it was cross-correlated with the earlier method. Thus, the results are comparable throughout the whole period.\nThe plasma concentrations of inhibin B and AMH were measured in serum by enzyme immunometric assays (Oxford Bio-Innovation reagents, Serotec, Oxford, UK for inhibin B and Immunotech reagents, Beckman Coulter Company, Marseille, France for AMH). The lower limit of detection was 10 pg/mL for inhibin B and 1 pmol/L for AMH. Their concentrations were compared to the normal values [9,12]\nData are expressed as means ± se. The differences between groups were analyzed by a Kruskall Wallis test, followed by Mann-Whitney tests. Correlations were made with the Spearman rank test. We also analysed the factors associated with abnormal gonadal function using standard statistical tools and Weka Data Mining software [13].\nWhen the assay method for a given hormone was changed during the study period, it was cross-correlated with the earlier method. Thus, the results are comparable throughout the whole period.\nThe plasma concentrations of inhibin B and AMH were measured in serum by enzyme immunometric assays (Oxford Bio-Innovation reagents, Serotec, Oxford, UK for inhibin B and Immunotech reagents, Beckman Coulter Company, Marseille, France for AMH). The lower limit of detection was 10 pg/mL for inhibin B and 1 pmol/L for AMH. Their concentrations were compared to the normal values [9,12]\nData are expressed as means ± se. The differences between groups were analyzed by a Kruskall Wallis test, followed by Mann-Whitney tests. Correlations were made with the Spearman rank test. We also analysed the factors associated with abnormal gonadal function using standard statistical tools and Weka Data Mining software [13].", "This retrospective single center study included 72 patients (38 boys, 34 girls) given HCT between 1976 and 2006 (median 1991, Table 1) and followed by one of us (R. Brauner) in a university pediatric hospital. The boys were younger than 15 years (8.2 ± 0.6 yr) at HCT and the girls younger than 13 years (7.0 ± 0.6 yr). Puberty began in 3 boys and 2 girls at HCT. They had reached pubertal age (over 13 years for boys and 11 years for girls) when their gonadal function was evaluated. The interval between HCT and evaluation was 8.3 ± 0.6 yr in boys and 6.9 ± 0.6 yr in girls.\nPatient characteristics\nThe initial diseases were malignant [acute lymphoblastic leukemia (n = 31), acute myeloid leukemia (n = 16), chronic myelogenous leukemia (n = 3), lymphoma (n = 5), neuroblastoma (n = 5), nephroblastoma (n = 1)], or non-malignant [severe aplastic anemia (n = 6), congenital immunodeficiency (n = 4), myelodysplasia (n = 1)]. Of the patients with malignant disease, 34 were given HCT in first remission and 27 in second or third remission. The patients were allografted (70%) or autografted (30%). The conditioning protocol for HCT included chemotherapy in all, TBI 12 Grays (Gy) as six fractions of 2 Gy over 3 consecutive days, or a single dose 10 Gy TBI, 5 or 6 Gy total lymphoid irradiation (TLI) as single 4-h exposures or chemotherapy alone (Table 1). The total chemotherapy doses were cyclophosphamide (120, 150 or 200 mg/kg according to the disease), melphalan (140 mg/m2) and/or busulfan (600 mg/m2). The other drugs were cytarabine in 12 (18 or 24 g/m2), etoposide in 5 (400 mg/m2), methotrexate in 4 and vincristine in 5.\nFifteen other boys and 11 girls were excluded because no samples were available for measuring inhibin B. Their characteristics were similar to those who were included. Another 75 patients seen during the same period and fulfilling these criteria were excluded because they had factors other than conditioning for HCT that might have interfered with their gonadal function: the initial disease [Fanconi's anemia (n = 19), Blackfan-Diamond anemia (n = 3), thalassemia (n = 3), drepanocytosis (n = 2), Seckel disease (n = 1)], central nervous system involvement or additional irradiation (n = 47).", "Informed consent for evaluation and treatment was obtained from the patients and their parents. The Ethical Review Committee (Comité de Protection des Personnes Ile de France III) stated that ''This research was found to conform to generally accepted scientific principles and research ethical standards and to be in conformity with the laws and regulations of the country in which the research experiment was performed\".\nWe recorded testicular dimensions [7] and pubic hair development in the boys [8]. Because the testicular dimensions may be altered by the tubular failure caused by the conditioning protocol, the pubertal stage was defined by the pubic hair development and the plasma testosterone concentration: P1 - below 0.5 ng/mL, P2 - 0.5-2 ng/mL, P3 - 2-3 ng/mL and P4-5 over 3 ng/mL [adapted from 9]. We recorded the age at breast development and the occurrence and progress of their menstruations in the girls [10]. Plasma samples were collected before the substitutive treatment except in 5 girls in whom the estradiol treatment was interrupted for at least 2 months (see below), and a long time after graft versus host disease. We measured the basal plasma concentrations of FSH, LH and testosterone in boys and estradiol in girls. Aliquots of plasma were frozen at -20°C and used to measure concomitant plasma inhibin B (in all) and AMH (31 boys and 25 girls) concentrations. We used the last sample taken from patients who had undergone more than one laboratory evaluation.\nNormal gonadal function was defined by the occurrence of spontaneous puberty in both sexes, plus regular menstruations in girls, and normal basal plasma FSH (< 9 IU/L) and LH (< 5 IU/L) concentrations. In boys, tubular failure was defined by an increased plasma FSH concentration, and Leydig cell failure by an increased plasma LH concentration with a normal (partial failure) or low (complete failure) plasma testosterone concentration. The normal basal testosterone concentration in adult boys is 3.5-8.5 ng/mL. In girls, ovarian failure was defined by increased plasma FSH and/or LH concentrations, which is partial when pubertal development is spontaneous and plasma estradiol is normal, and complete when pubertal development is partial or absent and plasma estradiol low.\nTwo boys with partial and the one with complete Leydig cell failure were given testosterone heptylate (25 mg i.m. every 14 days) at the age of around 13 years. Seven girls with partial and the 16 with complete ovarian failure were given oral ethinyl estradiol (2 μg/day) at the age of around 12 years. The doses were increased to adult levels, and associated with cyclical progestin in girls when they had finished growing.\nThe growth hormone (GH) secretion of the patients given TBI who had a decreased growth rate was evaluated by a stimulation test. The test was repeated in those with a low peak to decide on GH treatment, and in those whose growth rate remained low despite a normal GH peak [11]. Twenty-four patients were given GH. Plasma cortisol and prolactin concentrations were normal in all patients. The 32 patients with high plasma thyroid stimulating hormone concentrations after TBI or TLI were given thyroxin (50 μg/m2/day).", "When the assay method for a given hormone was changed during the study period, it was cross-correlated with the earlier method. Thus, the results are comparable throughout the whole period.\nThe plasma concentrations of inhibin B and AMH were measured in serum by enzyme immunometric assays (Oxford Bio-Innovation reagents, Serotec, Oxford, UK for inhibin B and Immunotech reagents, Beckman Coulter Company, Marseille, France for AMH). The lower limit of detection was 10 pg/mL for inhibin B and 1 pmol/L for AMH. Their concentrations were compared to the normal values [9,12]\nData are expressed as means ± se. The differences between groups were analyzed by a Kruskall Wallis test, followed by Mann-Whitney tests. Correlations were made with the Spearman rank test. We also analysed the factors associated with abnormal gonadal function using standard statistical tools and Weka Data Mining software [13].", "Gonadal function and the plasma inhibin B concentrations did not differ with the type of HCT (allograft or autograft), or of TBI (single 10 or fractionated 12 Gy). We therefore analysed all the data together.\n[SUBTITLE] 1. Boys [SUBSECTION] The plasma FSH and LH concentrations indicated that 10 (26%) boys had normal testicular function and 28 (74%) had abnormal function (Table 2). Of the latter, 16 (42%) had isolated tubular failure and 12 (32%) also had Leydig cell failure (11 partial and one complete). The mean plasma inhibin B and AMH concentrations were significantly higher in the boys with normal testicular function than in those with tubular failure, and among these latter higher in those with isolated tubular failure than in those with tubular and Leydig cell failures. Among the 3 boys given HCT after the onset of puberty, the two given TBI and cyclophosphamide had normal testicular function and the one given TBI, melphalan and vincristin had tubular failure.\nFeatures of testis function after HCT\nmean±se\nP*<0.0001 compared to normal \nP**=0.01 compared to tubular and Leydig cell failures\nP***<0.005 compared to normal testis function \nP****<0.02 compared to tubular and Leydig cell failure \nThe 10 boys with normal FSH and LH concentrations included 8 who were given TBI for acute lymphoblastic leukemia (n = 2), acute myeloid leukemia (n = 5), and chronic myelogenous leukemia (n = 1), one who was given TLI for severe aplastic anemia, and one who was given chemotherapy alone (aracytine and busulfan) at one year for acute lymphoblastic leukemia. All, except the 3 treated for acute lymphoblastic leukemia, were given cyclophosphamide alone.\nAll 16 boys who were given melphalan had increased plasma FSH, and 14/16 had decreased inhibin B concentrations (Figure 1). The melphalan was combined with TLI in one case and with TBI in the others (single 10 Gy in four and fractionated 12 Gy in eleven cases).\nDistributions of plasma inhibin B and antimüllerian hormone (AMH) in boys given hematopoietic cell transplantation according to the conditioning protocol. Broken lines to the 5th and 95th percentiles9,12; * indicates those given melphalan; a indicates the 2 boys with increased FSH (10 and 11 IU/L) and normal inhibin B.\nThe plasma AMH concentrations were normal in 25 patients and decreased in 6, all of whom had increased FSH and low inhibin B concentrations (Figure 1). The plasma inhibin B concentrations were normal in 10 and decreased in 28 boys. There were dissociations between the plasma FSH and inhibin B concentrations in 3 patients: one boy with normal FSH had low inhibin B (55 pg/mL); conversely two boys with increased FSH (10 and 11 IU/L) had normal inhibin B (79 and 95 pg/mL). These 3 patients included 2 with testicular volumes greater than 10 mL.\nThe plasma inhibin B concentrations were positively correlated with those of AMH and negatively correlated with those of FSH (Rho = 0.71, P < 0.0001 for both), but not with the age at HCT, interval between HCT and evaluation, testicular volume, which was negatively correlated with the plasma FSH concentrations (Rho = -0.46, P < 0.006).\nData Mining analysis showed that the factors associated with increased plasma FSH concentrations were melphalan treatment, no cyclophosphamide treatment, and a low inhibin B concentration: all 22 boys with an inhibin B concentration below 55 pg/mL had increased plasma FSH concentrations, while all 7 of the boys whose inhibin B concentration was above 100 pg/mL had normal plasma FSH and LH concentrations.\nThe plasma FSH and LH concentrations indicated that 10 (26%) boys had normal testicular function and 28 (74%) had abnormal function (Table 2). Of the latter, 16 (42%) had isolated tubular failure and 12 (32%) also had Leydig cell failure (11 partial and one complete). The mean plasma inhibin B and AMH concentrations were significantly higher in the boys with normal testicular function than in those with tubular failure, and among these latter higher in those with isolated tubular failure than in those with tubular and Leydig cell failures. Among the 3 boys given HCT after the onset of puberty, the two given TBI and cyclophosphamide had normal testicular function and the one given TBI, melphalan and vincristin had tubular failure.\nFeatures of testis function after HCT\nmean±se\nP*<0.0001 compared to normal \nP**=0.01 compared to tubular and Leydig cell failures\nP***<0.005 compared to normal testis function \nP****<0.02 compared to tubular and Leydig cell failure \nThe 10 boys with normal FSH and LH concentrations included 8 who were given TBI for acute lymphoblastic leukemia (n = 2), acute myeloid leukemia (n = 5), and chronic myelogenous leukemia (n = 1), one who was given TLI for severe aplastic anemia, and one who was given chemotherapy alone (aracytine and busulfan) at one year for acute lymphoblastic leukemia. All, except the 3 treated for acute lymphoblastic leukemia, were given cyclophosphamide alone.\nAll 16 boys who were given melphalan had increased plasma FSH, and 14/16 had decreased inhibin B concentrations (Figure 1). The melphalan was combined with TLI in one case and with TBI in the others (single 10 Gy in four and fractionated 12 Gy in eleven cases).\nDistributions of plasma inhibin B and antimüllerian hormone (AMH) in boys given hematopoietic cell transplantation according to the conditioning protocol. Broken lines to the 5th and 95th percentiles9,12; * indicates those given melphalan; a indicates the 2 boys with increased FSH (10 and 11 IU/L) and normal inhibin B.\nThe plasma AMH concentrations were normal in 25 patients and decreased in 6, all of whom had increased FSH and low inhibin B concentrations (Figure 1). The plasma inhibin B concentrations were normal in 10 and decreased in 28 boys. There were dissociations between the plasma FSH and inhibin B concentrations in 3 patients: one boy with normal FSH had low inhibin B (55 pg/mL); conversely two boys with increased FSH (10 and 11 IU/L) had normal inhibin B (79 and 95 pg/mL). These 3 patients included 2 with testicular volumes greater than 10 mL.\nThe plasma inhibin B concentrations were positively correlated with those of AMH and negatively correlated with those of FSH (Rho = 0.71, P < 0.0001 for both), but not with the age at HCT, interval between HCT and evaluation, testicular volume, which was negatively correlated with the plasma FSH concentrations (Rho = -0.46, P < 0.006).\nData Mining analysis showed that the factors associated with increased plasma FSH concentrations were melphalan treatment, no cyclophosphamide treatment, and a low inhibin B concentration: all 22 boys with an inhibin B concentration below 55 pg/mL had increased plasma FSH concentrations, while all 7 of the boys whose inhibin B concentration was above 100 pg/mL had normal plasma FSH and LH concentrations.\n[SUBTITLE] 2. Girls [SUBSECTION] Menarche had occurred spontaneously in 11 girls, 6 of whom had normal FSH and LH concentrations and 5 had partial ovarian failure.\nThe plasma FSH and LH concentrations indicated that 7 (21%) girls had normal ovarian function and 27 (79%) had ovarian failure (11 partial and 16 complete, Table 3). The mean plasma inhibin B concentrations were significantly higher in the girls with normal ovarian function than in those with ovarian failure, and among these latter higher in those with partial ovarian failure than in those with complete failure. The two girls given HCT after the onset of puberty suffered from ovarian failure.\nFeatures of ovarian function after HCT\nmean ± se\na: NS\nP* = 0.0003 compared to normal ovarian function\nThe 7 girls with normal FSH and LH concentrations included 3 who were given TBI for acute lymphoblastic leukemia (n = 2) or acute myeloid leukemia (n = 1), one who was given TLI for severe aplastic anemia and 3 who were given chemotherapy alone for congenital immunodeficiency (= 2) or nephroblastoma.\nAmong the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH and LH concentrations and 7/8 had low inhibin B with dissociation in one case.\nThe plasma AMH concentrations were very low, between 0 and 2 pmol/L, including in the 7 with normal ovarian function. The plasma inhibin B concentrations were normal in 6 and decreased in 28 girls (Figure 2). There were dissociations between the plasma FSH/LH and inhibin B concentrations in 2 patients: one girl given cyclophosphamide and busulfan without irradiation had normal FSH and LH and low inhibin B (22 pg/mL) at 12.2 years; conversely, one girl with increased FSH/LH had normal inhibin B (124 pg/mL) with few spontaneous menstruations.\nDistributions of plasma inhibin B in girls given hematopoietic cell transplantation according to the ovarian function and conditioning protocol. * indicates those given busulfan; a indicates the 2 girls with dissociations between gonadotropin and inhibin B concentrations.\nThe plasma inhibin B concentrations were negatively correlated with those of FSH (Rho = -0.6, P = 0.0006), but not with the age at HCT or the interval between HCT and evaluation.\nData Mining analysis showed that the factors associated with increased plasma FSH/LH concentrations were busulfan and a low inhibin B concentration: the inhibin B concentration was below 40 pg/mL in 26 of the 27 girls with increased LH/FSH concentrations, while it was above 40 pg/mL in 6 of the 7 girls (above 100 pg/mL in 5 girls) with normal plasma FSH and LH concentrations.\nMenarche had occurred spontaneously in 11 girls, 6 of whom had normal FSH and LH concentrations and 5 had partial ovarian failure.\nThe plasma FSH and LH concentrations indicated that 7 (21%) girls had normal ovarian function and 27 (79%) had ovarian failure (11 partial and 16 complete, Table 3). The mean plasma inhibin B concentrations were significantly higher in the girls with normal ovarian function than in those with ovarian failure, and among these latter higher in those with partial ovarian failure than in those with complete failure. The two girls given HCT after the onset of puberty suffered from ovarian failure.\nFeatures of ovarian function after HCT\nmean ± se\na: NS\nP* = 0.0003 compared to normal ovarian function\nThe 7 girls with normal FSH and LH concentrations included 3 who were given TBI for acute lymphoblastic leukemia (n = 2) or acute myeloid leukemia (n = 1), one who was given TLI for severe aplastic anemia and 3 who were given chemotherapy alone for congenital immunodeficiency (= 2) or nephroblastoma.\nAmong the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH and LH concentrations and 7/8 had low inhibin B with dissociation in one case.\nThe plasma AMH concentrations were very low, between 0 and 2 pmol/L, including in the 7 with normal ovarian function. The plasma inhibin B concentrations were normal in 6 and decreased in 28 girls (Figure 2). There were dissociations between the plasma FSH/LH and inhibin B concentrations in 2 patients: one girl given cyclophosphamide and busulfan without irradiation had normal FSH and LH and low inhibin B (22 pg/mL) at 12.2 years; conversely, one girl with increased FSH/LH had normal inhibin B (124 pg/mL) with few spontaneous menstruations.\nDistributions of plasma inhibin B in girls given hematopoietic cell transplantation according to the ovarian function and conditioning protocol. * indicates those given busulfan; a indicates the 2 girls with dissociations between gonadotropin and inhibin B concentrations.\nThe plasma inhibin B concentrations were negatively correlated with those of FSH (Rho = -0.6, P = 0.0006), but not with the age at HCT or the interval between HCT and evaluation.\nData Mining analysis showed that the factors associated with increased plasma FSH/LH concentrations were busulfan and a low inhibin B concentration: the inhibin B concentration was below 40 pg/mL in 26 of the 27 girls with increased LH/FSH concentrations, while it was above 40 pg/mL in 6 of the 7 girls (above 100 pg/mL in 5 girls) with normal plasma FSH and LH concentrations.", "The plasma FSH and LH concentrations indicated that 10 (26%) boys had normal testicular function and 28 (74%) had abnormal function (Table 2). Of the latter, 16 (42%) had isolated tubular failure and 12 (32%) also had Leydig cell failure (11 partial and one complete). The mean plasma inhibin B and AMH concentrations were significantly higher in the boys with normal testicular function than in those with tubular failure, and among these latter higher in those with isolated tubular failure than in those with tubular and Leydig cell failures. Among the 3 boys given HCT after the onset of puberty, the two given TBI and cyclophosphamide had normal testicular function and the one given TBI, melphalan and vincristin had tubular failure.\nFeatures of testis function after HCT\nmean±se\nP*<0.0001 compared to normal \nP**=0.01 compared to tubular and Leydig cell failures\nP***<0.005 compared to normal testis function \nP****<0.02 compared to tubular and Leydig cell failure \nThe 10 boys with normal FSH and LH concentrations included 8 who were given TBI for acute lymphoblastic leukemia (n = 2), acute myeloid leukemia (n = 5), and chronic myelogenous leukemia (n = 1), one who was given TLI for severe aplastic anemia, and one who was given chemotherapy alone (aracytine and busulfan) at one year for acute lymphoblastic leukemia. All, except the 3 treated for acute lymphoblastic leukemia, were given cyclophosphamide alone.\nAll 16 boys who were given melphalan had increased plasma FSH, and 14/16 had decreased inhibin B concentrations (Figure 1). The melphalan was combined with TLI in one case and with TBI in the others (single 10 Gy in four and fractionated 12 Gy in eleven cases).\nDistributions of plasma inhibin B and antimüllerian hormone (AMH) in boys given hematopoietic cell transplantation according to the conditioning protocol. Broken lines to the 5th and 95th percentiles9,12; * indicates those given melphalan; a indicates the 2 boys with increased FSH (10 and 11 IU/L) and normal inhibin B.\nThe plasma AMH concentrations were normal in 25 patients and decreased in 6, all of whom had increased FSH and low inhibin B concentrations (Figure 1). The plasma inhibin B concentrations were normal in 10 and decreased in 28 boys. There were dissociations between the plasma FSH and inhibin B concentrations in 3 patients: one boy with normal FSH had low inhibin B (55 pg/mL); conversely two boys with increased FSH (10 and 11 IU/L) had normal inhibin B (79 and 95 pg/mL). These 3 patients included 2 with testicular volumes greater than 10 mL.\nThe plasma inhibin B concentrations were positively correlated with those of AMH and negatively correlated with those of FSH (Rho = 0.71, P < 0.0001 for both), but not with the age at HCT, interval between HCT and evaluation, testicular volume, which was negatively correlated with the plasma FSH concentrations (Rho = -0.46, P < 0.006).\nData Mining analysis showed that the factors associated with increased plasma FSH concentrations were melphalan treatment, no cyclophosphamide treatment, and a low inhibin B concentration: all 22 boys with an inhibin B concentration below 55 pg/mL had increased plasma FSH concentrations, while all 7 of the boys whose inhibin B concentration was above 100 pg/mL had normal plasma FSH and LH concentrations.", "Menarche had occurred spontaneously in 11 girls, 6 of whom had normal FSH and LH concentrations and 5 had partial ovarian failure.\nThe plasma FSH and LH concentrations indicated that 7 (21%) girls had normal ovarian function and 27 (79%) had ovarian failure (11 partial and 16 complete, Table 3). The mean plasma inhibin B concentrations were significantly higher in the girls with normal ovarian function than in those with ovarian failure, and among these latter higher in those with partial ovarian failure than in those with complete failure. The two girls given HCT after the onset of puberty suffered from ovarian failure.\nFeatures of ovarian function after HCT\nmean ± se\na: NS\nP* = 0.0003 compared to normal ovarian function\nThe 7 girls with normal FSH and LH concentrations included 3 who were given TBI for acute lymphoblastic leukemia (n = 2) or acute myeloid leukemia (n = 1), one who was given TLI for severe aplastic anemia and 3 who were given chemotherapy alone for congenital immunodeficiency (= 2) or nephroblastoma.\nAmong the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH and LH concentrations and 7/8 had low inhibin B with dissociation in one case.\nThe plasma AMH concentrations were very low, between 0 and 2 pmol/L, including in the 7 with normal ovarian function. The plasma inhibin B concentrations were normal in 6 and decreased in 28 girls (Figure 2). There were dissociations between the plasma FSH/LH and inhibin B concentrations in 2 patients: one girl given cyclophosphamide and busulfan without irradiation had normal FSH and LH and low inhibin B (22 pg/mL) at 12.2 years; conversely, one girl with increased FSH/LH had normal inhibin B (124 pg/mL) with few spontaneous menstruations.\nDistributions of plasma inhibin B in girls given hematopoietic cell transplantation according to the ovarian function and conditioning protocol. * indicates those given busulfan; a indicates the 2 girls with dissociations between gonadotropin and inhibin B concentrations.\nThe plasma inhibin B concentrations were negatively correlated with those of FSH (Rho = -0.6, P = 0.0006), but not with the age at HCT or the interval between HCT and evaluation.\nData Mining analysis showed that the factors associated with increased plasma FSH/LH concentrations were busulfan and a low inhibin B concentration: the inhibin B concentration was below 40 pg/mL in 26 of the 27 girls with increased LH/FSH concentrations, while it was above 40 pg/mL in 6 of the 7 girls (above 100 pg/mL in 5 girls) with normal plasma FSH and LH concentrations.", "We have evaluated the specific effect of conditioning for HCT on the gonads. Any patient having other factors that might have interfered with their gonadal function was excluded, except the 27 given HCT in the second or third remission who were previously given chemotherapy without irradiation.\n[SUBTITLE] 1. Boys [SUBSECTION] Their plasma FSH and LH concentrations indicated that 26% of our boys had normal testicular function. These figures are higher than those reported by Sanders et al [2] (13%) and Van Casteren et al [14] who reported that the 10 boys given TBI 7.5 or 12 Gy before HCT had almost undetectable (< 26 pg/mL, n = 8) or 26-62 pg/mL (n = 2) inhibin B concentrations; the severity of gonadal dysfunction in this study may be partly due to the additional testicular radiotherapy given to 5/10 boys. The differences might also be explained by differences in the age at HCT and/or the interval between HCT and evaluation. However, these two factors did not influence testicular function in our study. The wide ranges of the age at HCT and the interval between HCT and evaluation, and the limited study population might partly explain the absence of any significant impact of these factors in our study.\nMelphalan seemed to be more gonadotoxic, giving rise to increased FSH (in all 16) and decreased inhibin B (in 14 of them) concentrations. This effect of melphalan was also seen by Crofton et al [15], who showed that inhibin B was normal before, during and after treatment in 16 boys given chemotherapy for solid tumors or acute lymphoblastic leukemia, except in one who had undetectable inhibin B after cyclophosphamide plus cisplatin plus melphalan. However conclusions cannot easily be drawn as the effect of the drug cannot be separated from that of irradiation.\nOur study confirms the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B that was reported by Cicognani et al [16]. They showed that all boys who have increased FSH (> 6.1 IU/L) or LH (> 5.8 IU/L) concentrations also had inhibin B concentrations below 112 pg/mL (-2 standard deviations). Unlike Lähteenmäki et al [17], we found no correlation between testicular volume and inhibin B (P = 0.06). They reported that patients with small testes (< 10 mL) after treatment for childhood malignancy had inhibin B concentrations below 42 pg/mL, and 6 out of 7 had FSH concentrations above 9 IU/L; patients with a testicular volume above 13 mL had inhibin B concentrations above 100 pg/mL. This difference might be explained by variations in the age at which the boys in our study were given HCT and evaluated.\nWe find that the plasma AMH concentration does not seem to be a good marker, as it normally decreases when testosterone increases at puberty.\nWe did not carry out a sperm analysis because our boys were too young. However, the reported relationship between inhibin B and the sperm count in normal men and in boys surviving a childhood malignancy suggests that 10/38 (based on FSH or inhibin B) or 7/38 (based on both) of the boys we studied will be normospermic. Thomson et al [18] showed that inhibin B was barely detectable in azoospermic patients whose germ cells had been destroyed, despite preservation of their Sertoli cells, as confirmed by testicular biopsy. Van Beek et al [19] showed that inhibin B is a better marker of spermatogenesis than is the FSH concentration in men given chemotherapy for Hodgkin's lymphoma as children. Multivariate analysis with inhibin B and FSH concentrations as determinants of sperm concentration showed that inhibin B was the only significant determinant; all the men with a plasma inhibin B concentration above 75 pg/mL were normospermic. Lähteenmäki et al [20] showed that the inhibin B and FSH concentrations and the testicular volume explain 44% of the variance in the sperm count. Van Casteren et al [14] found that their 5 patients who were normospermic had inhibin B concentrations of 125-234 pg/mL, while the corresponding values in the 9 azoospermic patients was 0-79 pg/mL, with a correlation between sperm count and inhibin B concentrations.\nTheir plasma FSH and LH concentrations indicated that 26% of our boys had normal testicular function. These figures are higher than those reported by Sanders et al [2] (13%) and Van Casteren et al [14] who reported that the 10 boys given TBI 7.5 or 12 Gy before HCT had almost undetectable (< 26 pg/mL, n = 8) or 26-62 pg/mL (n = 2) inhibin B concentrations; the severity of gonadal dysfunction in this study may be partly due to the additional testicular radiotherapy given to 5/10 boys. The differences might also be explained by differences in the age at HCT and/or the interval between HCT and evaluation. However, these two factors did not influence testicular function in our study. The wide ranges of the age at HCT and the interval between HCT and evaluation, and the limited study population might partly explain the absence of any significant impact of these factors in our study.\nMelphalan seemed to be more gonadotoxic, giving rise to increased FSH (in all 16) and decreased inhibin B (in 14 of them) concentrations. This effect of melphalan was also seen by Crofton et al [15], who showed that inhibin B was normal before, during and after treatment in 16 boys given chemotherapy for solid tumors or acute lymphoblastic leukemia, except in one who had undetectable inhibin B after cyclophosphamide plus cisplatin plus melphalan. However conclusions cannot easily be drawn as the effect of the drug cannot be separated from that of irradiation.\nOur study confirms the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B that was reported by Cicognani et al [16]. They showed that all boys who have increased FSH (> 6.1 IU/L) or LH (> 5.8 IU/L) concentrations also had inhibin B concentrations below 112 pg/mL (-2 standard deviations). Unlike Lähteenmäki et al [17], we found no correlation between testicular volume and inhibin B (P = 0.06). They reported that patients with small testes (< 10 mL) after treatment for childhood malignancy had inhibin B concentrations below 42 pg/mL, and 6 out of 7 had FSH concentrations above 9 IU/L; patients with a testicular volume above 13 mL had inhibin B concentrations above 100 pg/mL. This difference might be explained by variations in the age at which the boys in our study were given HCT and evaluated.\nWe find that the plasma AMH concentration does not seem to be a good marker, as it normally decreases when testosterone increases at puberty.\nWe did not carry out a sperm analysis because our boys were too young. However, the reported relationship between inhibin B and the sperm count in normal men and in boys surviving a childhood malignancy suggests that 10/38 (based on FSH or inhibin B) or 7/38 (based on both) of the boys we studied will be normospermic. Thomson et al [18] showed that inhibin B was barely detectable in azoospermic patients whose germ cells had been destroyed, despite preservation of their Sertoli cells, as confirmed by testicular biopsy. Van Beek et al [19] showed that inhibin B is a better marker of spermatogenesis than is the FSH concentration in men given chemotherapy for Hodgkin's lymphoma as children. Multivariate analysis with inhibin B and FSH concentrations as determinants of sperm concentration showed that inhibin B was the only significant determinant; all the men with a plasma inhibin B concentration above 75 pg/mL were normospermic. Lähteenmäki et al [20] showed that the inhibin B and FSH concentrations and the testicular volume explain 44% of the variance in the sperm count. Van Casteren et al [14] found that their 5 patients who were normospermic had inhibin B concentrations of 125-234 pg/mL, while the corresponding values in the 9 azoospermic patients was 0-79 pg/mL, with a correlation between sperm count and inhibin B concentrations.\n[SUBTITLE] 2. Girls [SUBSECTION] Their plasma FSH and LH concentrations indicate that 21% of our girls had normal ovarian function. These figures are similar to those reported by others [2,21,22]. However, the girls in our study with spontaneous puberty were no younger at HCT than were those with ovarian failure. Similarly, the FSH, inhibin B and AMH concentrations were not influenced by age or pubertal status at treatment. Fong et al [23] studied women treated for cancer during childhood and found that those with undetectable AMH were significantly older at diagnosis, only 2/17 had regular menstrual cycles, and more of them had been given body or abdomen radiotherapy than those with greater AMH concentrations. The preeminence of the role of the irradiation and of busulfan might partly explain why the age at HCT has no significant impact on ovarian function in our study.\nBusulfan seemed to be more gonadotoxic. Of the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH/LH concentrations and 7/8 had low inhibin B. This effect of busulfan was also seen by Teinturier et al [24], who showed that the 10 girls given busulfan (600 mg/m2) without TBI before autologous HCT all developed severe, persistent ovarian failure. They identified 26 of 29 girls in published reports given busulfan during prepuberty or puberty as having signs of ovarian failure.\nLarsen et al [25] used multiple linear regression analysis of 100 girls who had survived childhood malignancy to predict the total number of antral follicles per ovary. Ovarian irradiation, alkylating chemotherapy, older age at diagnosis and a long period off treatment all reduced the number of follicles. Among the 10 girls conditioned for HCT by TBI, the 3 with spontaneous cycles had smaller ovaries, fewer total follicles and higher FSH than the others.\nWe have therefore confirmed the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B, and agreement with the clinical evolution, while those of AMH were very low. The 7 girls with normal puberty and FSH and LH all had AMH concentrations similar to those with ovarian failure. Three [23,26,27] studies have evaluated the capacity of plasma AMH concentrations to assess the ovarian damage in adults surviving childhood malignancy. Bath et al [26] compared 10 girls treated for cancer during childhood, all with regular menstrual cycles, with 11 controls. The treated girls had smaller ovaries than the controls and significantly higher FSH and lower AMH concentrations, while the other hormone concentrations were unchanged. Thus, the AMH concentration indicates a depletion of ovarian reserve, despite regular menstrual cycles. Van Beek et al [27] showed that girls treated with MOPP (mechlorethamine, oncovin, prednisone, procarbazine) for Hodgkin's lymphoma during childhood developed into women with elevated FSH concentrations and decreased AMH concentrations. Three of the women with normal FSH had decreased AMH.\nTheir plasma FSH and LH concentrations indicate that 21% of our girls had normal ovarian function. These figures are similar to those reported by others [2,21,22]. However, the girls in our study with spontaneous puberty were no younger at HCT than were those with ovarian failure. Similarly, the FSH, inhibin B and AMH concentrations were not influenced by age or pubertal status at treatment. Fong et al [23] studied women treated for cancer during childhood and found that those with undetectable AMH were significantly older at diagnosis, only 2/17 had regular menstrual cycles, and more of them had been given body or abdomen radiotherapy than those with greater AMH concentrations. The preeminence of the role of the irradiation and of busulfan might partly explain why the age at HCT has no significant impact on ovarian function in our study.\nBusulfan seemed to be more gonadotoxic. Of the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH/LH concentrations and 7/8 had low inhibin B. This effect of busulfan was also seen by Teinturier et al [24], who showed that the 10 girls given busulfan (600 mg/m2) without TBI before autologous HCT all developed severe, persistent ovarian failure. They identified 26 of 29 girls in published reports given busulfan during prepuberty or puberty as having signs of ovarian failure.\nLarsen et al [25] used multiple linear regression analysis of 100 girls who had survived childhood malignancy to predict the total number of antral follicles per ovary. Ovarian irradiation, alkylating chemotherapy, older age at diagnosis and a long period off treatment all reduced the number of follicles. Among the 10 girls conditioned for HCT by TBI, the 3 with spontaneous cycles had smaller ovaries, fewer total follicles and higher FSH than the others.\nWe have therefore confirmed the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B, and agreement with the clinical evolution, while those of AMH were very low. The 7 girls with normal puberty and FSH and LH all had AMH concentrations similar to those with ovarian failure. Three [23,26,27] studies have evaluated the capacity of plasma AMH concentrations to assess the ovarian damage in adults surviving childhood malignancy. Bath et al [26] compared 10 girls treated for cancer during childhood, all with regular menstrual cycles, with 11 controls. The treated girls had smaller ovaries than the controls and significantly higher FSH and lower AMH concentrations, while the other hormone concentrations were unchanged. Thus, the AMH concentration indicates a depletion of ovarian reserve, despite regular menstrual cycles. Van Beek et al [27] showed that girls treated with MOPP (mechlorethamine, oncovin, prednisone, procarbazine) for Hodgkin's lymphoma during childhood developed into women with elevated FSH concentrations and decreased AMH concentrations. Three of the women with normal FSH had decreased AMH.", "Their plasma FSH and LH concentrations indicated that 26% of our boys had normal testicular function. These figures are higher than those reported by Sanders et al [2] (13%) and Van Casteren et al [14] who reported that the 10 boys given TBI 7.5 or 12 Gy before HCT had almost undetectable (< 26 pg/mL, n = 8) or 26-62 pg/mL (n = 2) inhibin B concentrations; the severity of gonadal dysfunction in this study may be partly due to the additional testicular radiotherapy given to 5/10 boys. The differences might also be explained by differences in the age at HCT and/or the interval between HCT and evaluation. However, these two factors did not influence testicular function in our study. The wide ranges of the age at HCT and the interval between HCT and evaluation, and the limited study population might partly explain the absence of any significant impact of these factors in our study.\nMelphalan seemed to be more gonadotoxic, giving rise to increased FSH (in all 16) and decreased inhibin B (in 14 of them) concentrations. This effect of melphalan was also seen by Crofton et al [15], who showed that inhibin B was normal before, during and after treatment in 16 boys given chemotherapy for solid tumors or acute lymphoblastic leukemia, except in one who had undetectable inhibin B after cyclophosphamide plus cisplatin plus melphalan. However conclusions cannot easily be drawn as the effect of the drug cannot be separated from that of irradiation.\nOur study confirms the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B that was reported by Cicognani et al [16]. They showed that all boys who have increased FSH (> 6.1 IU/L) or LH (> 5.8 IU/L) concentrations also had inhibin B concentrations below 112 pg/mL (-2 standard deviations). Unlike Lähteenmäki et al [17], we found no correlation between testicular volume and inhibin B (P = 0.06). They reported that patients with small testes (< 10 mL) after treatment for childhood malignancy had inhibin B concentrations below 42 pg/mL, and 6 out of 7 had FSH concentrations above 9 IU/L; patients with a testicular volume above 13 mL had inhibin B concentrations above 100 pg/mL. This difference might be explained by variations in the age at which the boys in our study were given HCT and evaluated.\nWe find that the plasma AMH concentration does not seem to be a good marker, as it normally decreases when testosterone increases at puberty.\nWe did not carry out a sperm analysis because our boys were too young. However, the reported relationship between inhibin B and the sperm count in normal men and in boys surviving a childhood malignancy suggests that 10/38 (based on FSH or inhibin B) or 7/38 (based on both) of the boys we studied will be normospermic. Thomson et al [18] showed that inhibin B was barely detectable in azoospermic patients whose germ cells had been destroyed, despite preservation of their Sertoli cells, as confirmed by testicular biopsy. Van Beek et al [19] showed that inhibin B is a better marker of spermatogenesis than is the FSH concentration in men given chemotherapy for Hodgkin's lymphoma as children. Multivariate analysis with inhibin B and FSH concentrations as determinants of sperm concentration showed that inhibin B was the only significant determinant; all the men with a plasma inhibin B concentration above 75 pg/mL were normospermic. Lähteenmäki et al [20] showed that the inhibin B and FSH concentrations and the testicular volume explain 44% of the variance in the sperm count. Van Casteren et al [14] found that their 5 patients who were normospermic had inhibin B concentrations of 125-234 pg/mL, while the corresponding values in the 9 azoospermic patients was 0-79 pg/mL, with a correlation between sperm count and inhibin B concentrations.", "Their plasma FSH and LH concentrations indicate that 21% of our girls had normal ovarian function. These figures are similar to those reported by others [2,21,22]. However, the girls in our study with spontaneous puberty were no younger at HCT than were those with ovarian failure. Similarly, the FSH, inhibin B and AMH concentrations were not influenced by age or pubertal status at treatment. Fong et al [23] studied women treated for cancer during childhood and found that those with undetectable AMH were significantly older at diagnosis, only 2/17 had regular menstrual cycles, and more of them had been given body or abdomen radiotherapy than those with greater AMH concentrations. The preeminence of the role of the irradiation and of busulfan might partly explain why the age at HCT has no significant impact on ovarian function in our study.\nBusulfan seemed to be more gonadotoxic. Of the 8 girls given busulfan, combined with TBI in two, 7 had increased plasma FSH/LH concentrations and 7/8 had low inhibin B. This effect of busulfan was also seen by Teinturier et al [24], who showed that the 10 girls given busulfan (600 mg/m2) without TBI before autologous HCT all developed severe, persistent ovarian failure. They identified 26 of 29 girls in published reports given busulfan during prepuberty or puberty as having signs of ovarian failure.\nLarsen et al [25] used multiple linear regression analysis of 100 girls who had survived childhood malignancy to predict the total number of antral follicles per ovary. Ovarian irradiation, alkylating chemotherapy, older age at diagnosis and a long period off treatment all reduced the number of follicles. Among the 10 girls conditioned for HCT by TBI, the 3 with spontaneous cycles had smaller ovaries, fewer total follicles and higher FSH than the others.\nWe have therefore confirmed the excellent correlation and concordance between the plasma concentrations of FSH and inhibin B, and agreement with the clinical evolution, while those of AMH were very low. The 7 girls with normal puberty and FSH and LH all had AMH concentrations similar to those with ovarian failure. Three [23,26,27] studies have evaluated the capacity of plasma AMH concentrations to assess the ovarian damage in adults surviving childhood malignancy. Bath et al [26] compared 10 girls treated for cancer during childhood, all with regular menstrual cycles, with 11 controls. The treated girls had smaller ovaries than the controls and significantly higher FSH and lower AMH concentrations, while the other hormone concentrations were unchanged. Thus, the AMH concentration indicates a depletion of ovarian reserve, despite regular menstrual cycles. Van Beek et al [27] showed that girls treated with MOPP (mechlorethamine, oncovin, prednisone, procarbazine) for Hodgkin's lymphoma during childhood developed into women with elevated FSH concentrations and decreased AMH concentrations. Three of the women with normal FSH had decreased AMH.", "In boys, the concordance between plasma FSH and inhibin B concentrations suggests that measuring inhibin B may help in counselling at pubertal age. Plasma AMH concentrations do not seem to be a good marker as they normally decrease when testosterone increases at puberty. All or most of the boys given melphalan, when combined with TBI or TLI, will be azoospermic.\nIn girls, there is also a good concordance between plasma FSH and inhibin B concentrations. Both were normal in 6/34 girls who had normal puberty and regular menstruations, but their very low AMH concentrations suggest that there is a major loss of primordial follicles.\nOne limitation of this study is the range of drugs used. The fact that the drug effect could not be separated from that of irradiation significantly limits our ability to draw any conclusion about the effects of melphalan or busulfan.", "AMH: antimüllerian hormone; FSH: follicle-stimulating hormone, growth hormone - GH; Gy: Grays; HCT: hematopoietic cell transplantation; LH: luteinizing hormone; TLI: total lymphoid irradiation; TBI: total body irradiation", "The authors declare that they have no competing interests.", "SL and AC-CS participated in the design of the study, data acquisition and analysis. ST and SBT carried out the immunoassays. PL and CT performed the statistical analyses. CT prepared the tables and the figures. HE, JM, AB and AF followed the patients and participated in the design of the study. RB directed the work and prepared the manuscript. All authors read and approved the final manuscript.", "The pre-publication history for this paper can be accessed here:\n\nhttp://www.biomedcentral.com/1471-2431/11/20/prepub\n" ]

[ null, "methods", null, null, null, null, null, null, null, null, null, null, null, null, null, null ]

[]

[ "Anti-Bacterial Agents", "Antiviral Agents", "Bacteria", "Bacterial Infections", "Fruit", "Humans", "Influenza A Virus, H5N1 Subtype", "Influenza B virus", "Influenza, Human", "Orthomyxoviridae", "Phytotherapy", "Plant Extracts", "Respiratory Tract Diseases", "Sambucus nigra" ]